Home

Technical Documentation

1. rx_array Handles of array objects for each Rx See Section 2 2 2 track Handles of track objects for each Rx See Section 2 2 3 pairing An index list of links for which channel are created The first row corresponds to the Tx and the second row to the Rx no_links Number of links for which channel coefficients are created read only Methods h_layout layout simpar Description Creates a new layout object Input simpar Handle of a simulation_parameters object See Section 2 2 1 h_parset h_cb create_parameter_sets initialize check_parfiles Description Creates parameter_set objects based on layout specification This function processes the data in the layout object First all tracks in the layout are split into subtracks Each subtrack corresponds to one segment Then then scenario names are parsed A parameter_set object is created for each scenario and for each transmitter For example if there are two terrestrial BSs each having urban LOS and NLOS users then 4 parameter_set objects will be created BS1 LOS BS2 NLOS BS2 LOS and BS2 NLOS The segments are then assigned to the parameter_set objects In the last step the parameter maps are created see Section 3 1 This can be disabled by setting initialize 0 Input initialize Enables 1 default or disables 0 the generation of the parameter maps If you want to adjust the parameters first use
2. 300 400 Track m Position dependant delay spread Delay Spread ns 0 100 200 300 400 500 600 Track m Figure 20 Results for the satellite channel tutorial Copyright Fraunhofer Heinrich Hertz Institute 90 eMail quadriga hbhi fraunhofer de QuaDRiGa v1 2 32 458 A TUTORIALS A 4 Drifting Phases and Delays Drifting is an essential feature of the channel model Drifting enables a continuous time evolution of the path delays the path phases the departure and arrival angles and the LSPs It is thus the enabling feature for time continuous channel simulations Although drifting was already available in the SCME branch of the WINNER channel model it did not make it into the main branch Thus drifting is not available in the WIM1 WIM2 or WIM model Here the functionality is implemented again This script focuses on the delay and the phase component of the drifting functionality Channel model setup and coefficient generation First we parameterize the channel model We start with the basic simulation parameters For the desired output we need two additional options we want to evaluate absolute delays and we need to get all 20 subpaths Normally the subpaths are added already in the channel builder s simulation_parameters s center_frequency 2 53e9 s sample_density 4 s use_subpath_output 1 s use_absolute_delays 1 Second we define a user track Her
3. Mhn Those are also drawn from a normal distribution as in 75 In order to fulfill all three we can combine two rotations one for the vertical and one for the horizontal component with a scaling operation We convert XPR and XPR to rotation angles ale and ql using 78 and calculate Meee to lo h pl a cos y tan y cosy m M lm l m Lm CPRim 79 l m siti v v 1 Tim COSY m JCPRim Due to the scaling with 1 V CPR MI will scale the power of the path Hence a normalization is needed to counteract this effect a linear _ V2 l m ve ll 3 Stochastic part for the elliptical component Elliptic polarization is obtained when there is a phase difference between the horizontal and the vertical component This is included by a scaling matrix elliptic __ exp JxKi m 0 Mi m 0 exp JKi m ep The phase shift is calculated to Kim X l m arccot V XPRim 82 where Xjm 1 1 is the positive or negative sign Copyright Fraunhofer Heinrich Hertz Institute 71 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 3 TECHNICAL DOCUMENTATION The polarization for the NLOS paths is a combination of five linear transformations First any change in the transmitter orientation is included by a rotation matrix M Jt1m s Then the influence of the scattering cluster is modeled by a combination of three operations A scaling operation that introduces a phase shi
4. gt 5 72 This plot shows the power of the first NLOS tap along the track The fading is significantly higher in the beginning and becomes much less strong towards the end phase unwrap angle squeeze d coeff 1 1 2 plot distance phase xlabel Distance from track start point ylabel Continuous phase 200 7 r j 7 7 5 10fF 0 15 2 g 200p A of S fal 3 25 3 5 400 z 30 S 5 a E 35f 600F 6 40f 45 800 50f 1000 fi fi fi fi 55 Ll fi 1 20 25 30 35 40 45 50 20 25 30 35 40 45 50 Distance from track start point Distance from track start point Figure 23 Drifting phases and Tx power vs Rx position drifting phases tutorial Without drifting the phases of the subpaths are approximated by assuming that the angles to the LBSs do not change However this only holds when the distance to the LBS is large Here the initial distance is small ca 5 m When the initial angles are kept fixed along the track the error is significant Here the phase ramp is negative indicating a movement direction towards the scatterer and thus a higher Doppler frequency However when the scatterer is passed the Rx moves away from the scatterer and the Doppler frequency becomes lower This is not reflected when drifting is turned off Note that with shorter delay spreads as e g in satellite channels the s
5. l track 1 generate circular 20 pi 0 4 Circular track with 10m radius 1l track 1 initial_position 10 0 1 5 4 Start east running nord 1 track 1 segment_index 1 40 90 h Segments 1 track 1 scenario UMal UMan UMal l track 2 generate linear 20 0 4 Linear track 20 m length 1 track 2 initial_position 10 0 1 5 Same start point 1 track 2 interpolate_positions 128 20 l track 2 segment_index 1 40 90 1 track 2 scenario UMal UMal UMan 1l visualize A Plot all tracks l track interpolate_positions s samples_per_meter 1 track compute_directions Now we create the channel coefficients Fixing the random seed guarantees repeatable results i e the taps will be at the same positions for both runs Note that the computing time is significantly longer when drifting is enabled Copyright Fraunhofer Heinrich Hertz Institute 95 eMail quadriga hhi fraunhofer de a PF OND 16 QuaDRiGa v1 2 32 458 A TUTORIALS Tx Position 207 q A Tx Antenna O Rx Position V_ Rx Antenna 15 Rx Track wn L BERLIN UMa NLOS BERLIN UMa LOS pad AFx1 J BERHNOUMB ESN O N UMa NLOS Y Position gt T 5b BERLIN UMa_LOS f 4 A I i i fi i 10 5 0 5 10 15 20 25 30 X Position Figure 25 Scenario setup for the time evolution tutorial RandStream setGlobalStream RandStream mti9937ar seed 2 p l creat
6. hold off axis 25 25 25 25 J legend Equal 3dB 4 xlabel SF_P dB ylabel SF_C dB title Shadow Fading Requested vs generated value Shadow Fading Requested vs generated value K Factor Requested vs generated value 25 i i r 30 l 20t Z ee 15 PAo aof Zee 10 poe pie ar ed 10 Pe J 5 ae ea 7 n m Ed Pag m ar Pj Z ol SA I B l RD A r a 2 EW 5t Pee poe 10 Eo a al ae ee 15 ae ey a Ba we 20t 2 20tl 7 Equal ae GE Equal ee 34B Ce 3dB 325 i i i i Ag l i i 20 1 0 10 20 230 20 z 0 20 30 SF dB KF dB P P Delay Spread Requested vs generated value 6 Delay Spread difference vs K factor 15 l T 7 7 i 7 7 d Lo 7 a 4p 1 I e ieee J me 4 1 ee eas 3 E ERA g v ae dea A n k a l A KO A 1 a a 0 5 E i 2f g P 4t q fo Equal Equal 10 Error 3dB f T T f Ll i 9 0 5 1 15 30 20 10 0 10 20 30 DS us KF dB P Figure 14 Comparison of input values and simulation results Copyright Fraunhofer Heinrich Hertz Institute 77 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 A TUTORIALS Next we repeat the same calculation for the K Factor Again we see that the values are almost identical 1 p_nlos sum mean abs coeff 2 end 2 3 2
7. A list of supported scenarios can be obtained by calling parameter_set supported_scenarios The scenario parameters are stored in the configuration folder config in the QuaDRiGa main folder The filenames e g MIMOSA_16 25_LOS conf also serves as scenario name probability The probability for which the scenario occurs This parameter must be a vector of the same length as there are scenarios Probabilities must be specified in between 0 and 1 The sum of the probabilities must be 1 By default or when probability is set to each scenario is equally likely seg_length_min seg_length_mu seg_length_std the minimal segment length in m The default is 10 m the median segment length in m The default is 30 m the standard deviation of the street length in m The default is 12 m set_speed speed Description Sets a constant speed in m s for the entire track This function fills the track movement_profile field with a constant speed value This helps to reduce computational overhead since it is possible to reduce the computation time by interpolating the channel coefficients Input speed The terminal speed in m s split_segment mi ma mu sig no_check Description Splits long segments in subsegments of the same type Input mi Minimum length of the subsegment in m default 10m ma Maximum length of the subsegment in m must be
8. Add folder button to add QuaDRiGa to your MATLAB Path Table 1 QuaDRiGa System Requirements Requirement Value Minimal required MATLAB version 7 12 R2011a Required toolboxes none Memory RAM requirement 1 GB Processing power 1 GHz Single Core Storage 50 MB Operating System Linux Windows Mac OS 1 2 General Remarks This document gives a detailed overview of the QuaDRiGa channel model and its implementation details The model has been evolved from the Wireless World Initiative for New Radio WINNER channel model de scribed in WINNER II deliverable D1 1 2 v 1 1 3 This document covers only the model itself Measurement campaigns covering the extraction of suitable parameters can be found in the WINNER documentation 3 4 or other publications such as 5 6 Furthermore the MIMOSA project 1 covers the model development and parameter extraction for land mobile satellite channels The QuaDRiGa channel model follows a geometry based stochastic channel modeling approach which allows the creation of an arbitrary double directional radio channel The channel model is antenna independent i e different antenna configurations and different element patterns can be inserted The channel parame ters are determined stochastically based on statistical distributions extracted from channel measurements The distributions are defined for e g delay spread delay values angle spread shadow fading and cross polarization
9. This class combines all functions to create and edit antenna arrays An antenna array is a set of single antenna elements each having a specific beam pattern that can be combined in any geometric arrangement A set of synthetic arrays that allow simulations without providing your own antenna patterns is provided see generate method for more details Properties name Name of the antenna array interpolation_ Method for interpolating the beam patterns method The default is linear interpolation Optional are e nearest Nearest neighbor interpolation QuaDRiGa optimized e linear Linear interpolation QuaDRiGa optimized Default e spline Cubic spline interpolation MATLAB internal function e nearest_int Nearest neighbor interpolation MATLAB internal function e linear_int Linear interpolation MATLAB internal function Note MATLAB internal routines slow down the simulations significantly polarization_basis The polarization basis of the pattern The polarization basis of the pattern cartesian Ludwig 1 az el Ludwig 2 Azimuth over Elevation el az Ludwig 2 Elevation over Azimuth polar spheric Ludwig 2 Polar Spheric DEFAULT You can specify the polarization basis of the pattern by setting the appropriate string By default QuaDRiGa requires a Polar Spheric basis If a different basis is specified an appropriate transforma tion will be carried out no_elements Number of antenna elements in the array
10. channel object Copyright Fraunhofer Heinrich Hertz Institute 46 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 2 SOFTWARE STRUCTURE freq_response fr bandwidth carriers isnapshot Description Input Transforms the channel into frequency domain and returns the frequency response bandwidth carriers i snapshot The baseband bandwidth in Hz The carrier positions There are two options 1 Specify the total number of carriers In this case carriers a scalar natural number gt 0 The carriers are then equally spaced over the bandwidth 2 Specify the pilot positions In this case carriers is a vector of carrier positions The carrier positions are given relative to the bandwidth where 0 is the begin of the spectrum and 1 is the end For example if a 5 MHz channel should be sampled at 0 2 5 and 5 MHz then carriers must be set to 0 0 5 1 The snapshot numbers for which the frequency response should be calculated By default i e if i_snapshot is not given all snapshots are processed Output freq_response The complex valued channel coefficients for each carrier in frequency domain The indices of the 4 D tensor are Rx Antenna Tx Antenna Carrier Index Snapshot c interpolate dist method Description Interpolates the channel coefficients and delays The channel builder creates o
11. 11 Change of environment Urban Forest This is the same as in point 2 The segment on the track gets assigned the scenario Satellite Forest and a third set of maps 15 21 is generated for the Satellite Forest segment The parameters are drawn from those maps new channel coefficients are calculated and the powers of the clusters are ramped up down 12 Turning off without change of environment NLOS Same as in point 4 Copyright Fraunhofer Heinrich Hertz Institute 19 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 2 SOFTWARE STRUCTURE 2 Software Structure 2 1 Overview QuaDRiGa is implemented in MATLAB using an object oriented framework The user interface is built upon classes which can be manipulated by the user Each class contains fields to store data and methods to manipulate the data An overview of the class structure is given in Section 2 It is important to keep in mind that all classes in QuaDRiGa are handle classes This significantly reduces memory usage and speeds up the calculations However all MATLAB variable names assigned to QuaDRiGa objects are pointers If you copy a variable i e by assigning b a only the pointer is copied a and b point to the same object in memory If you change the values of b the value of a is changed as well This is somewhat different to the typical MATLAB behavior and might cause errors if not considered properly Copyi
12. Olm flm at the transmitter Tx gets coupled with a random angle pair 6 1m at the receiver Rx see 3 Copyright Fraunhofer Heinrich Hertz Institute 63 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 3 TECHNICAL DOCUMENTATION Table 11 Offset Angle of the m Sub Path from 3 Sub path Offset angle Sub path Offset angle m m degrees m m degrees 1 2 0 0447 11 12 0 6797 3 4 0 1413 13 14 0 8844 5 6 0 2492 15 16 1 1481 7 8 0 3715 17 18 1 5195 9 10 0 5129 19 20 2 1551 3 4 Drifting After the path delays powers and angles are known for the initial position their values are updated for each snapshot of the segment Thus we get an evolution of the path parameters over a short time interval Drifting for 2 D propagation was already introduced in an extension of the SCM 8 However it was not incorporated into the WINNER model and no evaluation was reported Here this idea is extended towards 3 D propagation to incorporate time evolution into the new model Besides the initial delays path powers and angles drifting requires the exact position of each antenna element At the MT element positions need to be updated for each snapshot with respect to the MT orientation The following calculations are then done element wise The indices r t 1 m s denote the index of the Rx antenna element r and the Tx antenna element t the path number l the sub path numbe
13. a significant amount of energy comes from a single direction Thus the AS gets smaller which leads to a negative correlation between the DS and the KF e Link level A user terminal at a specific position black dot on the map in Figure 6 is assigned to a propagation scenario Depending on the position and the scenario it experiences a radio channel which is deter mined by the specific values of the seven LSPs Due to the autocorrelation properties small distances between users in the same scenario also lead to high correlations in the channel statistics e g a second terminal right next to the current user will experience a similar DS The second granularity level thus contains the specific values of the LSPs for each user position Generating those values can be seen as going backwards from the scenario wide distribution u of a LSP to individual measurement values for each MT e Path Level Last the individual components of the CIR are calculated This procedure takes the values of the LSPs into account and calculates the power and the delay of the channel coefficients The detailed procedure for this is described in the following sections The correlation maps are generated at a fixed sampling grid by successively filtering a random normal distributed sequence of numbers with a finite impulse response FIR filter The principle is depicted in Figure 7 The map is represented by a matrix B and one pixel of that matrix is Bys whe
14. x 2 a coupling 1i sqrt 2 1 1 1j 1j b rotate_pattern 90 x 2 b coupling 1 sqrt 2 1 13 1j 1j Place arrays in layout We place two of those arrays in a layout The scenario LOSonly has no NLOS scattering One can see this setup as a perfect anechoic chamber 1 layout 1 simpar show_progress_bars 0 1 simpar drifting_precision 0 rx_position 11 0 0 track no_snapshots 1 track ground_direction pi track scenario LOSonly sta array a rx_array b PRRPRPHH p l create_parameter_sets cb p get_channels cb pin zeros size cb pin Get array response We now sample the array response for each degree in the antenna array pat zeros a no_el a no_az 2 2 values a no_az fprintf Calculating mO 0 tStart clock 4 A Status message for n 1l a no_az mi ceil n values 50 if mi gt mO for m2 1 m1i m0 fprintf o end mO m1 end al a copy_objects al rotate_pattern a azimuth_grid n 180 pi z for m 1 a no_el a2 al copy_objects a2 rotate_pattern a elevation_grid m 180 pi y cb tx_array a2 c cb get_channels pat m n c coeff end end fprintf 5 0f seconds n round etime clock tStart Copyright Fraunhofer Heinrich Hertz Institute 108 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 A TUTORIALS 1 Calculating 000000000
15. Fraunhofer Heinrich Hertz Institute eMail quadriga hhi fraunhofer de 21 QuaDRiGa v1 2 32 458 2 SOFTWARE STRUCTURE coupling generate_multi no_elements spacing tilt optimize Description Generates a multi element array with electric tilt This function generates a vertically stacked multi element array of the given antenna object The element spacing is relative to the wavelength and an additional electric tilt can be applied The output is stored in the current antenna object For this method to work you need to define a single antenna element using array generate Then you can call array generate_multi to transform this element into a stacked multi element array The provided antenna object can only have one element The method returns an error if a multi element array is given Input no _elements The number of virtual antenna elements stacked in elevation z direction spacing The element spacing as a factor of the wavelength Default value 0 5 tilt An additional electric downtilt value in deg Default value 0 optimize If this parameter is set to 1 the optimal beamformer is calculated Otherwise the phases are calculated using geometric settings Default 0 geometric Output coupling The coupling matrix used to calculate the virtual antenna pattern h_array import_pattern fVi fHi azimuth_grid elevation_grid Description Converts antenna field patte
16. In this case it is 20 c p get_channels cn c merge 0 2 Generate coefficients Combine segments 21 00000000000000000000000000000000000000000000000000 1 Channels 00000000000000000000000000000000000000000000000000 Merging seconds seconds Evaluation of the data The next two plots show some basic evaluations of the generated coefficients The first plot shows the received power for the 4 MIMO links along the track between the LOS and NLOS segments and the cross pol discrimination between the MIMO links average path loss for LOS was set to 95 dB and for NLOS 113 dB The plot shows the differences The dist ind los Es for n 1 numel ind start t segment_index ind n if n numel ind try stop catch stop end else stop end los 1 cn no_snap t get_length cn no_snap find strcmp t scenario MIMOSA_10 45_L0S t segment_index ind n 1 t no_snapshots t segment_index ind n 1 los start stop end power reshape 10 log10 squeeze sum abs cn coeff 2 3 4 mi ma min reshape power 1 max reshape power 1 ar ones 1 cn no_snap ma ar los mi figure Position 100 a area dist ar set a i FaceColor 0 7 0 9 0 set a LineStyle none 100 1000 700 7 hold on plot dist power hold off 1 Copyright Fraunhofer Heinrich Hertz Institu
17. Increasing the number of elements creates new elements which are initialized as copies of the first element Decreasing the number of elements deletes the last elements from the array elevation_grid Elevation angles in rad were samples of the field patterns are provided The field patterns are given in spherical coordinates This variable provides the elevation sampling angles in radians ranging from 4 downwards to upwards azimuth_grid Azimuth angles in rad were samples of the field patterns are provided The field patterns are given in spherical coordinates This variable provides the azimuth sampling angles in radians ranging from 7 to m element_position Position of the antenna elements in local cartesian coordinates using units of m Fa The first component of the antenna pattern If the polar spheric polarization basis is used this variable contains the vertical or theta component of the electric field given in spherical coordinates This variable is a tensor with dimensions elevation azimuth element describing the theta component of the far field of each antenna element in the array Fb The second component of the antenna pattern If the polar spheric polarization basis is used this variable contains the horizontal or phi component of the electric field given in spherical coordinates This variable is a tensor with dimensions elevation azimuth element describing the phi component of th
18. KF sigma dB KF lambda meter Copyright Fraunhofer Heinrich Hertz Institute eMail quadriga hhi fraunhofer de 52 QuaDRiGa v1 2 32 458 2 SOFTWARE STRUCTURE xpr_mu dB The XPR is defined by the XPR Antenna see antenna pattern and the XPR xpr_sigma dB environment The parameters describe the statistical properties of the XPR environment For the XPR no correlation map is calculated and the XPR is updated once per segment Note For the LOS component no XPR environment is assumed only the XPR antenna is applied Hence the overall XPR depends also highly on the K factor DS_mu log10 s Statistical properties of the delay spread DS _sigma log10 s DS_lambda meter AS_A_mu log10 deg Statistical properties of the azimuth of arrival spread at the receiver AS_A_sigma log10 deg AS_A_lambda meter ES_A_mu log10 deg Statistical properties of the elevation of arrival spread at the receiver ES_A sigma log10 deg ES_A_lambda meter AS_D_mu log10 deg Statistical properties of the azimuth of departure spread at the transmitter AS_D sigma log10 deg AS_D_lambda meter ES_D_mu log10 deg Statistical properties of the elevation of departure spread at the transmitter ES_D _sigma log10 deg ES_D_lambda meter Cross correlations There are interdependencies between parameters For example if the K Factor is high the delay spread gets shorter since more power
19. Same as quick but the output contains the complex valued am plitude instead of the power sample_distance Distance between sample points in m default 10 m x_min x coordinate in m of the top left corner y max y coordinate in m of the top left corner x_max x coordinate in m of the bottom right corner y min y coordinate in m of the bottom right corner tx_power A vector of tx powers in dBm for each transmitter in the layout This power is applied to each transmit antenna in the tx antenna array By default if tx_power is not given 0 dBm are assumed rx_height Height of the receiver points in m default 0 m Output map A cell array containing the power map for each tx array in the layout The power maps are given in W and have the dimensions n_y_coords n_x_coords n_rx_elements n_tx_elements x_coords Vector with the x coordinates of the map in m y coords Vector with the y coordinates of the map in m randomize_rx_positions max_dist min_height max_height track_length Description Generates random Rx positions and tracks Places the users in the layout at random positions Each user will be assigned a linear track with random direction The random height of the user terminal will be in between min_height and max_height Input max_dist the maximum distance from the layout center in m Default is 50 m min_height the minimum user height in
20. and so on This is independent of the order in layout tx_name which might have a different order e Outputs are alphabetically sorted Input varargin A list of cell arrays containing the transmit antenna indices Output chan_out The split channel objects Copyright Fraunhofer Heinrich Hertz Institute 48 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 2 SOFTWARE STRUCTURE 2 3 Data Flow The data flow of the QuaDRiGa channel model is depicted in Fig 4 This figure shows how each of the processing steps which are described in detail in the following sections are linked together The lines show which parameters are exchanged and how often they are updated Black lines are for parameters that are either provided by the model users or which are given in the parameter table Those values are constant Blue values are updated once per segment and red values are updated once per snapshot User Input Variables Terminal trajectories Network layout Propagation scenarios Transmitter positions Antenna Parameters Speed profile Carrier frequency Trajectories 7 Scenarios Transmitter Positions Split terminal trajectories into segments Snapshot position Antenna patterns and array geometries F LOS direction Los Scenarios Snapshot position t Path Loss Parameters XPRu XPRo No clusters L Parameter Table Cluster wise azimuth spread Caoa Calculate path loss fo
21. eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 3 TECHNICAL DOCUMENTATION In order to account for the inter LSP correlation a matrix X is assembled containing all cross correlation values p between each two LSPs 1 PDS KF PDS SF PDS ASD PDS ASA PDS ESD PDS ESA PKF DS 1 PKF SF PKF ASD PKF ASA PKF ESD PKF ESA PSF DS PSF KF 1 PSF ASD PSF ASA PSF ESD PSF ESA X PASD DS PASD KF PASD SF 1 PASD ASA PASD ESD PASD ESA 14 PASA DS PASA KF PASA SF PASA ASD 1 PASA ESD PASA ESA PESD DS PESD KF PESD SF PESD ASD PESD ASA 1 PESD ESA PESA DS PESA KF PESA SF PESA ASD PESA ASA PESA ESD 1 The matrix square root X 2 of this matrix is multiplied with the values of the seven LSP maps Bps Besp In order to calculate the matrix square root X must be positive definite to get a unique real numbered solution 5 By z DS y xz DS x 15 By x EsD Bo ap Last the MTs are placed on the maps and the corresponding values for the LSPs are obtained by interpo lating the surrounding pixels of the map In this way initial LSPs for the following parts of the channel model are generated 3 2 Initial Delays and Path Powers Initial delays are drawn randomly from a scenario dependent delay distribution as r r o ln X 16 where X U 0 1 is an uniformly distributed random variable having values between 0 and 1 is the initial DS from the map and r is a proportionality factor see 3 The term r was introduced in 15 be
22. gt 2 mi default 30m mu Mean length of the subsegment mi lt mu lt ma default 15m sig Std of the length of the subsegment default 5m no_check Disable parsing of input variables default false visualize Description Plots the track Copyright Fraunhofer Heinrich Hertz Institute 35 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 2 SOFTWARE STRUCTURE 2 2 4 Class layout Objects of this class define the network layout of a simulation run Each network layout has one or more transmitters and one or more receivers Each transmitter and each receiver need to be equipped with an antenna array which is defined by the array class In general we assume that the transmitter is at a fixed position and the receiver is mobile Thus each receivers movement is described by a track Properties name Name of the layout simpar Handle of a simulation_parameters object See Section 2 2 1 no_tx Number of transmitters or base stations no rx Number of receivers or mobile terminals tx_name Identifier of each Tx must be unique tx_position tx_array rx_name rx_position Position of each Tx in global cartesian coordinates using units of m Handles of array objects for each Tx See Section 2 2 2 Identifier of each Tx must be unique Initial position of each Rx relative to track start in global cartesian coordinates using units of m
23. mus axis 0 20 0 1 grid on Simulated Delay Spread 1 r r r 0 i i 0 5 10 15 20 Distance from start point m Figure 37 DS along the track manual parameter selection close all disp QuaDRiGa Version simulation_parameters version QuaDRiGa Version 1 0 1 145 Copyright Fraunhofer Heinrich Hertz Institute 113 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 B DEPARTURE AND ARRIVAL ANGLES ADOPTED WINNER METHOD B Departure and Arrival Angles Adopted WINNER Method In the WINNER model the azimuth arrival and departure angels are modeled using a wrapped Gaussian distribution see 3 page 39 1 P i a 94 O 94 The wrapping is applied later by 97 when the discrete cluster angles are drawn from the statistics Since the above formula assumes a continuous spectrum whereas the channel model uses discrete paths we need to correct the variance by a function C L K This function ensures that the input variance og is correctly reflected in the generated angles The same approach was taken by the WINNER model However 3 does not explain how the correction values were obtained Generation of azimuth and elevation angles The individual angles are obtained by first normalizing the power angular spectrum so that its maximum has unit power We can thus omit the scaling factor 1 0gV2m The path powers P 21 are also normalized such th
24. rad AoA The initial azimuth of arrival angles for each path in rad EoD The initial elevation of departure angles for each path in rad EoA The initial elevation of departure angles for each path in rad xpr The initial cross polarization power ratio in dB for each sub path The dimensions correspond to the MT the path number and the sub path number pin The initial phases in rad for each sub path kappa The phase offset angle for the circular XPR in rad The dimensions correspond to the MT the path number and the sub path number random_pol Random phasors for the WINNER polarization coupling method The dimensions correspond to polarization matrix index 1 3 2 4 the subpath number and the MT subpath_coupling A random index list for the mutual coupling of subpaths at the Tx and Rx The dimensions correspond to the subpath index 1 20 the angle AoD AoA EoD EoA the path number and the MT Methods h_cb channel_builder h_parset Description Creates a new channel_builder object Input h_parset A parameter_set object Output h_cb A channel_builder object Copyright Fraunhofer Heinrich Hertz Institute 44 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 2 SOFTWARE STRUCTURE phi_d_tlm theta_d_lm lbs_pos fbs_pos calc_scatter_positions Description Calculates the positions of the scatterers This function calculates the positions of the scatterers and initia
25. ylabel Probability legend LOS NLOS 1 close all disp QuaDRiGa Version gt simulation_parameters version QuaDRiGa Version 1 0 1 145 Position dependent power Empirical PDF of the LOS and NLOS power r r 7 r 10 T T T T T T r r 85r 7 Los oF ME NLOs 90 8F J 95 7l 100 Sal 5 D z 105 z 5 110 4l J 115 3H J 120 2p 7 a i l oH i i i i 0 100 200 300 400 500 120 115 110 105 100 95 90 85 Track m P oil dB Position dependant delay spread Empirical PDF of the LOS and NLOS RMSDS 2 57 LOS 12l Los S E NLOS 2 2 S 3 1 5 2 amp 3 gt 6 o l a a 0 5 0 5 0 0 5 1 1 5 0 100 200 300 400 500 o us i Track m rth Figure 17 Results for the measurement based simulation tutorial Copyright Fraunhofer Heinrich Hertz Institute eMail quadriga hhi fraunhofer de 82 wn wn 10 11 12 13 14 15 16 19 QuaDRiGa v1 2 32 458 A TUTORIALS A 3 Generation of Satellite Channels This script demonstrates the parametrization of the channel model to generate time continuous sequences for a satellite scenario Setting up the Simulation Parameters First we set up the general simulation parameters We choose a center frequency of 2 1 GHz We also want to use drifting in order to get the correct delays and angles for the ti
26. 1 name p 2 name MIMOSA 10 45 LOS_Tx1 Select good state p S1 scenpar 4 Show parameter list ans NumClusters 8 r_DS 2 5000 PerClusterAS_D 6 2000e 07 PerClusterAS_A 12 PerClusterES_D 1 9000e 07 PerClusterES_A 7 LOS_scatter_radius 0 1000 LNS_ksi 3 xpr_mu 11 9000 xpr_sigma 5 5000 Copyright Fraunhofer Heinrich Hertz Institute 86 eMail quadriga hhi fraunhofer de yee Bee FOOMA NIAAA e wh Ns 00 NNN NN NNN WD h o i 39 won i QuaDRiGa v1 2 32 458 A TUTORIALS DS_mu 7 5000 DS_sigma 0 3000 AS_D_mu 4 6000 AS_D_sigma 0 1000 AS_A_mu 5000 AS_A_sigma 0 2000 ES_D_mu 5 1200 ES_D_sigma 0 1000 ES_A_mu 4000 ES_A_sigma 0 1000 F_sigma 3 6000 KF_mu 5 5000 KF_sigma 5 9000 DS_lambda 30 5000 AS_D_lambda 000 AS_A_lambda 31 5000 ES_D_lambda 000 wn ES_A_lambda 6 SF_lambda 35 KF_lambda 4 5000 asD_ds 0 asA_ds 0 6100 asA_sf 0 5600 asD_sf 0 ds_sf 0 4300 asD_asA 0 asD_kf 0 asA_kf 0 4400 ds_kf 0 4600 sf_k 0 3000 esD_ds 0 esA_ds 0 0500 esA_sf 0 1800 esD_sf 0 esD_esA 0 esD_asD 0 esD_asA 0 esA_asD 0 esA_asA 0 1500 esD_kf 0 esA_kf 0 0300 Note that the values are given for a log normal distribution Thus the RMSDS in nanoseconds follows from 10 p S1 scenpar DS_mu 1e9 ans 31 6228 Each parameter on that list can be changed by just assigning it a new value Here we
