Home

PDF - University of Twente Student Theses

1. PROJECTION i i UNLESS STATED METHOD _ OTHERWISE TOLERANCES 0 5 MM MATERIAL vA oo i SURFACE FINISH UNIVERSITEIT TWENTE x DRAWN CHECKED DATE SCALE TITLE NO NAME DRAWING NO FILE PART NAME Bar riaht 2 1 14 1 20 13 5 14 1 UNLESS STATED PROJECTION i METHOD OTHERWISE TOLERANCES 0 5 MM A MATERIAL perm SURFACE FINISH UNIVERSITEIT TWENTE i la DRAWN DATE 2 CHECKED SCALE 2 TITLE NO NAME DRAWING NO FILE PART NAME Bar riaht cap TOLERANCES 0 5 MM oD UNIVERSITEIT TWENTE DATE SCALE 2 TITLE NO NAME DRAWING NO FILE PART NAME rina bearinas 60 DRAWN ES ae PROJECTION METHOD TOLERANCES NS 5MM CHECKED MATERIAL SURFACE FINISH of UNIVERSITEIT TWENTE TITLE DRAWING NO FILE PART NAME DATE 2 SCALE 1 NO NAME baseplate 20 60 2 CD ST 7 ev i am i N 1 1 1 5 oO oi 0 RN RES 1 Rr 7 5 _7 5 30 PROJECTION i i UNLESS STATED METHOD _ OTHERWISE TOLERANCES 0 5 MM MATERIAL vr oo SURFACE FINISH UNIVERSITEIT TWENTE x DRAWN CHECKED
2. DATE SCALE TITLE NO NAME DRAWING NO FILE PART NAME sensorbasedual 2 1 Y S9 20 OL 09 2 1 DATE SCALE DRAWN CHECKED OTHERWISE UNLESS STATED TOLERANCES 0 5 MM PROJECTION METHOD Base vsa NO NAME FILE PART NAME TITLE DRAWING NO MATERIAL SURFACE FINISH UNIVERSITEIT TWENTE 20 7 5 15 15 60 UNLESS STATED PROJECTION i METHOD OTHERWISE TOLERANCES 0 5 MM MATERIAL X Q pee SURFACE FINISH UNIVERSITEIT TWENTE DRAWN DATE 2 CHECKED SCALE 1 TITLE NO NAME DRAWING NO FILE PART NAME motorblock 2 5 1 _ 60 3 L 2 25 _ M n al 8 xp A 5 A O 7 5 1 9 30 UNLESS STATED PROJECTION METHOD OTHERWISE TOLERANCES 0 5 MM MATERIAL vA o SURFACE FINISH UNIVERSITEIT TWENTE x DRAWN DATE 2 CHECKED SCALE 2 TITLE NO NAME DRAWING NO FILE PART NAME plate Appendix E E Matlab files E 1 File used to extract single data values M file to extract data from final measurements HOW TO USE THIS FILE First place all matlab
3. 2 where vio dowo is the linear velocity of a point on the ring gear with diameter dio and rotating with angular velocity and vg Bug is the linear velocity of a point on the sun gear rotating with angular velocity wg The angular velocity w7 of the frame can thus be expressed as a function of the angular velocity w o of the ring gear and of the angular velocity wg of the sun gear as dio W10 d de Wg 2 dg do 2 dg dg The planet carrier 6 and the sun gear 8 are rigidly connected However wg refers to the velocity of the planet carrier with respect to the the ring gear 7 i e the frame 0 of Figure 3 while wy and wg refer to an absolute frame of reference Using Equation 3 wg can be expressed as Ug 3 We Wg d d 2 ds d 10 1 ws The motion of the planet carrier carrying the pivot gear and the motion of the rotating frame of the VSA are thus linearly coupled i e ir bun C Level 3 The Actuation Stage where dio 2 dg dg dg 2 dg dg In the third level the internal actuators are coupled to the ring gear 10 and the sun gear 8 of the differential mechanism by the parts 11 to 13 in Figure 2 The first motor 12 engages the ring gear 10 while the second motor 13 engages a second smaller sun gear 11 which is fixed to the sun gear 8 via the shaft of the planet carrier 6 It is shown in Figure 5 The ring gear 10 and th
4. Uploading the program using Arduino is easily done by selecting the correct serial port in the tools menu and then clicking the upload button While programming you should see both the RX and TX led flash When not check either if the program works by opening the serial port and check the connections of the LEDS If it doesn t program something is either wrong with the serial to USB convertor or with programming the boot loader onto the ATmega328 If all works like expected the programming is done When opening the serial monitor the measurements of the selected channels should appear Expected results The expected precision of the 8AMP board depends on the resistors strain gauges sensors used at the inputs of the board It can be up to 13 bits precise depending on the thermal noise of the input The data collecting frequency can go up to more than 10000 Hz and mainly depends on the amount of inputs used and the communication speed BAUD RATE Usually the BAUD RATE is the bottle neck for the data collecting frequency ZZ CAN CHIFH 1 Appendix 1 Placement of all components lt co Q 3 SM SS BY 922772 VEE my p prz HERE N EN p or nr nr e t NEA 3 KIA 7 AAN R Lt Lt AD 7 MI ean aoa 4 2 IP Va eee VAN 224 972072272201 62424624 2 7 a r7 acaban a 072372722 V
5. A2 1 R5 1 C6 2x C4 e RESET A3 s aoc B1 1xCFB VZ gt ZN gt VZ B2 2x C2 1x12 2x C3 1x C6 B3 C1 2x LED C2 2x LED 1x12 1 C3 1x C6 1x reset D1 1x C3 1x C4 1x L2 D2 1 R1 1x LED Figure 13 Front view of the PCB displaying the positio 1x C4 components 1x C6 1x Riep power D3 1x R1 CFB is an optional capacitor that can be used to stabilize the voltage of the half bridge located on the PCB This is only necessary when the power source is rather noisy The orientation of the leds is important The correct orientation is shown on the PCB as well as in figure 13 Due to the design of the wiring this Ohm resistor doesn t necessarily have to be soldered After soldering these components the backside can be soldered On this side which is shown in figure 14 the bigger version is also located in the Appendix all small components will be soldered first to These are shown in table 5 Table 5 Sector Components A1 2 1x C4 A3 1x RA 1x C7 B1 1x C4 1x C5 B2 1x L2 1x C1 D 1x OSC B3 1x C5 1x C1 C1 d C2 1x C4 1x C6 C3 centre one is RXD resist Bottom one is TXD resist Top one is CLOCK resist 1 2 3 D2 D3 Figure 14 Back view of the PCB displaying the position of compon At this point it is useful to know whether u want to use the CAN i
6. Steady ring dataring 1 1 Steady sun datasun 1 1 finalRingAverageData number 1 numel dataring 2 dataring 2 finalSunAverageData number 1 numel datasun 2 datasun 2 This final part stores all data gathered in one array This array increases with each run of the program The data stored is as following 1 Torque while rising position 2 Velocity in rising direction can differ for the sun when both are in opposite direction 3 Torque while decreasing position 4 Velocity in falling direction can differ for the sun when both are in opposite direction 5 Steady state value 0 Then each run number is increased and as result the table will increase by one row finalSunData number 1 sungear risingAvg 1 finalSunData number 2 sungear risingAvg 2 finalSunData number 3 sungear_fallingAvg 1 finalSunData number 4 sungear_fallingAvg 2 92 finalSunData number 5 finalSunData number 1 finalSunData number 3 finalRingData number 1 finalRingData number 2 finalRingData number 3 finalRingData number 4 finalRingData number 5 finalRingData number 1 finalRingData number 3 if sortBy 1 steady sun sungear risingAvg 1 finalSunData number 5 if ringgear risingAvg 2 finalSunData number 3 finalSunData number 4 finalSunData number 1 finalSunData number
7. determining the dissi pation of the mVSA UT by determining the friction coefficients is not achieved An indication can be given for this coefficients but the measurement setup has to be improved to verify the correctness of these achieved coefficients It was attempted to use as less budget as possible and a measurement setup was build The accuracy is not good enough for the measuring dissipation in the mVSA UT but the frictional torque to be measured was at the low end of what was expected for the velocities used for measuring There is a big possibility that rebuilding some parts of the setup and building a better calibration setup will improve the accuracy of the torque sensors Testing this will take time but will not be very expensive Therefore it can be concluded that the side goal of building a low budget setup has been achieved The recommendations for eventual continuation of work is to improve the setup Some improvements were suggested in this report like printing a new base for the sensor creating a better calibration setup and using only pulleys and timing belts instead of gears Another improvement could be to use motors with smaller gearboxes These do provide less torque but from the measurements it can be seen that the 200m N m of torque from these motors and gearboxes is much more than necessary Finally since an indication of expected torque is now available if necessary the cantilevers of the sensors could be red
