Home
Ultimate Equalizer V9.0 Supplemental User Manual
Contents
1. When an attempt is made to load a driver file into the project file and the currently selected Sampling Rate or Buffer Size are different from the one stored in the driver s file you will be confronted with one or two messages and the file will not load Here are the error messages associated with possible mistakes UE Message x UE Message X A This File Has Different Buffer Size A This File Was Sampled at Different Frequency 26 Phase response used by DSP section is HBT derived A quick comparison see below between the bass resolution for all 4 buffers and 48kHz sampling frequency reveals that HBT process will clean up all unwanted environmental noise crept into the measurement results This is particularly evident in the frequency range below 10Hz where the measured SPL level at 1Hz is actually higher than the maximum SPL of the driver within it s operating frequency range Therefore SPL Phase response used by DSP section is HBT derived meaning that you must run linear frequency HBT before saving the driver file It is clearly observable that good results can be accomplished with 5 86Hz bass resolution for subwoofers that do not exhibit unduly rapid variations in the frequency response Also there is very little difference between drivers processed in logarithmic frequency scale and drivers processed with 1 47Hz bass resolution right down to 10Hz HBT Editor
2. 12 a7 52 57 62 68 73 73 a3 a9 34 39 104 103 115 120 125 130 135 w 146 151 15 6 161 167 72 177 182 188 133 138 23 Figure 8 Impulse response of a tweeter driver In the next step FFT is applied to ultimately obtain SPL and phase responses However the immediate result of the FFT is the driver s complex Transfer Function expressed as TF jw Re w j Im w The real and imaginary parts do not even resemble the final SPL and phase curves yet but are quite good candidates for smoothing Such operation is depicted on the figure below where all relevant variables are plotted in linear frequency scale The smoothing algorithm must take into account that at low frequencies data is very sparse but at high frequencies there is almost too much data So the frequency range over which the smoothing is calculated has to be progressively expanded as the frequency increases gl ee Ce ee 750 1500 2250 3000 3750 wa aa poog dk he a pooo ds en dee 12000 12750 i 20 EL bL Io teen Ei eno bai 20250 21000 21750 22500 23250 24000 24750 25500 26250 27000 27750 28500 29250 30000 30750 31 reen SPL before smoothing Red SPL after smoothing Linear Frequency Scale j YV an Dupi LAX ull AS f N WA JTT V i elil Figure 9 Complex Smoothing explained Red Re of TF after smoothing Green Im of TF after smoothing 9 The degree of smoothing is no longer expressed in dB oct
3. Also there is the issue of latency and therefore the size of UE Buffer 512 1024 2048 and 4096 and corresponding bass resolution If you are unable to make this determination the safest option is to perform full set of measurements Complete elimination of the re mapping of the frequency scales in the data and DSP processing results in some additional workload This could be summarized as follows l Measure drivers at all three sampling rates 44 1kHz 48kHz and 96kHz This is not a significant requirement because all that needs to be done is to set the sampling frequency to the next value and repeat the measurement Please SAVE each impulse response as you go and clearly mark what is the sampling frequency in this impulse response file As a safety feature the impulse response file actually saves the sampling frequency as well The IR files will be reused several times In the next step you would create driver s data file Once again the description here presents all options included First go to Preferences screen and set the size of the buffer choose from 512 1024 2048 or 4096 size This will determine the bass resolution and latency Also select the sampling rate which must be the same as the sampling rate of the impulse response you saved previously Go to MLS system and load driver s impulse response Now you should have the MLS system at the same frequency as the DSP section Next determine the windowing compl
4. Window HBT Editor Window 1 47Hz LOG Figure 39 Comparison between the bass resolution for all 4 buffers and 48kHz sampling frequency Real time SPL level adjustments F System Wide Preferences Audio Devices Preferred Input WASAPI Device out ot 8 Line 3 4 M Audio Delta 1010LT Line 5 6 M Audio Delta 1010LT Multichannel M Audio Delta 1010L Line 7 8 M Audio Delta 1010LT Record 01 02 Lynx AES16 SRC S PDIF M Audio Delta 1010L Graph Colour Scheme Red J Green Blue 4 Thickness Peeper Type E 2 2 pixels z 2 Dotted z M Dialogue Box Background I Filter SPL Curve I Driver SPL Curve M HBT SPL Curve M SPL Corrected Curve I Filter Phase Curve l Driver Phase Curve M HBT Phase Curve I Phase Corrected Curve I System SPL Corrected I System Phase Corrected M HBT Filter SPL Curve J Equalized Room SPL M Room EQ Curve SPL M Room EQ Phase M HBT Filter Phase Curve l Excess Group Delay M MLS In I MLS Ref EIRE FT Te TN EN LY a oe Le LLNS It tn e Preferred Output WASAPI Device 1 out of 9 S PDIF M Audio Delta 1010LT Line 1 2 M Audio Delta 1010LT Multichannel M Audio Delta 1010LT Line 5 6 M Audio Delta 1010LT Line 3 4 M Audio Delta 1010L J AUTOSTART I Player Mode REAR LOUDSPEAKERS LEFT RIGHT ARROW Output 3 ZZ Output 4 Output 5 i Output 6 i
5. at 96kHz and SAVE the impulse response into a file It cannot be stressed enough that the above approach is the best way to protect your valuable measurement efforts and collected information for all future combinations of sampling frequencies and buffer sizes used It is recommended to label your Project Files embedding the sampling frequency into the file name Fortunately the way the MLS system is operating it allows you to use impulse response collected at 96kHz for your 48kHz design and playback needs So at the end of the day one needs to measure only at two sample frequencies 96kHz and 44 1kHz Impulse responses measured at those two frequencies can now be processed with 4 different low frequency resolutions depending on the selected UE buffer size The room EQ example shown below illustrates how the low frequency resolution is affected by selecting different buffer sizes 48kHz sampling frequency 512 Buffer 11 72Hz bass resolution 1024 Buffer 5 86Hz bass resolution solution H Figure 2 Room EQ affected by low frequency resolution It is observable that 11 72Hz bass resolution is not sufficient to activate excess group delay detector necessary for room equalization to operate The 2 93Hz resolution is accurate enough to resolve where the non minimum phase regions are However you could use the 2 93Hz bass resolution to detect and exclude those regions from EQ but still run your system at lower ba