27. 7 mod an T 25 The relationship between path powers and angles is random Hence the resulting AS is undefined In the next step the actual AS is calculated This requires to calculate the power weighted mean angle This angle is subtracted from the angles gl and the wrapping around the unit circle is applied a second time The AS then follows from o are gt P exp iP 26 i gt oP o n mod 2r T 27 L 2 a YR Coi _ 7 P a 28 l 1 l 1 Copyright Fraunhofer Heinrich Hertz Institute 61 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 3 TECHNICAL DOCUMENTATION Now with oy being the initial AS from the map the angles gl are updated 3 Og 2 actual 29 o If og is larger than ae then 25 needs to be applied again in order for account for the periodicity of the angles However in this case some of the angles could be mapped to values around 0 This could lower the AS instead of increasing it as intended by the scaling operation A solution is to create new angles with a bias to the negative side of the circle 4 l 3 if 3 _ aa if lt 7 30 N n otherwise Pl fl a mod 2r T 31 This however changes the AS and the calculations 26 to 31 need to be repeated iteratively until the actual actual AS T applied converges either to the given value og or a maximum value Finally the LOS direction is pi oF
28. Output V The interpolated vertical field pattern H The interpolated horizontal field pattern CP The interpolated common phase field pattern dist The effective distances between the antenna elements when seen from the direction of the incident path The distance is calculated by an projection of the array positions on the normale plane of the incident path cp rotate_pattern deg rotaxis element usage Description Rotates antenna patterns Pattern rotation provides the option to assemble antenna arrays out of single elements By setting the element position property of an array object elements can be placed at different coordinates In order to freely design arbitrary array configurations however elements often need to be rotated e g to assemble a 45 crosspolarized array out of single dipoles This functionality is provided here Note Calling rotate_pattern will always remove the common phase from the field patterns Call estimate_common_phase before calling rotate_pattern to extract the common phase information Input deg The rotation angle in degrees ranging from 180 to 180 rotaxis The rotation axis specified by the character x y or z element The element numbers for which this interpolation is done is applied If no element number is given the interpolation is done for all elements in the array usage The optional parameter usage can limit the rota
29. a similarly strong path with a large delay ramps up Hence the DS can fluctuate randomly within the overlapping region To balance this out paths from both segments are paired in a way that minimize these fluctuations This is done by determining the values of the DS before and after the transition Then a target DS is calculated for each sub interval For example if the old segment yields a DS of 200 ns and the new segment has Copyright Fraunhofer Heinrich Hertz Institute 73 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 3 TECHNICAL DOCUMENTATION Transitions between Segments overlapping part merging area initial pos fo variable length segment 2 ge ee Variable Speeds initial pos segment 1 original snapshots m ow constant distance r xO a TA interpolated snapshot ae constant samplerate Figure 12 Top Illustration of the overlapping area used for calculating the transitions between segments step G Bottom Illustration of the interpolation to to obtain variable MT speeds step H 400 ns then the target DS will be 220 ns for the first sub interval 240 ns for the second and so on Then a combination of paths is searched that best matches the target DS for each sub interval 3 9 Postprocessing Variable Speeds In the real world MTs move at arbitrary speeds including accelerations and decelerations Provided that the sampling theorem is fulfilled we can interpolate th
30. channel does not change significantly from segment to segment we need to include correlation This is done by so called parameter maps see Section3 1 The maps ensure that neighboring segments do not have significantly different propagation characteristics For example measurements show that the shadow fading the average signal attenuation due to building trees etc is correlated over up to 100 m Hence we call all channel characteristics showing similarly slow changes LSPs With a segment length of 20 m two neighboring segments of the same state will have similar receive power To get the correct correlation QuaDRiGa calculates a map for the average received power for a large area The received power for two adjacent segments is then obtained by reading the values of the map This map based approach also contains cross correlations to other LSPs such as the delay spread For example a shorter delay spread might result in a higher received power Hence there is a positive correlation between power and delays spread which is also reflected in the maps To get a continuous time series of channel coefficients requires that the paths from different segments are combined at the output of the model In between two segments clusters from the old segment disappear Copyright Fraunhofer Heinrich Hertz Institute 14 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 1 INTRODUCTION AND OVERVIEW and new cluster appear This is model
31. cluster AS_A PerCluster deg The elevation angular spread of the 20 sub paths within one cluster ES_A LOS scatter_radius meter This parameter allows an additional spread of the 20 sub paths of the LOS compo nent by emulating scattering in the near field of the antennas EXPERIMENTAL LNS_ksi Normally cluster powers are taken from an exponential power delay profile This parameter enables an additional variation of the individual cluster powers around the PDP r_DS This parameter allows the mapping of delay spreads to delays and powers for the clusters See section 3 2 Copyright Fraunhofer Heinrich Hertz Institute eMail quadriga hhi fraunhofer de 53 to rar won o e H OOOO wwwwnsn nn nnnnnn 3 3 oS won e e e e e be wWwWwwwnwnw w ak wWONrFOOCMAN AA A QuaDRiGa v1 2 32 458 2 SOFTWARE STRUCTURE 2 4 2 Adding New Scenarios The scenario parameters are set in parameter_set scenpar Here you also have the option to change individual parameters by assigning new values The scenario UMal for example uses by default 8 clusters When the simulations should be done with 15 clusters one can change the settings by parameter_set BERLIN_UMa_LOS 15 4 New parameter set 4 Set new number of clusters p p scenpar NumClusters A list of currently supported scenarios is generated by parameter_set supported_scenarios The default settings
32. coefficient index ooa aa ee 59 KB Ricean K Factor logarithmic scale ooo a 115 L Number of paths 24 sa a soc tesa 58 awaa o RE ERR ppe 60 61 114 115 l Path index l Sq lO ving Dp ooa ae ee eh e 60 61 64 m Sub path index m 1 2 ee 20 oo aaa SRA Re ee 63 64 M Polarization coupling matrix ooa aa ee 66 N Normal distribution N u o with mean p and STD o 60 63 71 72 114 P POWGT o e pon ea a ae BR Bae be Re a e Ae we ee BS 60 61 72 114 115 r Receive antenna index r 1 2 a np o he aa bode ede eee dee bd ee ee eden 64 R Rotation matrix 2 a p ae a i a ok e ae i a e E a 69 r Vector pointing from the Tx position to the Rx position 64 65 70 fr Proportionality factor to trade between delays and path powers ooo ooa a a a 60 S Number of snapshots o socio got 24 5 sara a da doaia Da bopa aoe a a we 72 s Snapshot index s 1 2 9P cae eee ee Rae eR eee Rs ee ek wk as 64 t Transmit antenna index t 1 2 fig o oo eed ea ew Re ee ee 64 u Continuous uniform distribution U a b with minimum a and maximumb 60 66 v Speed in M S Gok we a h ee he EVRY ESE E Se ee eRe 56 X A random variable s se s S osere me o a e a ee 60 71 114 X Matrix containing the inter parameter correlation values aooo oa a e a a 60 Y A random variable se ms cw p goe a ka doi s kaoi GVE E E podar e a ee ee aa a 114 Z A random variable sos 8 26 4 eso fe ee ee ee ae eee ee
33. default all subtracks are returned Output subtracks A vector of track objects corresponding to the number of segments dist interpolate movement si method Description Interpolates the movement profile to a distance vector This function interpolates the movement profile The distance vector at the output can then be used to interpolate the channel coefficients to emulate varying speeds See also the tutorial Applying Varying Speeds Channel Interpolation in Section A 6 for more details Input si the sampling interval in seconds method selects the interpolation algorithm The default is cubic spline interpolation Op tional are e nearest Nearest neighbor interpolation e linear Linear interpolation e spline Cubic spline interpolation e pchip Piecewise Cubic Hermite Interpolating Polynomial e cubic Cubic spline interpolation Output dist Distance of each interpolated position from the start of the track in m interpolate_posit ions samples_per_meter Description Interpolates positions along the track This function interpolates the positions along the track such that it matches the samples per meter specifies in the simulation parameters The channel model operates on a position based sample grid That means that the channel_builder generates one CIR for each position on the track In practise however a time continuous evolution of the CIR is often needed This
34. drifting the delays are not updated and stay constant during the segment The position of the first scatterer is in close distance to the Rx only some m away 190 180 100 20 25 30 35 40 45 50 Distance from track start point Figure 22 Cluster delays vs Rx position drifting phases tutorial Copyright Fraunhofer Heinrich Hertz Institute 92 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 A TUTORIALS When moving the Rx first approaches the scatterer delay gets a bit smaller and then the distance increases again phase unwrap angle squeeze c coeff 1 1 2 plot distance phase xlabel Distance from track start point ylabel Continuous phase The second plot Fig 23 left shows the phases of the 20 subpaths of the first NLOS tap for the drifting case Note that the phases are not linear This comes from the close proximity of the scatterer to the initial Rx position The position of all 20 reflection points are calculated by the channel model Those position mark the position of the last bounce scatterer LBS When moving the Rx the distance to the LBS changes for each subpath and so does the phase Here the phase of each of the subpaths is calculated from the length of the path pow abs squeeze sum c coeff 1 1 2 plot distance 10 logi0 pow r xlabel Distance from track start point ylabel Tap power dB
35. envi ronment is described by 55 individual parameters These parameters are stored in configuration files that can be found in the subfolder named config in the main channel model folder The parameters and values can be describes as follows Copyright Fraunhofer Heinrich Hertz Institute 50 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 2 SOFTWARE STRUCTURE e No Clusters This value describes the number of clusters Each cluster is the source of a reflected or scattered wave arriving at the receiver Typically there are less clusters in a LOS scenario then in a NLOS scenario Note that the number of clusters directly influences the time needed to calculate the coefficients Path Loss PL A common path loss PL model for cellular systems is the log distance model where A and B are scenario specific coefficients The path loss exponent A typically varies between 20 and 40 depending on the propagation conditions the base station height and other influences They are typically determined by measurements d is the distance in units of meters between the transmitter and the receiver In other environments such as in satellite systems the PL does not depend on the distance but has a constant value In this case A would be 0 e Shadow Fading SF Shadow fading occurs when an obstacle gets positioned between the wireless device and the signal transmitter This interference causes significant reduction in signal stren
36. in h_parset position aso init_parameters force Description Generates the initial parameters This function creates the initial parameters for the channel builder If the parameters are already initialized no new update is performed The optional parameter force can be used to enforce an update even if the parameters are already given Input force Enforces 1 the generation of new parameters Default 0 Output aso An array with angular spread values for each terminal The rows are AoD AoA EoD EoA visualize_clusters i mobile i_cluster Description Plots the scattering clusters for a mobile terminal This method plots all scattering clusters for a given mobile terminal If i_cluster is not given then only the main paths are shown for all MPCs If i_cluster is given then also the subpaths are shown for the selected cluster The plot is in 3D coordinates You can rotate the image using the rotate tool Input i mobile The index of the mobile terminal within the channel builder object i_cluster The index of the scattering cluster Optional Copyright Fraunhofer Heinrich Hertz Institute 45 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 2 SOFTWARE STRUCTURE 2 2 7 Class channel Objects of this class are the output of the channel model They are created by the channel_builder By default channel coefficients are prov
37. initialize 0 then adjust the parameters in the parameter_set objects and call update_parameters manually check_parfiles Enables 1 default or disables 0 the parsing of shortnames and the validity check for the config files This is useful if you know that the parameters in the files are valid In this case this saves some execution time Output h_parset A matrix of parameter_set objects Rows correspond to the scenarios columns correspond to the transmitters See Section 2 2 5 h_cb A vector of channel_builder objects See Section 2 2 6 mem estimate_memory_usage verbose Description This function estimates the memory requirements for running the simulation specified by the layout This can help to determine if the computer is equipped with enough memory before starting complex simulations Input verbose Shows a detailed report of the memory requirements default 1 Output mem The estimate of the required memory in Bytes Copyright Fraunhofer Heinrich Hertz Institute 36 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 2 SOFTWARE STRUCTURE h_layout layout generate regular no_sites isd h_array Description generates a new multicell layout using a regular grid of BS positions Each BS has three sectors Input no _sites the number of sites This can be be 1 7 or 19 resulting in 3 21 or 57 sectors respectively isd
38. logi0 abs ifft h 1 72 figure imagesc pdp 1 256 caxis max max pdp 50 max max pdp 5 J colorbar cm colormap hot colormap cm end 1 1 set gca XTick 1 32 255 set gca XTickLabel 0 32 256 100e6 1e6 xlabel Delay mus set gca YTick 1 cn 2 no_snap 8 cn 2 no_snap set gca YTickLabel 0 cn 2 no_snap 8 cn 2 no_snap cn 2 no_snap 20 ylabel Distance from start point m title PDP for the linear track without drifting close all disp QuaDRiGa Version simulation_parameters version QuaDRiGa Version 1 0 1 145 Copyright Fraunhofer Heinrich Hertz Institute 98 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 A TUTORIALS PDP for the linear track with drifting ste al aed Distance from start point m 0 0 32 0 64 0 96 1 28 16 1 92 2 24 Delay us Distance from start point m 0 032 0 64 0 96 1 28 Delay us y PDP for the linear track without drifting y r r 1 6 1 92 2 24 Figure 27 Received power on the linear track time evolution tutorial 100 105 110 115 120 125 4 130 Copyright Fraunhofer Heinrich Hertz Institute eMail quadriga hhi fraunhofer de 99 QuaDRiGa v1 2 32 458 A TUTORIALS A 6 Applying Varying Speeds Channel Interpolation One new feature that makes the s
39. m Default is 1 5 m max_height the maximum user height in m Default is 1 5 m track_length the length of the linear track in m Default is 1 m Copyright Fraunhofer Heinrich Hertz Institute 39 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 2 SOFTWARE STRUCTURE pairs power set_pairing method threshold tx_power check_parfiles Description Determines links for which channel coefficient are generated This function can be used to automatically determine the links for which channel coefficients should be generated For example in a large network there are multiple base stations and mobile terminals The base stations however only serve a small area It the terminal is far away from this area it will receive only noise from this particular BS In this case the channel coefficients will have very little power and do not need to be calculated Disabling those links can reduce the computation time and the storage requirements for the channel coefficients significantly There are several methods to du this which can be selected by the input variable method Methods all Enables the simulation of all links power Calculates the expected received power taking into account the path loss the antenna patterns the LOS polarization and the receiver orientation If the power of a link is below the threshold it gets deactivated sf Same as power but this option also in
40. now be freely rotated in 3D coordinates while maintaining the polarization properties e New MATLAB implementation The MATLAB code was completely rewritten The implementations now fosters object oriented programming and object handles This increases the performance significantly and lowers the memory usage 1 3 Introduction to QuaDRiGa QuaDRiGa QUAsi Deterministic RadIo channel GenerAtor was developed to enable the modeling of MIMO radio channels for specific network configurations such as indoor satellite or heterogeneous configurations Besides being a fully fledged three dimensional geometry based stochastic channel model QuaDRiGa con tains a collection of features created in spatial channel model SCM and WINNER channel models along with novel modeling approaches which provide features to enable quasi deterministic multi link tracking of users receiver movements in changing environments The main features of QuaDRiGa are e Three dimensional propagation antenna modeling geometric polarization scattering clusters e Continuous time evolution e Spatially correlated propagation parameter maps e Transitions between varying propagation scenarios Copyright Fraunhofer Heinrich Hertz Institute 12 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 1 INTRODUCTION AND OVERVIEW The QuaDRiGa approach can be understood as a statistical ray tracing model Unlike the classical ray tracing approach it doesn t
41. objects See Section 2 2 6 h channel h cb get_channels_seg tx rx seg overlap Description Returns the channel coefficients for a single segment only This function can be used to obtain the channel coefficients for a single segment or single rx tx com bination only Thus the channel model can be run in streaming mode where updates are provided on the fly This can significantly reduce the memory requirements for long time sequences However the parameter maps still need to be generated for the entire scenario Features e Parameter maps will be deleted after the parameters were extracted to free memory before creating the channels Caching will be used to avoid multiple calculation of the same overlapping regions Preinitialization will be used to return the same coefficients for successive calls Input tx The index of the transmitter e g the BS rx The index of the receiver or track e g the BS seg The segment indices on the the track If it is not provided or empty the entire track is returned It is also possible to concat successive segments i e 1 3 or 3 5 etc overlap The opverlapping fraction for the channel merger Default is 0 5 Output h_channel A channel object See Section 2 2 7 h_cb A vector of channel_builder objects See Section 2 2 6 Copyright Fraunhofer Heinrich Hertz Institute 38 eMail quadriga hhi fraunhofer de QuaDRiGa
42. of discrete values e g 307 ns for segment 1 152 ns for segment 2 233 ns for segment 3 and so on This is done for all LSPs 3 The trajectory describes the position of the MT in the maps For each segment of the trajectory clusters are calculated according to the values of the LSPs at the map position The cluster positions are random within the limits given by the LSP For example a delay spread of 152 ns limits the distance between the clusters and the terminal 4 Each cluster is split into 20 sub paths and the arrival angles are calculated for each sub path and for each positions of the terminal on the trajectory Copyright Fraunhofer Heinrich Hertz Institute 15 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 1 INTRODUCTION AND OVERVIEW The antenna response for each of the arrival angels is calculated the same holds for the departure angles If there is more than one antenna at the transmitter and or receiver side the calculation is repeated for each antenna The phases are calculated based on the position of the terminal antennas in relation to the clusters The terminal trajectory defines how the phases change This results in the Doppler spread The coefficients of the 20 sup paths are summed the output are paths If there is more than one antenna and depending on the phase this sum results in a different received power for each antenna pair At this point the MIMO channel response is creat
43. part relative to the segment length When there are scenario transitions KF and PG change smoothly during a prede fined interval The length of that interval is a percentage of previous segment The parameter overlap adjusts this percentage ranging from 0 i e very hard step like change at the scenario boundary to 1 very smooth but long transition usage Changes the behavior of the method usage 0 Deletes all existing parameters from the track usage 1 Deletes all existing parameters from the track and generates new ones Existing LSPs will be overwritten usage 2 default Keeps existing parameters but generates missing ones This is useful when for example the effective path gain PG is provided but the other LSPs are unknown In this case the unknown gaps are filled with values which are generated from the provided scenario description check_parfiles check_parfiles 0 1 default 1 Disables 0 or enables 1 the parsing of shortnames and the validity check for the config files This is useful if you know that the parameters in the files are valid In this case this saves execution time verbose Enables 1 default or disables 0 the progress bar Output par The automatically generated parameters This variable contains structure of the LSPs with the following fields e ds The delay spread in s per segment e kf The Ricean K Factor in dB per snapshot e pg The effective path gain in dB excluding antenna ga
44. set the number of clusters for the LOS scenario to 7 Note that the default settings are stored in files in the sub folder config of the channel model folder Here the default settings can be permanently set After a change the parameters of the segments need to be updated This is done by calling the update_parameters method p Sl scenpar NumClusters 7 p update_parameters Parameters 00000000000000000000000000000000000000000000000000 24 seconds When update_parameter is called the specific parameters for each segment are generated E g each segment gets assigned a RMS Delay Spread and other values which are drawn from the statistics defined in scenpar For the LOS segments the individual RMSDS values for each segment are rmsds p S1 ds 1e9 average mean p S1l ds 1e9 Copyright Fraunhofer Heinrich Hertz Institute 87 eMail quadriga hhi fraunhofer de N wwwwwnnnnn or D NaO O QuaDRiGa v1 2 32 458 A TUTORIALS rmsds Columns 1 through 7 48 8391 6 8370 55 2154 48 7273 25 9658 29 7929 22 2436 Columns 8 through 11 68 7237 13 3159 85 5425 121 5296 average 47 8848 Generate channel coefficients Next we generate the channel coefficients This is a lengthy task The next line then combines the channels of the individual segments into a time continuous channel Here the parameter 0 2 sets the length of the overlap region between two segments
45. set up our antenna arrays We use the synthetic dipole antennas for this case Those antennas show perfect polarization characteristics First we generate a Element 1 Element 2 Element 3 Vertical Vertical Vertical Horizontal Horizontal Horizontal 18 15 12 9 6 3 0 19 16 13 10 7 4 l 19 16 13 10 7 4 l Attenuation dB Attenuation dB Attenuation dB Figure 31 Polarimetric dipole antenna patterns for different orientations Copyright Fraunhofer Heinrich Hertz Institute 104 eMail quadriga hhi fraunhofer de af wn nae wn eB won ee QuaDRiGa v1 2 32 458 A TUTORIALS single dipole with V polarization Then we create multiple copies of this antenna element and rotate them by 45 and 90 degrees respectively We then use the same array for the receiver 1 layout s h 1 tx_array generate dipole l tx_array set_grid 180 10 180 pi 180 l tx_array field_pattern_vertical A 1 tx_array field_pattern_vertical max max l tx_array set_grid 180 5 180 pi 180 tx_array copy_element 1 2 3 tx_array rotate_pattern 45 y 2 90 Create a new Layout create V polarized dipole 90 10 90 pi 180 Normalize tx_array field_pattern_vertical 90 pi 180 Duplicate the element two more times 45 degree polarization tx_array rotate_pattern 90 y 3 tx_array visualize rx_array 1 tx_array 90 degree polarization Plot
46. the array Use the same array for the Rr PRPPP HEH wee axxo Defining a track The third step defines the track Here we use a circle with 20 m diameter starting in the east traveling north We also choose a LOS scenario since we want to study the LOS polarization The transmitter is located 8 m north of the center of the circle at an elevation of 2 m tx_position 0 12 6 Tx position rx_position 20 O O 4 Start position for the Ra track track generate circular 40 pi 0 4 A circular track with radius 10 m track scenario BERLIN_UMa_LOS Chosse the Urban Macro LOS scenario track interpolate_positions s samples_per_meter h Interpolate positions visualize Plot the track PRPPPPH i T t T T T T f T F Tx Position E E EE E E E Tx Antenna O Rx Position Y Rx Antenna J Rx Track 4x1 w T s Y Position c 7 a 20 15 10 5 0 5 10 15 20 X Position Figure 32 Scenario layout Generating channel coefficients Now we have finished the parametrization of the simulation and we can generate the parameters We thus create a new set of correlated LSPs and fix the shadow fading and the K factor to some default values This disables the drifting for those parameters We need to do that since otherwise drifting and polarization would interfere with each other RandStream setGlobalStream RandStream mti9937ar seed 1 p 1 cr
47. the inter site distance between the BSs in m h_array the antenna array h_array is for one sector only It will be rotated to match the sector orientations and copied to all sites The broadside direction of the provided antenna must be 0 facing east Output h_layout The generated layout par h_parset generate_parameters overlap usage check_parfiles Description Generates large scale parameters and stores them in track par Normally parameters are handled by objects of the parameter_set class which are gener ated by calling layout create_parameter_sets Those objects then feed the parameters to the channel_builder However this method is rather inflexible when the user wants to manipulate the parameters directly As an alternative parameters can be provided in the property track par of the track class This allows the user to edit parameters without dealing with the parameter_set objects This function extracts the LSPs for the given scenario from the parameter_set class and stores them in track par Hence it automatically generates the LSPs and thus implements an easy to use interface for the parameter_set class Input overlap The length of the overlapping part relative to the segment length When there are scenario transitions KF and PG change smoothly during a prede fined interval The length of that interval is a percentage of previous segment The parameter overla
48. the second segment with a lot of fading goes back to the first while slowing down at the same time After staying constant for one second the channel starts running again speeding up towards the end of the track h ci fr 100 6 512 h squeeze h pdp 10 logi0 abs ifft h 1 72 figure imagesc pdp 1 256 caxis max max pdp 50 max max pdp 5 colorbar cm colormap hot colormap cm end 1 1 set gca XTick 1 32 255 set gca XTickLabel 0 32 256 100e6 1e6 xlabel Delay mus set gca YTick 1 ci no_snap 8 ci no_snap set gca YTickLabel O ci no_snap 8 ci no_snap ci no_snap 20 ylabel Time s Copyright Fraunhofer Heinrich Hertz Institute 102 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 A TUTORIALS close all disp QuaDRiGa Version simulation_parameters version QuaDRiGa Version 1 0 1 145 20 80 sail 25 16 85 90 g E 12 T3 95 3 10 2 10 100 5 E 8 12 5 105 E 6f 15l 110 ji 115 al 17 5 120 5 10 15 20 0 0 32 0 64 0 96 128 1 6 1 92 2 24 Time s Delay us Figure 30 Movement profile left and interpolated PDP right Copyright Fraunhofer Heinrich Hertz Institute 103 eMail quadriga hhi fraunhofer de a PF ON QuaDRiGa v1 2 32 458 A TUTORIALS A 7 Geometric Polarization Here we demonstrat
49. they are not provided or it they are incomplete they are completed with values from the LSP maps If the maps are invalid e g because they have not been generated yet new maps are created force 1 Creates new maps and reads the LSPs from those maps Values from layout track par are ignored Note that the parameters pg and kf will still be taken from layout track par when generating channel coefficients force 2 Creates dummy data for the maps and the LSPs Any existing maps will be deleted Data and maps will be declared as invalid and the next time when update_parameters is called new parameters are generated Values in layout track par will NOT be affected Copyright Fraunhofer Heinrich Hertz Institute 43 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 2 SOFTWARE STRUCTURE 2 2 6 Class channel_builder This class implements all functions that are needed to generate the channel coefficients It thus implements the core components of the channel model The class holds all the input variables as properties It s main function get_channels then generates the coefficients The procedure is summarized as follows The channel builder first generates a set of random clusters around each receiver This is done by drawing random variables for the delay the power and the departure and arrival angles for each cluster Each cluster thus represents the origin of a reflecte
50. us Figure 16 2D PDP of the simulated track Copyright Fraunhofer Heinrich Hertz Institute eMail quadriga hhi fraunhofer de 80 16 7 18 19 20 9 10 11 13 14 10 QuaDRiGa v1 2 32 458 A TUTORIALS The next plot shows the total received power along the path Fig 17 top left Green shaded ares are LOS The rest is NLOS dist 1 cn no_snap l track get_length cn no_snap ind find strcmp l track scenario S1l los for n 1 numel ind los los 1 track segment_index ind n l track segment_index ind n 1 end power 10 1logi0 sum reshape abs cn coeff 2 cn no_snap 1 4 ar zeros 1 cn no_snap ar los 200 figure a area dist ar set a i FaceColor 0 7 0 9 0 7 set a LineStyle none hold on plot dist power hold off title Position dependent power xlabel Track m ylabel Power dB axis 0 500 min power 5 max power 5 legend LOS P_ total 4 grid on The following plot Fig 17 top right shows the distribution PDF of the received power for both the LOS and NLOS segments bins 150 2 80 p_los hist power los bins cn no_snap 100 p_nlos hist power setdiff 1 cn no_snap los bins cn no_snap 100 figure bar bins p_los p_nlos axis 124 5 83 0 ceil max p_los p_nlos grid on colormap Cool title Empirical PDF of the ac LO
51. 0 sf_kf 0 3 61 esD_ds 0 4 0 5 62 esD_asD 0 4 0 5 63 esA_sf 0 8 64 esA_asA 0 4 65 66 A logi0 d B C logi0 f D log10 hBS E logio RMS 67 Two different values first before breakpoint last after breakpoint 68 Different SF coefficients 69 70 PL_model winner_los 71 72 PL_A1 26 73 PL_B1 25 74 PL_C1 20 75 PL_D1 0 76 PL_E1 0 77 PL_sigi 4 78 79 PL_A2 40 80 PL_B2 9 27 81 PL_C2 6 82 PL_D2 14 83 PL_E2 14 84 PL_sig2 6 You can create you own scenario by editing this file and saving it under a new filename in the config Folder The file ending must be conf The filename then is also the scenario name and the settings can be accessed from inside MATLAB as described above Copyright Fraunhofer Heinrich Hertz Institute 55 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 3 TECHNICAL DOCUMENTATION 3 Technical Documentation This chapter describes an extension of the WINNER model 3 where time evolution based on the ideas presented in 7 and 8 is incorporated A reference implementation in MATLAB is available as open source 9 The modeling approach consists of two steps A stochastic part generates LSPs and calculates random three dimensional 3 D positions of scattering clusters It is assumed that the base station BS is fixed and the mobile terminal MT is moving In this case scatterin
52. 00000000000000000 5 seconds ans WINNER UMa C2 NLOS_Tx1 WINNER UMa C2 LOS_Tx1 ans We set the number of clusters for the NLOS segments to 14 Currently it is not possible to have a different number of clusters for each segment i e it is not possible for the first NLOS segment to have 14 clusters and for the second to have only 10 p 1 scenpar NumClusters 14 In order to manually set the parameters we have to overwrite the original settings We do this here for the delay spread The automatically generated values are p 1 ds 1 p 1 ds 2 p 2 ds 1 p 1 ds 3 1e6 ans 0 2696 0 2433 0 0948 0 2094 Copyright Fraunhofer Heinrich Hertz Institute 111 eMail quadriga hhi fraunhofer de ew N He QuaDRiGa v1 2 32 458 A TUTORIALS We want to set the values of the four segments to 0 45 0 33 0 12 and 0 60 microseconds This is done by p 1 ds 1 0 45e 6 p 1 ds 2 0 33e 6 p 2 ds 1 0 12e 6 p 1 ds 3 0 60e 6 The K Factor and the shadow fading are read from the map during the generation of channel coefficients This would overwrite any manual values However this could be deactivated A drawback is that in this case the KF SF and PL are only updated once per segment This will result in a step like function of the output power There is currently no method the set the power manually on a per snapshot basis In the following example we want to