8. finalSunData number 5 ringgear risingAvg 1 ringgear risingAvg 2 ringgear fallingAvg 1 ringgear fallingAvg 2 93 finalRingData number 5 steady ring finalRingData number 3 ringgear risingAvg 1 finalRingData number 5 finalRingData number 1 ringgear_fallingAvg 1 finalRingData number 5 end end if invert 1 invert 2 4 tempi finalSunData number 1 4 temp2 finalSunData number 2 finalSunData number 1 finalSunData number 3 4 finalSunData number 2 finalSunData number 4 finalSunData number 3 tempi 4 finalSunData number 4 temp2 4 tempi finalRingData number 1 temp2 finalRingData number 2 vA finalRingData number 1 finalRingData number 3 finalRingData number 2 finalRingData number 4 finalRingData number 3 tempi 4 finalRingData number 4 temp2 end dataring 2 finalRingAverageData number 1 numel dataring 2 datasun 2 finalSunAverageData number 1 numel datasun 2 Aclearvars except finalRingAverageData finalRingData finalSunAverageData finals 4 EOF E 2 File used to extract the final matrices from all measurement data M file to take the data from the final table and convert it into the 4 matrices wanted if exist sortBy var else sortBy 0 end close all 94 4 Seperate the final table into the wanted torques and velocities if sortBy 1 finalRin
9. from which the best one for this application will be chosen The correct dimensions for the sensor will then be determined using the data sheet of the semiconductor strain gauges and FEM analysis This will ensure that the sensor is able to measure the expected range of strains using the maximum linear region of the strain gauges 4 1 1 Options research There are several different structures suitable for torque sensors Some very basic designs are the solid and hollow cylinders like shown in figure 4 2 One side of the sensor is connected to the motor and the other side to the mVSA UT Figure 4 2 Basic designs for torque sensors Solid cylinder is shown left hollow cylinder is shown right Then there are the more complex shapes in the shape of hub sprockets and hollow cruciform which are shown in figure 4 3 The hub sprocket has an inner cylinder and an outer cylinder One of them is connected to the motor and the other to the mVSA UT The hollow cruciform is connected likewise as the solid and hollow cylinder Figure 4 3 More advanced designs for torque sensors Left the hub sprocket is shown right the hollow cruciform One advantage of the basic designs is that the sensors are easily produced but for the range of torques in this project the cylindrical shape would have to be very small and the hollow shape very thin to assure full use of the sensitiv ity of the strain gauges making it very fragile to non torsional torques
10. is pulled high by default and can be pulled down using a jumper A led will blink j Figure 6 Part of the complete schematic when a reset is performed The MCU is programmed using a In Serial programmer For this a special connector is necessary The connections for this connector are shown in figure 8 The clock which is also used by the AD converter is also connected to a led showing whether the AD converter is converting or not and whether the ISP is programming or not The RESET MISO MOSI of Microcontroller the programmer connector are connected to the MCU The names in figure 8 are connected to the corresponding names in figures 6 and 7 So MOSI in figure 6 is connected with MOSI in figure 7 and MOSI in figure 8 LED8 amp 03 So MOSI in one figure does not connect to MISO in another figure Figure 7 Part of the complete schematic Finally the output of the 8AMP can either be the serial communication coming directly from the MCU or it can be CAN communication coming from an on board serial to CAN chip The schematic of this part is shown in figure 9 Since the serial or CAN communication is an option the choice can be made using O O resistors If there is no need for the CAN transceiver there is an extra O O resistor to shut down the power of the CAN transceiver This prevents extra unnecessary power consumption and noise For the rest leds are also connected to b
11. of 5 volts and assuming a precision of 13 bits is 0 625 mV which is considered accurate enough The micro controller chosen to operate the ADC and the multiplexer is the AT MEGA328PA This micro controller has been chosen mainly due to its compat ibility with the Arduino software With the Arduino bootloader being available for this type of micro controller the coding can be done in the easier to use Arduino environment and re uploading code can be done through serial com munication instead of using an In Serial Programmer An external clock source of 16 MHz is used to achieve double the processing power compared to the internal clock of 8 MHz of the ATMEGA328PA The ATmega328P uses a serial interface to communicate with either the user computer or another microprocessor To also have the option to use CAN communication the SN65HVD1050D from Texas Instruments has been selected as Serial to CAN transceiver This is an extra option of the board which will not be further discussed anymore Next to these components there will be some LEDs indicating whether the power is on a reset is initiated and if the communication is active 3 2 3 Design The complete schematic of the 8AMP is well explained in the manual in ap pendix B and thus will not be explained again here 1LSB stands for least significant bit In this case only the least significant bit is considered to be noise This is normal since the last bit always rounds of the f
12. used in for instance robot joints To achieve good sensitivity the stiffness of these shapes has to be low and non torsional torques have to be small If used in for instance a robot joint the Y sensitivity can be set lower making it stiff enough for a weight carrying joint Figure 4 4 Illustration of hub sprocket torque sensor design with two can tilevers and a strain gauge full bridge Comparing all advantages to the dis advantages the hub sprocket design clearly is the best design for this project Its disadvantage of the dras tic change in size is easy to overcome and the filtering of non torsial forces outweighs the more difficult production process 18 4 1 2 Sensor design using FEM analysis The sensor designed using FEM analysis is shown in figure 4 5 Its design has been extended with screw holes to connect to the other compositions of the setup As can be seen in the side view the inner circle is a bit 1 0mm more thin than the outer circle This allows the part connecting to the outer circle to be less difficult to produce The thickness of the cantilevers has been determined using FEM analysis with Ansys 12 0 Figure 4 5 The final design of the torque sensor Left the dimetric view is shown right the side view To determine this thickness it is important to know the maximum applied torques Unfortunately this is unknown for the mVSA The assumption is that it is at most 100m Nm Furthermo
13. 0 3856 48 6863 0 9030 0 3771 0 6602 0 3767 48 7165 1 3201 0 7404 0 7548 0 7404 48 8788 0 4309 0 7487 0 4398 0 7293 49 0899 1 1781 1 1020 1 0573 1 1034 48 6211 0 2857 1 1074 0 4324 1 1091 48 3084 0 7824 1 4913 1 0330 1 4937 48 5125 0 1178 1 4667 0 5066 1 4676 48 3572 101
14. 2 finalSunData number 5 finalSunData number 3 finalSunData number 1 finalRingData number 3 finalRingData number 4 finalRingData number 1 finalRingData number 2 finalRingData number 5 finalRingData number 3 finalRingData number 1 end end if sortBy 2 if sungear risingAvg 2 finalSunData number 3 finalSunData number 4 finalSunData number 1 finalSunData number 2 finalSunData number 5 finalSunData number 3 finalSunData number 1 finalRingData number 3 finalRingData number 4 finalRingData number 1 finalRingData number 2 sungear_fallingAvg 1 finalSunData number 5 ringgear risingAvg 1 ringgear risingAvg 2 ringgear_fallingAvg 1 ringgear_fallingAvg 2 steady_ring ringgear_risingAvg 1 finalRingData number 5 ringgear fallingAvg 1 finalRingData number 5 0 sungear_risingAvg 1 sungear_risingAvg 2 sungear_fallingAvg 1 sungear_fallingAvg 2 steady_sun sungear_risingAvg 1 finalSunData number 5 sungear_fallingAvg 1 finalSunData number 5 ringgear risingAvg 1 ringgear risingAvg 2 ringgear fallingAvg 1 ringgear fallingAvg 2 Steady ring ringgear risingAvg 1 finalRingData number 5 ringgear fallingAvg 1 finalRingData number 5 0 sungear_risingAvg 1 sungear_risingAvg 2 sungear_fallingAvg 1 sungear_fallingAvg 2 steady_sun sungear_risingAvg 1 finalSunData number 5 sungear_fallingAvg 1