6. but it offered as Level 1 to Level 8 smoothing MLS Frequency Domain C Ultimate_Equalizer Directional Measurements Centre Horizontal_HT Li ear_Phase_1 D_10kHz j x SPL Phase Mike Preamp Post Processing Overlaped Curves 125 180 120 160 115 140 poceo 110 120 105 100 fs 100 80 SPL Phase Mike Preamp Post Processing Overlaped Curves 125 180 120 160 115 140 fo 110 120 105 100 100 80 95 60 90 40 O Se EEEE EEE 80 0 CLO BE 65 60 60 80 pes S56 LL Fics RRR 45 140 35 180 i ARETY PETTE TIETEET ETET ETE H H E Y f F f li H H SE es ee E ee Se H H H dB deg 10 20 30 40 60 80 100 200 300 400 600 1k 2k 3k 4k 6k 8k 10k 20k 30k 40k Hz Figure 11 Final SPL with Level 8 smoothing As observable on the figures above Complex Smoothing in linear frequency scale is very effective in obtaining smooth SPL and phase curves It is also very effective in removing rapid phase fluctuations 180deg resulting in well defined phase transitions greatly assisting in extracting the minimum phase phase response Combination of FFT windowing and Complex Smoothing provides excellent results for SPL and phase smoothing It is recommended to apply FFT windowing to remove most of the time of flight from the measurements before applying Complex Smoothing This can be easily done by correctly placing the FFT window 10 Modelling of Pre ringing in Impulse Response Linear Phase
7. frequency re mapping 3 Figure 1 demonstrates that the artefacts of frequency re mapping algorithm result in increased base level noise around impulse response This noise level is at 100dB so it is unlikely that it would deteriorate normal to loud listening experience to any degree I have been using Ultimate Equalizer software for four years now and the listening experience has been outstanding However if there was a way to eliminate this issue it would be worth to pursue And it turned out that there is a way to do it A complete elimination of frequency mapping and remapping of the data from the signal processing path provides the most consistent way of generating final channel impulse responses The new signal processing path starts with the measurement system 1 MLS measurement system operates in linear frequency scale as usual 2 Next the FFT also linear frequency scale provides SPL and Phase of the measured system 3 SPL Phase smoothing traditionally performed and the SPL Phase in logarithmic frequency scale as fractions of a dB now has to operate on the FFT data in linear frequency scale Therefore Complex Smoothing was introduced more on this later New HBT algorithm operates in linear frequency scale Loudspeaker equalization is performed in linear frequency scale Room correction is performed in linear frequency scale Filter generation and design is performed in linear frequency scale Finally partitione
8. that each driver file is of the same sampling frequency and buffer size 4 Calibration files for microphone and pre amplifier should match the sampling rate and buffer size of the driver file One option to deal with this issues is as follows 1 The main 5 2HT surround sound system operates at 48kHz and buffer size of 1024 bins latency of 147ms 2 WAVE Player operates at 44 1kHz commercial CDs and buffer size of 1024 bins 3 WAVE Player also operates at 96kHz Hi Res file playback and buffer size of 2048 bins Buffer size 1s twice as big to keep the bass resolution the same as for 48kHz As you can see UE buffer equals 1024bins or 2048bins Therefore I ended up with only three driver files 44 1kHz 1024 buffer 48kHz 1024 buffer and 96kHz 2048 buffer per driver With this combination I can create any playback system that provides the level of performance described above and in other UE Manuals Project Files are then saved to HD and switching between them switching complete program profiles is accomplished the usual way via numeric section of the keyboard Every time Sampling Rate or Buffer Size are changed in Preferences screen the previous frequency scale of the DSP engine becomes invalid and therefore the SPL and phase plots are cleared in System Design screen so you ll need to replot them The Player Mode automatically re calculates all filtering parameters in the new frequency scale before starting playback
9. you enter 40 samples for timing offset Pre Finging Inspection iW Show Temporal Mask M IR in Logarithmic Scale RfromMLS System IR IR2 Advanced h Convolve Driver Show Minimum Phase Show Fine Time Scale An Samples Impulse Response Window Ultimate Equalizer Frequency Domain 16 Outputs Te File Loaded OE To Plot Simulating midrange tweeter section 32 Midrange tweeter sections are simulated next in the CAD system design screen and after frequency responses are plotted you can go to Optimize Export Impulse Response menu option and select IR1 IR2 Advanced when you enter 40 samples for timing offset Fre Ringing Inspection If Show Temporal Mask I IR in Logarithmic Scale IR from MLS System IRI The Advanced b Convolve Driver Show Minimum Phase Show Fine Time Scale Ap Samples 10 390 10 70 20 60 30 50 40 40 50 30 60 20 70 10 80 0 30 10 100 20 110 30 120 40 130 50 140 60 150 70 160 80 170 90 k dB Time of flight distance difference between drivers is as follows woofer midrange 28 6cm 40 samples 48kHz midrange tweeter 28 6cm 40 samples 48kHz Interestingly the short analysis above indicates that midrange and tweeter will have less margin well almost none under the stringent 70dB pre masker Data Files Compatibility Due to introduction of the linear frequency scale in data files the files have become very large a