53. 00000000000000000000000000000000000000000 J 17 seconds Plot For plotting we use the internal function of the array class We adjust the title of the figures accord ingly 1 d a copy_objects 2 d field_pattern_vertical pat 1 3 d field_pattern_horizontal pat 2 4 x d visualize 6 set x 1 11 String RHCP RHCP 7 set x 1 12 String RHCP LHCP 9 set x 2 11 String LHCP RHCP 10 set x 2 12 String LHCP LHCP 1 close all 2 disp QuaDRiGa Version simulation_parameters version 1 QuaDRiGa Version 1 0 1 145 Element 1 Element 2 RHCP RHCP LHCP RHCP RHCP LHCP LHCP LHCP Attenuation dB Attenuation dB Figure 34 RHCP LHCP antenna patterns Copyright Fraunhofer Heinrich Hertz Institute 109 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 A TUTORIALS A 9 How to manually set LSPs in QuaDRiGa This tutorial explains how to generate a time series of channel coefficients with manual selection of LSPs Setting general parameters We set up some basic parameters such as center frequency sample density and the position of the transmitter close all clear all set 0 defaultTextFontSize 14 4 Set default font size for the plots set 0 defaultAxesFontSize 14 s simulation_parameters Basic simulation parameters s center_frequency 2 185e9 Center Frequ
54. 2 p_los mean abs coeff 1 72 3 3 kf p_los p_nlos 5 figure 6 plot 35 35 35 35 k 7 hold on 8 plot 35 35 3 35 35 k 9 plot 35 35 3 35 35 k 10 plot 10 logi0 p kf 10 log1i0 kf 11 hold off 12 axis 30 30 30 30 14 legend Equal 3dB 4 15 xlabel KF_P dB 16 ylabel KF_C dB 17 title K Factor Requested vs generated value Now we repeat the calculation for the RMS delays spread 1 pow_tap abs coeff 2 2 pow_sum sum pow_tap 2 3 mean_delay sum pow_tap delay 2 pow_sum 4 ds sqrt sum pow_tap delay 2 2 pow_sum mean_delay 2 5 ds mean ds 3 7 figure 8 plot 0 0 1 2 0 0 1 2 k 9 hold on 10 plot 0 0 1 2 1 1 0 0 1 2 k 11 plot 0 0 1 2 0 0 1 2 1 1 k 12 plot p ds 1e6 ds 1le6 13 hold off 14 axis 0 1 5 0 1 5 J 15 16 legend Equal 10 Error 4 17 xlabel DS_P mus is ylabel DS_C mus 19 title Delay Spread Requested vs generated value The following plot shows the RMSDS of the requested and generated values in dB vs the K factor A value of 3 means that the RMSDS of the generated coefficients is twice a high as in the parameter_set P We see that for a K Factor of up to 30 dB the DS difference is sma
55. 2 32 458 2 SOFTWARE STRUCTURE sf kf get_sf_profile evaltrack i mobile Description This function returns the shadow fading and the K factor along the given track This function is mainly used by the channel builder class to scale the output channel coefficients The profile is calculated by using the data in the correlation maps and interpolating it to the positions in the given track Increasing the resolution of the maps also increases the resolution of the profile Input evaltrack A track object for which the SF and KF should be interpolated i mobile If simulation_parameters drifting precision is set to 3 then this parameter is required to select the Rx antenna array Output sf The shadow fading linear scale along the track kf The K factor linear scale along the track set_par name value Description Sets the parameters of all objects in parameter_set arrays This function sets all values of the parameter specified by name of the parameter_set array to the given value Example set_par ds 1e 9 sets all ds values to 1 ns Input name The fieldname that should be altered Supported are ds kf sf asD asA esD esA samples_per_meter map_extension and LSP_xcorr_matrix value The value that should be assigned If the LSP_xcorr_matrix is altered then the low
56. 43 0 457 0 490 0 503 0 507 0 510 0 517 0 520 0 523 0 537 20 0 Ww 0 423 0 467 0 553 0 592 0 610 0 623 0 647 0 658 0 666 0 701 Q 0 300 0 310 0 320 0 323 0 327 0 330 0 337 0 340 0 343 0 350 Copyright Fraunhofer Heinrich Hertz Institute eMail quadriga hhi fraunhofer de
57. 60 66 References 1 E Eberlein T Heyn F Burkhardt S Jaeckel L Thiele T Haustein G Sommerkorn M Kaske C Schneider M Dominguez and J Grotz Characterisation of the MIMO channel for mobile satellite systems acronym MIMOSA TN8 2 final report Fraunhofer Institute for Integrated Circuits IIS Tech Rep v1 0 2018 2 S Jaeckel L Raschkowski K Borner L Thiele F Burkhardt and E Eberlein QuaDRiGa Quasi Deterministic Radio Channel Generator User Manual and Documentation Fraunhofer Heinrich Hertz Institute Tech Rep v1 1 0 248 2014 3 P Ky sti J Meinila L Hentil et al IST 4 027756 WINNER II D1 1 2 v 1 1 WINNER II channel models Tech Rep 2007 Online Available http www ist winner org 4 P Heino J Meinil P Ky sti et al CELTIC CP5 026 D5 3 WINNER final channel models Tech Rep 2010 Online Available http projects celtic initiative org winner 5 C Schneider M Narandzic M Kaske G Sommerkorn and R Thoma Large scale parameter for the WINNER II channel model at 2 53 GHz in urban macro cell Proc IEEE VTC 710 Spring 2010 6 M Narandzic C Schneider M Kaske S Jaeckel G Sommerkorn and R Thoma Large scale pa rameters of wideband MIMO channel in urban multi cell scenario Proc EUCAP 11 2011 Copyright Fraunhofer Heinrich Hertz Institute 8 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458
58. 80 west and 270 south The LOS delay stays constant since the distance to the Tx is also constant Copyright Fraunhofer Heinrich Hertz Institute 97 eMail quadriga hhi fraunhofer de wn wn QuaDRiGa v1 2 32 458 A TUTORIALS However the power of the LOS changes according to the scenario Also note that the NLOS segment has significantly more taps due to the longer delay spread Next we create the same plot for the linear track Fig 27 Note the slight increase in the LOS delay and the high similarity of the first two LOS segments due to the correlated LSPs Segment change is at around 6 m h cn 2 fr 100 e6 512 h squeeze h pdp 10 logi0 abs ifft h 1 72 figure imagesc pdp 1 256 caxis max max pdp 50 max max pdp 5 J colorbar cm colormap hot colormap cm end 1 1 set gca XTick 1 32 255 set gca XTickLabel 0 32 256 100e6 1e6 xlabel Delay mus set gca YTick 1 cn 2 no_snap 8 cn 2 no_snap set gca YTickLabel 0 cn 2 no_snap 8 cn 2 no_snap cn 2 no_snap 20 ylabel Distance from start point m title PDP for the linear track with drifting Last we plot the same results for the linear track without drifting Fig 27 right Note that the LOS delay is not smooth during segment change There are two jumps at 6 m and again at 13 5 m h dn 2 fr 100e6 512 h squeeze h pdp 10
59. EPARTURE AND ARRIVAL ANGLES ADOPTED WINNER METHOD Table 12 Correction values from 3 for different numbers of paths L 4 5 8 10 11 12 14 15 16 20 C 0 779 0 860 1 018 1 090 1 123 1 146 1 190 1 211 1 226 1 289 Table 13 Comparison of the correction functions KF Number of paths L dB W Q 4 5 8 10 11 12 14 15 16 20 11 7 W 0 779 0 860 1 018 1 090 1 123 1 146 1 190 1 211 1 226 1 289 Q 0 765 0 822 0 904 0 923 0 929 0 935 0 935 0 935 0 935 0 943 8 0 Ww 0 895 0 988 1 169 1 252 1 290 1 316 1 366 1 391 1 408 1 480 Q 0 790 0 820 0 857 0 870 0 880 0 890 0 977 1 020 1 070 1 250 4 0 W 0 917 1 012 1 198 1 283 1 322 1 349 1 401 1 425 1 443 1 517 Q 0 713 0 777 1 047 1 213 1 277 1 340 1 427 1 470 1 500 1 613 0 0 W 0 860 0 949 1 123 1 203 1 239 1 265 1 313 1 336 1 353 1 422 Q 0 830 0 990 1 277 1 380 1 420 1 460 1 520 1 550 1 570 1 637 3 1 W 0 779 0 860 1 018 1 090 1 123 1 146 1 190 1 211 1 226 1 289 Q 0 926 1 029 1 221 1 295 1 325 1 354 1 391 1 409 1 425 1 481 4 0 W 0 752 0 831 0 983 1 053 1 085 1 107 1 149 1 170 1 184 1 245 Q 0 930 1 020 1 190 1 257 1 283 1 310 1 343 1 360 1 373 1 420 8 0 W 0 625 0 690 0 817 0 875 0 901 0 920 0 955 0 972 0 984 1 035 Q 0 820 0 870 0 967 1 003 1 017 1 030 1 057 1 070 1 077 1 103 12 0 W 0 508 0 561 0 664 0 711 0 733 0 748 0 776 0 790 0 800 0 841 Q 0 627 0 653 0 707 0 727 0 733 0 740 0 760 0 770 0 773 0 793 16 0 W 0 431 0 476 0 563 0 603 0 621 0 634 0 658 0 670 0 678 0 713 Q 0 4
60. OS power num2str mean sum abs c coeff 1 1 2 end 2 3 4 LOS power 0 52851 NLOS power 0 22249 The LOS power is almost constant when the Rx is south of the Tx However in close proximity at 90 the power is lowered significantly This comes from the 2 m elevation of the Tx When the Rx is almost under the Tx the radiated power of the Dipole is much smaller compared to the broadside direction The average power of the LOS is thus also lowered to 0 56 W The average sum power if the 7 NLOS components Tx Vertical Rx Vertical Tx Vertical Rx 4 Tx vertical Rx 45 0 97 p ue Tx 45 Rx vertical 0 87 0 8 4 2 2 0 7 H p S 2 2 5 5 0 6 05 2 z 04 wn a 0 3 E Q 4 02F 0 1 WNL AE als PIELER a CICA oF J 0 1 f f f f 0 1 f f 1 1 f 1 f 50 100 150 200 250 300 350 0 50 100 150 200 250 300 350 Position degrees Position degrees Tx 45 Rx 45 Tx 90 Rx 0 45 90 0 9 0 9 BRO i Rx 45 0 8 al 0 8 Rx 90 2 2 07 4 S S Z Z 0 6 s s vo o E 5 05 z z 0 4 a a 7 7 0 3 0 2 0 1 0 la 0 1 1 1 1 1 1 1 Ll 0 1 1 1 1 1 1 1 Ll 0 45 90 135 180 225 270 315 360 0 50 100 150 200 250 300 350 Position on circle degrees Position degrees Figure 33 Results from the geometric polarization tutorial Copyright Fraunhofer Heinrich Hertz Institute 106 eMa
61. Quasi Deterministic Radio Channel Generator User Manual and Documentation Aa Quadriga The Next Generation Radio Channel Model Document Revision v1 2 32 458 March 23 2015 Fraunhofer Heinrich Hertz Institute Wireless Communications and Networks Einsteinufer 37 10587 Berlin Germany e mail quadriga hhi fraunhofer de http www quadriga channel model de A Fraunhofer Heinrich Hertz Institute QuaDRiGa v1 2 32 458 Contributors Editor Fraunhofer Heinrich Hertz Institute Wireless Communications and Networks Einsteinufer 37 10587 Berlin Germany Contributing Authors Stephan Jaeckel Leszek Raschkowski Kai Borner and Lars Thiele Fraunhofer Heinrich Hertz Institute Frank Burkhardt and Ernst Eberlein Fraunhofer Institute for Integrated Circuits TIS Grants and Funding This work was supported by e the European Space Agency ESA in the Advanced Research in Telecommunications Systems ARTES programme under contract AO 1 5985 09 08 NL LvH Acronym MIMOSA 1 http artes esa int projects mimosa characterisation mimo channel mobile satellite systems e the German Federal Ministry of Economics and Technology BMWi in the national collaborative project In telliSpektrum under contract 01ME11024 http www intellispektrum de e the European Commission co funded the project METIS as an Integrated Project under the Seventh Framework Programme for research and development FP7 http www metis2020 com e th
62. Receiver start position Beenie Movement track Height 10 0 10 lt iti0D im 5 10 X PO Received power along the track Old model New model Normalized received power O N A O CO me Position on circle Figure 10 Example showing the effect of the new polarization rotation method The same expression is found in the antenna pattern 58 of the WINNER model where the complex value A eJ from the Jones vector can be identified with the generally also complex valued component F 0 of the antenna pattern Likewise Azel can be identified with Fl 0 This implies that the polarization coupling matrix M in 59 is a Jones matrix and that the Jones calculus could apply also to the WINNER model In general M can be seen as a transformation operation that maps the incoming signal on the polarization plane to an outgoing signal If the coefficients are real valued then linear transformations such as rotation scaling shearing reflection and orthogonal projection as well as combinations of those operations are possible If the coefficients are complex valued then the matrix shows characteristics of a Mobius transfor mation Such transformations can map straight lines to straight lines or circles and vice versa Since the Jones calculus allows the use of complex coefficients it can transform linear polarized signals into circular or elliptical polarized signals and elli
63. References 12 13 14 15 16 17 18 19 20 21 22 23 24 25 H Xiao A Burr and L Song A time variant wideband spatial channel model based on the 3gpp model Proc IEEE VCT 06 Fall 2006 D Baum J Hansen and J Salo An interim channel model for beyond 3G systems Proc IEEE VCT 05 Spring vol 5 pp 3132 3136 2005 Online Available http www quadriga channel model de S Jaeckel K B rner L Thiele and V Jungnickel A geometric polarization rotation model for the 3 D spatial channel model IEEE Trans Antennas Propag vol 60 no 12 pp 5966 5977 2012 A Algans K Pedersen and P Mogensen Experimental analysis of the joint statistical properties of azimuth spread delay spread and shadow fading IEEE J Sel Areas Commun vol 20 no 3 pp 523 531 2002 K Bakowski and K Wesolowski Change the channel IEEE Veh Technol Mag vol 6 pp 82 91 2011 S Szyszkowicz H Yanikomeroglu and J Thompson On the feasibility of wireless shadowing corre lation models IEEE Trans Veh Technol vol 59 pp 4222 4236 2010 M Gudmundson Correlation model for shadow fading in mobile radio systems IET Electron Lett vol 27 no 23 pp 2145 2146 November 1991 3GPP TR 25 996 v10 0 0 Spatial channel model for multiple input multiple output MIMO simula tions Tech Rep 3 2011 K Peder
64. S parameter set selected for bad state e The name selects the related configuration file For the given example the files MIMOSA_10 45_LOS conf and MIMOSA_10 45_NLOS conf are selected Table 9 Parameter sets provided together with the standard software LOSonly One LOS Path only no Shadow Fading no Path Loss WINNER_UMa_C2_LOS WINNER_UMa_C2_NLOS WINNER Urban Macrocell For typical terrestrial base stations deployed above rooftop in densely populated urban areas The max cell radius is about 1 km WINNER_UMi_B1_LOS WINNER_UMi_B1_NLOS WINNER Urban Microcell For typical terrestrial pico base stations deployed below rooftop in densely populated urban areas The max cell radius is about 200 m WINNER_SMa_C1_LOS WINNER_SMa_C1_NLOS WINNER Sub Urban Macrocell For typical terrestrial base stations deployed above rooftop in sub urban areas The max cell radius is about 10 km WINNER _Indoor_A1_LOS WINNER_Indoor_Al_NLOS WINNER Indoor Hotspot For typical indoor deployments such as WiFi or femto cells WINNER_UMa2Indoor_C4_LOS WINNER_UMa2Indoor_C4_NLOS WINNER Urban Macrocell to Indoor For users within buildings that are connected to a terrestrial base station deployed above rooftop in densely populated urban areas WINNER_UMi2Indoor_B4_LOS WINNER_UMi2Indoor_B4_NLOS WINNER Urban Microcell to Indoor For users within buildings that are connected to terrestrial pico ba
65. S and NLOS power xlabel P_ total dB ylabel Probability legend LOS NLOS 1 The next plot shows the RMS delay spread along the path Again shaded ares are for the LOS segments pow_tap squeeze sum sum abs cn coeff 2 1 2 pow_sum sum pow_tap 1 mean_delay sum pow_tap cn delay 1 pow_sum ds sqrt sum pow_tap cn delay 2 1 pow_sum mean_delay 2 ar zeros 1 cn no_snap ar los 10 figure a area dist ar set a i FaceColor 0 7 0 9 0 7 set a LineStyle none hold on plot dist ds 1le6 hold off ma 1e6 max ds 0 1 max ds axis 0 500 O ma title Position dependant delay spread xlabel Track m ylabel Delay Spread dB legend LOS sigma_ tau 1 grid on Copyright Fraunhofer Heinrich Hertz Institute 81 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 A TUTORIALS The final plot Fig 17 bottom right shows the distribution PDF of the RMS delay spread for both the LOS and NLOS segments bins 0 0 03 3 ds_los hist ds los 1e6 bins cn no_snap 100 ds_nlos hist ds setdiff 1 cn no_snap los 1e6 bins cn no_snap 100 figure bar bins ds_los ds_nlos axis 0 1 5 0 ceil max ds_los ds_nlos grid on colormap Cool title Empirical PDF of the LOS and NLOS RMSDS xlabel sigma_ tau mus
66. Section 3 2 with different values of K included Based on those values the actual RMS angular spread glcall is calculated using equations 26 27 and 28 The correction function follows from comparing gamel with og However two aspects need to be considered 1 Due to the randomization of the angles in 96 we have to take the average angle over a sufficiently large quantity 1000 realizations of oo This value is denoted as as Copyright Fraunhofer Heinrich Hertz Institute 114 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 B DEPARTURE AND ARRIVAL ANGLES ADOPTED WINNER METHOD 2 There is a nonlinear relationship between the angular spread in the simulated data aeS and the initial value og This comes from the logarithm in 95 and the modulo in 97 However for small values the relationship can be approximated by a linear function The maximum angular spread Oa is defined as the point where the error between the corrected value OEK and ie is 10 For a range of typical values L 2 42 and K 8 20 20 C L K can be numerically calculates as max avg 1 oS Oo T C L K f Oy 00 toy s 100 Th 0 Og where the og dependency of a a4 comes from the individual angles Generated Value 2507 unity k N a Q o T T Angular Spread in Data Ss O 50p f i i i i i i i 0 20 40 60 80 100 120 140 160 180 Requested Angular Spread o b
67. The three vector elements represent the x y and z component cos 0 cos c 6 cos sing 62 sin 0 Dipole antenna 0 tilt Dipole antenna 20 tilt vertical pattern i SES 20 15 10 5 0 Attenuation dB Figure 11 Example patterns for a dipole antenna Copyright Fraunhofer Heinrich Hertz Institute 68 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 3 TECHNICAL DOCUMENTATION A 3x3 matrix R can now be used to describe the orientation change in 3 D space The example in Figure 11 was tilted by 20 around the x axis of the coordinate system The corresponding matrix is 1 0 0 R 20 0 cos 20 sin 20 63 0 sin 20 cos 20 By multiplying R with 62 the orientation change is included in the vector c 0 R c 0 9 64 The transformed pattern F is needed in spherical coordinates Thus c is transformed back to spherical coordinates This results in the new angles 0 Q 0 arcsin ct 0 4 65 0 0 6 arctang cy 0 6 cf 8 66 ci cf and cf are the x y and z component of c respectively The coefficients of the rotated pattern are now obtained by reading the original pattern F at the transformed angles s W O 4 FIA 10 6 pagg reen Since the patterns are usually sampled at a fixed angular grid interpolation is needed here As a standard computationally inexpensive pro
68. al we now have 14 maps seven for LOS and another seven for NLOS The parameters for calculating the channel coefficients are drawn from the second seven maps We get a set of channel coefficients with different properties e g more multipath components lower K Factor etc A smooth transition between the coefficients from the first segment and the second is realized by the ramping down the powers of the clusters of the old segment and ramping up the power of the new This is implemented in step 4 Post processing Copyright Fraunhofer Heinrich Hertz Institute 17 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 1 INTRODUCTION AND OVERVIEW 10 NLOS LOS Change This is essentially the same as in point 2 However since the third segment is also in the scenario Satellite LOS Urban no new maps are generated The parameters are extracted from the same map as for the starting segment Turning off without change in reception condition LOS QuaDRiGa supports free 3D trajectories for the receiver Thus no new segment is needed the terminal stays in the same segment as in point 3 However we assume that the receive antenna is fixed to the terminal Thus if the car turns around so does the antenna Hence the arrival angles of all clusters including the direct path change This is modeled by a time continuous update of the angles delays and phases of each multipath component also known as drifting Due the
69. alized by updating the delays the departure and arrival angels the polarization the shadow fading and the K Factor based on the position of the terminal e Scenario transitions When the mobile terminal MT moves through the fading channel it may pass through several different scenarios QuaDRiGa supports smooth transitions between adjacent channel segments This is used to emulate long term time evolution and allows the simulation of e g handover scenarios e Variable speeds for mobile terminals QuaDRiGa supports variable speeds including accelerating and sowing down of mobile terminals e Common framework for LOS and NLOS simulations In WINNER line of sight LOS and non line of sight NLOS scenarios were treated differently QuaDRiGa used the same method for both scenarios types This reduces the model complexity and enables freely configurable multicell scenarios E g one MT can see two BSs one in LOS and another in NLOS e Geometric polarization The polarizations for the LOS and for the NLOS case is now calculated based on a ray geometric approach e Improved method for calculating correlated large scale parameters LSPs The WINNER model calculates maps of correlated parameter values using filtered random numbers QuaDRiGa uses the same method but extends the map generation algorithm to also consider diagonal movement directions and to create smoother outputs e New functions for modifying antenna patterns Antenna patterns can
70. ameter refers to a set of specific properties of the propagation channel Those are the delay spread the K factor the shadow fading the cross polarization ratio and four angular spread values Those properties can be extracted from channel sounding data If a large amount of channel measurements is available for a specific propagation scenario and the LSPs can be calculated from those channels statistics of the LSPs e g their distribution and correlation properties can be obtained A complete set of such statistical properties forms a parameter table that characterizes the scenario mobile terminal MT ong ee oh ee ee eR de ee EEE eee he Ee ES 56 Mobile terminals MTs are mobile receivers with one or more receive antennas They are usually assigned to a serving BS which delivers data to the terminal multipath component MPG 5 6 24 d e5 ee AE BR eR eR Eee ee hee a eS 56 Synonym for path Patli s s shy ta ai ache Soe ee a Gd ae ee Re Be es ey ee ee aA 61 64 72 73 A path describes the way that a signal takes from the transmitter to the receiver In the channel model there is usually a direct or LOS path and several indirect or NLOS paths Indirect paths involve one or more scattering events which are described by clusters However paths do not describe single reflections but combine sub paths that can not be separated in the delay domain Usually the channel model uses 6 25 paths to describe
71. an be established When the receiver is horizontal red line however there are two points where the two dipoles are aligned For the 45 dipole the same behavior can be observed but with roughly half the power close all disp QuaDRiGa Version simulation_parameters version QuaDRiGa Version 1 0 1 145 Copyright Fraunhofer Heinrich Hertz Institute 107 eMail quadriga hhi fraunhofer de 10 11 14 QuaDRiGa v1 2 32 458 A TUTORIALS A 8 Visualizing RHCP LHCP Patterns The internal algorithms of the channel model only work with linear polarization The antenna patterns are thus only stored in H V polarization This script defines two circular patch antennas and places them in an environment It then rotates the transmit antenna degree by degree and thus samples all azimuth and elevation angles The channel model is set up to record the channel response and thus record the RHCP LHCP response like in a measurement in an anechoic chamber Set up the array We set up a patch antenna with an opening angle of 120 We then copy that patch and rotate it by 90 around the x axis to create an X Pol array We then set the coupling to 90 phase to transmit circular polarized waves resolution 10 4 in Degrees w array custom 120 120 0 set_grid 180 resolution 180 pi 180 90 resolution 90 pi 180 a copy_element 1 2 w b a copy_objects a rotate_pattern 90
72. ange in the environment without a change in the environment type higher density of buildings but still the environment remains urban 9 Stopping at traffic lights NLOS 10 Houses have the same characteristics as before but are further away from the street urban environment with different reception characteristics 11 Change of environment Urban Forest 12 Turning off without change of environment NLOS Each simulation run in QuaDRiGa is done in three and an optional fourth step Set up tracks scenarios antennas and network layout Generate correlated LSPs Calculate the channel coefficients optional Post processing Poe Those steps also need to be done for the above scenario However different aspects of the track are handled in different parts of the model Additionally the QuaDRiGa model supports two operating modes for handling the LSPs 1 The first default mode generates the correlated LSPs automatically based on a scenario specific parameter set This is done in step 2 and involves so called parameter maps 2 The manual mode does not generate LSPs automatically Here the user has to supply a list of parameters to the model The step 2 thus to be implemented by the user Steps 1 3 and 4 are identical for both modes The following list describes the modeling of the observed effects along the track when using the automatic mode 1 1 Start Environment Urban LOS reception of satellite signal Each segment al
73. arameters Looo00000000000000000000000000000000000000000000000 31 seconds Channels Looo00000000000000000000000000000000000000000000000 15 seconds Merging 00000000000000000000000000000000000000000000000000 2 seconds Copyright Fraunhofer Heinrich Hertz Institute 79 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 A TUTORIALS Tx Position A Tx Antenna O Rx Position WV Rx Antenna Rx Track RtMOSA_ 10 45 _ NOATEN IOSA 10 45 NLOS SANAR IMOSA_10 45_NLOS MIMOSA_10 45_NLOS 200 fe 500 Y Position 300 X Position Figure 15 Scenario setup for the comparison of simulated and measured data Results First we plot the PDP vs distance from the start point see Fig 16 h cn 1 fr 20e6 256 pdp squeeze sum sum abs ifft h 3 2 1 2 pdp 10 logi0 pdp figure imagesc pdp end 1 1 1 192 cm colormap hot colormap cm end 1 1 caxis max max pdp 60 max max pdp 5 J colorbar title Time variant power delay profile set gca XTick 1 32 192 set gca XTickLabel 0 32 192 20e6 1e6 xlabel Delay mus ind sort cn no_snap cn 1 no_snap 10 1 set gca YTick ind set gca YTickLabel round sort 500 ind 3 descend ylabel Distance m Time variant power delay profile Distance m 0 1 6 3 2 4 8 6 4 8 Delay
74. arameters We do not need drifting here since no time varying channels are generated close all clear all set 0 defaultTextFontSize 14 set 0 defaultAxesFontSize 14 s simulation_parameters s center_frequency 2 53e9 s sample_density 2 s use_absolute_delays 1 s drifting_precision 0 We have one transmitter and 250 receiver positions Each receiver gets a specific channel However the receivers LSPs will be correlated We use omni directional antennas at all terminals Copyright Fraunhofer Heinrich Hertz Institute 75 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 A TUTORIALS layout s no_rx 250 randomize_rx_positions 200 1 5 1 5 1 200 m radius 1 5 m Re height track set_scenario BERLIN_UMa_NLOS 1 tx_position 3 25 4 25 m tx height 1 tx_array generate omni l rx_array 1 tx_array 1l visualize 0 view 33 60 Tx Position A Tx Antenna O Rx Position V_ Rx Antenna Rx Track Y Position 200 200 X Position Figure 13 Distribution of the users in the scenario We set up the scenario such that there is no XPR I e all vertical polarized paths will remain vertical after a reflection The same result would be achieved with a perfectly X polarized array at the receiver and summing up the power over all elements We further increase the KF to have a wider spread This allows us to stu
75. ard physical copy function While the standard copy command creates new physical objects for each element of obj in case obj is an array of object handles copy_objects checks whether there are object handles pointing to the same object and keeps this information par generate array_type element Ain Bin Cin Din Description Generates predefined arrays Array Types omni dipole half wave dipole patch custom parametric xpol rhcp dipole lhcp dipole lhcp rhcp dipole An isotropic radiator with vertical polarization A short dipole radiating with vertical polarization A half wave dipole radiating with vertical polarization A vertically polarized ideal patch antenna with 90 opening in azimuth and eleva tion An antenna with a custom gain in elevation and azimuth E g a generate custom 1 90 90 0 1 creates an array with 90 opening in az imuth and elevation and 0 1 rear gain An antenna following the function F A VB 1 B cos exp D Two elements with ideal isotropic patterns vertical polarization The second element is tilted by 90 Two crossed dipoles with one port The signal on the second element horizontal is shifted by 90 out of phase The two elements thus create a right hand circular polarized RHCP signal Two crossed dipoles with one port The signal on the second element horizontal is shifted by 90 out of phase The t
76. arization effects Zins 1 XPRa Zoh M 60 J 1 XPRy Zw Zhh T 60 where Z exp j U 7 7 is a random phase component However this does not account for all effects contributing to the polarization state of a MIMO radio link Thus a different method for calculating M based on linear transformations is proposed in the following 3 5 1 Relation between the Polarization Model and the Jones Calculus Jones invented a simple method to calculate polarization effects in optics 18 In his work he described the polarization state of a ray of light by the so called Jones vector Any object that changes the polarization or the intensity of the light is represented by a 2x2 Jones matrix It was found that the product of the Jones matrix of the optical element and the Jones vector of the incident light accurately describes the polarization state of the resulting ray Generally this calculus can be used for any electromagnetic wave It is especially interesting for the SCM and WINNER models where the paths are handled similarly like optical rays In the Jones calculus the Jones vector contains the x and y components of the amplitude and phase of the electric field traveling in z direction E t pjut A el ey O Ase we r Jones vector Copyright Fraunhofer Heinrich Hertz Institute 66 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 3 TECHNICAL DOCUMENTATION Transmitter position Simulation O
77. array to face sky wards b rotate_pattern 90 y b visualize 4 Plot the pattern of the Rxa Antenna Element 1 Element 2 Vertical Vertical Horizontal Horizontal 12 9 6 3 0 3 6 12 9 6 3 0 3 6 Attenuation dB Attenuation dB Figure 19 Antenna patterns for the satellite channel tutorial Copyright Fraunhofer Heinrich Hertz Institute 85 eMail quadriga hhi fraunhofer de 16 NH H U Ne Oo A oO MND wwn nnn nnn bd rar QuaDRiGa v1 2 32 458 A TUTORIALS Setting up the Layout In this step we combine the track the antennas and the position of the satellite into a simulation layout A layout object contains all the geometric information that are necessary to run the simulation First we define the position of the satellite Since the model uses Cartesian coordinates we have to transform the position of the satellite first 1 layout s 4 Create a new layout Z Choose a random satellite position Astra 2 seen from Berlin 4 The distance only needs to be big enough to ensure insignificant changes 4 in the reception angle on the ground sat_el 28 4 4 Elevation angle sat_az 161 6 Azimuth angle South 180 degree rx_latitude 51 4 Latitude of the Ra 4 Approzimate the satelite distance for GEO orbit dist_x 35786 rx_latitude 90 6384 h km dist_y 1 rx_latitude 90 6384 km sat_dist sqrt dist_x 2 dist_y 2 hk km sat_dist sat_dist 1e3 m Transfor
78. at the strongest peak with unit power corresponds to an angle 0 All other paths get relative departure or arrival angles depending on their power 1 Tb a sel Pi C L K y ep cart 5 l Aa Next two random variables X and Y are drawn where X 1 1 is the positive or negative sign and Y N 0 zay introduces a random variation on the angle The angles gil are then updated to oP Xil Y 96 If the power P of a path is small compared with the strongest peak its angle g might exceed 7 In this case it is wrapped around the unit circle by a modulo operation o oP r mod an T 97 In case of elevation spreads the possible range of elevation angles goes from 7 2 to 7 2 In this case the values gh need an additional correction This is done using 36 The positions of the Tx and Rx are deterministic and so are the angles of the LOS component This is taken into account by updating the values of the angles in order to incorporate the LOS angle A 98 PI g 4008 99 Finally the NLOS cluster paths are split into 20 sub paths to emulate intra cluster angular spreads Calculation of Cy L K The correction function C L K takes the influence of the KF and the varying number of clusters on the angular spread into account To approximate the function random powers P and angles are generated with the correction function set to Cy 1 The powers are calculated as described in
79. atellite orbit QuaDRiGas reference coordinate system is on the surface of the earth In order to use QuaDRiGa for satellite links the satellite position must be set Normally this position is given in azimuth and elevation relative to the users position This function takes a satellite orbital position and calculates the corresponding transmitter coordinates Input rx_latitude The receiver latitude coordinate on the earth surface in deg Default is 52 5 sat_el Satellite elevation in deg Default is 31 6 sat_az Satellite azimuth in deg given in compass coordinates Default is 180 south sat_height Satellite height in km relative to earth surface Default is 35786 GEO orbit tx_no The tx_no in the layout object for which the position should be set Default is 1 Output pos The satellite positions in the metric QuaDRiGa coordinate system visualize Description Plots the layout Copyright Fraunhofer Heinrich Hertz Institute eMail quadriga hhi fraunhofer de 40 QuaDRiGa v1 2 32 458 2 SOFTWARE STRUCTURE 2 2 5 Class parameter_set This class implements all functions that are necessary to generate and manage correlated LSPs It also provides interfaces for the channel builder LSPs are the shadow fading the Ricean K Factor the RMS delay spread and the four angles elevation and azimuth at the transmitter and receiver This class implements some core functions of the channel model and the user does nor