15. For easy reading these two values are split into distinct parameters ungear_risingAvg calculateAverage sun sunpos sun risingStart sun risingEnd ungear_fallingAvg calculateAverage sun sunpos sun fallingStart sun fallingEnc inggear_risingAvg calculateAverage ring ringpos ring risingStart ring risingk inggear_fallingAvg calculateAverage ring ringpos ring fallingStart ring falli Torque sun rise sungear_risingAvg 1 Velocity sun rise sungear risingAvg 2 Torque sun fall sungear_fallingAvg 1 Velocity sun fall sungear_fallingAvg 2 Torque ring rise ringgear risingAvg 1 Velocity ring rise ringgear_risingAvg 2 Torque ring fall ringgear fallingAvg 1 Velocity ring fall ringgear fallingAvg 2 91 E 4 Function steadyValue Torquevalues startpointsRising startpointsFalling 4 determines the steadystate value of the torquesensors with velocity 0 4 of both inputs 4 The function returns the steady state value on position 1 1 in the 4 return array The function returns all the steady state data points 4 collected in position 2 This can be used to verify that the steady 4 state points are all valid 4 The data points are stored in an array which increases with the variable number which increases in every run dataring steadyValue ring ring risingStart ring fallingStart datasun steadyValue sun sun risingStart sun fallingStart
16. PORTD B01110000 Channel 4 111 break 112 113 case 5 114 PORTD B10010000 Channel 5 115 break 116 117 case 6 118 PORTD B10110000 Channel 6 119 break 120 121 case 7 122 PORTD B11010000 Channel 7 123 break 124 125 case 8 126 PORTD B11110000 Channel 8 127 break 128 129 130 Serial print the data Easier to display on screen The serial write is 131 however faster 132 shiftData data Ox1FFF 133 Serial print shiftData DEC 134 Serial print t 135 136 137 138 Every loop trough all channels a newline is printed 139 Serial print Vn 140 141 142 143 End loop a Using the Arduino environment after programming and locking the boot loader section using the ISP the program can be programmed into the MCU Of course the computer has to be connected to the PCB first This can be done connecting an RS232 serial to USB converter This is not included in the BOM in this manual How to connect this converter to the PCB is described inside the datasheet of the converter Again watch carefully not to connect the RX and TX outputs to the wrong pins of the converter When not sure use 1 kQ resistors in between If wrongly connected the MCU won t break When you are not able to program the MCU using Arduino after uploading the boot loader correctly the RX and TX are wrongly connected
17. at point given by T Ius Ro Fs FY i 2 P If no friction is present between the pivot pin and the k the force acting on the pivot pin is always perpendicul the lever arm From a balance of the forces along the le and from the balance of the torques around the pivot pin reaction force on the lever in P is given by Fr 4 F p The perpendicular force on the output FY p can calculated with the angle 6 05 0 between the I and the crank The resulting output torque acting arounc output axis of rotation is given by d d Hin cos sing F gt p and therefore the output stiffness is 004 B Level 2 The Differential Mechanism The second level of the mVSA UT differentially cou the two degrees of freedom of the variable stiffness me nism namely the neutral output position 0 and the posi of the planet carrier 0g By using a differential drive internal actuators can contribute to the output power It second set of planetary gears made up by the parts amp 10 of Figure 2 and illustrated in Figure 4 The sun gear 8 is fixed to the shaft of the planet ca 6 The three planet gears 9 connect the sun 8 with rotating frame 1 velocity of the frame wy 1S given Dy w7 tdo 2 where vg is the linear velocity of the shaft of the planet gears dg is the diamater of the sun gear and dy the diameter of the planets Note that v o 98
18. be AVRISP MKII Choose the device to be the ATMEGA328P Press apply and the press Read to read the device ID If this gives an error first consult the error codes below Table 6 The error codes of the LED of the AVRISP LED Color Description Red ldle Notarget power Green ldle With target power Red blinking Short circuit on target Upgrade mode If the programmer does not give the green light indicating it is connected properly there is something wrong with the connections assuming the board is designed as described in the previous chapters and the on board connections aren t loose The connectors of the programmer input can either be loose or the connections of the socket of the 6 pin connector of the AVRISP could be reflected Also a possibility is that the MISO and MOSI connections are switched If all this is not the case check the power supply This can be read using the Read target voltage button in the AVR programmer in AVR studio If the board has been programmed before it could be the case that accidently some fuses were wrongly changed The fuse settings are described below since here it is assumed the microcontroller is not programmed yet Fuse settings Assuming all connections are correct and the programmer is working now the fuses have to be programmed in order to use the external clock instead of the internal clock The fuses can be programmed by going to the AVR programming in AVR studio Then sele
19. by the elongation of the springs f output deflection 01 and pivot position x Consequently way the springs are attached should not only depend constructional considerations but also on the resulting ou stiffness profiles The output torques and stiffnesses have been analysec several different spring configurations shown in Figur using both linear and torsional springs Configuration represents the design that was chosen in the mVSA where two linear springs are attached at right angles to end of the lever In configuration 6 b the springs are attached to the end of the lever but their equilibrium p lies in the center of the ring gear In configuration Mr d e Fig 6 Different spring configurations 6 a 6 b and 6 c show linear springs with different attachement points on the lever and frame 6 d and 6 e show torsional springs the springs are connected at right angles to the center of the lever Configurations 6 d and 6 e use torsional springs between the lever and output and between the lever and frame respectively Figure 7 shows the output stiffness in the equilibrium position i e 0 0 as a function of the pivot point position of the lever where x l d7 2 with respect to Equation 1 The stiffness calculations for configurations 6 a 6 b and 6 c follow Section III A where the torque on the output shaft is obtained by an equilibrium of forces and moments of the lever The lever is constraine
20. functions in the same folder and make this the working folder Load each measurement one at a time After loading run this m file you should see two matrices appearing in the command window After running load a new measurement Again run the file Now you should see the matrices of last time with one extra row Etc etc etc When all measurement values wanted are taken open file finalDataExtract m and run this file Do not clear the workspace during this process You can use clc and close all if exist number var else number 0 end if exist sortBy var else sortBy 0 end 88 Number increases to create a data array containing data of all measurements close all number number 1 First import the data from the measurements Sun and ring are the torque values sunpos and ringpos are the position values sun sensor2data signals values 3 r ing sensoridata signals values 3 sunpos sensor2data signals values 4 r ingpos sensoridata signals values 4 Plot options figure plot sun r hold on 1 ring figure 1 sunpos r hold on plot ringpos k Determine the start and endpoints to take the measurements The motors cannot run forever forward so the measurements are cycles of the motors moving forward waiting moving backward and