10. 444 4 4 4 444444 4 4 4 4 44 44 44 4 LEII YY tbetere tft tt Y tttti tht ao SSa eS pulse Response Window a sp i pulse Resp wi bP eet eee See ESSE e maam Ss Se Geese Td ape pa ae me he ESntt The next example shows equalization implementation of a large 18 subwoofer This is where the bass resolution is of importance 6 Example of Equalizing 18 Subwoofer The subwoofer was measured using the in built MLS system and the impulse response is shown below I MLS Impulse Response Ref 14 95 In 3 31 Bin 189 Scroll O0 2000 Figure 4 Subwoofer impulse response After FFT was applied we obtain SPL and phase responses shown on the figure below The figure also presents HBT curves derived using logarithmic scale HBT algorithm 3 HBT Editor wingow lolx HBT Editor Window A gp sssssag OASE AAE AOAOINA PIETOSA AASA AOAO IIIT SOOSIDA BOISI O O OS E O S O E S OE E E E E A S E E S A O E A S A E OER E S O ES O E E E A AA AAIE IOON Figure 5 Measured SPL and phase but the figure bottom also presents HBT curves derived using linear scale HBT algorithm and bass resolution of 5 86Hz It is observable that the 5 86Hz resolution is obviously quite coarse but seems quite sufficient to proceed with equalization as it approximates the subwoofer s SPL down to 11 7Hz quite well 7 HBT Edito
11. Also if we implement again a peaking filter with gain of 30dB and Q factor of 10 and hope that convolving it with the driver would eliminate pre ringing no the pre ringing will manifest itself again Figure 26 Pre and post ringing of a peaking filter convolved with a driver 18 Resampling to 44 1kHz MLS measurement system includes a simple re sampler to convert impulse response measured at 48kHz to 44 1kHz sampling rate Prior to using MLS system with re sampling to 44 1kHz you must select 44 1kKHz sampling rate in the Preferences screen Resampling to 44 1kHz From 48kHz IF lf Upsample Decimate Linear Interpolation There are two resampling options available 1 Up sample Decimate Here the 48000 sampling data is up sampled by 147 therefore in now runs at 7056000 sampling frequency and in the next step it is decimated by 160 therefore now the data is running at 44100 Hz 2 Linear Interpolation Here the 44100 data is extracted from 48000 data buffer by linear interpolation between two most suitable 48000 data points It is strongly recommended to try both variants to see which one works better for the given impulse response MLS Impulse Response Ref 14 28 In 3 32 Bin 188 Scroll 0 2000 Figure 27 Original Impulse Response in 48kHz MLS Frequency Domain C Ultimate_Equalizer Directional Measurements Centre Horizontal_HT Linear_Phase_100_10
12. Mode of operation is the primary mode of operation of the Ultimate Equalizer Since the impulse response of a linear phase system is symmetrical it is therefore recommended to inspect impulse response pre and post ringing levels The inspection is carried out using controls of the Optimize Export Impulse response dialogue box as shown below Optimize Export Impulse Response i x l Select Transfer Function to 4 IR Optimization 4 IR V ndows summary W Method 1 Outl Wide Window L F Mode Output Port 2 l Wlethod2 Cute Wide Window L P Wlode Output Port 3 Outs Wide Window L P Wode Output Port 4 ute Wide Window L P Wlode Output Port 3 Outs Wide Window L P Wlode ute Wide Window L P Wlode uth Wide Window L P Wlode Dut Wide Window L P Wlode Dutt Wide Window L P Wlode Outlo Wide Window L F Mode Duti l Wide Window L F Mode Outl2 Wide Window L F Mode Outls Wide Window L F Mode Dutl4 Wide Window L P Mode Outls Wide Window L F Mode Outl Wide Window L F Mode 2 oelect FET Window T tetm Blackman H atris Cosine Hide Fhase Below W Show Reference SPL Use Longer FFT Clear SPL Response Impulse Response Window b Select File Format al gt Raw Impulse Response UE format IR Mag Gain 6 miDSP ASCII Format CSV E E E IF Time Scale 2220 sare Impulse Response Done Pre Funging Inspection Show Temporal Mask Rin Logarithmic Scale IR fom MLS
13. System IRI 1TR2 Advanced b Corrobre Driver Show Minimum Phase Show Fine Time Scale T Samples Figure 12 Controls for estimating pre ringing Before we examine impulse response characteristics it is beneficial to highlight two important factors associated with hearing the sound pre masking and post masking Source http zone n1 com reference en X X help 373398B 01 svaconcepts svtimemask Time Temporal Masking Simultaneous masking describes the effect when the masked signal and the masking signal occur at the same time Human hearing is sensitive to the temporal structure of sound and masking also can occur between sounds that are not present simultaneously Pre masking is when the test tone occurs before the masking sound Post masking is when the test tone occurs after the masking sound The following figure shows the time regions of pre masking simultaneous masking and post masking in relation to the masking signal 11 60 5 Pre Simultaneous Post masking 50 0 50 100 150 9 50 100 150 200 Time After Masking Delay Time ty ms Onset At ms Level of Test Tone dB Figure 13 Slopes of temporal masking Post masking is a pronounced phenomenon that corresponds to decay in the effect of the masking signal Pre masking is a more subtle effect caused by the fact that hearing does not occur instantaneously because sounds require some time to sense As indicated in the figure above researchers typ
14. Ultimate Equalizer V9 0 Supplemental User Manual September 2015 New Features Implemented in Ultimate Equalizer 9 0 Digital 1 10 11 12 13 All internal calculations are now performed in linear frequency scale This is a very large change and includes HBT running in linear frequency mode All displayed plots are presented in Log Log scale as before Program is streamlined and is lighter Several unused options were removed MLS system operates at maximum frequency resolution of 262144 MLS sequence length and has three sampling frequencies only 44 1kHz 48kHz 96kHz Phase response used by all DSP sections of the program is HBT derived only Impulse response design and optimization functionality now includes display of IR in logarithmic scale with pre and post masker limits and allows for final convolution of filter s impulse response with driver itself to obtain full channel impulse response Data and project files are very large as they are required to store frequency phase data in linear frequency scale up to 131072 data points Guiding Filter method for assisting in minimum phase extraction has been facilitated by introducing a copying function for SPL phase curves from MLS system and selection of filters Currently the 8 partition version of UE is available The 16 partition version running internally in linear frequency scale is being considered in the future Complex Smoothing