80. between 0 no overlap and 1 ramp along the entire segment A value of 0 disables the merging process and the channel segments are simply concatenated A value of 1 constantly merges the channels The default setting is 0 5 optimize The channel merger tries to automatically optimize the pairing of the taps i e one tap if the old segment ramps down and one of the new ramps up This is enabled by default but it is computing intensive For quicker results it can be disabled by setting optimize to 0 verbose Enables 1 default or disables 0 the progress bar Output c An array of channel objects containing the merged coefficients and delays chan _out split_tx varargin Description Splits channel arrays based on transmit antenna indices This function can be used to split large transmit antenna arrays into smaller arrays For example this can be used to calculate the channels for individual sectors at a BS Example A channel array has channels from three base stations BSs The first and second BS have two sectors each with two antennas However the sector antennas are merged into one array The third BS has only one sector To split the channels into five sectors the following command can be used cs c split 1 2 3 4 1 2 3 4 1 2 Notes e The order of the inputs must match the transmitters in alphabetical order i e the first input corresponds to Tx01 the second to Tx02
81. can be interpolated from the position based grid provided that the spatial sample theorem is not violated i e the channel needs to be sampled at least twice per half wave length In order to do that enough sample points are needed along the track INTERPOLATE_POSITIONS calculates the missing sample points and places them equally spaced along the track This corresponds to a constant speed when evaluating the output CIRs The required value for samples_per_meter can be obtained from the simulation_parameters object Input samples_per_meter the samples per meter e g from simulation_parameters samples_per_meter Copyright Fraunhofer Heinrich Hertz Institute 34 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 2 SOFTWARE STRUCTURE set_scenario scenario probability seg_length_min seg_length_mu seg_length_std Description Assigns random scenarios and creates segments This function can be used to create segments along the trajectory and assign scenarios to the segments If there are less than 3 input arguments i e only scenario and or probability is given then no segments will be created To create segments with the default settings call set_scenario scenario Input scenario A cell array of scenario names Each scenario synonym for propagation en vironment is described by a string e g MIMOSA _16 25_LOS or WIN NER_SMa_C1_NLOS
82. catterers are placed closer to the Rxs initial position This will amplify this effect Hence for correct time evolution results drifting needs to be turned on Copyright Fraunhofer Heinrich Hertz Institute 93 eMail quadriga hhi fraunhofer de wd QuaDRiGa v1 2 32 458 A TUTORIALS pow abs squeeze sum d coeff 1 1 2 5 72 plot distance 10 logi0 pow r xlabel Distance from track start point ylabel Tap power dB close all disp QuaDRiGa Version simulation_parameters version QuaDRiGa Version 1 0 1 145 100 0 i 0 4 Q 5t 100 SS S 1 l S 200 SSS 10 2 SS 2 5 300F E 1 g 15t 5 S J 2 g 400 F SS J 5 E 500 Ss e 8 E 600F 5 25L 700 NS 30F 800 900 i fi fi fi fi 35 fi fi L fi fi 20 25 30 35 40 45 50 20 25 30 35 40 45 Distance from track start point Distance from track start point Figure 24 Phases and Tx power vs Rx position without drifting drifting phases tutorial 50 Copyright Fraunhofer Heinrich Hertz Institute eMail quadriga hhi fraunhofer de 94 QuaDRiGa v1 2 32 458 A TUTORIALS A 5 Time Evolution and Scenario Transitions The channel model generates the coefficients separately for each segment In order to get a time continuous output these coefficients have to be combined This is a featur
83. cause g is influenced by both the delays T and the powers P r is usually calculated from measurement data Next the delays are normalized such that the first delay is zero and then they are sorted r sort frf min am 17 The NLOS path powers are drawn from a single slope exponential power delay profile PDP depending on the DS c and a random component Z N 0 3 The term is a scenario dependent coefficient emulating an additional shadowing process It is obtained from measurements T 1 z p exp n2 1040 18 The power of the first path is further scaled according to the initial KF from the map and path powers are normalized so that their sum power is one L 2 1 EE 19 l 2 2 _ pil P3 Pol 20 L P Pr 21 l 1 Copyright Fraunhofer Heinrich Hertz Institute 60 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 3 TECHNICAL DOCUMENTATION The scaling of the powers with the KF changes the DS Hence in the last step this is corrected by calculating the actual delay spread using the scaled powers and normalizing the delays in order to obtain the desired RMS delay spread in the PDP The DS after applying 21 is calculated to L L g g ectual SOP 3 T gt P n 22 l 1 l 1 With o being the initial DS from the map path delays note Or 2 a g actual a 23 3 3 Departure and Arrival Angles Four angles are calculated for each propagation pat
84. cedure linear interpolation can be used Alternatively more advanced techniques based on the effective aperture distribution function EADF are possible 19 Polarization rotation The second step takes the polarization into account Due to the rotation of the primary antenna the alignment with the probe changes The antenna patterns are defined in a polar spherical polarization basis However the rotation is defined in Cartesian coordinates Thus we need to perform the polarization rotation in a Cartesian basis as well The transformation from the polar spherical polarization basis to the Cartesian polarization basis is given by 20 FE 0 sinOcos sin e Fwvl 9 snOsin cos Ta 68 FI 6 cos 0 SLL moa F 0 6 The transformation matrix T O is both orthogonal and normalized to unity Hence the inverse transfor mation matrix is equal to the matrix transpose The rotated pattern F is obtained by using the interpolated pattern F and transforming it to a Cartesian polarization basis Then this pattern is rotated using the rotation matrix R and the resulting pattern is transformed back to the polar spherical basis F T 0 9 R T O F 6 69 M The entire process can be described by a 2x2 polarization rotation matrix M Due to the fact that the radiated energy in both polarizations remains constant and only the alignment with the probe changes this matrix is a r
85. ch path There are two different options If the delays are identical on the MIMO links i e individual_delays 0 then delay is a 2 D matrix with dimensions Path Snapshot If the delays are different on the MIMO links then delay is a 4 D tensor with dimensions Rx Antenna Tx Antenna Path Snapshot initial_position The snapshot number for which the initial LSPs have been generated Normally this is the first snapshot However if the user trajectory consists of more than one seg ment then initial_position points to the snapshot number where the current segment starts For example If initial_position is 100 then snapshots 1 99 are overlapping with the previous segment tx_position Position of each Tx in global cartesian coordinates using units of m rx_position The receiver position global cartesian coordinates using units of m for each snapshot no_rx Number of receive elements read only no_tx Number of transmit elements read only no_path Number of paths read only no_snap Number of snapshots read only Methods h_channel channel Ccoeff Cdelay Cinitial_position Description Creates a new channel object Input Ccoeff The complex valued channel coefficients for each path Cdelay The delays for each path Cinitial_position The snapshot number for which the initial LSPs have been generated Output h_channel A
86. change of the arrival angles and the path lengths the terminal will also see a change in its Doppler profile Stopping at traffic light LOS QuaDRiGa performs all internal calculations at a constant speed However a stop of the car at a traffic light is realized by interpolating the channel coefficients in an additional post processing step step 4 Here the user needs to supply a movement profile that defines all acceleration deceleration or stopping points along the track An example is given in section A 6 Since the interpolation is an independent step it makes no difference if the mobile terminal is in LOS or NLOS conditions Turning off with change of reception condition LOS NLOS This is realized by combining the methods of point 2 scenario change and point 4 turning without change The scenario change is directly in the curve Thus the LOS and the NLOS segments have an overlapping part where the cluster powers of the LOS segment ramp down and the NLOS clusters ramp up The update of the angles delays and phases is done for both segments in parallel Crossing side Street NLOS short LOS gt NLOS This is modeled by two successive scenario changes NLOS LOS and LOS NLOS For both changes a new set of clusters is generated However since the parameters for the two NLOS segments are extracted from the same map they will be highly correlated Thus the two NLOS segments will have similar properties Stru
87. channel angles and phases fading and K Factor coefficients over a short segment 4 G Transitions Postprocessing Analysis between segments Figure 5 Steps for the calculation of time evolving channel coefficients Compared to the WINNER model changes were made in the gray shaded boxes Time evolution requires a more detailed description of the mobility of the terminals This is done by assigning tracks i e ordered lists of positions to each MT In reality this may include accelerations decelerations and MTs with different speeds e g pedestrian and vehicular users However to minimize the computational overhead and memory requirements channel coefficients are calculated at a constant sample rate that fulfills the sampling theorem fr gt 2 Bp 4 max Afp 4 TEH 3 C where Bp is the width of the Doppler spectrum Afp is the maximum frequency change due to the velocity v and A is the carrier wavelength Thus the appropriate sampling rate is proportional to the maximum speed of the MT Since it is sometimes useful to examine algorithms at different speeds it is unfortunate to fix the sampling rate in advance as the speed is then fixed as well To overcome this problem channel coefficients are now calculated at fixed positions with a sampling rate fg measured in samples per meter In its normalized form it is known as sample density SD A time series for arbitrary
88. cludes the shadow fading Therefore the LSPs have to be calculated LSPs get then stored in layout track par This method is the most accurate The actual power in the channel coefficients can be up to 6 dB higher due to multipath effects Input method Link selection method Supported are all power and sf see above threshold If the Rx power is below the threshold in dBm the link gets deactivated tx_power A vector of tx powers in dBm for each transmitter in the layout This power is applied to each transmit antenna in the tx antenna array By default if tx_power is not given 0 dBm are assumed check_parfiles Disables 0 or enables 1 default the parsing of shortnames and the validity check for the config files This is useful if you know that the parameters in the files are valid In this case this saves execution time Output pairs An index list of links for which channel are created The first row corresponds to the Tx and the second row to the Rx An identical copy gets assigned to layout pairing power A matrix containing the estimated receive powers for each link in dBm Rows cor respond to the receiving terminal columns correspond to the transmitter station For MIMO links the power of the strongest MIMO sublink is reported pos set_satellite_pos rx_latitude sat_el sat_az sat_height tx_no Description Calculates the Tx position from a s
89. cluster and scatterer are used synonymously A cluster describes an area where many scattering events occur simultaneously e g at the foliage of trees or at a rough building wall In QuaDRiGa each scattering cluster is approximated by 20 individual scatterers Each one is modeled by a single reflection The 20 signals can be resolved in spatial domain where they have a typical angular spread of 1 6 However they cannot be resolved in delay domain Therefore in the output of the channel model these 20 signals also named sub paths are combined into a single signal which is represented by a path The difference to Rayleigh fading models which use wide sense stationary uncorrelated scattering WSSUS taps instead of paths is that each path has a very limited angular spread 1 6 which also results in a narrow Doppler spectrum The terms path multipath component MPC and tap are also used synonymously in the QuaDRiGa documentation To emulate a rich scatting environment with a wider angular spread many scattering clusters are created QuaDRiGa supports up to 42 clusters However depending on the angular spread and the amount of diffuse scatting which is approximated by discrete clusters in QuaDRiGa typical values are around 10 cluster for LOS propagation and 20 clusters for non LOS The positioning of the clusters is controlled by the environment angular spread and the delay spread The environment angular spread has values of around 20 90 a
90. cription Creates a new parameter_set object Input scenario The scenario name for which the parameters should be loaded A list of supported scenarios can be obtained by calling parameter_set supported_scenarios positions The list of initial positions for which LSPs should be generated check_parfiles check_parfiles 0 1 default 1 Disables 0 or enables 1 the parsing of shortnames and the validity check for the config files This is useful if you know that the parameters in the files are valid In this case this saves execution time Output h_parset A parameter_set object copy_objects Description A modified version of the standard physical copy function While the standard copy command creates new physical objects for each element of obj in case obj is an array of object handles copy_objects checks whether there are object handles pointing to the same object and keeps this information angles get_angles Description Calculates the departure and arrival angles of the LOS path between Tx and Rx Output angles A matrix containing the four angles Azimuth of Departure at the Tx AoD row 1 Azimuth of Arrival at the Rx AoA row 2 Elevation of Departure at the Tx EoD row 3 Elevation of Arrival at the Rx EoA row 4 The number of columns corresponds to the number of rx positions h_channel h_cb get_channels Description Calculate the cha
91. ctural change in the environment without a change in the environment type higher density of buildings but still the environment remains urban This is not explicitly modeled However the Satellite NLOS Urban map covers a typical range of parameters E g in a light NLOS area the received power can be some dB higher compared to an area with denser buildings The placement of light dense areas on the map is random Thus different characteristics of the same scenario are modeled implicit They are covered by the model but the user has no influence on where specific characteristics occur on the map when using the automatic mode An alternative would be to manually overwrite the automatically generated parameters or use the manual mode In order to update the LSPs and use a new set of parameters a new segment needs to be created I e here an environment change from Satellite NLOS Urban to the same Satellite NLOS Urban has to be created Thus a new set of LSPs is read from the map and new clusters are generated accordingly Stopping at traffic lights NLOS This is the same as in point 5 Structural change in environment Houses have the same characteristics as before but are further away from the street urban environment with different reception characteristics Same as point 8 Copyright Fraunhofer Heinrich Hertz Institute 18 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 1 INTRODUCTION AND OVERVIEW
92. d and scattered signal The clusters are then represented as taps in the final CIR The random variables fit the distributions and correlations defined by the parameter_set object Next antenna dependent parameters are extracted for each user Those depend on the position of the terminal its orientation and the equipped antennas The polarization rotation of the NLOS taps is modeled by a random variable which fits to the distribution defined by the parameter_set The LOS polarization is calculated from the geometric orientation of the antennas A core function here is the interpolation of the antenna patterns which results in a specific H and V value for each subpath The core function then generates the coefficients themselves This is done for each antenna element and for each snapshot separately and also includes the Doppler shift of each subpath Finally the K factor and the shadow fading are applied and a all the data is returned as an channel object Properties name Name of the channel_builder object par The parameter_set object for this channel builder taus The initial delays for each path in s Rows correspond to the MTs columns to the paths pow The normalized initial power squared average amplitude for each path Rows correspond to the MTs columns to the paths The sum over all columns must be 1 AoD The initial azimuth of departure angles for each path in
93. d of arrival in deg for each receiver position xpr The cross polarization ratio linear scale for each receiver position ds_map The RMS delay spread map in log10 s kf_map The Ricean K Factor map logarithmic scale sf_map The shadow fading mpa logarithmic scale asD_map The azimuth spread of departure map in log10 deg asA_map The azimuth spread of arrival map in log10 deg esD_map The elevation spread of departure map in log10 deg esA_map The elevation spread of arrival map in log10 deg map _extension map_extent map size samples_per_meter map _valid LSP_xcorr_matrix LSP_matrix_isOK map _x_coord map _y_coord Distance in m that is added to each direction when generating maps Extent of the mpas in x and y direction xmin xmax ymin ymax in m Number of map pixels in x and y direction n_x_samples n_y_samples Resolution of the decorrelation maps in samples m This value is obtained from simulation_parameters map_resolution Indicates if maps contain valid data The Cross correlation matrix for the LSPs Determines if the XCorr matrix is positive definite The x coordinates in m for each pixel of the maps The y coordinates in m for each pixel of the maps Copyright Fraunhofer Heinrich Hertz Institute eMail quadriga hhi fraunhofer de 41 QuaDRiGa v1 2 32 458 2 SOFTWARE STRUCTURE Methods h_parset parameter _set scenario positions check_parfiles Des
94. del software Copyright Fraunhofer Heinrich Hertz Institute eMail quadriga hhi fraunhofer de 21 QuaDRiGa v1 2 32 458 2 SOFTWARE STRUCTURE 2 2 Description of Classes Properties and Methods In the following all properties and methods of the QuaDRiGa classes are described For the methods input and output variables are defined and explained There are three types of methods Standard methods require an instance of a class They are printed in black without the class name par generate_parameters overlap usage check_parfiles verbose Static methods can be called directly from the command line without creating an instance of the class first They are printed in blue h_array mse mse_pat array import_pattern fVi fHi correct_phase accuracy max_num_elements azimuth_grid elevation_grid verbose The constructor is a special method that is called when the class name is used as a function e g when calling a array dipole There is only one constructor for each class They are printed in blue h_array array array_type phi_3dB theta_3dB rear_gain Copyright Fraunhofer Heinrich Hertz Institute 22 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 2 SOFTWARE STRUCTURE 2 2 1 Class simulation_parameters This class controls the simulation options and calculates constants for other classes Properties sample_density The numb
95. dy the parameters at a wider range when evaluating the results p l create_parameter_sets 0 p plpar p scenpar xpr_mu 100 h Disable XPR p scenpar xpr_sigma 0 p scenpar KF_mu 5 4 Increase KF Range p scenpar KF_sigma 15 p scenpar DS_mu 1log10 0 6e 6 4 Median DS 600 ns p scenpar DS_sigma 0 3 4 300 1200 ns range p update_parameters c p get_channels coeff squeeze cat 1 c coeff delay permute cat 3 c delay 3 1 2 Parameters 00000000000000000000000000000000000000000000000000 J 5 seconds Channels 00000000000000000000000000000000000000000000000000 J 8 seconds Copyright Fraunhofer Heinrich Hertz Institute 76 eMail quadriga hhi fraunhofer de 10 QuaDRiGa v1 2 32 458 A TUTORIALS Results and discussion In the following four plots we extract parameters from the generated coefficients and compare them with the initial ones which were generated by the parameter set object P The values in P can be seen as a request to the channel builder and the values in the generated coefficients C as a delivery We first calculate the SF from the channel data by summing up the power over all 20 taps We see that the values are almost identical sf sum mean abs coeff 2 3 2 figure plot 35 35 35 35 k hold on plot 35 35 3 35 35 k plot 35 35 3 35 35 k plot 10 logi0 p sf 10 logi0 sf
96. e GreenTouch consortium within the funded project LSAS Channel Modelling http www greentouch org Acknowledgements The authors thank G Sommerkorn C Schneider M Kaeske Ilmenau University of Technology IUT Ilmenau Germany and V Jungnickel Heinrich Hertz Institute HHI Berlin Germany for the fruitful discussions on the QuaDRiGa channel model and the manuscript of this document How to Cite this Document 2 S Jaeckel L Raschkowski K Borner L Thiele F Burkhardt and E Eberlein QuaDRiGa Quasi Deterministic Radio Channel Generator User Manual and Documentation Fraun hofer Heinrich Hertz Institute Tech Rep v1 2 32 458 2015 Copyright Fraunhofer Heinrich Hertz Institute 2 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 Contents Contents 1 Introduction and Overview 11 1 1 Installation and System Requirements 0 000 ee ee 11 1 2 General Remarks aoaaa a 11 L Introduction to QuaDRiGa spoe se wa ew a a g be Pe ws Bm ee Ge a ee 12 1A Continuous time evoluti n s a s sus ee Bee Be ed ee en he ee ee a ate 14 Lb QuaDRiGa Program FloW go sg ie as ee eg re Ba ce Ve ae ee ee Ge 15 1 6 Description of modeling of different reception conditions by means of a typical drive course 16 2 Software Structure 20 2 OWervieW s e cwd Baca wee a a a ER eg Re Re RS ee RO ee 20 2 2 Description of Classes Properties and Methods 0 020002 eee 22 22 1 Class simu
97. e angles should be distributed around the LOS direction we need to subtract 0 05 before scaling the angles and add it again after scaling them 5 _ 98 4 _ gLos LOS 0 actual 6 4 0 37 Copyright Fraunhofer Heinrich Hertz Institute 62 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 3 TECHNICAL DOCUMENTATION This update however might shift some angles outside the allowed interval especially if there is already a strong bias due to the LOS path which might happen e g in satellite scenarios when the satellite has a high elevation Hence angles outside the allowed range are replaced with 25 e ito lt F 38 o otherwise actual As for the azimuth angles equations 35 to 38 must be applied in an iterative fashion until o converges either to the given value og or a maximum value Maximal achievable angular spread As already mentioned there is a upper limit for the AS due to the wrapping around the unit circle For two equally strong paths the maximum azimuth spread is achieved when the paths face opposite directions Hence in order for 26 to be zero the two angles must be 5 and In this case the AS 28 becomes 7 2 127 For three paths the maximum azimuth spread would be 170 and for four paths 220 With increasing KF however the maximum AS decreases since more power is allocated to the LOS path For example with a KF of 10 dB the maximum azimuth spread is only 57 H
98. e channel coefficients to include such effects This is illustrated in the bottom part of Figure 12 The white dots represent the snapshots at a constant distance However the sample points gray stars can have unequal spacing e g for an accelerated movement Each sample point in the time domain given in units of seconds has a corresponding position on the MT trajectory in units of meters The amplitudes and phases of the channel coefficients are interpolated separately using cubic spline interpolation The path delays are interpolated with a piecewise cubic hermite interpolating polynomial Copyright Fraunhofer Heinrich Hertz Institute 74 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 A TUTORIALS A Tutorials In the following we provide a variety of tutorials that can get you started with QuaDRiGa You can also use the MATLAB Help to access these files A 1 Network Setup and Parameter Generation The channel model class parameter_set generates correlated values for the LSPs The channel builder then uses those values to create coefficients that have the specific properties defined in parameter_set One important question is therefore Can the same properties which are defined in parameter_set also be found in the generated coefficients This is an important test to verify if all components of the channel builder work correctly Channel model setup and coefficient generation We first set up the basic p
99. e far field of each antenna element in the array Fc The third component of the antenna pattern Currently it is only used when the antenna pattern is using a cartesian polarization basis Copyright Fraunhofer Heinrich Hertz Institute 25 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 2 SOFTWARE STRUCTURE coupling Coupling matrix between elements This matrix describes a pre or postprocessing of the signals that are fed to the antenna elements For example in order to transmit a left hand circular polarized LHCP signal two antenna elements are needed They are then coupled by a matrix 5 J2 3 The rows in the matrix correspond to the antenna elements the columns to the signal ports In this example the antenna has one port i e it is fed with one input signal This signal is then split into two and fed to the two antenna elements where the second element radiates the signal with 90 phase shift In a similiar fasion it is possible to create fixed beamforming antennas and include crosstalk between antenna elements By default coupling is set to an identity matrix which indicates perfect isolation between the antenna elements iscompressed Indicates if the array is compressed It is possible to compress the antenna array in memory to save storage space and relay the memory requirements for large arrays This property indicates it the array is compressed no_az Number of azimut
100. e originally described in the documentation of the WIM2 channel model although it was never implemented Since this component is needed for time continuous simulations it was implemented here This script sets up the simulation and creates such time continuous CIRs Channel model setup and coefficient generation First we set up the channel model close all clear all set 0 defaultTextFontSize 14 set 0 defaultAxesFontSize 14 simulation_parameters center_frequency 2 53e9 Sample_density 4 use_absolute_delays 1 nnn DN Second we create a more complex network layout featuring an elevated transmitter 25 m and two receivers at 1 5 m height The first Rx moves along a circular track around the receiver The second receiver moves away from the Tx Both start at the same point Note that each track is split into three segments The first Rx goes from an LOS area to a shaded area and back The second track also starts in the LOS area Here the scenario changes to another LOS segment and then to an NLOS segment The LOS LOS change will create new small scale fading parameters but the LSPs will be highly correlated between those two segments 1 layout s 4 New layout 1l no_rx 2 4 Two receivers 1 tx_array generate dipole 4 Dipola antennas at all Ra and Tz l rx_array 1 tx_array 1 tx_position 3 25 4 Elevate Tx to 25 m UMal BERLIN_UMa_LOS UMan BERLIN_UMa_NLOS
101. e same for all NLOS paths I e vertically polarized waves remain vertically polarized after scattering On the other hand a value of Inf dB means that the polarization is turned by 90 In case of 0 dB the axis is turned by 45 i e the power of a vertically polarized wave is split equally into a H and V component The following table gives an overview of the parameters in the config files They get converted into a structure parameter_set scenpar Parameter Unit or type Description plpar model Text string Selects the model for average path loss For satellite applications the pathloss is PL_model defined by the satellite distance and is assumed to be constant for the reception are For terrestrial cases pathloss models like Hata or others e g WINNER pathloss models can be selected plpar A dB For satellite applications this parameter defines the average path loss and is equiv PL_A alent to the mu of the lognormal distribution of the shadow fading Parameters in structure parameter_set scenpar Large Scale Parameters Those parameters describe how the large scale parameters vary within a propaga tion environment SF_sigma dB Those parameter describe the slow fading implemented as Lognormal distribu SF_lambda meter tion and filtered see map generation according to the de correlation distance lambda KF mu dB Statistical properties of the K factor
102. e standard copy command creates new physical objects for each element of obj in case obj is an array of object handles copy_objects checks whether there are object handles pointing to the same object and keeps this information correct_overlap overlap Description Corrects positions of the segment start to account for the overlap After the channel coefficients are calculated adjacent segments can be merged into a time continuous output The merger assumes that the merging interval happens at the end of one segment before a new segments starts In a reality however the scenario change happens in the middle of the overlapping part and not at the end of it This function corrects the position of the segment start to account for that Input overlap The length of the overlapping part relative to the segment length It can have values in between 0 no overlap and 1 ramp along the entire segment generate linear track_length direction Description Creates a linear track with given length and direction Direction describes the travel direction along the track in rad in mathematical sense i e 0 means east pi 2 means north pi means west and pi 2 south If track_length or direction is not specified then the default track is 1 m long and has a random direction Input track_length The track length in m Default length is 1 m direction specifies the driving direction i
103. e the merging took place In those areas the power is scaled linearly That means that for example in between 7 5 and 10m the power ramps up from 97 to 82 dBm power squeeze sum abs c coeff 2 3 power 10 1l0g10 power figure dist t get_length plot dist power title Simulated Power xlabel Distance from start point m ylabel Received Power dBm axis 0 20 110 80 grid on Copyright Fraunhofer Heinrich Hertz Institute 112 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 A TUTORIALS Simulated Power 80 r r r Received Power dBm l Kej wn 105 5 10 15 20 Distance from start point m 110 0 Figure 36 Power along the track manual parameter selection The last plot shows the DS along the path The results reflect the settings of 0 45 0 33 0 12 and 0 60 quiet well As for the power there is an overlap in between the segments For example in between 7 5 and 10m the DS drops from 0 33 to 0 12 microseconds Additional fluctuations are caused by small scale fading coeff squeeze c coeff delay c delay pow_tap abs coeff 72 pow_sum sum pow_tap mean_delay sum pow_tap delay pow_sum ds sqrt sum pow_tap delay 2 pow_sum mean_delay 2 figure plot dist ds 1e6 title Simulated Delay Spread xlabel Distance from start point m ylabel RMS DS
104. e the polarization rotation model that calculates the path power for polarized antenna arrays We do this by setting up the simulation with different H V polarized antennas at the transmitter and at the receiver Then we define a circular track around the receiver When the receiver moves around the transmitter it changes its antenna orientation according to the movement direction In this way all possible departure and elevation angles are sampled Depending on the antenna orientation the polarizations are either aligned e g the Tx is V polarized and the Rx is V polarized they are crossed e g the Tx is V polarized and the Rx is H polarized or the polarization orientation is in between those two The generated channel coefficients should reflect this behavior Setting up the simulation environment First we have to set up the simulator with some default settings Here we choose a center frequency of 2 1 GHz We also want to use drifting in order to get the correct angles for the LOS component and we set the number of transmitters and receivers to one close all clear all set 0 defaultTextFontSize 14 set 0 defaultAxesFontSize 14 s simulation_parameters 4 Set the simulation parameters s center_frequency 2 1e9 4 Center frequency 2 1 GHz s use_polarization_rotation 1 s samples_per_meter 360 40 pi 4 One sample per degree s drifting_precision 1 Setting up the antenna arrays In the second step we