21. programming guide for the 8AMP using the Arduino Bootloader 3 2 1 Requirements To create a good design the requirements have to be known The requirements that are needed for the torque sensor setup are 1 At least two inputs able to measure differential voltages or at least four inputs to measure non differential voltages 2 A data collecting frequency of at least 100 Hz for each channel and the data has to be converted to digital signals 3 An amplifier and low pass filter are required The amplifier gain has to be adaptable depending on the range of different input voltages 4 The board itself has to provide power to the strain gauges The first requirement is defined by the chosen approach 2 3 The second is due to the fact that low frequency signals below 20 Hz should be clearly visible in the measurement results The third requirement is made because the expected strains to be measured and thus the expected voltage range is not well known If it is the case that for example only sth of the total voltage range of the analog to digital converter ADC is used which will decrease the accuracy of the ADC by at least 2 bits this should be easily changed The last requirement is necessary to ensure the voltages applied to the strain gauges are matched The electronics created in this project to measure the strain gauge bridges will also be used in another project the manipulator of the Airobots group 9 10 This manip
22. replaced by other types are listed in table 1 below Table 1 Components that cannot be replaced by other types Part of the BOM Notation Description Component name Package Datasheet Possible manufacturers Quantity one board Optional ATMEGA328P Microcontroller ATMEGA328P AU 32TQFP Link ATMEL 1 unit 8bit MCU INA118 High speed INA118UB SOIC8 NARROW Link TEXAS INSTRUMENTS 1 Amplifier ADS7279 14 bit ADC ADS7279IPW 16TSSOP Link TEXAS INSTRUMENTS 1 DG408DY 81 DG408DY SOIC16 NARROW Link INTERSIL 1 multiplexer MAXIM VISHAY SILICONIX CAN CHIP CAN SN65HVD1050D SOIC8 NARROW Link TEXAS INSTRUMENTS 0 1 transceiver None Molex MOLEX 53048 04 4 pins connector Link MOLEX 1 connector 90 degrees angle None Molex MOLEX 53047 04 4 pins connector Link MOLEX 1 connector straight 16 MHz OSC Oscillator Crystal resonator 5 0 x 3 2 mm Link TXC 1 9 The crystal resonator can be changed as long as the footprint matches When alternating the frequency it is at own risk This could give certain compatibility errors for instance when using the arduino bootloader as described in chapter Programming When changing the resonator keep in mind that the C1 values need to match the specifications of the resonator The second group components that can be freely chosen as long as the package is the same is listed below in table 2 Table 2 Components that can be replaced by other
23. studio Select the AVR Programming option in sub menu tools Then choose the correct settings and read the device ID If the ID can be read everything is still correct At this point the HEX file of the boot loader can be uploaded to the MCU In the tab Memories search for the boot loader file in your directories Select the correct boot loader and program it to the flash memory If this doesn t work something is wrong with the fuse settings or the lock bits The lock bits should not be set yet but if they are change them to No lock for all three options This is OxFF in hexadecimal code After uploading the boot loader change the lock bits to the following settings v LB This is to lock the complete memory This should be set to NO LOCK v BLBO This locks the memory used for applications This should be set to NO LOCK Y BLB1 This locks the memory used for the boot loader This should be set to LPM SPM disabled Main program After doing this the MCU is ready to be programmed using the Arduino environment The code to run the 8 channel multiplexer board is displayed below It uses one of the AD conversion methods described in the datasheet of the ADS7279 the Read While Sampling method Selecting channels is done using the truth table given in the datasheet of the DG408DY The code M lt 7 1 Code for the 8 channel amplifier 2 3 Initializes the data integers used furthe
24. the 8AMP has to be redesigned try to keep the following rules 1 Separate analog and digital signals If possible use the bottom two layers for e g digital signals and the top two layers for analog signals Separate analog and digital power and ground wiring at the input of the PCB Make sure that the wiring on different layers is oriented in perpendicular directions as much as possible Try to make signal wires as straight as possible and as short as possible The more curves and the longer the wire the more EMF will afflict the signal on this wire Don t use 90 degree angles when wiring Instead try to use round angles or two 45 degree angles Use ground planes in the following order Analog ground Analog voltage Digital ground Digital voltage This will filter EMF radiation Use and place ferrite beads and filtering capacitors close to power inputs of components This will filter noise coming from components like the MCU and multiplexer Don t make wires too small but don t make them too thick either Materials The materials needed to solder the board are indicated below The notations in the left columns are the notations used on the board The components are divided in two separate groups components that cannot be replaced by other types due to the footprint and location of certain pins and components that can be replaced by other types with the same package The first group of components those which cannot be
25. the HET RC DS 001MG and 002MG the only difference between those two being internal transmission These servo motors have been tomized for continuous rotation and provided with di angular position reading The current motors are sti enough to provide the nominal output torque of 1 Nn the actuator the speed of the mVSA UT however can be increased by using faster motors A DC motor control board based on a ATMEGA micro controller and a Polulu DC driver perform joint position control of the motors at 1kHz The user is interf with the low level motor control by means of a stan computer with a rate of 10 ms On the high level the defines the position references to the pivot pin position the undeflected output position that are then converted motor reference positions which are sent to the low motor controller In addition to the sensors for the motor position magnetic sensors are placed in the housing for initiali the positions of the rotating frame and the planet car This is necessary to identify the output and pivot posit because of the ability of continuous rotation One sens placed in the housing and picks up the position of a ma up the position OT a magnet placed on the end of the shaft of the planet carrier that protrudes from the bottom of the housing Since the actuator is compliant the position of the output shaft is not only determined by the motor positions but can be deflected by applying a torque To measure the outpu