15. be your 4 drivers in the rear speakers in the Preferences screen see example below Reset Nominated Drivers To 0dB Gain REAR LOUDSPEAKERS LEFT RIGHT ARROW Output 3 9 Output 4 Output 5 1 Output 6 Nominated Drivers 2dB DOWN Nominated Drivers 2dB UP Adjusting Volume of 4 Nominated Drivers Subwoofers SUBYYOOFERS INSERT DELETE Sub 1 Sub Sub 3 CABS is Sub 4 CABS lt Subwoofer 2dB Subwoofer 2dE 29 Adjusting the BBM levels This adjustment control the level of low frequency filtered from all left channels and all right channels signal mixed into left and right subwoofers BBM 2dB BBM 2dB Page Up 2dB Page Down 2dB Adjusting Volume of 4 Nominated Drivers Front Tweeters This particular adjustment option came to being after auditioning many movies and concerts There seems to be a subtle difference well sometimes quite pronounced in the amount of treble recorded during the final studio mix Some video material has excessive brilliance and sibilance and these can be controlled to large extent by controlling the level of tweeters in front speakers Brilliance The 6kHz to 16kHz range controls the brilliance and clarity of sounds Too much emphasis in this range can produce sibilance on the vocals Sibilant Essy Exaggerated s and sh sounds in singing caused by a rise in the response around 6 to 10 kHz Often heard on radio In order to b
16. d convolution is obviously performed in linear frequency scale CY as So this is how the internal mathematics is now working From the user s perspective all plots and graphs are generated by remapping the data linearly processed in each and every stage onto the standard Log Log scale Therefore there is basically very little visual difference between the older versions of the program and the latest linear scale version All the above changes have far reaching consequences If the frequency re mapping is to be completely removed from the data and signal processing path then we must maintain the same sampling rate of the data from the measurement right down to the final partitioned convolution stage This approach also requires that measurements are made at the same sampling frequency as the partitioned convolution will be run So for instance if you intend to use the playback function for WAV files at 44 1kHz then you need to provide SPL Phase data measured by MLS system at 44 1kHz to start with Or if your goal is to run 96kHz Hi Res WAV files then your measured SPL Phase needs to be sampled at 96kHz The recommended measurement approach for collecting the impulse responses is as follows 1 Measure the driver using sampling frequency of 44 1kHz and SAVE the impulse response into a file 2 Then simply repeat the measurement instantly at 48kHz and SAVE the impulse response into a file 3 Finally repeat the measurement instantly
17. e Mode And here is the frequency and phase responses of the subwoofer 120 160 115 140 110 120 105 100 100 S 95 E 30 4 55 Z 50 O 75 20 r 40 65 60 60 80 55 100 50 120 45 140 40 160 dB deg 10 20 230 40 60 0 100 200 300 400 B00 Tk Figure 22 Frequency and phase responses of the equalized subwoofer The above level of performance was accomplished with the low frequency resolution of 5 86Hz Buffer 1024 and 48kHz sampling When the raw SPL and phase do not contain rapid peaks and valleys it is clearly possible to equalize the low frequency driver to excellent standard 16 Case 2 Equalized Tweeter driver Impulse responses the green curve is the same tweeter measured and processed in linear frequency scale Red with convolved loudspeaker All curves fit very well under the strict 70dB pre masker Impulse Response Window Figure 23 Impulse response of the complete tweeter channel red Impulse Response SPL Window Figure 25 Tweeter channel SPL Phase response flat phase response and equalized driver 17 Are there any filters that pre ring unacceptably Yes peaking filters such as the one depicted below Here we have a Q Parametric filter with gain of 30dB and Q factor of 10 Figure 26 Pre and post ringing of a peaking filter This filter rings so much that even the minimum phase version blue curve will exceed our 70dB post masker limit
18. e able to adjust those level during playback you need to nominate the outputs to which the tweeters are connected In the example below these are Out 1 Out 3 and Out 5 Output 7 is a dummy one and nothing is connected to it The levels are controlled via Home and End keys on the keyboard in 1dB Steps TWEETERS HOME END Tweeter 1 Tweeter 2 Tweeter 3 5 e Tweeter 4 iii alien i i Home Tweeters 1dB End Tweeters 1dB 30 Ultimate Equalizer Frequency Domain 16 Outputs E xj E Ultimate Equalizer Frequency Domain 16 Outputs Figure 41 Tweeter at 10dB and tweeter at 10dB level E Ultimate Equalizer Frequency Domain 16 Outputs Front Left Figure 42 Commonly used voicing with 6dB difference between 50Hz and 10kHz Summation of two impulse responses with time delay Finally a simulation of a typical 3 way system where the crossover frequencies are selected as 500Hz and 5000Hz The filter if 24dB oct LR design The sampling rate is 48kHz and impulse responses are shown in decibels 31 3 Ultimate Equalizer Frequency Domain 16 Outputs Chee eee be we eee wee doe BO ee bs fi File Loaded E To Plot Simulating woofer midrange section Woofer midrange sections are simulated first in the CAD system design screen and after frequency responses are plotted you can go to Optimize Export Impulse Response menu option and select IR1 IR2 Advanced when