105. e we choose a linear track with a length of 30 m The track start 20 m east of the transmitter and runs in east direction thus linearly increasing the distance from the receiver 1 layout s 1 tx_position 3 25 l track generate linear 30 0 l track initial_position 20 0 0 1l track scenario WINNER_UMa_C2_L0S8 l track interpolate_positions s samples_per_meter 1l visualize Tx Position 257 gi A Tx Antenna O Rx Position 20L J V_ Rx Antenna Rx Track gS s Z OfFx1 oa WINNER UMa 2 LOS 4 gt ee J 10 J 154 J 20H 4 254 J i i i fi i i 0 10 20 30 40 50 X Position Figure 21 Scenario setup for the drifting phases tutorial Now we generate the LSPs In order to get repeatable results we set a specific random seed This is a MATLAB internal function and is not a feature of the channel model We also set the shadow fading and K factor to 1 and disable the path loss model Copyright Fraunhofer Heinrich Hertz Institute 91 eMail quadriga hhi fraunhofer de ao Pw Ny QuaDRiGa v1 2 32 458 A TUTORIALS RandStream setGlobalStream RandStream mt1i9937ar seed 5 p l1 create_parameter_sets p parameter_maps 2 3 0 4 Fix SF and KF to 0 p plpar 4 Disable path loss model p update_parameters Parameters 00000000000000000000000000000000000000000000000000 1 seconds Now we generate the c
106. e_parameter_sets disp Drifting enabled s drifting_precision 1 RandStream setGlobalStream RandStream mti9937ar seed 1 c p get_channels cn c merge disp Drifting disabled s drifting_precision 0 RandStream setGlobalStream RandStream mti9937ar seed 1 d p get_channels dn d merge Parameters 00000000000000000000000000000000000000000000000000 1 seconds Drifting enabled Channels Looo00000000000000000000000000000000000000000000000 36 seconds Drifting disabled Channels 00000000000000000000000000000000000000000000000000 5 seconds Results and discussion Now we plot and discuss the results We start with the power of the LOS tap along the circular track and compare the outcome with and without drifting Fig 26 left degrees 0 cn 1 no_snap 1 cn 1 no_snap 360 los_pwr_drift 10 1log10 squeeze abs cn 1 coeff 1 1 1 72 los_pwr_nodrift 10 1log10 squeeze abs dn 1 coeff 1 1 1 72 figure plot degrees los_pwr_drift hold on plot degrees los_pwr_nodrift r hold off a axis axis 0 360 a 3 4 xlabel Position on circle degree ylabel Power of the LOS component title Power of the LOS component for the circular track legend Drifting No drifting 4 Copyright Fraunhofer Heinrich Hertz Institute 96 eMail quadriga hhi fraunhofer de N He Q
107. ears Oe tose 47 6 ws Srei Qr l m s z 48 ss lates Since we assume a static scattering environment we use the same departure angles for all Tx elements The phases and path delays however depend on the total path length dr tim To obtain this value we need to calculate the vector b m from the vectors r and aj at r s 1 bim r am 49 drl m s bi ml eee 50 Finally we calculate the phases w and path delays 7 20 Wr im s ps Caer mod A 51 1 20 Trls 20 c D dr lms 52 LOS drifting The direct component is handled differently since we have to update the angles at both the Tx and the Rx sides The angles are updated for each combination of Tx Rx antenna elements based on the position of the element in 3 D coordinates Prts T ers 53 Pias arctang rr tsy Trte 54 r Ofi arcsin eta 55 ee Erts Pis arctanz rr tsy Tr t s 0 56 r Sis arcsin i 57 j Erts Copyright Fraunhofer Heinrich Hertz Institute 65 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 3 TECHNICAL DOCUMENTATION The vector ry t points from the location of the Tx element t to the location of the Rx element r at snapshot s The phases and delays are determined by the length of this vector and are calculated using 51 and 52 where d j m s is replaced by r t s 3 5 Antennas and Polarization Next antenna patterns polarization and phases are combined in order to calculate
108. eate_parameter_sets 4 Create parameter sets p parameter_maps 2 3 4 Fis KF to 3 dB p parameter_maps 3 0 4 Fina SF to 0 dB p plpar 4 Disable path loss model p update_parameters c cb p get_channels 4 Get the channel coefficients Parameters 00000000000000000000000000000000000000000000000000 1 seconds Channels Looo00000000000000000000000000000000000000000000000 3 seconds Copyright Fraunhofer Heinrich Hertz Institute 105 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 A TUTORIALS Results and Evaluation We now check the results and confirm if they are plausible or not We start with the two vertically polarized dipoles at the Tx and at the Rx side The model creates 8 taps which is the default for the Urban Macro LOS scenario Without path loss and shadow fading SF 1 the power is normalized such that the sum over all taps is 1W and with a K Factor of 3 dB we get a received power of 0 67W for the LOS component The remaining 0 33W are in the NLOS components The results can be seen in Fig 33 top left New figure Plot the graph Set the asis Add description figure plot abs squeeze c coeff 1 1 72 axis 0 360 0 1 1 xlabel Position degrees ylabel LOS Power linear scale title Tx Vertical Rx Vertical Sx x Add title disp LOS power num2str mean abs c coeff 1 1 1 72 4 disp NL
109. ed The channel coefficients of adjacent segments are combined merged This includes the birth death process of clusters Additionally different speeds of the terminal can be emulated by interpolation of the channel coefficients The channel coefficients together with the path delays are formatted and returned to the user for further analysis 1 6 Description of modeling of different reception conditions by means of a typical drive course This section describes some of the Key features of the model using a real world example A detailed introduction with a variety of tutorials test cases and interface descriptions then follows in section A The later part of the document then focusses on the mathematical models behind the software and the assumptions made a atellit RHCP signal Figure 2 Typical driving course From home to woodland parking site on the village outskirts The different effects along the track can be summarized as follows Ole Ne Start Environment Urban LOS reception of satellite signal LOS NLOS Change NLOS LOS Change Turning off without change in reception condition LOS Stopping at traffic light LOS Turning off with change of reception condition LOS gt NLOS Crossing side Street NLOS short LOS NLOS Copyright Fraunhofer Heinrich Hertz Institute 16 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 1 INTRODUCTION AND OVERVIEW 8 Structural ch
110. ed by merging the channel coefficients of adjacent segments The active time of a scattering cluster is confined within the combined length of two adjacent segments The power of clusters from the old segment is ramped down and the power of new clusters is ramped up within the overlapping region of the two segments The combination clusters to ramp up and down is modeled by a statistical process Due to this approach there are no sudden changes in the LSPs For example if the delay spread in the first segment is 400 ns and in the second it is 200 ns then in the overlapping region the delay spread DS slowly decreases till it reaches 200 ns However this requires a careful setup of the segments along the used trajectory If the segments are too short sudden changes cannot be excluded This process is described in detail in Section 3 8 1 5 QuaDRiGa Program Flow For a propagation environment e g urban suburban rural or tree shadowing typical channel characteris tics are described by statistics of the LSPs Those are the median and the standard deviation of the delay spread angular spreads shadow fading Ricean K Factor as well as correlations between them Additional parameters describe how fast certain properties of the channel change i e the decorrelation distance Those parameters are stored in configuration files which can be edited by the model user Normally the parameters are extracted from channel measurements A detailed descripti
111. ee BS 69 70 0 Elevation angle of arrival EoA in rad ooa 61 64 g3 Elevation angle of departure EoD in rad ooo aaa 61 64 C A coefficient describing additional shadowing within a PDP 60 a Vector pointing from the position of the LBS to the Rx position 65 70 B Bandwidth in units of Hz sos se ae 84 eR oe RE ae ES RE ee ES EY 56 B LSP map represented as a matrix with real valued coefficients 58 b Vector pointing from the Tx position to the position of the LBS 65 c Speed of Light o s ae smo riea Gua a a oaa b Re a RE aha a e hae Ge 64 65 c Representation of the departure or arrival angle in Cartesian coordinates 68 Copyright Fraunhofer Heinrich Hertz Institute 7 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 References Co The scenario dependent cluster wise RMS angular spread in degree 63 d Length of a propagation path in m 2 44 645 wale a 64 73 dy Decorrelation distance in m where the autocorrelation falls below e7 58 ers Vector from the Rx position to Rx antenna element r at snapshots 65 F Polarimetric antenna response 1 a a a 66 69 fs Sampling Rate in samples per meter 2 2 eee ee ee 57 fr Sampling Rate in samples per second 1 0 eee 56 g Channel coefficient in time domain 2 a 00000 eee 66 K Ricean K Factor linear scale 2 a a 60 114 k Filter
112. ee Se a a a ee ee ke be ee ee ws a 56 64 Time evolution describes how the propagation channel changes or evolves with time In the channel model two effects are used to describe this time dependency drifting and the birth and death of scattering clusters during the transition between segments The propagation environment is considered static and thus the model includes time evolution only when the receiver is moving USET erh I Sess Oe Se a a SO ee ee ek ee oe Be ee ts Be ee 58 Synonym for mobile terminal List of Symbols y Polarization rotation angle for the linear NLOS polarization in rad o ooo aaa 71 A Wavelength in units of m ic ooo a 56 65 o Azimuth angle in rad can be used for f or onoo 61 62 66 114 115 gp Azimuth angle of arrival AGA in frad og ek a a SG SE we oe a 61 63 64 ot Azimuth angle of departure AoD in rad 2 0 2 0 020 61 63 64 db The offset angle between the path angle of the mt sub path in degree 2 2 63 yY Phase of a path in frad ooa bow de oe ROS hed ad dH 65 72 p Correlation coefficient 2 60 Tg The RMS angular spread in rad o0 a a 62 114 115 Or The RMS delay spread in units of s 2 2 ee ee ee ee 57 60 61 T Delay of a MPC in units of s 2 42 as bod ek Sa ee ek ee ee A 60 61 64 65 0 Elevation angle in rad 0 can be used for 0f or 0 o oo aaa a 62 66 0 Polarization rotation angle in rad 44 4 ga ey ee DR ee ee ee we
113. ency s sample_density 4 4 4 samples half wave length 1 layout s 4 Create Layout Setting up a user track QuaDRiGa needs the positions of transmitter and receiver e g for calculating the polarization or the arrival and departure angels The positioning information of the Tx and Rx is essential also when the LSPs are not calculated The following generates a linear track with 20 m length having a direction The track is further split into 4 segments of 5 m length The splitting is done by calling the method split_segment of the track object The the first two arguments of that function are the minimum length of a segment 1 m and the maximum length of the segment 6 m Each existing segment that is longer then the maximum length is split into subsegments The length of those segments is random where the third and fourth parameter determine the mean and the STD of the length of new subsegment Hence t split segment 1 6 5 0 splits all segment longer than 6 m into subsegments of 5 m length Each segment gets assigned a scenario This is also essential since many parameters such as the number of clusters the XPR etc are scenario specific Hence they are the same for the entire scenario Here we set the first the segments to NLOS the third to LOS and the last to NLOS Last we set a random starting position for the track in the layout 1 tx_position 0 0 25 4 Set Tx position t track linear 20 4 L
114. er of samples per half wave length Sampling density describes the number of samples per half wave length To fulfill the sampling the orem the minimum sample density must be 2 For smaller values interpolation of the channel for variable speed is not possible On the other hand high values significantly increase the computing time significantly A good value is around 4 samples_per_meter Samples per meter This parameter is linked to the sample density by SD Ja 2 fon g where fc is the carrier frequency in Hz SD is the sample density and c is the speed of light drifting_precision Precision of the drifting functionality drifting_precision 0 This method applies rotating phasors to each path which emulates time varying Doppler characteris tics However the large scale parameters departure and arrival angles shadow fading delays etc are not updated in this case This mode requires the least computing resources and may be preferred when only short linear tracks up to several cm are considered and the distance between transmitter and receiver is large The phases at the antenna arrays are calculated by a planar wave approximation drifting precision 1 default When drifting is enabled all arrival angles the LOS departure angle delays and phases are updated for each snapshot using a single bounce model This requires significantly more computing resources but also increases the accuracy of the results Drif
115. er triangular part of the matrix is ignored and replaced by a transpose of the upper triangular matrix scenarios file names file_dir parameter_set supported_scenarios parse_shortnames Description Returns a list of supported scenarios Input parse_shortnames This optional parameter can disable 0 the shortname parsing This is significantly faster By default shortname parsing is enabled 1 Output scenarios A cell array containing the scenario names Those can be used in track scenario file names The names of the configuration files for each scenario file_dir The directory where each file was found You can place configuration file also in you current working directory update_parameters force Description Generates the LSP maps and updates the parameters for all terminals This function calculates correlated large scale parameters for each user position Those parameters are needed by the channel builder class to calculate initial parameters for each track or segment which are then evolved into time varying channels By default update_parameters reads the values given in the track objects of the layout If there are no values given or if parts of the values are missing the correlation maps are generated to extract the missing parameters Input force Changes the behavior of the function force 0 default Tries to read the parameters from layout track par If
116. escription Uncompresses an array See array compress visualize element Description Create a plot showing the element configurations Input element The element numbers for which the plot os created If no element number s are given a plot is created for each element in the array Copyright Fraunhofer Heinrich Hertz Institute 30 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 2 SOFTWARE STRUCTURE 2 2 3 Class track One feature of the channel model is the continuous evolution of wireless channels when the terminal moves through the environment A track describes the movement of a mobile terminal It is composed of an ordered list of positions During the simulation one snapshot is generated for each position on the track Along the track wireless reception conditions may change e g when moving from an area with LOS to a shaded area This behavior is described by segments or states A segment is a subset of positions that have similar reception conditions Each segment is classified by a segment index i e the center position of the segment and a scenario The scenario must be one of the supported scenarios in class parameter_set Properties name Name of the track initial position Position offset will be added to positions This position is given in global cartesian coordinates x y and z component in units of m The initial position normally refers t
117. ew Radio wide sense stationary wide sense stationary uncorrelated scattering cross polarization ratio Copyright Fraunhofer Heinrich Hertz Institute eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 Glossary Glossary base station BS eo se bea 4362444 2 eC ede pokok SRE eoa aS PRESSE LSE EES 56 The term base station BS refers to a fixed transmitter which utilizes one or more transmit antennas to serve one or more MTs BSs might further use sectors to increase the capacity Usually BSs operate independent of each other which might lead to inter BS interference if they use the same time and frequency resource Synonym for sector Cluste 2 ch RAHA RS A EB RR SEY RRA EES we eH RG lee wR SS 70 73 A cluster describes an area where many scattering events occur simultaneously e g at the foliage of trees or at a rough building wall In the channel model each scattering cluster is approximated by 20 single reflections Each of those reflections has the same propagation delay Ari Se ee Se ad Ee Been ee oe ee he oe a ee ee a 64 Drifting occurs within a small area about 20 30 m diameter in which a specific cluster can be seen from the MT Within this area the cluster position is fixed Due to the mobility of the terminal the path length resulting in a path delay and the arrival angels change slowly i e they drift large scale parameter LSP 224 6642 ee ee Ae Ew RAE EE Se eS OE es 57 The term large scale par
118. fix the received power to 102 97 82 and 99 dBm K Factors are taken from the map p 1 plpar 4 Disable path loss for NLOS p 2 plpar Z Disable path loss for LOS p 1 sf 10 7 0 1 102 97 99 4 Set power for NLOS segments p 2 sf 10 7 0 1 82 4 Set power for LOS segments p 1 map_valid false 4h Disable automatic overwrite for NLOS p 2 map_valid false Disable automatic overwrite for LOS Calculate channel coefficients Now we calculate the coefficients and the received power along the path The following command calculate the channel coefficients We then check the number of clusters that have been produced for each segment cm p get_channels 4 Calculate the channel coefficients cat 1 cm no_path 4 Display the number of paths Channels 00000000000000000000000000000000000000000000000000 7 seconds ans 14 14 14 8 The first three segments have 14 clusters This has been set earlier The last LOS segment has 15 clusters One can see that the three NLOS segment come first followed by the LOS segment The order has been scrambled The following command sorts and combines the segments into one fading sequence c cm merge We now plot the power along the path You can see the power levels of around 102 97 82 and 99 dBm which have been set earlier Each segment has a transition area e g from 2 5m to 5m from 7 5m to 10m and from 12 5m to 15m wher
119. ft between the vertical and horizontal component to obtain an elliptic XPR a reflection operation and a rotation operation to obtain the desired linear XPR Last the change in the receiver orientation is included by a second rotation matrix M 0 1m s The complete polarization transfer matrix notes T n 1 0 n as Os l ll Mironis M Orima Mim 0 1 Mel M Dr tm s 83 If the same XPR is used for V and H component the equation can be simplified to Vim Brims Vm 84 MNEOS _ cos Yim SLY exp JKim 0 cost sin Dt rtlm s sin ae cosy 0 exp JKim sin Vt Ccosvts Obtaining initial XPR values In the model the simplified option 2 a is used to keep the compatibility with the WINNER parameter tables where the same XPR value is applied to both the V and H component Statistics for the values for XPR and XPR are extracted from measurements When those parameters are calculated from measured data they are usually averaged over different propagations paths Thus the XPR value from the parameter tables is a LSPs having a scenario dependent distribution i e it depends on the positions of the MT However here we need to draw a values xPRE B for individual MPCs If this is done using 75 then the value for XPR obtained from the generated cau coefficients will effectively become zero due to the averaging over several MPC and snapshots Therefore a two step approach is used here First a value xP is drawn fr
120. ftware 2 2 0 e 21 QuaDRiGa Data Plow o poq saci gon aaia eo Goa ke Pee ew EA ee A 49 Steps for the calculation of time evolving channel coefficients 56 Principle of the generation of channel coefficients based on correlated LSPs 58 Map based 2 D autocorrelation shaping using FIR filters 204 59 Maximal achievable angular spread depending on the K factor 63 Scatterer positions and arrival angles 2 a a ee 64 Example showing the effect of the new polarization rotation method 67 Example patterns for a dipole antenna 1 e 68 Illustration of overlapping segments and variable MT speeds 008 4 74 Distribution of the users in the scenario 2 2 a 76 Comparison of input values and simulation results 0 0 0 e el 77 Scenario setup for the comparison of simulated and measured data 80 2D PDP of the simulated track 2 naoa aa ee 80 Results for the measurement based simulation tutorial 0 2 204 82 Receiver track for the satellite channel tutorial 0 020002 0000 84 Antenna patterns for the satellite channel tutorial 0 0 00 000 eee eee 85 Results for the satellite channel tutorial 2 0 0 0 0 200000 eee eee 90 Scenario setup for the drifting phases tutorial 2 2 2 ee 91 Cluster delays vs Rx position drifting phases tutorial 000004 92 Drifting phases and Tx powe
121. g clusters are fixed as well and the time evolution of the radio channel is deterministic Different positions of the MT lead to different arrival angles delays and phases for each multipath component MPC Longer sequences are generated by transitions between channel traces from consecutive initializations of the model This allows the MTs to traverse different scenarios e g move from indoors to outdoors Figure 5 gives an overview of the modeling steps The user needs to configure the network layout i e the positions of the BSs antenna configurations downtilts the positions and trajectories of the MTs and the propagation scenarios The channel coefficients are then calculated in seven steps which are described in detail in Sections 3 1 to 3 8 Major extensions concerning 3 D propagation are made in steps C and D Time evolution is incorporated in steps D and G and a new 3 D model of the polarization 10 is introduced in step E In order to integrate these extensions some changes are made in the other parts of the model as well Input variables A Calculation of B Calculation of C Calculation of network layout correlated large scale gt initial delays and gt departure and terminal trajectories parameter maps path powers arrival angles propagation scenario antenna patterns j F Application of E Calculation of D Drifting of delays path gain shadow polarized
122. gh08 32 Elevation angles Since the elevation angles can only have values in between F and 4 the calculation method differers slightly from the method used for the azimuth angles Again the same method is used for the EoD and the EoA Hence is used instead of 0 or 6 Likewise the corresponding AS is denoted as Og As for the azimuth angles a random list of angles is created for the NLOS paths only following a Gaussian normal distribution with zero mean and a variance corresponding to the given AS from the LSP maps The LOS angle is defined to be 0 o 0 and ofl N 0 02 33 Next the LOS direction is applied This makes sure that the departure and arrival elevation are spread around the LOS path oP g 4 gt0s 34 The so obtained angels need to be mapped to the interval 3 3 This is done by a modulo operation which wraps the angles around the unit circle and an additional operation that mirrors the angles at the poles of the unit sphere e g an elevation angle of 91 is mapped to 89 92 to 88 and so on This ensures that the already known azimuth angles are not changed g ap m mod 27 7 35 f m0 ifo gt z o 4 ola ifo lt 2 36 g otherwise As for the azimuth angles the resulting AS is undefined Hence we calculate the actual elevation spread ee using equations 26 27 and 28 Then with og being the initial elevation spread from the map the angles ol are updated Since th
123. ght Bil X with X N 0 1 9 L4 dy dpx4 BH gt ak By k x 10 k 0 L4 dy dpx4 Be E gt ap By z k 11 k 0 Next the second filter 8 is applied on the diagonals of the map at first from the top left to the bottom right and then from the bottom left to the top right L4 dy dpx4 B S J bp Bes 12 k 0 L4 dy dpx4 B Jo Bye 13 k 0 After the 2 D autocorrelation shaping is done the extension space is removed and values of the remaining map are scaled to have the desired distribution u o The same procedure is repeated for all seven LSPs However the decorrelation distance d as well as u for each LSP can be different Filter b a Eas SOS Sa oo b Ror Ae wt bi R 20 z So Ses 20 Es f SA re p 40 eee 40 P La 40 iad wf on fa 60 60 60 80s c gt so T 80 AS q g 4 r fi hd 3 4 lt rT seal tacos ne Hepa Phe att 1 s 00 Se a E EA AT S EA 00 lt 1 aa 100 positions EW Wey CARET 4 et lt a 12040 ide feces eases x pee UD iad pie ly 12049 z i ERI ETE ETE Dena y j g iE Mind Sg ets ht f h e aof ae TETEE dd it 4 140 S EON EA DOR LEA ATANIN i s 20 40 60 80 100 120 20 40 60 80 100 120 20 40 60 80 100 120 F il 60 80 100 120 20 40 60 80 100 120 ilter Figure 7 Map based 2 D autocorrelation shaping using FIR filters Copyright Fraunhofer Heinrich Hertz Institute 59
124. gle spread in deg per segment at the receiver e xpr The NLOS cross polarization in dB per segment An identical copy of this variable is assigned to track par h_parset A matrix of parameter_set objects Rows correspond to the scenarios columns correspond to the transmitters See Section 2 2 5 Copyright Fraunhofer Heinrich Hertz Institute 37 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 2 SOFTWARE STRUCTURE h_channel h_parset h_cb get_channels sampling_rate check_parfiles Description Calculate the channel coefficients This is the most simple way to create the channel coefficients This function executes all steps that are needed to generate the channel coefficients Hence it is not necessary to use use the parameter_set or channel_builder objects Input sampling_rate channel update rate in s This parameter is only used if a speed profile is provided in the track objects Default value 0 01 10 ms check_parfiles Enables 1 default or disables 0 the parsing of shortnames and the validity check for the config files This is useful if you know that the parameters in the files are valid In this case this saves some execution time Output h_channel A vector channel objects See Section 2 2 7 h_parset A matrix of parameter_set objects Rows correspond to the scenarios columns correspond to the transmitters See Section 2 2 5 h_cb A vector of channel_builder
125. gth because the wave is shadowed or blocked by the obstacle It is modeled as log normal distributed with two parameters The standard deviation o defines the width of the distribution i e the power value in dB above or below the distance dependent PL and the decorrelation distance A This parameter defines how fast the SF varies when the terminal moves through the environment E g a value of 87 means that when the terminal moves 87 m in any given direction then the correlation of the value at this distance with the value at the initial position is e 0 37 Delay Spread DS The root mean square RMS delay spread is probably the most important single measure for the delay time extent of a multipath radio channel The RMS delay spread is the square root of the second central moment of the power delay profile and is defined to be NS i Pe or 5D a So DoAen 2 P P l 1 l 1 with P is the total received power P the cluster power and 7 the cluster delay In order to generate the coefficients QuaDRiGa has to generate delays for each of the multipath clusters I e the total lengths of scattered paths have to be defined This generation of delays is governed by value of the DS in a specific environment The DS is assumed to be log normal distributed and defined by two parameters Its median value u and its STD Thus a values of DS of 6 69 corresponds to 204 ns o then defines the range of possible values E g DS 0 3 lead
126. h The 2 D WINNER model 3 introduced the azimuth angle of departure AoD and the azimuth angle of arrival AoA In 3 D coordinates we also need the elevation angle of departure EoD 0f and the elevation angle of arrival EoA 0 The angles share similar calculation methods but have different ASs ogo Oga Cga and oga The individual angles are generated by first drawing random angles which are assigned to the already known path powers In order to obtain the correct ASs a scaling operations is used that readjusts the angles This approach is different from the WINNER model where the angels are mapped to the already known powers using a wrapped Gaussian distribution 3 16 A summary of the WINNER method together with a description of how it can be used in the new model is given in Appendix B Azimuth angles Here the calculation method for the azimuth angles is described The same calculation method is used for the AoD and the AoA Hence is used instead of or Likewise the corresponding AS is denoted as ag At first a random list of angles is created for the NLOS paths only following a Gaussian normal distribution with zero mean and a variance corresponding to the given AS from the LSP maps The LOS angle is defined to be 0 oy 0 and 45l MCO 03 24 The so obtained angels need to be mapped to the interval 7 7 This is done by a modulo operation which wraps the angles around the unit circle gl of
127. h values no _el Number of elevation values Methods h_array array array_type phi_3dB theta_3dB rear_gain Description Creates a new array object See array generate for a description of the input parameters and the list of supported antenna types append_array a Description Appends an antenna array to the existing one This method append the antenna array given in a to the existing array object If the polarization basis or the sampling grid do not match appropriate conversations or interpolations are performed Input a The array object which is appended to the current array object gain_dBi pow_max calc_gain element Description Calculates the gain of the antenna array Output gain_dBi Normalized Gain of the antenna pow_max Maximum power in main beam direction change_pol_basis new _basis Description This method can be used to change the polarization basis of an antenna pattern By default QuaDRiGa uses the polar spheric basis However the antenna patterns can be given in other bases as well Input new_basis The basis in which the pattern should be transformed A list of supported bases can be obtained by array supported_pol_basis combine_pattern center_frequency Description Calculates a virtual pattern of the given array When the inputs of an array are coupled i e fed with the same signal the
128. hannel coefficients The first run uses the new drifting module the second run disables it Note that drifting needs significantly more computing resources In some scenarios it might thus be useful to disable the feature to get quicker simulation results s drifting_precision 1 RandStream setGlobalStream RandStream mti9937ar seed 2 c p get_channels s drifting_precision 0 RandStream setGlobalStream RandStream mti9937ar seed 2 d p get_channels Channels Loo000000000000000000000000000000000000000000000000 12 seconds Channels Looo00000000000000000000000000000000000000000000000 J 7 seconds Results and discussion The following plots represent the results of the test figure distance 20 1 c no_snap l track get_length c no_snap plot distance c delay 1 1e9 b hold on plot distance d delay 1 1e9 b plot distance c delay 2 1e9 r plot distance d delay 2 1e9 r hold off xlabel Distance from track start point ylabel Delay ns The first plot Fig 22 shows the delay of the LOS tap blue and the delay of the first NLOS tap red vs distance The solid lines are from the channel with drifting the dashed lines are from the channel without The LOS delay is always increasing since the Rx is moving away from the Tx However the increase is not linear due to the 25 m height of the Tx Without
129. hannel merger which creates long time evolving sequences out if the snipes produced by the channel builder Addi tional function such as the transformation into frequency domain can help the user to further process the data An overview of the model software is depicted in Fig 3 The unified modeling language UML class diagram of the QuaDRiGa channel model gives an overview of all the classes methods and properties of the model The class diagram serves as a reference for the following descriptions which also lists the methods that implement a specific functionality Copyright Fraunhofer Heinrich Hertz Institute 20 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 2 SOFTWARE STRUCTURE lt lt input gt gt array name interpolation method polarization_basis iscompressed no_elements azimuth grid element_position lt lt input gt gt simulation_parameters sample_ density samples_per_meter t drifting precision t use polarization rotation t use absolute delays use angular mapping use map algorithm show progress bars center_frequency map_resolution append_array calc_gain change pol _basis combine pattern compress copy_element copy_objects generate generate_multi import_pattern interpolate normalize gain rotate pattern set_grid sub_array supported pol _basis uncompress visualize set_speed name simpar no
130. i9937ar seed 5 p l1 create_parameter_sets Parameters 00000000000000000000000000000000000000000000000000 2 seconds Channel generation and results Next we generate the channel coefficients Note that the initial sample density is 2 5 We then interpolate the sample density to 20 It would take ten times as long to achieve the same result with setting the initial sample density to 20 The interpolation is significantly faster It is done by first setting the speed to 1 m s default setting and then creating a distance vector which contains a list of effective sampling points along the track c p get_channels cn c merge t set_speed 1 dist t interpolate_movement s wavelength 2 20 ci cn interpolate dist t get_length spline Channels Looo00000000000000000000000000000000000000000000000 5 seconds The next plot shows the power of the first three taps from both the original and the interpolated channel plotted on top of each other The values are identical except for the fact that the interpolated values blue line have 5 times as many sample points Copyright Fraunhofer Heinrich Hertz Institute 100 eMail quadriga hhi fraunhofer de e WNE QuaDRiGa v1 2 32 458 A TUTORIALS Tx Position 407 7 A Tx Antenna O Rx Position 20F Y Position gt Ed i i i 1 i 0 10 20 30 40 50 60 70 80 X Position Figure 28 Scenari