26. 0 0075 0 022 Figure 4 7 Equivalent strain diagram of the sensor design when 1 Nm is applied to the outer ring while the inner ring is fixed In figure 4 6 the equivalent von Mises Stress is shown when the maximum torque of 0 5 N m is applied The maximum value is 177 M Pa which will not cause de formation The equivalent Elastic Strain is shown in figure 4 7 The maximum 2376p 27 is within the maximum operating strain 20 In figure 4 8 the equivalent elastic strain in the sensor is shown when a torque of 100m Nm is applied As can be seen the maximum strain is 4754277 Al though this strain indicates the full linear region of the strain gauges is not used making the cantilevers more thin will decrease the safety factor Furthermore the rounded of dimensions simplify the design 0 00010565 5 2848e 5 5 0335e 8 0 0 015 0 03 m 0 0075 0 022 Figure 4 8 Equivalent strain diagram of the sensor design when 100 mNm is applied to the outer ring while the inner ring is fixed FEM analysis has also been used to verify the maximum strain is approximately linear to the applied torque This is the case but this analysis is not added in this report The values of the strains above agree to this To utilize the maximum strain values the strain gauges should be glued to the positions with maximum strain These positions are determined from the figures given in this section 4 2 Setup design To use the sensor for mea
27. 1 sun fallingEnd sun falling 2 y Zn ann 4 Function ringVSsun positionArrayRing positionArraySun risingStartpoints 4 checks whether the ringgear is rotating in the same or in the opposite 4 direction of the sungear In case the direction is opposite this function interchanges the sungears rising and falling start and end 4 points to assure that the values of velocity of the ring and sun are correct 90 invert ringVSsun ringpos sunpos ring risingStart if invert 1 temp sun fallingStart temp2 sun fallingEnd sun fallingStart sun risingStart sun fallingEnd sun risingEnd sun risingStart temp sun risingEnd temp2 end S S r r Functionality to calculate the average torques in the steady velocity modes The function calculateAverage torqueArray startPointsArray endPointsArray sums up all the values of torque between startPoints and endPoints and takes the average of this As second value it returns the velocity belonging to these these values The velocity is also an average between all start and end points and is taken by taking the position at the end point and subtract the position at the start point from this and then dividing this between the number of ticks inbetween the start and end points The total is then multiplied by a constant that relates time to ticks
28. 1 3997 1 4559 2 1582 1 2464 48 3991 1 7288 1 6477 2 1767 1 6480 48 3369 0 5985 0 0 6395 0 48 3725 1 9167 0 4275 1 2521 0 4215 48 4609 2 0454 0 8320 1 4482 0 8309 48 3345 2 0783 1 4751 1 8266 1 2372 48 2182 2 4555 1 6381 1 9092 1 6393 48 3421 1 2422 0 4203 1 0934 0 4200 48 0046 1 6043 1 2348 1 8124 1 2356 48 1647 1 8315 1 6435 2 0397 1 6313 48 2278 0 7547 0 0 7690 0 48 3238 2 0705 0 4179 1 2953 0 4181 48 4587 2 3617 0 8245 1 1582 0 8253 48 7161 2 1685 1 6271 2 2048 1 6279 48 0805 2 2790 1 2288 1 7177 1 2300 48 2929 0 4810 0 4161 0 4248 0 4162 47 6766 1 6318 0 9820 1 1682 0 8232 47 8868 1 4429 1 6223 2 2224 1 6231 48 4365 0 8484 0 0 8541 0 48 2276 1 6822 0 4146 1 6612 0 4148 48 0907 2 4512 0 9385 1 2664 0 8214 48 6726 1 9587 1 2195 1 9999 1 2236 48 1023 2 1301 1 6124 2 4641 1 6164 47 7909 1 1146 0 9739 1 4245 0 8177 48 2460 1 6244 1 6095 1 9353 1 6096 48 2453 Table F 2 Final measurement data for the ring gear input T2 02 T2 OMEGQ2 rever 2 avg 0 0947 0 0 2640 0 48 6514 0 3849 0 0 3743 0 48 6739 0 3087 0 0 5297 0 48 5442 0 9335 0 3800 0 6021 0 3541 48 7651 1 2246 0 3738 0 5555 0 3736 48 9135 0 2049 0 4604 0 2083 0 3833 48 4549 0 2096 0 3785 0 0112 0 3791 48 5726 0 3136 0 4390 0 0259 0 3750 48 5176 0 4107 0 3721 0 0213 0 3720 48 5124 0 3337 0 8092 0 9050 0 8090 48 3894 1 0998 0 7716 0 4714 0 7611 48 8800 0 7882 0 7508 0 9116 0 7508 48 4921 1 0714 0
29. 8885 0 8919 0 7449 48 6807 1 3201 0 7404 0 7548 0 7404 48 8788 0 2780 0 7588 0 0482 0 7587 48 7947 0 1375 0 7429 0 2566 0 7433 48 8567 0 0433 0 7416 0 3077 0 7369 48 8433 0 5286 1 2036 0 9123 1 2043 48 4638 0 6577 1 1327 1 2539 1 1329 48 2786 0 7066 1 1163 1 1407 1 1184 48 3721 1 1781 1 1020 1 0573 1 1034 48 6211 0 8913 1 1103 1 3691 1 1112 48 2881 0 3600 1 1274 1 0483 1 1284 48 2933 0 2756 1 3285 0 4890 1 1152 48 2882 0 2408 1 0992 0 2285 1 0983 48 4396 0 7188 1 5960 1 0716 1 5958 48 4843 0 7253 1 4979 1 3396 1 4994 48 2830 0 8856 1 6929 1 5371 1 4840 48 2362 0 8604 1 4691 1 6466 1 4739 48 1795 1 4668 1 4563 1 3631 1 4595 48 5511 0 1076 1 7593 0 4563 1 4767 48 4876 0 0603 1 4538 0 3250 1 4540 48 5112 100 Verification data Table F 3 Final verification data for the sun gear input T2 022 T2 OMEGQ2 rever 2 avg 2 0555 0 8941 1 4223 0 8951 48 4248 0 3070 0 0 5888 0 48 2874 1 7448 0 4258 1 2854 0 4268 48 3000 2 4247 1 2519 1 3079 1 2516 48 6360 2 4555 1 6381 1 9092 1 6393 48 3421 1 2791 0 8293 1 6639 0 8076 48 2726 2 1685 1 6271 2 2048 1 6279 48 0805 1 1572 1 2268 2 1239 1 2282 48 5698 0 7546 0 4129 1 1528 0 4135 47 1877 1 1611 1 2178 1 9284 1 2187 48 4501 Table F 4 Final verification data for the ring gear input T2 W2 T2 OMEGQ2 rever 2 avg 0 3278 0 0 2366 0 48 7596 0 5372 0 4097 0 4621 0 4100 48 7202 0 6855 0 3846 0 5152
30. Also non torsional components will not be permitted since they can influence mea surements The hub sprocket design is more advanced requiring more advanced production methods It can be build with any amount of cantilevers but since the torques are assumed to be very small two cantilevers will be sufficient This sensor in combination with a strain gauge full bridge has the advantage that bending moments normal forces and perpendicular forces can be mostly filtered from the strain gauge readings The sensor with a strain gauge full bridge is shown in figure 4 4 When a normal force in the z direction or a non torsial bending moment is applied all strain gauges will elongate an equal amount thus there is no difference in the output When a perpendicular force is applied pointing down in the y direction the top two strain gauges will shorten while the two bottom gauges elongate In the ideal case nothing will happen since both half bridges change the same way In the less ideal case if both half bridges are not completely balanced these changes can be mostly removed by calibration The disadvantage of this shape the drastic change in size between the outer ring and the inner ring is no problem in this setup The hollow cruciform shape again has a certain amount of cantilevers that are bent due to applied torques How ever this shape requires even more advance production methods This shape of torque sensors are more often
31. ISP MKII serial programmer link A driver for the AVRISP has to be installed This driver comes with the programmer Also AVR studio is used in this manual to program the microcontroller AVR studio is available here First the wires to connect the ISP programmer to the PCB have to be made The pin out of the AVRISP is given in figure 15 This is also the pin out of the socket The MISO MISO 1 2 VCC and MOSI connections have to be connected to the MOSI and MISO of the programmer connector of the PCB In SCK 3 4 MOSI this case in contradiction to the connections on board the MISO of the socket connects to the MOSI of the PCB RESET 5 6 GND and vice versa To ensure no components get broken it is advised to place 1 resistors in between the socket and the input connector of the PCB In that way When Figure 15 Pin out of the AVRISP MKII wrongly connecting MISO and MOSI the components won t break Connect the SCK to the SCLK of the programmer connector and the reset to the reset The ground has to be connected to the input ground which can be done directly at the source The same for VCC The AVRISP is powered through the USB The VCC is only to verify power To check whether the programmer is correctly connected the LED on the AVRISP can be consulted The meaning of the color codes are displayed in table 6 To check if the programmer works start up AVR studio Then go to menu Tools and click AVR programming Choose the Tool to