19. ex smoothing and extra dBs to be added and run the FFT The resulting SPL and phase should be smooth and phase should not contain any spurious transitions Next go to File Editor screen and set the HBT parameters You will notice that there are two versions of HBT linear frequency HBT and logarithmic frequency HBT The reason for this is that logarithmic frequency HBT 3100 bins calculates much faster than linear frequency HBT up to 65536 bins so you can use logarithmic frequency HBT to play with the HBT settings but once you are happy with those you MUST use linear frequency HBT Only the SPL Phase data calculated via linear frequency HBT 1s saved to driver s data file IMPORTANT Save the driver s file clearly marking the buffer size and sampling frequency in the driver file name The above description applies to measurements not requiring multiple impulse responses to be saved Woofers for example will require port and cone close mike measurements and then Im distance SPL measurements ALL of those must be conducted at three sampling rates Then you need to combine relevant impulse responses and create all required driver files 25 Sampling UE Buffer 44 1kHz 512 1024 2048 4096 48kHz 512 1024 2048 4096 96kHz 512 1024 2048 4096 A total of 12 driver files per driver This is when planning ahead pays off 3 When the Project File is created you ll need to load driver files that are mutually compatible so
20. ically can measure pre masking for only about 20 ms Post masking is the more dominant temporal effect and can be measured for 100 ms following the cessation of the masking sound Both the threshold in quiet and the masked threshold depend on the duration of the test tone Researchers must know these dependencies when investigating pre and post masking because they use short duration test signals to 99 perform these measurements Modelling Regime Firstly it is observable that slopes of temporal masking in Figure 13 are plotted using lin log scale that 1s time scale is linear in milliseconds and level of test tone is in logarithmic scale decibels Since impulse responses which we are going to examine are typically calculated and plotted in linear scale it is necessary to use the same vertical scale units and type dBs Y decibels of a variable X are expressed as Figure 14 Y A logl0 X B where A B are suitably chosen screen display constants Now in order for the slope of temporal masking display similar characteristics as on Figure 1 the X variable Figure 15 X exp T C D Where T time from T T1 to T T2 and C and D are suitably chosen constants Expressions on Figure 14 and Figure 15 can be used to create temporal masks if we were to display impulse responses in logarithmic scale 12 Calibration of the Simulation Process Mathematical formulas describing pre and post masking slopes on F
21. igure 13 were not available Therefore visual inspection of Figure 13 resulted in adjusting coefficients A B C and D to obtain plots of both maskers similar to the ones on Figure 13 green curve on Figure 16 right Please note that pre and post masking 1s shown from OdB to 50dB on the vertical scale on Figure 13 For calibrating the peak level of the impulse response a 1000Hz 12dB oct Butterworth filter was used and the corresponding impulse response is shown below 9 Ultimate Equalizer Frequency Domain 16 Outputs E 3 Wind wO 0 x j i Original Pre Masker Impulse Response O i x arpa ge i New Pre Masker and sol i Post Masker levels 40 0 shifted down to 70dB ont 20 30 40 60 80100 200 300400 600 1k 2k 3k 4k 6k 8 amp k 10k 20k 30k 40kHz Figure 16 Calibrating Simulation Process for IR 70 180 dB deg 10 Figure 16 depicts calibration process for this simulation Impulse Response IR green curve is plotted with 10dB div horizontal scale resolution IR of such system has a peak at OdB level see green peak at Figure 16 right The floor of IR is clipped at 170dB level observe no noise at all IR is now calculated and shown in logarithmic scale Also since the IR can be negative as it wiggles along the time scale the negative values are shown as absolute values of the IR in C language Y 20log10 fabs IR This is why the IR looks differently from what you would typically see
22. iles W Uselt Import Cal Mike File Uselt Import Cal Preamp File show Cal Files Reset Cal Files Cutofffiequency z000 Hz Geinfinn Ofano Create Cal Files Export Cal Files Bees iwe BSpPorse Cancel Dione Figure 35 Calibration File simulator used as filter generator LP 12dByYoct LP 18dByYoct Butteraorth LP 24dByYoct Parametric Now position mouse pointer exactly on the SPL curve in the Mike Preamp TAB and press left mouse button The SPL curve will be highlighted in pink colour and the SPL and phase curves will be copied into Overlapped Curves TAB Next position mouse pointer again exactly on the SPL curve in the SPL Phase TAB and press left mouse button The SPL curve will be highlighted in pink colour and the SPL and phase curves will be copied into Overlapped Curves TAB cy Domain Copied Curve 18 Sefler s Lett SPL Phase Mite Preamp Pest Processing Overleped Curves a Dn 4 co a w w mW o w Figure 36 SPL of the driver and filter SPL highlighted in pink Examining the phase response of the filter and measured driver and particularly the 360deg phase transition it 1s evident that in order to get the measured phase response phase response 360deg transition to overlap filter s phase transition a small 12usec delay needs to be added to the phase response of the measured SPL curve 23 MLS Frequency Domain C Ultimate_Equalizer Directio
23. in MLS systems In order to increase confidence level in the simulation the levels the pre and post masker shown on Figure 13 were dropepd by 20dB The new much more stringent masker levels are now plotted in pink on Figure 16 right and are extended down to 70dB range From now onwards the gain in the system must be kept constant for all impulse responses We will now proceed to inspect several impulse responses for potential audibility of pre ringing Case 1 Equalized subwoofer Since the pre masker and post masker are time limited phenomenon it may be prudent to examine the slowest and heaviest driver in the system the subwoofer The driver in this example is really big 18 McCauley subwoofer mounted in a 300 litre venter enclosure The equalization and filtering transfer function is depicted on the figure below 13 Impulse Response SPL Window Figure 17 Correction filter for subwoofer It is noticeable that the phase response of the correction filter is inverse This is because we are developing a linear phase subwoofer Also the correction characteristics are evident below 340H z as the filter transfer function 24dB oct LR filter is also correcting driver s SPL phase below 340Hz Impulse response of the correction filter is shown below on Figure 18 Please note that the filter alone has developed significant pre ringing and clearly fails the 70dB pre masker level It is also observable that the i