131. ided in time domain as a list of delays and complex valued amplitudes However this class also implements certain methods to postprocess the channel data Those include e Transformation into frequency domain e Interpolation in time domain to change the terminal speed and sampling rate e Combining channel traces into longer segments including birth and death of clusters Properties name Name of the channel object This string is a unique identifier of the channel object The channel_builder creates one channel object for each MT each Tx and each segment They are further grouped by scenarios propagation environments The string consists of four parts separated by an underscore Those are e The scenario name from track scenario e The transmitter name from layout tx_name e The receiver name from layout rx_name e The segment number After channel merge has been called the name string consists of e The transmitter name from layout tx_name e The receiver name from layout rx_name version Version number of the QuaDRiGa release that was used to create the channel object individual_delays Indicates if the path delays are identical on each MIMO link 0 or if each link has a different path delay 1 coeff The complex valued channel coefficients for each path The indices of the 4 D tensor are Rx Antenna Tx Antenna Path Snapshot delay The delays for ea
132. il quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 A TUTORIALS is 0 26 W This mainly come from the XPR which leakes some power from the vertical into the horizontal polarization and thus reduces the received power on the vertically polarized Dipole Next we study two cases Either the Tx is vertical polarized and the Rx is at 45 or vise versa Fig 33 top right figure 4 New figure plot abs squeeze c coeff 2 1 1 72 4 Tx vertical Ra 45 degree hold on plot abs squeeze c coeff 1 2 1 2 r 4 Tx 45 degree Ra vertical hold off axis 0 360 0 1 1 legend Tx vertical Rx 45 circ Tx 45 circ Rx vertical xlabel Position degrees ylabel LOS Power linear scale title Tx Vertical Rx 46 cire The receiver changes its direction in a way that it always has the same orientation towards the Tx However due to the displacement of the Tx the radiated power towards the Tx becomes minimal at around 90 This minimum is visible in both curves blue and red However the pole of the 45 slanted dipole now points to a different direction which explains the difference in the two lines When the Rx is at 45 and the Tx is vertical the pole is in the right half if the circle resulting in a lower received power When the Rx is Vertical and the Tx is 45 the minimum power is achieved in the left half of the circle Next we evaluate the two dipoles which are ro
133. imulations more realistic is the function to apply arbitrary speed and movement profiles e g accelerating breaking or moving at any chosen speed These profiles are defined in the track class The profiles are then converted into effective sampling points which aid the interpolation of the channel coefficients Channel model setup First we set up the simulation parameters Note the sample density of 2 5 which enables very fast simulations even with drifting simulation_parameters s s center_frequency 2 53e9 s sample_density 2 5 s use_absolute_delays 1 Second we define a track It has a length of 20 m starts at 10 m east of the transmitter and consists of three segments LOS NLOS LOS The positions are interpolated to match the sample density defined above The track is then plugged into a network layout with one transmitter at position 0 0 25 Both transmitter and receiver are equipped with dipole antennas The last three lines create the LSPs t track linear 20 0 t initial_position 60 0 1 5 t interpolate_positions 128 20 t segment_index 1 40 90 t scenario WINNER_UMa_C2_LOS WINNER_UMa_C2_NLOS gt WINNER_UMa_C2_LO0S interpolate_positions s samples_per_meter ct layout s tx_array generate dipole rx_array 1 tx_array tx_position 3 25 track t PREP PH 1l visualize RandStream setGlobalStream RandStream mt
134. inear track 20 m length t interpolate_positions s samples_per_meter Interpolate to sample density t split_segment 1 6 5 0 4 Split in 4 segments Un WINNER_UMa_C2_NLOS Ul WINNER_UMa_C2_L08S t scenario Un Un U1 Un Z Set scenarios l randomize_rx_positions 500 0 0 0 Random start position tmp 1 rx_position l track t l rx_position tmp l visualize Manual setting of the parameters Now we initialize the parameter set objects The method 1 create_parameter_sets splits the track into smaller sub tracks one for each segment It further extracts the scenario informations Each scenario gets its own parameter set object So we get an Copyright Fraunhofer Heinrich Hertz Institute 110 eMail quadriga hhi fraunhofer de e WNE e WNE QuaDRiGa v1 2 32 458 A TUTORIALS 250 gt F Be Position A Tx Antenna 200 4 i O Rx Position Y Rx Antenna Rx Track 150 100F DNNE Y Position 100F 150 200 250 0 50 100 150 200 250 300 350 400 450 500 550 X Position Figure 35 Scenario overview manual parameter selection array of two parameter sets The first element p 1 has three positions NLOS segments and the second has one position LOS segment p l create_parameter_sets 1 name no_positions p 1 p 1 p 2 p 2 name no_positions Parameters 000000000000000000000000000000000
135. initial channel coefficients for each snapshot of a segment The antennas patterns are defined in spherical coordinates Az over E with a polar spherical polarization basis see 17 F o poa le s The coordinate system has two angles and two poles The elevation angle 0 is measured relative to the pole axis A complete circle will go through each of the two poles similar to the longitude coordinate in the world geodetic system WGS The azimuth angle moves around the pole similar to the latitude in WGS The electric field is resolved onto a polar spherical polarization basis The first component of the radiated field Fl represents the case where the probe i e the reference antenna which is used to measure the pattern is polarized in direction Likewise for the second component F the probe is polarized in direction The channel model requires the directional antenna gains from both the Tx and Rx antennas at the pre viously calculated departure and arrival angles If in practice F 8 is sampled once per degree in both azimuth and elevation direction interpolation is needed in order to obtain the exact antenna response F 6 1m s r lm s The initial channel coefficient then notes 1 T d d oe F OF pia Or ives g M t 1 m s i Bir tans De ma 59 The polarization is changed along the propagation path This is captured by the matrix M The SCM WINNER and COST models use random coefficients to handle pol
136. ins per snapshot e asD The azimuth angle spread in deg per segment at the transmitter e asA The azimuth angle spread in deg per segment at the receiver e esD The elevation angle spread in deg per segment at the transmitter e esA The elevation angle spread in deg per segment at the receiver e xpr The NLOS cross polarization in dB per segment An identical copy of this variable is assigned to track par Copyright Fraunhofer Heinrich Hertz Institute 33 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 2 SOFTWARE STRUCTURE len dist get_length Description Calculates the length of the track in m Output len Length of a track in m dist Distance of each position snapshot from the start of the track in m subtracks get subtrack i segment Description Splits the track in subtracks for each segment state This function returns the subtracks for the given segment indices When no input argument is provided all subtracks are returned After defining segments along the track one needs the subtrack that corresponds only to one segment to perform the channel calculation This new track can consist of two segments The first segment contains the positions from the previous segment the second from the current This is needed to generate overlapping channel segments for the merging process Input i segment A list of indices indicating which subtracks should be returned By
137. into Paths oaoa aa e 72 3 7 Path Gain Shadow Fading and K Factor osaa 0 00000 eee eee 73 3 8 Transitions between Segments sao cracca cadranno somadas aneao dan 73 3 9 Postprocessing Variable Speeds o s coe coe ga Be ioia Be a Re a eaaa a ew SZ 74 A Tutorials 75 A 1 Network Setup and Parameter Generation ooo a e a ee 75 A 2 Simulating a Measured Scenario a oaoa ee 79 A 3 Generation of Satellite Channels 2 0 0 0 00 0 ee ee 83 AA rittine Phases and Delays spi sosta To ice ee Sec Be oe e Oe EE ee gee De 91 A b Time Evolution and Scenario Transitions s 2 6 66448 Apa be ee ee pee ee 95 A 6 Applying Varying Speeds Channel Interpolation 0 200000000 0s 100 A 7 Geometric Polarization s s o e oc 4448 66 e 84 eee Re ee ee ee es 104 AS Visualizing RHCP LHCP Patterns os 4406 c 4 eho oe ee eS eee aw eee eb eed 108 A 9 How to manually set LSPs in QuaDRiGa 2 20 02 0000 02 eee ee 110 B Departure and Arrival Angles Adopted WINNER Method 114 Copyright Fraunhofer Heinrich Hertz Institute 3 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 List of Tables List of Figures OONrDoKBWNMH k ga pt OD WWWWWWWWWNNNNNMNNNNNMNNR KF RRP RF KF ee CONDOR WNF THO OONDTBWNrF OO AON DTK WwW bd Simplified overview of the modeling approach used in QuaDRiGa 13 Typical driving Course s si a ve eee Bae ES ce ae Die ee de ee ee Pe 16 UML class diagram of the model so
138. is coming from the direct component This is expressed by the cross correlations parameters They can vary between 1 and 1 Negative values denote inverse correlation e g a high K Factor implies a low delay spread Positive Value implies a positive correlation such as a high K Factor also implies a high shadow fading Cross Correlations are used during the map generation ds_kf Correlation of delay spread and K Factor asA_ds Correlation of delay spread and azimuth of arrival angle spread esA_ds Correlation of delay spread and elevation of arrival angle spread ds_sf Correlation of delay spread and shadow fading asA_kf Correlation of K Factor and azimuth of arrival angle spread esA_kf Correlation of K Factor and elevation of arrival angle spread sf_kf Correlation of K Factor and shadow fading esA_asA Correlation elevation of arrival angle spread and azimuth of arrival angle spread asA_sf Correlation of shadow fading and azimuth of arrival angle spread esA_sf Correlation of shadow fading and elevation of arrival angle spread Cluster Parameter Those parameters influence the generation of the scattering clusters and the dis tribution of the sub paths within each cluster NumClusters Integer The number of clusters generated For multipath rich environments typically more clusters are used If the LOS component is dominant a lower number of clusters is sufficient PerCluster deg The azimuth angular spread of the 20 sub paths within one
139. itial arrival and departure angles and the traveled distance from the start point However since the Rx moves along a circular track the angles change continuously which is not correctly modeled The phase at the end of the first segment does not match the phase at the beginning of the second When adding both components artifacts appear as can be seen in the red curve Next we plot the power delay profiles for both tracks We calculate the frequency response of the channel and transform it back to the time domain by an IFFT Then we create a 2D image of the received power at each position of the track We start with the circular track h cn 1 fr 100 e6 512 h squeeze h pdp 10 logi0 abs ifft h 1 72 figure imagesc pdp 1 256 caxis max max pdp 50 max max pdp 5 J colorbar cm colormap hot colormap cm end 1 1 set gca XTick 1 32 255 set gca XTickLabel 0 32 256 100e6 1e6 xlabel Delay mus set gca YTick 1 cn 1 no_snap 8 cn 1 no_snap set gca YTickLabel 0 cn 1 no_snap 8 cn 1 no_snap cn 1 no_snap 360 ylabel Position on circle degree title PDP for the circular track with drifting The X axis Fig 26 right shows the delay in microseconds and the Y axis shows the position on the circle For easier navigation the position is given in degrees 0 means east starting point 90 means north 1
140. l Copyright Fraunhofer Heinrich Hertz Institute 84 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 A TUTORIALS Finally we interpolate the track to the given sample density 2 samples per half wave length and plot the track see Fig 18 t interpolate_positions s samples_per_meter t visualize Defining Antenna Arrays In the third step we set up our antenna arrays for the transmitter at the satellite and the receiver We use synthetic dipole antennas for this case Two dipoles are crossed by an angle of 90 degree The signal is then split and fed with a 90 degree phase shift to both elements generating RHCP and LHCP signals 4 Create a patch antenna with 120 degree opening a array custom 120 120 0 Copy element 1 to element 2 the resulting antenna array has two elements both dipoles a copy_element 1 2 4 Rotate the second pattern by 90 degree around the z axis a rotate_pattern 90 x 2 Set the coupling between the elements The Tz signal for the first element is shifted by 90 degree out of phase and put on the second element The signal for the second element is shifted by 90 degree and copied to the first element Both antennas thus radiate a RHCP and a LHCP wave coupling 1 sqrt 2 1 1 1j 1j de SS xN w 4 Create a copy of the array for the receiver b a copy_objects b coupling 1 sqrt 2 1 13 1j 1j Rotate the receive antenna
141. lation parameters su e Sas a a Re Ae i ee Gree doe be 23 222 Class aray lt ganie ws as asin ech Bo EBS PA eee ee eee oe See oe 25 223 Class track jade l e e eh a EL eS ee a ae eee of Oe S 31 224 Class layout 2424 6 4 8 64 40 4 Pew he ee tea Bi oe SEA Ee ee ee Pee 36 2 2 5 Class parameteret sa c sbe sacs eoi a aated eaa MR eR ee EE Ee ee 41 2 2 6 Class channel builder 2 2 sovs sane 20a6 a Pa pS eu oe RAE eG oR E a 44 220 Glass channel sss os de ase h Sar Boa Se RO we Wao Bde PS ee ee Ql 46 Ron Data EON e i eeo a oe its oe Yes ve a Ses Gt gsr Re Gs tue tte age ne Gage as es es Ge A 49 2 4 Scenario Specific Parameters 2 20 0 ee 50 2 4 1 Description of the Parameter Table 0 0000000 eee 50 24 2 Adding New Scenarios ne seor 8 eka ee a ee ewe eee E a 54 3 Technical Documentation 56 3 1 Correlated Large Scale Parameter Maps 2 0 2 a 57 3 2 Initial Delays and Path Powers cisse e g a k a a bok eee a we Re a a a we eG 60 3 3 Departure and Arrival Angles aoaaa ee 61 o4 Drifting 644 6 pu se eb oo a ea e a ee te Ye Bowe ee ee ee eee a 64 3 5 Antennas and Polarization s s 4 lt a0 2 544 4 444408 48 tee Re Hee ae ane 66 3 5 1 Relation between the Polarization Model and the Jones Calculus 66 3 5 2 Changing the Orientation of Antennas ooo o a a a 68 3 5 3 Constructing the Polarization Transfer Matrix aoaaa a 000002 eae 70 3 6 Combining Sub Paths
142. lizes the drifting module The output variables are the NLOS Tx angles for the precomputation of the Tx array response Input i mobile The index of the mobile terminal within the channel builder object Output phi dlm The departure azimuth angles for each subpath theta_dtm The departure elevation angles for each subpath lbs_pos The position of the last bounce scatterer fbs_pos The position of the first bounce scatterer h_channel get_channels Description Generates the channel coefficients This is the main function of the channel_builder Output h_channel An array of channel objects h_channel channel_builder get_los_channels h_parset Description Generates channel coefficients for the LOS path only This function generates static coefficients for the LOS path only This includes the following properties e antenna patterns e orientation of the Rx if provided e polarization rotation for the LOS path e plane wave approximation of the phase e path loss e shadow fading No further features of QuaDRiGa are used i e no drifting spherical waves time evolution multipath fading etc This function can thus be used to acquire quick previews of the propagation conditions for a given layout Input h_parset A parameter_set object see Section 2 2 5 Output h_channel A channel object see Section 2 2 7 The output contains one coefficient for each position
143. ll less than 3 dB 1 figure 2 plot 35 35 0 0 k 3 hold on 4 plot 35 35 3 3 k 5 plot 35 35 3 3 k 6 plot 10 logi0 p kf 10 logi0 ds p ds 7 hold off 8 axis 30 30 6 6 10 legend Equal 3dB 3 11 xlabel KF_P dB 12 ylabel DS_P DS_C dB 13 title Delay Spread difference vs K factor 1 close all 2 disp QuaDRiGa Version simulation_parameters version 1 QuaDRiGa Version 1 0 1 145 Copyright Fraunhofer Heinrich Hertz Institute eMail quadriga hhi fraunhofer de 78 NNN ONA AR wWONH OO wwwnwnwnwwwwwnn nv nv nvnnn co J YN H O C W e H O e e W N w ye o e w Ne QuaDRiGa v1 2 32 458 A TUTORIALS A 2 Simulating a Measured Scenario This script recreates a measured drive test from the Park Inn Hotel at Berlin Alexanderplatz The transmit ter was at the rooftop of the hotel while the mobile receiver was moving south on Grunerstra e A simplified version of the scenario is recreated in the simulation where the scenarios along the track were classified by hand Channel model setup and coefficient generation First we set up the channel model set 0 defaultTextFontSize 14 set 0 defaultAxesFontSize 14 RandStream setGlobalStream RandStream mti9937ar seed 1 close all clear all s
144. lormap cm end 1 1 set gca XTick 1 32 255 set gca XTickLabel 0 32 256 100e6 1e6 xlabel Delay mus set gca YTick 1 ci no_snap 8 ci no_snap set gca YTickLabel 0O ci no_snap 8 ci no_snap ci no_snap 20 ylabel Time s Now we create a movement profile It is defined by a set of value pairs in track movement_profile The first value represents the time in seconds the second value the position on the track Here we start at a position of 7 m ie in the second NLOS segment We then go back to the beginning of the track This takes 5 seconds Then we wait there for 1 second and go to the end of the track which we reach after additional 14 seconds The next step is to interpolate the sample points This is done by the interpolatemovement method It requires the sample interval in s as an input argument Here we choose an interval of 1 ms which gives us 1000 samples per second Fig 30 left illustrates the results t movement_profile 0 7 5 0 6 0 20 20 dist t interpolate_movement 1e 3 ci cn interpolate dist t get_length nsnap ci no_snap time O nsnap 1 t movement_profile 1 end nsnap 1 figure plot time dist r xlabel Time s ylabel Position on track m The last plot Fig 30 right shows the PDP of the interpolated channel with the movement profile applied The channel starts in
145. m the WINNER II model 3 However it was neither implemented nor tested Our implementation requires that parts of the segments are overlapping as depicted in the top of Figure 12 The lifetime of scattering clusters is confined within the combined length of two adjacent segments The power of paths from the old segment is ramped down and the power of new paths is ramped up within the overlapping region of the two segments Hence this process describes the birth and death of scattering clusters along the trajectory Outside the overlapping region all paths of the segment are active The overlapping region is further split into sub intervals to keep the computational overhead low During each sub interval one old path ramps down and one new path ramps up The power ramps are modeled by a squared sine function wl sin 5 wll 93 Here w i is the linear ramp ranging from 0 to 1 and w is the corresponding sine shaped ramp with a constant slope at the beginning and the end This prevents inconsistencies at the edges of the sub intervals If both segments have a different number of paths the ramp is stretched over the whole overlapping area for paths without a partner For the LOS path which is present in both segments only power and phase are adjusted Paths need to be carefully matched to minimize the impact of the transition on the instantaneous values of the LSPs For example the DS increases if a path with a small delay ramps down and
146. m angles to Cartesian coordinates sat_x sat_dist cosd sat_el cosd sat_az 90 sat_y sat_dist cosd sat_el sind sat_az 90 sat_z sat_dist sind sat_el x We also turn the antenna of the satellite so it points to the receiver a rotate_pattern sat_el y a rotate_pattern 270 sat_az z 4 Set the satellite position in the layout 1 tx_position sat_x sat_y sat_z l track t 4 Set the track for the receiver 1l tx_array a 4 Set the t array 1 rx_array i ion 4 Set the rav_array Setting up scenario parameters Next the large scale parameters are set The first line calls create_parameter_sets a built in function that processes the data in the layout and returns a new pa rameter_set object p p is an array with two elements One of them contains all the parameters for the good state LOS and one for the bad state NLOS p 1 create_parameter_sets 0 Each parameter set has two different kinds of parameters One for the scenario and one for the current state For example a scenario might have an average RMS Delay spread of 158 ns plus a certain variance which defines a range for the RMSDS In addition to that there are cross correlations with other parameters such as the angular spread at the transmitter All those parameters are stored in the scenpar property For the good state that parameters are S1 stremp p
147. mally not need to interact with it However if parameter tables need to be changed here is the place to do so Properties name Name of the parameter_set object simpar Handle of a simulation_parameters object See Section 2 2 1 tx_array Handles of array objects for each Tx See Section 2 2 2 rx_array Handles of array objects for each Rx See Section 2 2 2 rx_track Handles of track objects for each Rx See Section 2 2 3 scenario Name of the scenario text string scenpar The parameter table See Section 2 4 plpar Parameters for the path loss See Section 2 4 no_positions positions tx_position Number of receiver positions associated to this parameter_set object Note that each segment in longer tracks is considered a new Rx position The list of initial positions for which LSPs are generated This variable is obtained from the properties track initial_position and layout rx_position The transmitter position obtained from the corresponding layout tx_position ds The RMS delay spread in s for each receiver position kf The Ricean K Factor linear scale for each receiver position sf The shadow fading linear scale for each receiver position asD The azimuth spread of departure in deg for each receiver position asA The azimuth spread of arrival in deg for each receiver position esD The elevation spread of departure in deg for each receiver position esA The elevation sprea
148. me continuous simulation A sample density of 2 5 ensures that the channel coefficients can be interpolated to different playback speeds later on close all clear all s simulation_parameters 4 Basic simulation parameters s center_frequency 2 185e9 s sample_density 0 25 RandStream setGlobalStream RandStream mti9937ar seed 1 Creating a random Track and defining states along the track Next we generate a simulation track A track describes the movement of a mobile terminal It is composed of an ordered list of positions During the simulation one snapshot is generated for each position on the track Later on the generation of the track is done by the state sequence generator Here we implement a simple version of the sequence generator to generate a random track We first create a set of streets with different length We assume a normal distribution of the street length where the parameters mu and sigma were fitted from random distances between two crossings in central Berlin measured with Google earth street_length_mu 187 4 Average street length in m street_length_sigma 83 min_street_length 50 turn_probability 0 4 4 The prob that the car turns at a crossing curve_radius 10 A The curve radius in m diro rand 2 pi 4 Random start direction For the given parameters we calculate a list of points along the track that resemble the street grid and the turns at crossings p
149. ments the continuous time evolution with smooth transitions between seg ments Each segment of a track is split in two parts an overlapping area with the previous segment and an exclusive part with no overlapping Each segment is generated independently by the channel builder However the distance dependent autocorrelation of the large scale parameters was considered when the parameters were drawn from the corresponding statistics Transition from segment to segment is carried out by replacing taps of the previous segment by the taps of the current segment one by one The modeling of the birth death process is done as published in the documentation of the WIM2 channel model The route between adjacent channel segments is split into sub intervals equal to the minimum number of taps in both overlapping segments During each sub interval the power of one old tap ramps down and one new tap ramps up Power ramps are modeled by a modified sinus function to allow smooth transitions Taps from the old and new segments are coupled based on their power If the number of clusters is dif ferent in the channel segments the weakest clusters are ramped up or down without a counterpart from the new old segment The merging is only done for the NLOS components since the LOS component has a deterministic behavior The LOS component is thus just scaled in power Input overlap The length of the overlapping part relative to the segment length It can have values in
150. ministic part The deterministic component is the same as for the LOS polarization However the wave travel direc tion ry ts used to calculate the departure and arrival angles for LOS propagation must be replaced by at l m s 1 the vector pointing from the position of the transmitter to the first bounce scatterer FBS and a jm s i e the vector pointing from the position of the LBS to the Rx position Furthermore the calculation must be repeated for each sub path Copyright Fraunhofer Heinrich Hertz Institute 70 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 3 TECHNICAL DOCUMENTATION 2 a Linear polarization with same XPR for V and H component Individual values of the XPR for each subpath are drawn from a normal distribution XERE ISN XPR XPR 75 In order to model the polarization change due to scattering XPR we can follow the approach from 25 and calculate an additional NLOS rotation matrix peoe as line m m cos sin M ae 2 vu vh Yim Yim 76 Mhv Mhh SIN Yl m COS Y1 m Following the notations in 26 we get Mol Mhh cos 71 m 2 lm Mn monl2 sin Yim co Yim arccot V XPRimn 78 Yim 2 b Linear polarization with different XPR for V and H component If the XPR is different for the vertical and horizontal component as suggested by 23 26 then we get three parameters z Mol Imah Mo u a XPRiin Mrl XPRi m Mon CPR m
151. n rad The default is random generate circular track_length direction Description Creates a circular track with given length and starting direction Input track_length The circumference of the circular track in m Default is 62 8 m direction The starting point on the circle in rad Positive values define the travel direction as counter clock wise and negative values as clock wise E g 0 sets the start point in the east of the circle traveling north 27 sets it in the east traveling south The default is random Copyright Fraunhofer Heinrich Hertz Institute eMail quadriga hhi fraunhofer de 32 QuaDRiGa v1 2 32 458 2 SOFTWARE STRUCTURE generate street track_length direction street_length_min street_length_mu street_length_std curve_radius turn_probability Description Emulates a drive route through a city grid The mobile terminal starts at point 0 going into a specified direction The trajectory grid is build from street segments The length of each street is specified by the parameters street_length_min street_length_mu and street_length_sigma At the end of a street i e at a crossing the terminal turns with a probability specified by turn_probability The change of direction is in between 75 and 105 degrees either left or right The radius if the curve is given by c
152. n it is possible to combine the elements of the array This function calculates the virtual pattern by using the QuaDRiGa simulator Input center_frequency The center frequency in Hz ratio compress Description Stores the array in compressed form If there are many similar elements in an array the memory and storage requirements might be high Therefore it is possible to compress the array to save storage space This is done as follows e Patterns are stored in ploar spheris polarization basis e If multiple elements have the same patterns the pattern is stored only once e Patterns are stored in single precision e If there are complex valued patterns with no imaginary part they are converted to real values It is recommended to call compress before saving an array to disk Decompressing is done automat ically when needed Output ratio The compression factor in percent Copyright Fraunhofer Heinrich Hertz Institute 26 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 2 SOFTWARE STRUCTURE copy_element source target Description Creates a copy of an antenna element Input source Index of the array object that should be copied The value must be scalar integer and greater than 0 and it can not exceed the array size target Target can be a scalar or vector with elements gt 0 copy_objects Description A modified version of the stand
153. nd is typically much larger than the per cluster angular spread However even with many clusters the Doppler spread is narrower in QuaDRiGa than when assuming pure Rayleigh fading This is also in line Copyright Fraunhofer Heinrich Hertz Institute 13 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 1 INTRODUCTION AND OVERVIEW with measurement results It can be observed in the field that the main components arrive from selected angles and the classical Doppler spectrum s Jakes or Butterworth filter shaped characteristics are only valid as long term average and not valid for a short time interval To summarize e A typical propagation environment requires 8 20 clusters e Internally each cluster is represented by 20 sub paths resulting in 160 400 sub paths in total e Each sub path is modeled as a single reflection e The 160 400 sub paths are weighted by the antenna response The 20 sub paths for each cluster are summed up which results in 8 20 paths e For a MIMO system with multiple antennas at the transmitter and receiver each path has as many channel coefficients as there are antenna pairs Hence at the output there are npgin NR2 NT Channel coefficients 1 4 Continuous time evolution QuaDRiGa calculates the channel for each defined reception point To generate a time series a continuous track of reception points can be defined The arrival angles of the sub paths play a crucial for the time evol
154. ne snapshot for each position that is listed in the track object When the channel sampling theorem is not violated i e the sample density is gt 2 then the channel can be interpolated to any other position on the track This can be used e g to emulate arbitrary movements along the track For more information see track movement_profile track interpolate_movement or the tutorial Applying Varying Speeds Channel Interpolation in Section A 6 Input dist A vector containing distance values on the track The distance is measured in m relative to the beginning of the track Alternatively dist can be given as a 3 D tensor with dimensions Rx Antenna Tx Antenna Snapshot In this case interpolation os done for each antenna element separately method Selects the interpolation algorithm The default is linear interpolation Optional are e linear Linear interpolation optimized for speed e spline Cubic spline interpolation of the channel coefficients and piecewise cubic hermite polynomial interpolation for the delays Output c A channel object containing the interpolated coefficients and delays for each entry in dist Copyright Fraunhofer Heinrich Hertz Institute 47 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 2 SOFTWARE STRUCTURE c merge overlap optimize verbose Description Combines channel segments into a continuous time evolution channel The channel merger imple
155. ng a QuaDRiGa object can be done by b a copy e User input The user inputs Point 1 in the programm flow are provided through the classes simulation_parameters array track and layout simulation_parameters defines the general settings such as the center frequency and the sample density It also enables and disables certain features of the model such as polarization rotation sub path output and progress bars array combines all functions needed to describe antenna arrays track is used to define user trajectories states and segments layout is a object including the tracks and antenna properties together with further parameters such as the satellite position Internal processing All the processing is done by the classes parameter_set and channel_builder parameter_set is responsible for generating LSPs for the cluster generation It also holds the parameter maps needed for generating auto and crosscorrelation properties of the parameters pa rameter_set implements point 2 of the program flow channel_builder creates the channel coefficients This includes the cluster generation and the MIMO channels It implements steps 3 7 of the program flow e Model output The final two steps 8 and 9 of the program flow are implemented in the class channel Objects of this class hold the data for the channel coefficients The class also implements the c