32. SMD 0402 5mCd Orange SMD 0603 VISHAY TLMO1000 GSO8 2mA 1 8V Link 16000 SMD 0402 7 5 mCd Yellow SMD 0603 VISHAY TLMY1000 GS08 LED 0603 Link 16000 SMD 0402 YELLOW 2mA 1 8V 7 5 mCd The total price of all parts placed on this PCB is approximately 55 to 60 anno 2012 Pay extra attention to R3 RA and R5 In order to decrease the measurement noise these components should be made of Thin Film material This material gives less noise than Thick Film material and also has more accurate resistor values Solder Guide Before starting this guide a warning should be given There is one error on the PCB There is an error in text on the board which denotes which resistors have to be placed to use either serial or CAN communication The location of this text is given in figure 12 The text on board is placed like this 9 RXD 10 TXD 11 CAN 12 CAN But the correct order has to be this will also be mentioned later 13 RXD Figure 12 Back view of the PCB showii 14 CAN error in the silk layer 15 CAN 16 TXD To solder the PCB it is best to start with soldering the flattest components More thick components are done at the end We start at the top layer shown in figure 13 This figure can be found in bigger format in the Appendix First to solder on the top side are the following components in table 4 Table 4 Sector Components A1 2x R3 1x C6
33. This indicates that the misalignment of the left and right standers is quite big One possibility to improve this is to rebuild the bar holders using the 3d printer The base of the setup is now build from 3 parts Creating this as one part using Rapid prototyping could improve the setup This could reduce the position dependency by inducing better calibration results The position dependency could also be caused by the NET F T sensor When calibrating it is visible that the NET F T sensor itself shows a little position dependency as well when rotating The bars of the setup are not perfectly straight and due to the removal of one bearing the NET F T sensor is a little bit tilted during calibration By supporting the NET F T sensor when calibrating an effort can be made to reduce the position dependency This requires an additional part for the calibration 5 4 Data extraction To extract the desired data from the measurements like the one shown in figure 5 5 an algorithm has been written in Matlab R2012a This algorithm is given in appendix E It is shortly explained below For more information the code has been properly documented The results are as said before composed out of 4 different stages The first one is actuating in the desired way Then there is some waiting period After that opposite actuation and again waiting The measurements of the desired and opposite actuation are considered two independent measurements To ext
34. University of Twente EEMCS Electrical Engineering Control Engineering Characterization of the mVSA UT H W Han Wopereis BSc Report Committee Prof dr ir S Stramigioli Dr R Carloni Dr ir P Breedveld M Fumagalli PhD Dr ir R J Wiegerink August 2012 Report nr 019CE2012 Robotics and Mechatronics EE Math CS University of Twente P O Box 217 7500 AE Enschede The Netherlands Contents 1 Introduction 4 EL Project goal cu eu RECTE une ee ACE d ett 5 1 2 Reporto tlne 2 2 1 2 8 bt a ben rese EROR E 5 2 Requirement and approach analysis 6 2 1 Approach options analysis 2 2 llle 6 2 2 General 8 2 3 Detailed approach 8 3 Electronic part of the measurement setup 10 Sel Strain gauges sy sy sk b e SE ae ee a 2 10 3 2 8 channel amplifier board 11 3 21 Requirements 11 3 2 2 Analysis s fena Sela Pedy ede pug REED 12 B23 DESIGN Hud este RE dele ane 13 3 3 Motors and motor drivers 15 3 4 Arduino ATMEGA2560 15 4 Mechanical part of the measurement setup 16 4 1 Sensor analysis i se ttv aa set S Rao d RR EAS 16 4 11 Options research 17 4 1 2 Sensor design using FEM analysis 19 4 2 Setup design 24 uen e aoe uU E RUP ak Sene et 21 4 2 1 Materia
35. a an atat Dp ty 7 3 2 E Appendix 4 Wiring diagram of the second layer Appendix 5 Wiring diagram of the third layer Appendix C C Data sheets Table C 1 Links to data sheets of components 2012 Component Url DG408DY www maxim ic com datasheet index mvp id 1003 t al ATMEGA328 http www atmel com devices atmega328 aspx INA118 http www ti com lit ds symlink ina118 pdf ADS7279 http www ti com lit ds symlink ads7279 pdf UBR232R http www ftdichip com Support Documents DataSheets Modules DS_UB232R pdf NET F T http www ati ia com products ft ft_models aspx id Mini40 76 Appendix D D Solid works drawings M4 ga PROJECTION Han Wopereis DATE 2 i UNLESS STATED DRAWN METHOD OTHERWISE TOLERANCES 0 5MM CHECKED Matteo Fumagalli SCALE 2 pes 20 UNIVERSITEIT TWENTE MATERIAL Aluminium SURFACE FINISH TITLE Sensor DRAWING NO FILE PART NAME Sensor 0 008 210 g6 0 016 2 2 1 M4 PROJECTION i i UNLESS STATED METHOD _ OTHERWISE TOLERANCES 0 5 MM MATERIAL vA SURFACE FINISH UNIVERSITEIT TWENTE x DRAWN DATE 2 CHECKED SCALE 1 TITLE NO NAME DRAWING NO FILE PART NAME Bar left
36. a second sun gear 11 that is also fixed to the shai the planet carrier 6 In the remainder of the Section we describe these r levels of the mechanical structure and the mVSA 1 operating principles in more detail A Level 1 The Variable Stiffness Mechanism The first level of the mVSA UT realizes the variable ness mechanism and includes the parts 1 to 7 in Figui It is illustrated in Figure 3 The operating principle of mVSA UT relies on a variable transmission between internal springs 3 and the output crank 1 by connec them via a lever 2 with variable length The variable 1 length is achieved by moving the pivot pin 4 along a in the lever A linear motion of the pivot pin is accompli Fig 3 Level 1 The Variable Stiffness Mechanism The output shaft rotates around the central axis of the mechanism and is connected to the lever arm with a crank On the other side of the lever arm two springs connect the lever to the frame which can continuously rotate The dotted spring is virtual and represents the equivalent linear spring of the two springs with a planetary gear train where the diameter d of the ring gear 7 is twice as large as that of the pivot gear 5 The ring gear forms part of a rotating frame that defines the undeflected output position The reference frame in 0 is fixed to it and its position is denoted by 07 The pivot gear is actuated by the planet carrier 6 which position relative to
37. and fourth movement The torques are much smaller at these rotational velocities than the esti mated top limit of 100m Nm but inside the lower limit of 1mm Nm The first item is probably caused due to the setup There is a small amount of position dependency in the setup which contributes to the problem Also in steady state the timing belt which is the connection between the motor and the sensor is still a bit stressed T his applies a torque to the sensor which at some point turns to zero by static friction The second and third item seem to be caused by position dependency of the sensor It can be seen especially on the torque measurement of the sun gear that the same shape appears 2 5 3 times which corresponds to the amount of rotations made by the sensor This position dependency is quite big 2 5mN m but could not yet be removed by calibration 1 cannot be assumed that the measurement results of actuating in one mode are inverse equal to the measurement results of actuating in exactly the opposite mode 31 The last item is unfortunate but not strange the rotational velocity is not incredible fast so the expected frictional torque is also low 43 0 5000 10000 15000 Figure 5 5 Graph showing one of the measurements The torques are in mNm The black line is the torque input of the ring gear the grey line is the torque input of the sun gear The 4 sections labeled are a Desir
38. asured with the strain gauge bridges 29 1 0 2000 4000 6000 8000 10000 12000 14000 Figure 5 3 Graph showing the applied torques of the calibration Note that there is an offset of around 50 mNm 1 1 8800 8900 9000 9100 9200 9300 9400 9500 9600 9700 9800 Figure 5 4 Graph showing the torques of both sensors after calibration Note that there is an offset of around 50 mNm As can be seen the torque measured with both sensors are approximately equal However the NET F T sensor has less higher frequency peeks in the torque readings This is probably caused by mechanical damping in the NET F T sensor This damping removes spikes and overshoots from the measurements which are not removed in the self built sensor There is not much that can easily be done to remove this difference However the differences are small and therefore these do not necessarily form a problem Furthermore it can be seen that both the torque sensors always have a torque applied of more than 0 The steady state of the sensor is located around 50 mNm So a torque reading of approximately 50 mNm is actually a torque of 0 mNm This offset does not have to be removed in the calibration to get the correct measurement results 30 5 2 Measurements and analysis To determine the unknown wanted matrices given in equation 2 2 in section 2 3 the same approach is used as for calibrating the setup using the Pseudo Inverse This mean