24. kHz_CalFile trg ol x SPL Phase Mike Preamp Post Processing Overlaped Curves 125 180 120 160 115 140 110 120 105 100 100 80 Figure 29 SPL and phase calculated from original 48kHz impulse response 19 E MLS Impulse Response Ref 14 28 In 3 32 Bin 177 Scroll 0 2000 MLS Frequency Domain C Ultimate_Equalizer Directional Measurements Centre Horizontal_HT Linear_Phase_100_10kHz CalFile trg SPL Phase Mike Preamp Post Processing Overlaped Curves 125 180 115 140 110 120 105 100 100 20 2 MLS Frequency Domain C Ultimate_Equalizer Directional Measurements Centre Horizontal_HT Linear_ Phase 100 10kHz CalFile trg SPL Phase Mike Preamp Post Processing Overlaped Curves 125 180 120 160 115 140 110 120 105 100 100 80 95 30 85 80 5 70 5 e o 5 5o 5o 40 35 dB Figure 32 Resampling Decimation red vs Linear Interpolation beige 20 Copying function for SPL phase curves from MLS system and filters The copying function primarily is designed to assist in developing minimum phase phase response of the measured driver in MLS system Dome tweeter example In this example we will examine minimum phase response of a popular Hi Fi dome tweeter driver The dual channel MLS system is actually designed to provide minimum phase response of the measured driver within the error of one sample time Here is how it works When the loo
25. ll delay or manipulate HBT slopes to get the measured and HBT calculated phase into alignment It is important to determine the start of the impulse response not the peak as various drivers have different rising slopes of the impulse response therefore the peak will be located at various distances from the start of the impulse So first we need to find the peak of the impulse response MLS Impulse Response Ref 7 84 In 1886 90 Bin 339 Scroll O0 2000 the peak is In 1886 90 located at Bin 339 21 Next we move the cursor to the left of the impulse response one sample time and each time we monitor the In vale MLS Impulse Response Ref 7 04 In 987 70 Bin 338 Scroll 0 2000 MLS Impulse Response Ref 14 45 In 147 21 Bin 337 Scroll 0 2000 MLS Impulse Response Ref 9 20 In 4 94 Bin 336 Scroll 0 2000 the In 4 94 and is very close to zero therefore we determine that the start of the impulse response is Bin 336 Finally we need to move the start of the FFT window 9 sample times for this system it 1s 9 sample times to the left from the start of the impulse response 336 9 327 Now the start of our FFT window is located at Bin 327 See figure below IMLS Impulse Response Ref 6 64 In 0 39 Bin 327 Scroll 0 2000 Figure 33 Start of the FFT window 9 bins to the left Next we need to obtain the SPL and phase of the driver using FFT And here is the result A MLS Freque
26. mpulse response is very asymmetrical not what you would expect from a linear phase system Impulse Response Window JecevescccecsocsoscocsocepfbcccoscocsonsoccoeseossesesGpeccosccscossecsossoccossonscosscounccococoecossoseedpecsocooscossoccoscosooccoesonceccsesoossccsesonscossapocssecsossocsocsossoccscsotcoococcosccessosscscee dbo cocccccoscoccoeMppocccccecoosscesessnscossccsossccsdpooscccsescoccossenccocecoosscossosseccessossescossesedpeocoocosssoseccoesooscecossonscosscsoossocsossoccoesdpocccoesnncoesoccoescecooe Mpeccoscescesseccoscoscossies ae ieee Si pocccccccceccnccnccccccccoscccsossoecossocsossodpoecoecooccecccocccoscocoocsoscosccesoccoces PacocccccececcccccsccscocccccnscccccccccconsccsoccooseodpecoasonccesoeccocccccecoceccccoccoseoseeMMpecoccoccoo ges 128 0ms Figure 18 Impulse response of the correction filter green in decibels 10dB div This is a very common misconception when discussing linear phase filters Simply because this is NOT the subwoofer channel impulse response it s missing the actual subwoofer transfer function 14 The figures presented above require the following check boxes to be ON Convolre Driver mhow Whirnmum Phase Now we can convolve the correcting filter with the actual subwoofer and then examine the resulting subwoofer channel impulse response this is what you will be listening to Impulse Response Window 7 Pre ringing of orrecting filter alone 128 0ms Figure 19 Pre ringi
27. nal Measurements Centre Horizontal_HT Linear_Phase_100_10kHz CalFile trg SPL Phase Mike Preamp Post Processing Overlaped Curves 125 180 120 160 115 140 110 120 105 100 100 95 60 g0 40 85 20 80 0 75 20 70 40 65 60 60 80 55 100 50 120 45 140 40 160 35 180 Poets i i i a e pl i i H i E p i pi y i H dB deg 1 20 30 40 60 80 100 200 300 400 600 1k 2k 3k 4k 6k 8k 10k 20k 30k 40k Hz Figure 37 Phase response 360deg transitions now overlap accurately We can now run HBT with the high side asymptotic slope of 51dB oct to see how the whole picture works out Hilbert Bode Transform High Pass Tail Low Pass Tail Stop 100 00 Start 3700 0 Hz PN a Slope 18 0 Slope 51 0 dB oct Linear Frequency HBT V Show Reference Phase E PERONA T OO EE A i TE em ee PO Use As Mike Cal File Use As Preamp Cal File vol AEE EO IRF niniscnscacosnsene ETETA ma os EERE EEEIEE EIE Figure 38 Final linear frequency HBT calculated We observe a perfect alignment of measured and HBT derived phase responses assuming 51dB oct asymptotic slope of the guiding filter HBT SPL blue curve HBT Phase red curve 24 General workflow and data flow It will pay dividends if you plan ahead and determine upfront what is the UE going to be used for The purpose of this is to determine what sampling rates will be necessary in the future usage of the program