156. nnel coefficients Output h_channel A vector channel objects See Section 2 2 7 h_cb A vector of channel_builder objects See Section 2 2 6 dist get_distances Description Calculates the distances between Rx and Tx Output dist A vector containing the distances between each Rx and the Tx in m pl scale_sf get_pl evaltrack i mobile Description Calculates the path loss The path loss model is specified in the configuration files and in parameter_set plpar Input evaltrack A track object for which the PL should be calculated If evaltrack is not given then the path loss is calculated for each Rx position Otherwise the path loss is calculated for the positions provided in evaltrack i_mobile The Rx index If it is not given the PL is evaluated for all Rx positions If evaltrack is given and if simulation_parameters drifting precision is set to 3 then this parameter is required to select the Rx antenna array default 1 Output pl The path loss in dB scale_sf In some scenarios the SF might change with increasing distance between Tx and Rx Hence the shadow fading provided by the parameter map has to be changed accordingly The second output parameter scale_sf can be used for scaling the logarithmic SF value from the map Copyright Fraunhofer Heinrich Hertz Institute 42 eMail quadriga hhi fraunhofer de QuaDRiGa v1
157. nrich Hertz Institute 23 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 2 SOFTWARE STRUCTURE use_polarization_ Select the polarization rotation method rotation i use_polarization_rotation 0 Uses the polarization method from WINNER No polarization rotation is calculated use_polarization_rotation 1 Uses the new polarization rotation method where the cross polarization ratio XPR is modeled by a rotation matrix No change of circular polarization is assumed use_polarization_rotation 2 default Uses the polarization rotation with an additional phase offset between the H and V component of the NLOS paths The offset angle is calculated to match the XPR for circular polarization use_polarization_rotation 3 Uses polarization rotation for the geometric polarization but models the NLOS polarization change as in WINNER use_absolute_delays Returns absolute delays in channel impulse response CIR By default delays are calculated such that the LOS delay is normalized to 0 By setting use_absolute_delays to 1 or true the absolute path delays are included in channel delays at the output of the model use_angular_ Selects the angular mapping method mapping use_angular_ mapping 1 Maps the path powers to arrival angles by a wrapped Gaussian distribution This method is adopted from the WINNER model However the generated angles show high correlations if the K Factor is large
158. nt arrival angles Copyright Fraunhofer Heinrich Hertz Institute 64 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 3 TECHNICAL DOCUMENTATION OF m Pm Those angles are transformed into Cartesian coordinates to obtain cos Ofm COS OF r alm A _ a a Alm sin Pim COS Oe i 2 J arm 42 pi a sin Om This vector has unit length and points from the initial Rx position towards the scatterer Next the length of the vector aj is obtained Since the drifting at the MT is modeled by a single reflection Tx Rx and LBS form a triangle We know d r and 4j and can thus apply the cosine theorem to calculate the length aj m between the Rx and LBS bim lr lanm 2 r larm cos aim 43 dj larm rl larm 2 larm r aim 44 d r id oes 45 2 di fap In 44 the term cos agm is substituted with rT m rl since we need the angle between azm and r the vector pointing from the Rx position towards the Tx Now we can calculate the vector a jm s for the Rx antenna element r at snapshot s The element position includes the orientation of the antenna array with respect to the moving direction of the Rx Hence the vector e points from the initial Rx location to the r antenna element at snapshot s ar l m s alm Er s 46 We obtain an update of the arrival angles by transforming a j m back to spherical coordinates Se arctang ar Fon
159. o setup for the speed profile tutorial pwr_orig 10 1log10 squeeze abs cn coeff 1 1 1 3 72 nsnap cn no_snap dist_orig O nsnap 1 t get_length nsnap 1 pwr_int 10 lo0og10 squeeze abs ci coeff 1 1 1 3 72 figure plot dist_orig pwr_orig r Linewidth 2 hold on plot dist pwr_int b hold off axis min dist max dist min pwr_orig pwr_orig gt Inf max pwr_orig pwr_orig gt Inf 10 xlabel Distance from start point m ylabel Power dB Fig 29 right shows the power delay profile PDP for the interpolated channel As defined in the track object it starts with a LOS segment going into a shaded area with significantly more multipath fading at around 4 seconds and then back to LOS at around 13 sec 80 2 5 85 o0 it 7 57 i 95 2 10 4 100 5 12 5 J 105 fal 110 115 17 5 120 0 5 10 15 20 0 032 0 64 0 96 128 16 1 92 2 24 Distance from start point m Delay us Figure 29 Received power and 2D PDP for the speed profile tutorial Copyright Fraunhofer Heinrich Hertz Institute 101 eMail quadriga hhi fraunhofer de none WN QuaDRiGa v1 2 32 458 A TUTORIALS h ci fr 100 e6 512 h squeeze h pdp 10 logi0 abs ifft h 1 72 figure imagesc pdp 1 256 caxis max max pdp 50 max max pdp 5 J colorbar cm colormap hot co
160. o the starting point of the track If the track has only one segment it is also the position for which the LSPs are calculated The initial position is added to the values in the positions variable no snapshots Number of positions on the track positions Ordered list of position relative to the initial position QuaDRiGa calculates an instantaneous channel impulse response also called snapshot for each po sition on the track movement_profile Time in sec vs distance in m for speed profile QuaDRiGa supports variable terminal speeds This is realized by interpolating the channel coefficients at the output of the model The variable track movement_profile describes the movement along the track by associating a time point with a distance point on the track An example is t movement_profile 0 7 5 0 6 0 20 20 dist t interpolate_movement 1le 3 ci cn interpolate dist t get_length See also the tutorial Applying Varying Speeds Channel Interpolation in Section A 6 for more details no_segments Number of segments or states along the track segment_index Starting point of each segment given as index in the positions vector scenario Scenarios for each segment along the track This variable contains the scenario names for each segment as a cell array of strings A list of supported scenarios can be obtained by calling parameter_set supported_scena
161. of those scenarios are stored in config files which are located in the config folder of the QuaDRiGa source path The UMal config file for example looks like this Config File for scenario UMal WINNER Urban Macro LOS See CELTIC CP5 026 D5 3 WINNER Final Channel Models and IST 4 027756 WINNER II D1 1 2 v 1 1 WINNER II Channel Models Stephan Jaeckel Fraunhofer Heinrich Hertz Institute Wireless Communication and Networks Einsteinufer 37 10587 Berlin Germany DE RC HE HE HE JE JE JE FE e mail stephan jaeckel hhi fraunhofer de NumClusters 8 r_DS 2 5 SF_sigma 65 SF_lambda 45 LNS_ksi 3 F_mu 7 KF_sigma 3 KF_lambda 12 DS_mu 7 39 DS_sigma 0 63 DS_lambda 40 AS_D_mu 1 AS_D_sigma 0 25 AS_D_lambda 15 PerClusterAS_D 6 AS_A_mu 1 7 AS_A_sigma 0 19 AS_A_lambda 15 PerClusterAS_A 12 ES_D_mu 0 7 ES_D_sigma 0 2 ES_D_lambda 15 A DE pp 7 PerClusterES_D 3 ES_A_mu 0 95 ES_A_sigma 0 16 ES_A_lambda 15 4 DS 3 pp 73 PerClusterES_A 7 xpr_mu xpr_sigma 4 Adjustments have 8 4 been made to keep xcorr matriz positive definite asD_ds 0 4 asA_ds 0 7 0 8 asA_sf 0 5 asD_sf 0 4 0 5 Copyright Fraunhofer Heinrich Hertz Institute 54 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 2 SOFTWARE STRUCTURE 55 ds_sf 0 4 56 asD_asA 0 3 57 asD_kf 0 1 58 asA_kf 0 2 59 ds_kf 0 4 6
162. oint 0 4 The start point always at 0 0 m 1 4 A counter for the points for n 1 3 4 We simulate 3 street segments 4 Get a random street length drawn from the distribution defined above street_length randn street_length_sigma street_length_mu while street_length lt min_street_length street_length randn street_length_sigma street_length_mu end 4 Get 3 points along the street point m 1 point m exp 1j diro street_length 0 1 point m 2 point m exp 1j diro street_length 0 9 point m 3 point m exp 1j diro street_length m m 3 4 At a crossing the car could change its direction This is modeled here if rand lt turn_probability dirn diro sign rand 0 5 pi 2 randn pi 12 point mt 1 point m curve_radius exp 1ij diro exp 1j dirn diro dirn m m 1 end end Copyright Fraunhofer Heinrich Hertz Institute 83 eMail quadriga hhi fraunhofer de wwe QuaDRiGa v1 2 32 458 A TUTORIALS Next we create a track object and pass the points along the track We then use the internal interpolation functions to interpolate the track to 1 point per meter t track 4 Create a track object t positions real point imag point zeros 1 numel point t interpolate_positions 1 4 Interpolate to 1 point per meter We now assemble a rudimentary state sequence generator that generates different states along the track We first define the distributi
163. om the a normal distribution XPRIS N XPR XPR2 85 This value represents the average XPR over all MPCs at the receiver positions Then in a second step the XPR for the individual MPCs is drawn using XPRI instead of XPR This maintains the original spread XPR in the generated channel coefficients XPRS XPRS XPR3 86 3 6 Combining Sub Paths into Paths Each of the 20 sub path gets initialized with a random phase 4 In addition a deterministic phase compo nent rim 51 is obtained from the total length of the propagation path Both components are combined to oe exp jY m rt lms 87 The initial channel coefficients for each sub path including the polarization effects are obtained by 59 Now the 20 sub paths get combined into a single path However due to the random initial phases a simple sum will result in a random path power since it is unpredictable if the phase components add up constructively or destructively This is compensated by normalization where the sum of the complex phases is calculated separately over all S snapshots of a segment Then the combined path weight is calculated as 2 Ip tls Ayo ee s AD vk s 88 2 B 2 a 89 s 1 m 1 where P is the initial path power from Section 3 2 Copyright Fraunhofer Heinrich Hertz Institute 72 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 3 TECHNICAL DOCUMENTATION 3 7 Path Gain Shadow Fading and K Factor No
164. on and P Vainikainen Parameterization of the COST 2100 MIMO channel model in indoor scenarios Proc EUCAP 11 pp 3606 3610 2011 Y Zhou S Rondineau D Popovic A Sayeed and Z Popovic Virtual channel space time processing with dual polarization discrete lens antenna arrays IEEE Trans Antennas Propag vol 53 pp 2444 2455 Aug 2005 Copyright Fraunhofer Heinrich Hertz Institute 9 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 References 26 C Oestges B Clerckx M Guillaud and M Debbah Dual polarized wireless communications From propagation models to system performance evaluation IEEE Trans Wireless Commun vol 7 no 10 pp 4019 4031 2008 27 M Hata Empirical formula for propagation loss in land mobile radio services EEE Trans Veh Technol vol 29 no 3 pp 317 325 1980 Copyright Fraunhofer Heinrich Hertz Institute 10 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 1 INTRODUCTION AND OVERVIEW 1 Introduction and Overview 1 1 Installation and System Requirements The installation is straightforward and it does not require any changes to your system settings If you would like to use QuaDRiGa just extract the ZIP File containing the model files and add the source folder from the extracted archive to you MATLAB Path This can be done by opening MATLAB and selecting File Set Path from the menu The you can use the
165. on are indicated by the arrows The lower figure shows the LOS power along the track The dashed curve comes from the WINNER approach The new model solid line calculates the change of the polarization due to the antenna orientation and adjusts the polarization accordingly 3 5 2 Changing the Orientation of Antennas Orientation changes are desirable in many cases e g when tilting BS arrays or changing the orientation of mobile terminals An example is depicted in Figure 11 The left side of the figure shows a dipole pattern that has only a vertical component and is in line with the global coordinate system GCS The right side shows the same antenna tilted by 20 around the z axis In order to maintain alignment with the GCS the antenna pattern was transformed The resulting pattern also has a horizontal component The antenna response 58 can now be obtained by reading the polarimetric beam pattern at the given angles 0 Interpolation of the antenna pattern An antenna pattern 58 is given in spherical coordinates as a function of the elevation angle 0 and the azimuth angle When the orientation of the antenna element changes the field pattern has to be read at different angles 0 which include the effect of the orientation change Rotations in 3 D are easier in Cartesian coordinates Therefore the original angle pair 0 is transformed into a vector c that describes the arrival or departure angles in Cartesian coordinates
166. on of the model steps can be found Section 3 1 The user of the model needs to configure the network layout This includes Setting the transmitter position e g the BS positions or the satellite orbital position Defining antenna properties for the transmitter and the receiver Defining the user trajectory Defining states or segments along the user trajectory Assigning a propagation environment to each state Defining the user trajectory states along the user trajectory and related parameters is performed by the state sequence generator SSG In the current implementation different SSGs are available e Manual definition of all parameters by the user e g definition of short tracks e Statistical model for the journey A simple model mainly designed for demonstration and testing purpose is included in the tutorial satellite_channel e Derive trajectory and state sequence from the measurement data 2 Configuration files define the statistical properties of the LSPs For each state also called scenario a set of properties is provided Typically two configurations files are used e One for the good state also called LOS scenario e The other for the bad state NLOS scenario For each state QuaDRiGa generates correlated maps for each LSP For example the delay spread in the file is defined as log normal distributed with a range from 40 to 400 ns QuaDRiGa translates this distribution in to a series
167. on parameters of the segment length and then calculate the segments themselves The two possible states are MIMOSA_10 45_LOS which stands for LOS or good state and MIMOSA_10 45_NLOS for NLOS or bad state segment_length_mu 30 4 Average segment length in m segment_length_sigma 12 4 Standard deviation in m min_segment_length 10 4 Minimum segment length in m Now we define the segments the states along the track ind 1 while ind lt t no_snapshots 4h Each scenario has a 50 probability if rand lt 0 5 t scenario t no_segments MIMOSA_10 45_LOS else t scenario t no_segments end gt MIMOSA_10 45_NLOS 4 Get the length of the current segment segment_length randn segment_length_sigma segment_length_mu while segment_length lt min_segment_length segment_length randn segment_length_sigma segment_length_mu end segment_length round segment_length 4 Segment length ind ind segment_length 4 Start of next segment if ind lt t no_snapshots 4 Exception for the last segment t no_segments t no_segments 1 t segment_index t no_segments ind end end Track layout Track Oo Segment start 250 4 200 7 E 150 e 3 3 3 sh IMOSA_10 45_LOS 1007 3 IMOSA_10 45_ OS SMIMOSA_10 45_NLOS 50 MIMOSA_10 45_LOS oL MOSA_10 45_LOS 250 200 150 100 50 0 X direction m Figure 18 Receiver track for the satellite channel tutoria
168. ong the track gets assigned an environment In the QuaDRiGa terminology this is called a scenario E g the first segment on the track is in the Satellite LOS Urban scenario The selection of the scenario is done during the first step set up tracks scenarios antennas and network layout QuaDRiGa itself does not supply functions to perform the setting up of tracks and scenarios automatically However external scripts can be used to perform this task An example can be found in section A 3 A RHCP LHCP signal is defined in the antenna setup After the model setup the automatic mode generates a set of LSPs for this segment I e the second step of the model calculates one value for each of the 7 LSPs using the map based method Thus a set of seven maps is created for the scenario Satellite LOS Urban Those maps cover the entire track Thus the same maps are used for all Satellite LOS Urban segments of the track The third step then calculates a time series of fading coefficients for the first segment that have the properties of the LSPs from the map E g if one calculates the RMS DS from the coefficients one gets the same value as generated by the map in step 2 2 LOS gt NLOS Change A scenario change is defined along the track E g the second segment along the track gets assigned the scenario Satellite NLOS Urban Now a second set of maps is generated for all Satellite NLOS Urban segments So in tot
169. or varying speeds is Copyright Fraunhofer Heinrich Hertz Institute 56 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 3 TECHNICAL DOCUMENTATION then obtained by interpolating the coefficients in a postprocessing step o fr 4 Is max v Ac 4 gt fs SD P gt 2 5 Longer time evolving channel sequences need to consider the birth and death of scattering clusters as well as transitions between different propagation environments This is addressed by splitting the MT trajectory into segments A segment can be seen as an interval in which the LSPs do not change considerably and where the channel keeps its wide sense stationary WSS properties Thus the length of a segment depends on the decorrelation distances of the LSPs Hence it might be beneficial to limit the segment length to the average decorrelation distance In the WINNER urban macro cell UMa scenario this would be 22 m for LOS and 48 m for NLOS propagation Channel traces are then generated independently for each segment In Section 3 8 on Page 73 those individual traces are combined into a longer sequence that includes the birth and death of scattering clusters 3 1 Correlated Large Scale Parameter Maps The positions of the scattering clusters are based on seven LSPs RMS delay spread DS Ricean K factor KF Shadow fading SF Azimuth spread of departure ASD Azimuth spread of arrival ASA Elevation spread of departure ESD Elevation
170. ormal distributed For example the median log normal delay spread DS in an urban cellular scenario is 6 89 which corresponds to a DS of o 128 ns With a standard deviation of DS 0 5 typical values may lie in between 40 and 407 ns The same principle applies for the other six LSPs The decorrelation distance e g DS 40 m describes the distance dependent correlation of the LSP If e g two mobile terminals in the above example are 40 m apart of each other their DS is correlated with a correlation coefficient of e 0 37 Additionally all LSPs are cross correlated A Copyright Fraunhofer Heinrich Hertz Institute 57 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 3 TECHNICAL DOCUMENTATION Parameter maps Linear transformation to impose White Gaussian inter parameter correlation noise generator Bos Cir os Ciy Aps Besp Cr Cy Agsp 2D autocorrelation shaping i Local values for an individual MT position Initial delays and cluster powers x 6 Z 4 2 gt 0 i I bd 0 200 400 600 Delay ns Figure 6 Principle of the generation of channel coefficients based on correlated LSPs typical example is the dependance of the angular spread AS e g the azimuth spread of arrival on the Ricean K factor KF With a large KF e g 10 dB
171. ormalization is done Default All elements Output gain_dBi Normalized gain of the antenna Copyright Fraunhofer Heinrich Hertz Institute eMail quadriga hhi fraunhofer de 28 QuaDRiGa v1 2 32 458 2 SOFTWARE STRUCTURE V H CP dist interpolate azimuth elevation element Description Interpolates the field pattern Note Interpolation of the beam patterns is very computing intensive It must be performed several thousands of times during a simulation run It is highly recommended to use linear interpolation for this task since this method is optimized for speed Spline interpolation calls the MATLAB internal interpolation function which is more than 10 times slower To enable linear interpolation set the interpolation_method property of the array object to linear Remark There are additional input parameters specified in the mat File that are not in the list below Those parameters correspond to the properties of the array class Passing those variables during the function call takes less time than reading them from the object properties This is used internally in channel_builder get_channels but is irrelevant here Input azimuth A vector of azimuth angles in rad elevation A vector of elevation angles in rad element The element numbers for which this interpolation is done is applied If no element number is given the interpolation is done for all elements in the array
172. otation matrix having the form MO cos sind 70 sinv cos Copyright Fraunhofer Heinrich Hertz Institute 69 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 3 TECHNICAL DOCUMENTATION where the polarization rotation angle J follows from sin 0 cos sin cos cos sin 0 sin R sinOsin 71 cos 0 cos O sing sin cos sind cos R sinOsin 72 0 cos O V arctan sin V cos J 73 3 5 3 Constructing the Polarization Transfer Matrix Polarization of direct LOS path In the LOS polarization model the probe is replaced by the receive antenna Both the transmitter and the receiver can have different orientations e g due to a downtilt at the BS and a given movement direction at the MT In addition a reflection operation is needed to transform the outgoing polarization direction at the transmitter into an incoming direction at the receiver Thus a combination three linear transformations two rotations and one reflection is sufficient to construct the polarization transfer matrix of the LOS path Los _ gr oros 1 9 g r MESI m 099 0 et M oRe cos gto S _ sin gto S cos okos in gos en sin gies cos gle a sin gol Zo fel Model for the indirect NLOS paths For the NLOS components the transmitted signal undergoes some diffraction reflection or scattering before reaching the receiver Following the common Fresnel formula in elect
173. owever if the angles are mapped randomly to the path powers the realistically achievable ASs are lower Figure 8 shows simulation results for the maximum AS depending on the KF For NLOS propagation the achievable azimuth spread is around 100 and the elevation spread is around 65 When the requested AS is larger than the maximum angle then the angles are adjusted in a way that the AS at the output of the model is close to the maximum AS This is illustrated in the right part of Figure 8 120 ee i oO z gy Unity o Re 100 100 Azimuth spread Hm dB a Eick ees ai Elevation sprea Bi 80 E 80 Hi a 4 ala i a a a 60 Zz 60 Q l pe 3 40 a AO A i i i i i a 20 Azimuth spread pp 20 A jeBlevation spread O o a o 20 15 10 5 0 5 10 15 20 0 20 40 60 80 100 120 K Factor dB Requested angular spread deg Figure 8 Maximal achievable angular spread depending on the K factor Subpaths Finally the NLOS paths are split into 20 sub paths to emulate intra cluster ASs The LOS path has no sub paths T Co m m 4 _ fol gt l 39 Q Q 180 or l gt 39 T Co m 61 O 180 or l gt Here m is the sub path index cg is the scenario dependent cluster wise RMS AS and is the offset angle of the m sub path from Table 11 cg and are given in degrees here Furthermore each of the 20 angle pairs
174. p adjusts this percentage ranging from 0 i e very hard step like change at the scenario boundary to 1 very smooth but long transition usage Changes the behavior of the method usage 0 Deletes all existing parameters from the track usage 1 Deletes all existing parameters from the track and generates new ones Existing LSPs will be overwritten usage 2 default Keeps existing parameters but generates missing ones This is useful when for example the effective PG is provided but the other LSPs are unknown In this case the unknown gaps are filled with values which are generated from the provided scenario description check_parfiles check_parfiles 0 1 default 1 Disables 0 or enables 1 the parsing of shortnames and the validity check for the config files This is useful if you know that the parameters in the files are valid In this case this saves execution time Output par The automatically generated parameters This cell array contains a parameter structure of the LSPs for each receiver with the following fields e ds The delay spread in s per segment e kf The Ricean K Factor in dB per snapshot e pg The effective path gain in dB excluding antenna gains per snapshot e asD The azimuth angle spread in deg per segment at the transmitter e asA The azimuth angle spread in deg per segment at the receiver e esD The elevation angle spread in deg per segment at the transmitter e esA The elevation an
175. ptical polarized signals to linear polarized signals This implies that using 60 with complex coefficients results in a completely random polarization behavior when XPR and XPRy are small i e when the off diagonal elements are large When XPR is large and the off diagonal elements are close to zero then 60 describes a scaling operation Here a different method to model the polarization based on a combination of linear transformations is proposed In the following M will be calculated explicitly for the LOS component taking the orientation of the antennas into account For the NLOS components additional operations are used to convert the XPR value from the parameter tables into Jones matrices for the linear and elliptical polarization component An example showing the effect of the new method is depicted in Figure 10 The upper part shows the model setup and the lower part the results Both the transmitter and the receiver are equipped with dipole antennas that were initially slanted by 45 around the y axis The transmitter is placed 5 m above ground and 5 m north of the center The receiver moves counterclockwise around the transmitter with its Copyright Fraunhofer Heinrich Hertz Institute 67 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 3 TECHNICAL DOCUMENTATION antenna orientation changing in accordance with the movement direction The orientations of transmitter and receiver including the movement directi
176. r m and the snapshot number s within the current segment respectively The scatterer positions are kept fixed for the time it takes a MT to move through a segment Hence the angles 6 seen from the BS do not change except for the LOS angle which is treated separately Based on this assumption the angles 0 as well as the path delay only change with respect to the last bounce scatterer LBS Hence if the BS array size is small compared to the BS MT distance it is sufficient to consider only a single scatterer the last bounce scatterer LBS for the NLOS paths last bounce scatterer i fe a m oh os Din yy NN i initial Rx location lt lt a a Ty IK eation Rx location at snapshot s Figure 9 Illustration of the calculation of the scatterer positions and updates of the arrival angles NLOS drifting The position of the LBS is calculated based on the initial arrival angles and the path delays Then the angles and path lengths between the LBS and the terminal are updated for each snapshot on the track This is done for each antenna element separately Figure 9 illustrates the angles and their relations The first delay is always zero due to 17 Hence the total length of the It path is dj T c l r 41 where r is the distance between the Tx and the initial Rx location and c is the speed of light All sub paths have the same delay and thus the same path length However each sub path has differe
177. r each snapshot Drifting Path Loss Snapshot position Interpolate KF and SF maps Draw random initial phases Generate XPR Generate channel coefficients X pol power ratios Drifting angles of arrival Generate drifting delays Overlapping area Connect successive Scaled channel Apply path loss K factor channel traces coefficients and shadow fading Merged channel coefficients Generation of Correlated Large Scale Parameters Per cluster SF std Drifting SF Drifting KF Delay factor No clusters Angular Spread K Factor Delay spread K Factor Generate initial Generate cluster delays powers Cluster powers Generate drifting AoAs Drifting delay for each path Channel coefficients Initial delay for each path y Constant values Apply speed profile P One update per segment t hot or Sea EX Figure 4 QuaDRiGa Data Flow Copyright Fraunhofer Heinrich Hertz Institute 49 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 2 SOFTWARE STRUCTURE 2 4 Scenario Specific Parameters The large scale parameters LSPs are defined by the parameter files which can be found in the folder config of the QuaDRiGa Core source folder The parameters are processed as follows e The states and segments are identified by a name Examples are S1 S2 gt MIMOSA_10 45_LOS parameter set selected for good state gt MIMOSA_10 45_NLO
178. r than 0 dB use_angular_mapping 2 default This method generates random angles for the paths The angular spread is maintained by a scaling operation The output angles have a more natural distribution However there is an upper limit for the angular spread of roughly 100 degree in NLOS conditions use_map _algorithm Selects the parameter map generation algorithm use_map algorithm 1 Uses the algorithm from the WINNER model use_map algorithm 2 Default Uses a modified version of the WINNER algorithm that also filters the diagonal directions show_progress_bars Show a progress bar on the MATLAB prompt center_frequency Center frequency in Hz map_resolution Resolution of the decorrelation maps in samples m version Version number of the current QuaDRiGa release constant speed_of_light Speed of light constant wavelength Carrier wavelength in m read only Methods h_simpar simulation_parameters Description Creates a new simulation_parameters object with default settings set_speed speed_kmh sampling _rate_s Description This method can be used to automatically calculate the sample density for a given mobile speed Input speed_kmh speed in km h sampling _rate_s channel update rate in s Copyright Fraunhofer Heinrich Hertz Institute 24 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 2 SOFTWARE STRUCTURE 2 2 2 Class array
179. r vs Rx position drifting phases tutorial 93 Phases and Tx power vs Rx position without drifting drifting phases tutorial 94 Scenario setup for the time evolution tutorial 2 0 ee 96 Received power on the circular track time evolution tutorial 0 0 97 Received power on the linear track time evolution tutorial 00 0 99 Scenario setup for the speed profile tutorial oaoa a a ee 101 Received power and 2D PDP for the speed profile tutorial oaoa a a 101 Movement profile left and interpolated PDP right o oaa 103 Polarimetric dipole antenna patterns for different orientations 0 104 Scenario layout 4 2 sora ad dami p Ee ee GRE a e Ewe EE ee a 105 Results from the geometric polarization tutorial 2 2 ee 106 RHCP 7 LAG antenna Patterns eos ece gow Goh oo eh ee Ee ee ee ees Bw 109 Scenario overview manual parameter selection 0 00 2 e ee eee 111 Power along the track manual parameter selection 00 00000 0 113 DS along the track manual parameter selection 0 00000 0 eee ee eee 113 Visualization of the angular spread correction function C L K 0 115 List of Tables 1 QuaDRiGa System Requirements 2 a a 11 Parameter sets provided together with the standard software 08 50 11 Offset Angle of the m Sub Path from 3 i44454 4 4554404044045 500 e584 64 12 Correction value
180. ratio For each channel segment the channel parameters are calculated from the distributions Specific channel realizations are generated by summing contributions of rays with specific channel parame ters like delay power angle of arrival and angle of departure Different scenarios are modeled by using the same approach but different parameters The basic features of the model approach can be summarized as follows e Support of freely configurable network layouts with multiple transmitters and receivers Scalability from a single input single output SISO or multiple input multiple output MIMO link to a multi link MIMO scenario Same modeling approach indoor outdoor and satellite environments as well as combinations of them Support of a frequency range of 2 6 GHz with up to 100 MHz RF bandwidth Support of multi antenna technologies polarization multi user multi cell and multi hop networks Smooth time evolution of large scale and small scale channel parameters including the transition be tween different scenarios e High accuracy for the calculation of the polarization characteristics Copyright Fraunhofer Heinrich Hertz Institute 11 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 1 INTRODUCTION AND OVERVIEW The QuaDRiGa channel model largely extends the WINNER model to support several new features that were originally not included These are e Time evolution Short term time evolution of the channel coefficients is re
181. re y is the row index and z is the column index The first pixel B1 is in the top left or north west corner of the map The FIR filter coefficients are calculated from the decorrelation distance d in units of meters The distance dependent correlation coefficient follows an exponential function pla exp 2 6 A Copyright Fraunhofer Heinrich Hertz Institute 58 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 3 TECHNICAL DOCUMENTATION with d as the distance between two positions 14 We now calculate two sets of filter coefficients one for the horizontal and vertical directions and one for the diagonal elements This is done by taking 6 and substituting the distance d by the relative distance dpx of two pixels 1 k dpx 7 poe a a T ak oR exp dh 7 1 k V2 dpx bk a et 8 k is the running filter coefficient index The exponential decay function is cutted at a maximum distance of 4 d and normalized with vd The map size is therefore determined by the distribution of the users in the scenario plus the length of the filter function This is also illustrated in Figure 7 where the user terminals are placed inside the black square The extension space is needed to avoid filter artifacts at the edges of the map The map is initialized with random normal distributed numbers Then the filter 7 is applied in vertical direction running from top to bottom and in horizontal direction from left to ri