39. ations such as wearable devices prostheses or exoskeletons which require the mechanisms to be portable For the purpose of having compactness it is important to use small motors and to exploit their joint efforts i e their torques for both changing the output position and the output stiffness This requirement suggests to connect the internal actuated degreed of freedom in a differential configuration The third requirement follows the elaboration presented in 11 and 17 by means of a port based approach The mechanism should realize a kinematic structure such that the apparent output stiffness can be changed without injecting energy into or extracting energy from the internal elastic el ements This property guarantees that all the energy supplied by the internal actuated degrees of freedom can be used to do work on the output without being captured in the internal springs This characteristic is satisfied only if the mechanical design is based on a lever arm of variable effective length As extensively analyzed in 12 a variable transmission ratio between the internal elastic elements and the actuator output can be realized by using a lever arm if the position of one of the three elements attached to the lever arm is varied i e by moving the pivot point 12 14 by changing the application point of the output force 11 18 or by varying the attachment points of the internal springs 13 It has been shown in 12 that moving
40. ct the device and tool and press apply Then read the device ID and choose the tab Fuses on the left The following fuse settings can be set but be careful while doing and double check all settings BODLEVEL This is the brown out detection This will make sure that the microcontroller turns off when the power drops below a certain value Set to 2 7 Volts Y RSTDISBL Disables reset Never check this option Since the ISP won t work without the reset So unchecked v DWEN This enables the debug wire Unnecessary in this application so unchecked Y SPIEN This enables the serial program downloading This is necessary for ISP so leave this checked Y WDTON Not important unchecked v EESAVE Also not important unchecked v BOOTSZ The boot size is 2048 3800 this is default v BOOTRST Unchecked Y CKDIV8 This is the clock divider When checked it will divide the clock speed by 8 thus lowering the speed This could be useful in some cases but in this case it is not useful So uncheck this Y CKOUT This will give the clock as an output on pin PBO This is not necessary for the 8 channel multiplexer board Unchecked Y SUT CKSEL This is an important fuse Leave this one as it is in first place INTRCOSC 8M 6CK 14CK 65ms Click program to set the fuses After changing the fuses the last fuse can be changed If the programmer still works it means that you changed the fuses correctly The SUT CKSEL se
41. d to follow the output crank and the pivot pin causing an elongation of the springs and thus a spring force The stiffness is then obtained by taking the partial derivative of the output torque 7 with respect to the output deflection 01 as given by Equation 2 In the case of the torsional springs in 6 d and 6 e the torque on the output is calculated with an equilibrium of forces and moments of the lever that experiences a torque 05 01 c and 05 c respectively where c is the torsional stiffness of the spring Il n 22 FEN 3 Config d Ne Config b Config c 0 dz 2 dz 2 Fig 7 Output stiffness K as function of the linear pivot position x for different spring configurations Fig 8 Plots of the output torque a and the output stiffness b as fun of and 06 for the mVSA UT Note that the plots are bounded due t singularity at 06 0 All profiles show an asymptote for x dr i e theo cally infinite stiffness at this point This is because the pin is then coincident with the connection between the 1 and the output crank However only configurations 6 a and 6 e show a m tone stiffness that has no local minimum between 0 d The other curves do show a minimum around 4 are steeper for x gt 47 2 Their shape after the minin resembles that of the stiffness for configuration 6 a were horizontally compressed The configurations 6 b and 6 d only seem to utilize half the leng
42. e These measurements are not included in the report because of the excessive amount of paper that would be needed However the final table containing all average values of all separated measure ments is added in appendix F The final Pseudo Inverse matrices taken from this table is shown below in equation 5 4 0 4863 0 2630 0 0819 de 5 4 ns EEE The torques in this are mNm the rotational velocities in rps of the input of the mVSA UT which is a 13 teeth gear The values in matrix A are in mNm and the values in matrix B are in The directions are of course non dimensional The graphs showing the torque values calculated using the Pseudo Inverse compared to the actual measured torque values for both the sun and ring inputs are shown in figures 5 10 The velocities of the ring and sun gears inputs belonging to these torque values are shown in figure 5 11 Figure 5 10 The two graphs showing the fits of the actual measured torques and using the Pseudo inverse in mNm Left are the torques of the ring gear Right are the torques of the sun gear The black lines are the measured values the grey lines are the calculated values 36 OHH 4 05 1 T 1 1 T J 1 5 X 7 10 20 30 40 50 60 70 Figure 5 11 The combination of input velocities belonging to the graph in figure 5 10 The black l
43. e sun gear 11 are driven by the motor pinions 12 and 13 with the transmission where 419 are the angular velocities of the two motors 412 and dia the diameters of the pinions and 411 the diameter of the sun gear 11 Note that the sun gear 11 does not have in general the same diameter as the sun gear 8 This allows for more flexibility of the mounting space or transmission ratio di dio qa 4 8 Fig 5 Level 3 The Actuation Stage The two motors 12 and perform the rotation of the two small gears thus actuating 412 and The two motors are fixed to the base but allow the mVSA UT to continuous rotation Because w11 wg the transmission between the actuz and the differential mechanism expressed in Equation 4 be written in matrix form as 610 ge d 2 Wg ge 913 leading to the overall transmission of mVSA UT d d wr ds Th hil P da Given the diameters of the gears used in the prototype internal transmission becomes 0 197 0 209 0 197 0 556 IV DESIGN EVALUATION In this Section the output stiffness characteristics of proposed actuation system are analysed and design cho regarding the spring configuration motivated The output stiffness as described in Equation 2 depi on the way the internal springs connect the lever to frame of the variable stiffness mechanism shown in Fi 3 It is determined