28. ncy Domain C Ultimate_Equalizer Directional Measurements Centre Horizontal_HT Linear_Phase_100_10kHz_CalFile trg T i 0 xi SPL Phase Mike Preamp Post Processing Overlaped Curves 125 180 120 160 115 140 Fo 110 120 105 100 100 80 95 60 30 40 85 20 80 0 SO Gl aie et 70 40 cla cs Saas 60 80 55 100 50 120 fester 45 140 40 160 35 180 3 3 3 3 i H i H EO l pa i i H H es ee ee i i H dB deg 10 20 30 40 60 80 100 200 300 400 600 1k 2k 3k 4k 6k 8k 10k 20k 30k 40k Hz Figure 34 Resulting SPL and phase responses To increase the level of confidence on the phase response it is desirable to overlap band pass filter SPL phase responses comparable with the loudspeaker amplitude response This will provide us with filter s phase response which one would use as an additional guidance for the locations of the 360deg transitions of the filter and the measured phase they should be very close 22 Since this is a tweeter example and we are only interested in finding the high frequency tail then a simple low pass filter located at 27kHz with 48dB oct slope can be used It is observable that the slope of the filter is slightly slower than the measured response sO we assume 51dB oct as the asymptotic slope of the measured driver SPL Also the 360deg transitions of both filter and the measured SPL are very close Calibration File Data X High Pass Section HF 6dByoct Calibration F
29. ng of the complete subwoofer channel red As shown on the Figure 19 above the linear phase impulse response red is now almost exactly symmetrical and the pre ringing has dropped by 10 20dB This is very significant improvement and now the total channel impulse response fits comfortably under the very strict 70dB pre masker And here is the complete transfer function of the subwoofer channel As you would expect the SPL is equalized right up to 340Hz and the phase response is now a flat line Impulse Response SPL Window z Figure 20 Complete transfer function of the subwoofer channel 15 Next we can perform some measurements on the newly equalized subwoofer 2ms wide pulses separated by 350ms space were used as the source signal On the 2ms pulse the minimum phase subwoofer version delivered a more of a thump instead of a pop or a click This is perhaps not surprising as the post ringing of the pulse extended tol30ms and far exceeded the 30ms memory effect of the auditory system Here the driver filter and vented enclosure added it s own combined signature It is also observable that the minimum phase version of the subwoofer has converted the clearly asymmetrical pulse into a much more symmetrical bi polar pulse with post ringing This is clearly visible on the screen shots below es o e tees eee sivl fy Norma i A Auto Lvl Auto Normal Figure 21 2ms Impulse in Linear Phase Mode and Minimum Phas
30. of SPL Phase implemented in MLS system see inside for details Impulse response resampling from 48kHz to 44 1kHz MLS measurements conducted with 96kHz sampling can be used with 48kHz DSP sampling without re measuring see inside for details Dynamic SPL adjustments for 12 nominated outputs This User Manual is intended to provide information about new features in version 9 and should be read in conjunction with the full User Manual for version 5 6 7 and 8 and available from Bodzio Software Pty Ltd website Linear Frequency Scale for All Internal MLS DSP Processes Loudspeaker design is typically performed in logarithmic frequency scale the Log Log scale and is very well suited to the way we perceive sounds and music Starting with the measurement process as soon as the impulse response is collected and FFT applied the resulting SPL and phase are calculated mapped onto logarithmic frequency scale From then onwards driver equalization all crossover design functions impedance compensation and so on are performed and visualized in the Log Log frequency scale Even the HBT algorithm was operating in logarithmic frequency scale and was fitting very well into the design process Obviously after all DSP functions like driver equalization and room corrections were derived there was a need to re map the results back into the linear frequency scale so that partitioned convolution could operate properly at evenly spaced f
31. p test is performed you will notice that you will get flat phase response of the signal channel when you place the start of the FFT window at 10 samples before the peak of the impulse response why This is because the reference channel is also automatically windowed with the fixed start of the FFT window also at 10 samples in front of the IR start The loop test simply measured the true minimum phase phase response of the sound card However each PC MLS system must be examined individually for the Reference Impulse response first Reference Impulse Response MLS Impulse Response Ref 41463 95 In 95 36 Bin 60 Scroll 0 2000 Peak of 41463 at bin 60 MLS Impulse Response Ref 6584 96 In 1 25 Bin 59 Scroll 0 2000 Peak 1 of 3584 at bin 59 MLS Impulse Response Ref 3762 23 In 1 57 Bin 58 Scroll 0 7000 Peak 2 of 3762 at bin 58 the IR has gone large negative now For this system bin 59 is the start of the impulse response Therefore the start of the FFT window for the Reference impulse response is 10 sample times from the peak or 9 sample times from the start of the impulse response Now we can apply the same technique to the loudspeaker measurement and place the start of the FFT window 9 samples ahead of the start of the IR and we ll obtain minimum phase phase response of the loudspeaker straight away with one sample time error To eliminate this small uncertainty error we have to add subtract sma
32. r Dialogue E3 Hilbert Bode Transform Dist 1 00000 m Delay 2 90698 ms High Pass Tail Low Pass Tail i g 2 Stop 10 00 Start 2650 0 Hz I Include Distance In PHASE Plots I Include Distance In AMPLITUDE Plots Slope 24 0 Slope 48 0 dBfoct Pin jo Watt Efficiency 90 0 dB Linear Frequency HBT Refresh Help IV Show Reference Phase i i Run 16383 of 16384 Use As Mike Cal File Show HBT as g Use As Preamp Cal File Log Frequency HBT Figure 6 HBT parameters for the subwoofer Buffer 1024 5 Ultimate Equalizer Frequency Domain 16 Outputs Fiter 1 SP HET 1 SPL ee Final 1 SPL i Fro ater Buffer 4096 Ultimate Equalizer Frequency Domain 16 Outputs iter 2 sp bier gi TE SAL Figure 7 SPL of all relevant EQ curves for two buffer sizes 1024 and 4096 8 Complex Smoothing Traditional smoothing applied to SPL and Phase curves involves fractional octave smoothing like 1 12oct or 1 60ct and so on The smoothing is performed in logarithmic frequency scale and is easily applied to SPL curve and significantly more difficult to phase response curve However now the SPL Phase data never makes it to the logarithmic scale for processing therefore the smoothing needs to be applied differently The example below shows an impulse response of a tweeter driver 100 120 1 140 160 180 0 5 T 05 10 16 2 26 31 36