182. rios If there is only one transmitter i e one base station the cell array has the dimension 1 x no_segments For multiple transmitters the rows of the array may contain different scenarios for each transmitter For example in a multicell setup with three terrestrial base stations the propagation conditions may be different to all BSs The cell arrays than has the dimension 3 x no_segments par Manual parameter settings This variable contains a structure with LSPs This can be used for assigning LSPs directly to the channel builder e g when they are obtained from measurements The structure contains the following fields ds The delay spread in s per segment kf The Ricean K Factor in dB per snapshot pg The effective path gain in dB excluding antenna gains per snapshot asD The azimuth angle spread in deg per segment at the transmitter asA The azimuth angle spread in deg per segment at the receiver esD The elevation angle spread in deg per segment at the transmitter esA The elevation angle spread in deg per segment at the receiver xpr The NLOS cross polarization in dB per segment If there is only a subset of variables e g the angle spreads are missing then the corresponding fields can be left empty They will be completed by the parameter sets See also the method track generate_parameters on how to fill this structure automatically ground_direction Azimuth orientation of
183. rns into a QuaDRiGa array object Input Vi fHi azimuth_grid elevation_grid The field pattern s for the vertical polarization given in spherical coordinates The first dimension corresponds to the elevation angle ranging from 90 to 90 degrees The second dimension is for the azimuth angle ranging from 180 to 180 degrees The third dimension belongs to the element number The default resolution is 1 degree hence the default size of fVi is j181x361x1j If a different resolution is given the optional variables azimuth_grid and elevation_grid must be defined The field pattern s for the horizontal polarization given in spherical coordinates fHi can be empty if no horizontal response is given If it is given then fHi must have the same size as fVi A vector specifying the azimuth sampling points of the patterns in units of radians raging from pi to pi This value only needs to be defined if the patterns do not have the default size A vector specifying the elevation sampling points of the patterns in units of radians raging from pi 2 to pi 2 This value only needs to be defined if the patterns do not have the default size Output h_array The QuaDRiGa antenna array object generated from the field patterns gain_dBi normalize_gain element Description Normalizes all patterns to their gain Input element A list of elements for which the n
184. rodynamics the polarization direction can be changed at the boundary surface between two dielectric media T Svantesson 21 provided a method for modeling the polarization of a reflected wave where the polarization coupling is a function of several geometric parameters such as the orientation of the scatterers However these parameters are not generally available in the SCM In addition to that only metallic reflec tions keep the polarization unchanged Reflections at dielectric media can cause changes of the polarization being a function of the complex valued dielectric constant of the media and of the angle of incidence Hence not only the polarization angle might change but also the polarization type In order to address this issue studies of the polarizations effects in individual scattering clusters in several outdoor and indoor scenarios were done 22 24 The published results indicate that in many cases scattering preserves the polarization quiet well However since only the powers of the elements in the polarization coupling matrix were analyzed no conclusions can be drawn on how elliptic the polarization of the scattered wave becomes The polarization coupling matrix M for the NLOS components can be described by a combination of linear transformations Hence it is possible to take advantage of the existing findings of the XPR The NLOS polarization model consists of three parts A deterministic part and two stochastic parts 1 Deter
185. s from 3 for different numbers of paths 0 116 13 Comparison of the correction functions s osoo e e a e 116 Copyright Fraunhofer Heinrich Hertz Institute 4 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 List of Acronyms List of Acronyms 2 D 3 D AoA AoD AS ASA ASD BS CIR DS EoA EoD ESA ESD FBS FIR GCS iid KF LBS LHCP LOS LSP MIMO MIMOSA MPC MT NLOS PDP PG PL RHCP Rx SCM SF SISO SSG STD Tx UML WGS WINNER WSS WSSUS XPR two dimensional three dimensional azimuth angle of arrival azimuth angle of departure angular spread azimuth spread of arrival azimuth spread of departure base station channel impulse response delay spread elevation angle of arrival elevation angle of departure elevation spread of arrival elevation spread of departure first bounce scatterer finite impulse response global coordinate system independent and identically distributed Ricean K factor last bounce scatterer left hand circular polarized line of sight large scale parameter multiple input multiple output MIMO over satellite multipath component mobile terminal non line of sight power delay profile path gain path loss right hand circular polarized receiver spatial channel model shadow fading single input single output state sequence generator standard deviation transmitter unified modeling language world geodetic system Wireless World Initiative for N
186. s to typical values in the range of 10 9 3 102 ns to 1076 69 0 3 407 ns As for the shadow fading the decorrelation distance DS defines how fast the DS varies when the terminal moves through the environment The delay spread o is calculated from both the delays 7 and the path powers P I e lager delay spreads g can either be achieved by increasing the values of and keeping P fixed or adjusting P and keeping 7 fixed In order to avoid this ambiguity an additional proportionality factor delay factor r is introduced to scale the width of the distribution of 7 r is calculated from measurement data See Sec 3 2 for more details Ricean K Factor KF Rician fading occurs when one of the paths typically a line of sight signal is much stronger than the others The KF K is the ratio between the power in the direct path and the power in the other Copyright Fraunhofer Heinrich Hertz Institute 51 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 2 SOFTWARE STRUCTURE scattered paths As for the DS the KF is assumed to be log normal distributed The distribution is defined by its median value KF and its STD KF The decorrelation distance KF defines how fast the KF varies when the terminal moves through the environment Angular Spread The angular spread defines the distribution of the departure and arrival angles of each multipath component in 3D space seen by the transmitter and receiver respectivel
187. se stations deployed below rooftop in densely populated urban areas BERLIN_UMa_LOS BERLIN UMa NLOS Terrestrial Urban Macrocell parameters extracted from measurements in Berlin Ger many MIMOSA_10 45_LOS MIMOSA_10 45_NLOS MIMOSA Satellite to Mobile Parameters for Urban Propagation Elevation range from 10 to 45 Parameters were extracted from terrestrial measurement using a high attitude platform MIMOSA_16 25_LOS MIMOSA_16 25_NLOS MIMOSA Satellite to Mobile Parameters for Urban Propagation Elevation range from 16 to 25 Parameters were extracted from terrestrial measurement using a high attitude platform MIMOSA_25 35_LOS MIMOSA_25 35_NLOS MIMOSA Satellite to Mobile Parameters for Urban Propagation Elevation range from 25 to 35 Parameters were extracted from terrestrial measurement using a high attitude platform MIMOSA_35 45_LOS MIMOSA_35 45_NLOS MIMOSA Satellite to Mobile Parameters for Urban Propagation Elevation range from 35 to 45 Parameters were extracted from terrestrial measurement using a high attitude platform 2 4 1 Description of the Parameter Table The QuaDRiGa channel model is a generic model That means that it uses the same method for generating channel coefficients in different environments E g the principal approach is exactly the same in a cellular network and in a satellite network The only difference is the parametrization for both cases Each
188. sen P Mogensen and B Fleury Power azimuth spectrum in outdoor environments Elec tronics Letters vol 33 no 18 pp 1583 1584 1997 G F Masters and S F Gregson Coordinate system plotting for antenna measurements AMTA Annual Meeting amp Symposium 2007 R C Jones A new calculus for the treatment of optical systems i description and discussion of the calculus Journal of the Optical Society of America vol 31 pp 488 493 July 1941 M Narandzic M Kaske C Schneider M Milojevic M Landmann G Sommerkorn and R Thoma 3D antenna array model for IST WINNER channel simulations Proc IEEE VTC 07 Spring pp 319 323 2007 S Gregson J McCormick and C Parini Principles of Planar Near Field Antenna Measurements IET 2007 T Svantesson A physical MIMO radio channel model for multi element multi polarized antenna systems Proc IEEE VTC 01 Fall vol 2 pp 1083 1087 2001 L Materum J Takada I Ida and Y Oishi Mobile station spatio temporal multipath clustering of an estimated wideband MIMO double directional channel of a small urban 4 5 GHz macrocell EURASIP J Wireless Commun Netw no 2009 804021 2009 F Quitin C Oestges F Horlin and P De Doncker A polarized clustered channel model for indoor multiantenna systems at 3 6 GHz IEEE Trans Veh Technol vol 59 no 8 pp 3685 3693 2010 J Poutanen K Haneda L Liu C Oestges F Tufvess
189. simulation_parameters Basic simulation parameters s center_frequency 2 185e9 s sample_density 2 s use_absolute_delays 1 t track linear 500 135 pi 180 4 Track with 500m length direction SE t initial_position 120 120 0 4 Start position t interpolate_positions 1 hk Interpolate to 1 sample per meter t segment_index 1 45 97 108 110 160 190 215 235 245 280 295 304 330 400 430 l 4 Set segments states S1 MIMOSA_10 45_LOS Sn MIMOSA_10 45_NLOS t scenario Sn S1 S5n S1 S5n Sn Sn S1 S5n S1 5n 81 Sn Sn Sn Sn t interpolate_positions 3 1 layout s 1 tx_position 0 0 125 l track t Set the position of the Te Set the ra track ax Generate Tx antenna 30 deg Tilt point southwards 1l tx_array array rhcp lhcp dipole tx_array rotate_pattern 30 y 1 tx_array rotate_pattern 90 z ja x x de Ra Antenna point skywards l rx_array array rhcp lhcp dipole l rx_array rotate_pattern 90 y axe l visualize view 33 45 lnk 1 tx_position 1l track positions 1 track segment_index 2 l track initial_position hold on plot3 1Ink 1 1nk 2 1nk 3 hold off Generate channel coefficients Next we calculate the channel coefficients p cb 1 create_parameter_sets 0 p 2 scenpar NumClusters 14 p update_parameters c cb get_channels cn c merge 0 2 P
190. spread of arrival ESA eS oe ONS Their distribution properties are directly obtained from measurement data e g 3 6 11 If some MTs or segments are close to each other their LSPs will be correlated and they will experience similar propagation conditions This is modeled by means of two dimensional 2 D maps as illustrated in Figure 6 Our method for generating these maps is adopted from 12 The maps are initialized with values obtained from an independent and identically distributed i i d zero mean Gaussian random process with desired variance The pixels are then subsequently filtered to obtain the desired autocorrelation function i e a decaying exponential function with a specific decorrelation distance In contrast to 12 the maps are filtered also in the diagonal direction to get a smooth evolution of the values along the MT trajectory Advanced methods going beyond this approach for generating such maps are discussed in 13 Once the maps are generated initial LSPs for each segment are obtained by interpolating the maps to match the exact position of the MT The granularity of each LSP can be described on three levels the propagation scenario level the link level and the path level e Propagation scenario level The magnitude variance and the correlation of a LSP in a specific scenario e g urban macro cell indoor hotspot or urban satellite are calculated from measurement data Normally LSPs are assumed to be log n
191. tated by 45 When moving around the circle the Tx stays fixed and the Rx rotates Subsequently at one position we will have both dipoles aligned and at another position both will be crossed When they are crossed the received power will be 0 and when they are aligned the power will match the first plot two vertical dipoles This can be seen in the following figure figure 4 New figure plot abs squeeze c coeff 2 2 1 2 Linewidth 1 axis 0 360 0 1 1 set gca XTick 0 45 360 xlabel Position on circle degrees ylabel LOS Power linear scale title Tz 45 cire Re 45 cire In the last figure we rotated the transmit antenna by 90 It is thus lying on the side and it is horizontally polarized For the Rx we consider three setups Vertical blue line 45 green line and 90 red line Note that the Tx is rotated around the y axis At the initial position 0 the Rx 45 and 90 is rotated around the x axis This is because the movement direction figure 4 New figure plot abs squeeze c coeff 3 1 72 axis 0 360 0 1 1 legend Re O cire Raz 6 cire Rzi SO cire xlabel Position degrees ylabel LOS Power linear scale title x lt SO ciee Be Oieiee 6icite 9O ecire When the receiver is vertical blue line both antennas are always crossed There is no position around the circle where a good link c
192. te eMail quadriga hhi fraunhofer de 88 40 e WNE 10 16 QuaDRiGa v1 2 32 458 A TUTORIALS xlabel Track m ylabel Received Power per MIMO LINK dB axis 0 t get_length mi ma legend L0S tP tits Ria PUe eit P Teer 4 box on title Received power along the track The next plot shows the RMS delay spread along the path for the first MIMO element Again shaded ares are for the LOS segments pow_tap abs squeeze cn coeff 1 1 72 pow_sum sum pow_tap 1 mean_delay sum pow_tap cn delay 1 pow_sum ds sqrt sum pow_tap cn delay 2 1 pow_sum mean_delay 2 ar zeros 1 cn no_snap ar los 10000 figure Position 100 100 1000 700 a area dist ar set a i FaceColor 0 7 0 9 0 7 set a LineStyle none hold on plot dist ds 1le9 hold off ma 1e9 max ds 0 1 max ds axis 0 t get_length O ma xlabel Track m ylabel Delay Spread ns legend LOS sigma_ tau 1 title Position dependant delay spread close all disp QuaDRiGa Version simulation_parameters version QuaDRiGa Version 1 0 1 145 Copyright Fraunhofer Heinrich Hertz Institute 89 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 A TUTORIALS Received power along the track Received Power per MIMO LINK dB
193. the propagation channel scattering cluster w sa s somat d mcr a eR ee ee ee a 56 Synonym for cluster Copyright Fraunhofer Heinrich Hertz Institute 6 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 List of Symbols BCCTOTVO g eget aes Ge deuns G8 ee ae bdo War Geer See GR Se A oe we ee oe Gy See A 57 In this thesis the term scenario refers to a specific propagation environment such as Urban macro cell Urban satellite Indoor hotspot etc Usually each propagation environment can be further split into LOS and NLOS propagation e g Urban macro cell LOS and Urban macro cell NLOS both of which might have very different properties In the channel model each scenario is fully specified by a parameter table SEGMEN in i ic de RN ee ed ee ae eS ee e a a 57 64 73 Segments are parts of a user trajectory in which the LSPs do not change considerably and where the channel keeps its WSS properties Typical segment lengths are 5 30 m It is assumed that within a segment the scattering clusters are fixed SUb path 244 686 Pe RE ORDA a a eRe REG ee bbe dea Ped a as 63 72 A sub path is the exact way that a signal takes from the transmitter to the receiver It contains at least one reflection However normally the channel model uses two scatterers resulting in two reflections to create a sub path 20 sub paths are combined to a path The LOS path has no sub paths time evolution oe i aoa e be be Oe
194. the terminal antenna for each snapshot This variable can be calculated automatically from the positions by the function track compute_directions Copyright Fraunhofer Heinrich Hertz Institute 31 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 2 SOFTWARE STRUCTURE height_direction Elevation orientation of the terminal antenna for each snapshot closed Indicates that the track is a closed curve Methods turn_probability h_track track track_type track_length direction street_length_min street_length_mu street_length std curve_radius Description Creates a new track object See track generate for a description of the input parameters and the list of supported track types compute_directions Description Calculates ground and height orientations from positions This function calculates the orientations of the transmitter based on the positions If we assume that the receive antenna array is fixed on a car and the car moves along the track then the antenna turns with the car when the car is changing direction This needs to be accounted for when generating the channel coefficients This function calculates the orientation based on the positions and stored the output in the ground_direction and height_direction field of the track object copy_objects Description A modified version of the standard physical copy function While th
195. ting is required when non linear tracks are generated or the distance between transmitter and receiver is small below 20 m The phases at the antenna arrays are calculated by a planar wave approximation drifting precision 2 The arrival angles the LOS departure angle delays and phases are updated for each snapshot and for each antenna element at the receiver spherical wave assumption The phases at the transmitter are calculated by a planar wave approximation This increases the accuracy for multi element antenna arrays at the receiver However the computational complexity increases as well drifting_precision 3 EXPERIMENTAL This option also calculates the shadow fading path loss and K factor for each antenna element at the receiver separately This feature tends to predict higher MIMO capacities since is also increases the randomness of the power for different MIMO elements drifting precision 4 This option uses spherical waves at both ends the transmitter and the receiver This method assumes a single bounce model and no mapping of departure and arrival angles is done Hence departure angular spreads are effectively ignored and results might be erroneous drifting precision 5 This option uses spherical waves at both ends the transmitter and the receiver This method uses a multi bounce model where the departure and arrival angels are matched such that the angular spreads stay consistent Copyright Fraunhofer Hei
196. tion procedure either to the pattern or polarization Possible values are e 0 Rotate both pattern polarization default e 1 Rotate only pattern e 2 Rotate only polarization Output cp The common phase of the field pattern set_grid azimuth_grid elevation_grid Description Sets a new grid for azimuth and elevation and interpolates the pattern This function replaces the properties azimuth_grid and elevation_grid of the antenna object with the given values and interpolates the antenna patterns to the new grid Input azimuth_grid Azimuth angles in rad were samples of the field patterns are provided The field patterns are given in spherical coordinates This variable provides the azimuth sampling angles in radians ranging from 7 to 7 elevation_grid Elevation angles in rad were samples of the field patterns are provided The field patterns are given in spherical coordinates This variable provides the elevation sampling angles in radians ranging from 4 downwards to 5 upwards sub_array mask Description generates a sub array with the given array indices This function creats a copy of the given array with only the selected elements specified in mask Input mask A list of element indices Copyright Fraunhofer Heinrich Hertz Institute 29 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 2 SOFTWARE STRUCTURE uncompress D
197. tx no rx tx_name tx_position tx_array rx_name rx_position rx_array track pairing no_links create parameter sets estimate memory _usage generate generate parameters get_channels get_channels_ seg power_map randomize_rx_positions set_pairing set_satallite pos visualize Implements the state merger and the functions to interpolate a specific speed profile name version individual_ delays coef f delay initial position tx_position rx_position merges with interpolate merge split_tx lt lt output gt gt channel movement_profile no_segments segment_index scenario par ground direction height_direction closed compute directions copy objects correct_overlap generate generate parameters get_length get_subtrack interpolate_ movement interpolate positions set_scenario set_speed split_segment visualize map_extent map_size samples_per_ meter map_valid LSP_xcorr_matrix l LSP_matrix_isOK loreates map_x_ coord l map_y_coord I I I copy_objects get_angles get_channels get_distances get_pl t get_sf_profile t set_par supported_ scenarios static t update_parameters Creates calc_scatter_positions get_channels get_los_channels init_parameters visualize _clusters Figure 3 UML class diagram of the mo
198. uaDRiGa v1 2 32 458 A TUTORIALS Power of the LOS component for the circular track PDP for the circular track with drifting 85 r i 90 90 i AU al Mt 95 3 Wid ee oN aie 2 Hie As Ad hh S 95 Vi Ng iy D 100 g KA hc iy 5 Hg il ii 2 gt 100 Wh yi W g ios a iN iy ll 5 P K Wi i Z 110 2 105 il ii i J g 4 5 l i 225 E ia g 4 3 110 4 E i E 120 E 4 115f a i 2 i 125 i Drifting No drifting i f 30 120 i i ERN 0 50 100 150 200 250 300 350 0 032 064 0 96 1 28 16 1 92 22 Position on circle A Delay us Figure 26 Received power on the circular track time evolution tutorial When drifting is enabled Fig 26 left blue curve the channel output after merging is time continuous The variations along the track come from the drifting K Factor and the drifting shadow fading When drifting is disabled these parameters are not updated and kept fixed at their initial value At the end of each segment both channels are cross faded I e the power of the output of the first segment ramps down and the power of the second segment ramps up Since drifting guarantees a time continuous evolution of the phase this ramping process is also time continuous and no artifacts are visible in the blue curve Without drifting however the phases are approximated based on their initial values the in
199. urve_radius The track is set up in a way that prevents driving in circles Input track_length the length in m Default length is 1000 m direction specifies the driving direction in rad of the first segment in mathematical sense 0 means east pi 2 means north The default value is random street_length_min the minimal street length in m The default is 50 m street_length_mu the median street length in m The default is 187 m This value was obtained from measurements in Berlin Germany street_length std the standard deviation of the street length in m The default is 83 m This value was obtained from measurements in Berlin Germany curve_radius the curve radius during a turn in m The default is 10 m turn_probability the probability of a turn at a crossing Possible values are in between 0 and 1 The default is 0 5 par generate _parameters overlap usage check_parfiles verbose Description Generates large scale parameters and stores them in par This function extracts the LSPs for the given scenario from the parameter_set class and stores them in track par Hence it automatically generates the LSPs and thus implements an easy to use interface for the parameter_set class Since the track class does not handle transmitter positions a default position of 0 0 25 is assumed Please refer to layout generate_parameters for a more detailed description Input overlap The length of the overlapping
200. use an exact geometric representation of the environment but distributes the positions of the scattering clusters the sources of indirect signals such as buildings or trees randomly A simplified overview of the model is depicted in Figure 2 For each path the model derives the angle of departure the angle between the transmitter and the scattering cluster the angle of arrival the angle between the receiver and the scattering cluster and the total path length which results in a delay 7 of the signal For the sake of simplicity only two paths are shown in the figure Scatterer NLOS pelos be Dinka eee a I LOS Path rf th dro Los A TT leng a Receiver LOS Transmitter d Power Impulse Response Signal delay Figure 1 Simplified overview of the modeling approach used in QuaDRiGa Terrestrial and Satellite scenarios can be modeled For Satellite to Earth communication the angle of departure is identical for all clusters The concept behind the model allows also the modeling of scenarios such as e Earth to satellite e Satellite systems with complementary ground components CGC Using several transmitters at dif ferent positions and simulating all propagation paths in one setup is supported The analysis of these scenarios was not in the scope of the MIMO over satellite MIMOSA project This feature is not tested and especially no parameter sets are available yet In the following the terms cluster scattering
201. ution because the phase changes are calculated deterministically based on the arrival angles This results in a realistic Doppler spectrum The temporal evolution of the channel is modeled by two effects e drifting and e birth and death of clusters Drifting see Section 3 4 occurs within a small area about 20 30 m diameter in which a specific cluster can be seen from the MT Within this area the cluster position is fixed Due to the mobility of the terminal the path length resulting in a path delay and arrival angels change slowly Longer time evolving channel sequences need to consider the birth and death of scattering clusters as well as transitions between different propagation environments We address this by splitting the MT trajectory into segments A segment can be seen as an interval in which the LSPs e g the delay and angular spread do not change considerably and where the channel keeps its wide sense stationary WSS properties Thus the length of a segment depends on the decorrelation distances of the LSPs We propose to limit the segment length to the average decorrelation distance Typical values are around 20 m for LOS and 45 m for NLOS propagation In the case where a state does not change over a long time adjacent segment must have the same state For example a 200 m NLOS segment should be split into at least 4 NLOS sub segments A set of clusters is generated independently for each segment However since the propagation
202. v1 2 32 458 2 SOFTWARE STRUCTURE map x_coords y_coords power_map scenario usage sample_distance x_min y_max x_max y_min tx_power rx_height Description Calculates a power map for the given layout This function calculates receive power values in W on a square lattice at a height of rx_height above the ground for the given layout This helps to predict the performance for a given setup Input scenario The scenario for which the map shall be created There are four options 1 A string describing the scenario A list of supported scenarios can be ob tained by calling parameter_set supported_scenarios 2 cell array of strings describing the scenario for each transmitter in the layout 3 A parameter_set object This method is useful if you need to edit the parameters first For example call p parameter_set UMal to load the parameters Then edit p scenpar or p plpar to adjust the settings 4 Aa array of parameter_set objects describing the scenario for each trans mitter in the layout usage A string specifying the detail level The following options are implemented e quick Uses the antenna patterns the LOS path and the path gain from the scenario e sf Uses the antenna patterns the LOS path the path gain from the scenario and a shadow fading map detailed Runs a full simulation for each pixel of the map very slow phase
203. w we apply the path gain PG the shadow fading SF and the KF Hata 27 presented a simple model for macro cellular settings where the PG scales with the logarithm of the distance d in units of meters between BS and terminal PGB A logio dim B 90 where A and B are scenario specific coefficients that are typically determined by measurements The path gain exponent A often varies between values of 20 and 40 depending on the propagation conditions the BS height and other factors Combining PG and SF results in the effective path gain PGI The values for the SF and the KF are obtained from the LSP map by an interpolation of the surrounding pixels at the position of the st snapshot The KF at the initial position is already included due to the scaling in 21 Thus we have to take this into account and scale the power accordingly PGi v p00 Pas ese i P GE E i 91 Ko K B Kg for l 1 fje PGE Vis tls m Ir t l s otherwise In the above equations K and gF are the interpolated values for the KF and the SF from the map Ko is the KF at the initial position PGE is the path gain without SF at the MT position 90 and P is the power of the LOS path 21 3 8 Transitions between Segments The calculations in Sections 3 2 to 3 7 were done independently for each segment of the MT trajectory Here we combine those segments into a long time evolving sequence of channel coefficients The idea comes fro
204. wo elements thus create a left hand circular polarized LHCP signal Two crossed dipoles For input port 1 the signal on the second element is shifted by 90 out of phase For input port 2 the the signal on the second element is shifted by 90 out of phase Port 1 thus transmits a LHCP signal and port 2 transmits a RHCP signal ula2 Unified linear arrays composed of 2 omni antennas vertical polarization with 10 cm element distance ula4 Unified linear arrays composed of 4 omni antennas vertical polarization with 10 cm element distance ula8 Unified linear arrays composed of 8 omni antennas vertical polarization with 10 cm element distance Input array_type One of the above array types element The element numbers for which this functions is applied If no element number is given the function creates a new array and delete the old elements in the array Ain a The parameter A for the parametric array type b The 3dB beam width in azimuth direction for the custom array type Bin a The parameter B for the parametric array type b The 3dB beam width in elevation direction for the custom array type Cin a The parameter C for the parametric array type b The isotropic gain linear scale at the back of the antenna for the custom array type Din The parameter D for the parametric array type Output par The parameters A B C and D for the parametric antenna type Copyright
205. y Each path gets assigned an azimuth angle in the horizontal plane and an elevation angle in the vertical plane Thus we have four values for the angular spread 1 Azimuth spread of Departure AsD 2 Azimuth spread of Arrival AsA 3 Elevation spread of Departure EsD 4 Elevation spread of Arrival EsA Each one of them is assumed to be log normal distributed Hence we need the parameters u o and A to define the distributions These spreads are translated into specific angles for each multipath cluster Additionally we assume that clusters are the source of several multipath components that are not resolvable in the delay domain Thus these sub paths do not have specific delays but they have different departure and arrival angles Thus we need an additional parameter cg for each of the four angles that scales the dimensions of the clusters in 3D space See Sec 3 3 for details Cross polarization Ratio XPR The XPR defines how the polarization changes for a multipath component I e the initial polarization of a path is defined by the transmit antenna However for the NLOS components the transmitted signal undergoes some diffraction reflection or scattering before reaching the receiver The XPR in dB is assumed to be normal distributed where u and o define the distribution We translate the XPR in a polarization rotation angle which turns the polarization direction A XPR value where a value of Inf means that the axis remains th
206. yi Figure 38 Visualization of the angular spread correction function C L K Left Surface plot of C L K for different values of L and K Right Scatter plot of the initial angular spread og vs the output of the model with correction Correction of the angular spread in the WINNER model The proposed correction function for the WINNER model see 3 pp 39 works as follows The individual angles 95 are calculated by a 2 74 Pi C y i 5 101 where C depends on the numbers of paths see Table 12 The KF is corrected by a polynomial of third grade With the constant coefficients in 101 and the factor of 2 in the square root of 95 the correction function C y L K of the WINNER model is CWINNER L K C 1 1035 0 028 K 0 002 K 0 0001 K 102 A comparison the both functions for different values of L and K is given in Table 13 In the second column the letter W indicates the value for the WINNER model and the letter Q indicates the value of the adopted function The polynomial has a value of 1 at KF values 11 65 3 11 and 28 54 At those points the WINNER correction function is independent of the KF The corresponding rows are highlighted in the table Generally both functions are similar They agree best as KF values around 14 3 and 12 but show differences at other values Copyright Fraunhofer Heinrich Hertz Institute 115 eMail quadriga hhi fraunhofer de QuaDRiGa v1 2 32 458 B D

Technical Documentation

Contents

Download Pdf Manuals

Related Search

Related Contents