33. requency samples This scheme worked well undisturbed for a long time Major benefits were 1 Small data files Typically 2000 3000 data points on the 20Hz 20kHz logarithmic frequency scale provide sufficient accuracy and resolution for all design work 2 Independent sampling frequencies for the measurement system and DSP playback This is because the re mapping of the frequency scales provided convenient buffer between the MLS and DSP This is an important factor providing a great deal of convenience Driver measurement files containing SPL Phase data could be shared with no restrictions as the frequency re mapping functionality always provided reasonable recovery of the frequency and phase responses needed for the CAD systems However recently Bodzio Software has been probing into the impulse response performance issues including plotting the amplitude of the final impulse response of a channel in decibels This allowed the visualization of the noise level performance of the impulse response Several impulse responses of simple filters generated initially in logarithmic frequency scale and then mapped onto linear frequency scale have been collected and then compared those with filters generated directly in linear frequency scale The base level noise around the peak of the impulse response was always lower in linear frequency scale by as much as 50dB Noise of HBT derived EQ filters measured loudspeakers was also lower in linear frequenc
34. s the must hold 65536 data points for 8 partition version and anticipated 131072 data points for 16 partition version of the software Driver files are now 4 1Mb and Project files are almost 600Mb in size The new data files are not compatible with older versions of the program 33
35. ss resolution of 5 86Hz The dependency of the bass resolution on the buffer size has been highlighted before in previous UE manuals UE system latency is the other affecting factor if you intend to use the software as a DSP processor for watching visual media 5 In general there are following latency options 36ms MP UE Buffer 512 Delta Buffer 64 Lynx Buffer 128 UE 8P 75ms LP UE Buffer 512 Delta Buffer 64 Lynx Buffer 128 UE 8P 65ms MP UE Buffer 1024 Delta Buffer 128 Lynx Buffer 128 UE 8P 147ms LP UE Buffer 1024 Delta Buffer 128 Lynx Buffer 128 VE 8P 130ms MP UE Buffer 2048 Delta Buffer 128 Lynx Buffer 256 UE 8P 295ms LP UE Buffer 2048 Delta Buffer 128 Lynx Buffer 256 UE 8P When equalizing 5 1HT or 7 1HT system one needs to observe maximum tolerable audio video latency which is 185ms It is therefore observable that a number of system configurations listed above will lead to a significantly lower latency than the 185ms threshold The following two examples clarify further the need to select latency bass balance suitable for your needs The first example illustrates the impulse response and SPL performance of a 50Hz 24dB oct filter implemented with difference buffer sizes 50Hz 24dB oct Butterworth Filter 512 Buffer 1024 Buffer SS SSS 4 of 4 444 4 4 0 94 n np LEILI ot SSF of oP bbe 4 444
36. xX b i m SUBWOOFERS INSERT DELETE b b b b b b b b b b b b b b b b b b d 28 Sampling Hz Buffer Size 44100 512 samples __ 48000 1024 samples 96000 2048 samples 4096 samples V Invert MLS Data Channel M Invert MLS Reference Channel Delay MLS Output By samples M Inputs 1 2 P 8 Outputs J Standard I Inputs 1 12 M 16 Outputs V BBM I Inputs 1 8 8 SPDIF J Summed L R Inputs 1 6 J Render Timing Output Queue Capture Timing COMASS TWEETERS HOME END Sub 1 ia Tweeter 1 Sub 2 Tweeter 2 Sub 3 CABS is Tweeter 3 Sub 4 CABS Tweeter 4 Figure 40 Preferences screen with outputs selectable for adjustments Important Note When you first open the Preferences screen the residual data displayed in these four fields may contain random value Please assign your output ports to these data fields immediately Adjusting Master Volume Master Volume control is performed remotely by pressing ARROW UP DOWN keyboard keys The SPL steps in 3dB increments decrements Master Volume 3dB UP Master Volume 3dB DOWN Adjusting Volume of 4 Nominated Drivers Rear Speakers This is a very handy and often used feature It appears that rear loudspeakers are not mastered in the recording studios at the same level Thus some movies concerts are lacking the sound coming from the rear speakers and others are correctly edited To remedy this problem UE allows you to nominate 4 output ports these will
37. y scale by 10 15dB As a result of these findings the Ultimate Equalizer program was to re written and a linear frequency scale is now being used for all MLS and DSP functions from start to finish while re mapping into logarithmic scale only for the purpose of providing standard display of all curves in Log Log scale This is completely the opposite to what the previous approach was Obviously some of the benefits of the original scheme are lost data files are very large and since the frequency re mapping is no longer there the MLS system must run at the same sampling frequency as the DSP system Next the HBT algorithm had to be re written to operate in linear frequency scale so now there are two versions of HBT We can now compare the impulse responses of filters with and without frequency re mapping 2 1000Hz 24dB oct Butterworth filter Logarithmic to Linear Mapping All processing in linear frequency scale vs na is tid Po PI P2 P3 P4 P5 PS P7 100Hz 24dB oct Butterworth filter Logarithmic to Linear Mapping All processing in linear frequency scale PA P5 tin eee tl E E ee Be ee tess 4 5000Hz 48dB oct Butterworth filter Logarithmic to Linear Mapping All processing in linear frequency scale Ims 21ims 2 7m Sns 5000Hz 48d B oct Butterworth filter Logarithmic to Linear Mapping All processing in linear frequency scale Figure 1 Impulse responses of filters with left and without right
Download Pdf Manuals
Related Search
Related Contents
Descargar - ergosistem Mise en page 1 - Cezam Auvergne 取扱説明書 趣 DONGLE USB W-MBUS AMR Range 取扱説明書・お料理ノート1 Altec Lansing IML237EAMCLR docking speaker Polar Cardio GX: Temas de Ayuda Manual del Usuario IBM eServer 326m Method and apparatus to retrieve information in a network tiptel - Sipgate Copyright © All rights reserved.
Failed to retrieve file