44. ed actuation b wait c opposite desired actuation d wait 0 5000 10000 15000 Figure 5 6 Graph showing the number of rotations of the inputs compared to the time shown on the x axis The time can be calculated in seconds by taking the value on the x axis and dividing it by 200 The total time for this graph is 75 seconds The black line is the torque input of the ring gear the grey line is the torque input of the sun gear To verify whether the measured torques are really due to the mVSA UT and not caused by the measurement setup some measurements where done without the mVSA UT connected The measurement result that is complementary to the one shown in figure 5 5 is shown in figure 5 7 The same actuation is used but for half the amount of time 32 1 0 2000 4000 6000 8000 10000 12000 Figure 5 7 The complementary measurement without the mVSA UT connected at the same velocity as 5 5 Grey is the sun gear input The position is given MN 1 1 1 1 0 2000 4000 6000 8000 10000 12000 Ec Figure 5 8 The position measurement without the mVSA UT connected at the same velocity as 5 5 Grey is the sun gear input The velocities of the first rising edge are not as desired The reason for this is unknown This is however When comparing the empty measurement to the measurement with the mVSA UT connected it can clearly be seen that the same steady state no velocity offset error rema
45. electable number of channels can be used The measured signals are outputted as digital 14 bit values with less than 3 bits of noise The 8AMP is designed to multiplex between the chosen input channels The selected channel is then amplified sampled and converted to a digital value This value is processed by an Atmega328 datasheet in Materials and then sent using either CAN or serial communication Choosing the amplifier gain is done by changing a certain resistor value Choosing the communication method is done by using 0 ohm resistors With its size smaller than 12 cm and total weight lower than 5 grams the board is perfectly suitable to measure strain in small or moving applications but can also be adapted for measurements with thermo couples and RTD sensors The latter is not described in this manual Contents Overview oi ten eon as or terria tu rs Co ee aA i Ae eda hans 2 SEM ELE EIE ILC ILC ONERE 3 PCB 7 Materials ceni edo niit ee LEE 9 S ld r Guidet sis cess M 11 Prograrnmiligiu 14 Programming anne 14 FUSE SETUNE S 15 Bootloader lock Ditssz eiie ten e Ire BB ehe ne 16 Main 1 1 pee une na 16 Expected re
46. ensions made a compact con struction of the mechanism possible Still their load ratings exceed the nominal bearing loads which ensures mechanical robustness as well as low friction and precision VI ACTUATION AND INSTRUMENTATION In this Section we present some details about the internal actuators and sensors used in the mVSA UT The design of the mVSA UT is modular so that different motors can be used as the internal actuators This results Static Structural ANSYS Equivalent Stress Type Equivalent von Mises Stress Unit Pa Tevet 1 12 12 2011 19 21 1 69e8 Max 1 502268 1 314468 1 126768 9 388967 7 511187 5 633307 2 0712e5 Min B Static Structural ANSYS Equivalent Stress Type Equivalent von Mises Stress Unit Pa 12 12 2011 15 05 3 9163e8 Max 3 481168 3 04668 2 6108e8 2 175768 1 740608 1 3054 8 8 7028e7 4 3514e7 210 42 Min Fig 9 Results of the FEM analysis of the carrier output and piv in freedom of design choices concerning the output and strength as well as dimensions motor types or strategies for the actuator The chosen motors must me carry the correct pinion 13 teeth and modulus 0 5 1 and have the correct distance between their axes to prop engage the lower sun and ring gear They can be fixe the base of the housing e g with an adapter piece th screwed to the housing The current prototype of the mVSA UT is equipped two micro servo motors
47. esigned to be more accurate The position dependency problem should be solved before doing this 39 Bibliography 1 Viactors group Official viactors website August to juli 2011 2012 2 Ieee international conference on robotics and automation conference icra 2011 shanghai In AwAS II A Novel Actuator with Adjustable Stiffness Based on Variable Ratio Lever Concept October 2011 Ieee int conf on robotics and automation icra 2011 In The DLR FSJ Energy based design of a variable stiffness joint pages 5082 5089 2011 Ieee rsj international conference on intelligent robots and systems taipei taiwan In VSA HD From the Enumeration Analysis to the Prototypical Implementation pages 3676 3681 2011 International conference of robotics and automation icra 2011 In VSA CubeBot A modular variable stiffness platform for multi degrees of freedom systems pages 5090 5095 2011 Icra 2012 In The mVSA UT a Miniaturized Differential Mechanism for a Continuous Rotational Variable Stiffness Actuator 2012 Icra 2012 In The vsaUT II a Novel Rotational Variable Stiffness Actuator page Submitted 2012 An Introduction to Measurements using Strain Gages pages 1 15 Hot tinger Baldwin Messtechnik GmbH Darmstadt 1989 Airobots group Official airobots website 2011 2012 Arvid Keemink Design realization and analysis of a manipulation system for uavs Master s thesis Januari 2012 Pololu Robotics and Mec
48. f the 8 channel amplifier board SAMP To connect the 8AMP directly to a PC a Serial To USB transceiver is required The transceiver chosen in this project is the UB232R module by FTDI Chip The link to the data sheet of this manual can be found in appendix C 14 3 3 Motors and motor drivers The motors that will be used initially are two MAXON RE10 DC motors with GP10A four stage 256 1 reduc tion gears and with MEnc10 12 ticks magnetic encoders which are readily available but broken The encoders of these motors were both broken but could be repaired by removing and reconnecting the magnetic sen sors These magnetic sensors either had loose connections or where bent resulting in bad encoder readings The motors can provide over 1 5 mNm of continuous torque when ap plying 6 volts which results in more Figure 3 2 A photograph of the Pololu than 200 mNm of torque at the out 755 motor driver completely assembled put of the gearbox This is expected to be more than sufficient to drive the measurement setup The rotational velocity of the output shafts of these motors is however quite low lt 8rpsifVm 6V due to the reduction gears But since these motors are available these will be used The motor drivers used to supply power to these motors are two Pololu 755 High Power Motor Drivers Figure 3 2 displays a completely soldered board This discrete MOSFET H bridge motor driver provides bidirectional control for o
49. fication factor of 5 V V The resistor R5 can be changed to change the gain factor of the INA118 The formula for the gain also given in the datasheet is 50000 Gain 14 a n R5 Keep in mind when choosing the gain factor The maximum range of the measurement is not O to 5 volts This is due to the amplifier The INA118 guarantees an minimum range of output of V 0 35 to V4 1 This is not the linear region The linear region guaranteed is V 1 1 to V 1 although it typically is much bigger In case a bigger linear range is necessary other components like the INA122 This will however decrease the speed in which measurements can be taken because of the lower slew rate The LEDs are not mentioned in the BOM above These are optional and will consume power when used Below certain advised LEDs and matching resistors are given in table 3 Choices for colors e g are up to the user All LEDs on the PCB are denoted by the functions they lit up to These functions are power reset TXD RXD and clock All matching resistors are denoted by R Table 3 The BOM of usable LEDS Color Package Description Datasheet Matching resistance 5V Package input power suggestion resistor Green SMD 0603 VISHAY TLMG1100 GS08 20mA 2 1V Link 5000 SMD 0402 15mCd Red SMD 0603 VISHAY TLMS1000 GS08 2mA 1 8V 4 Link 16000 SMD 0402 mCd Blue SMD 0603 VISHAY TLMB1100 GS08 20mA 3 9V Link 2000
50. g to part 8 e Part 8 This part connects the sensor and the second shaft e Part 10 This is the second shaft It does not contain a hole in the center like part 4 The SolidWorks drawings of all custom made components can be found in ap pendix D The Bill Of Materials BOM for the used bearings gears and pulleys can be found in subsection 4 2 1 The bottom plate to which the complete setup will be connected has slits to allow the different subsystems to move closer and further away from each other to allow the sizes of the gears and pulleys to change 25 4 2 1 Materials and components All designed and custom build parts in this setup including the sensor are made of Aluminium alloy 6082 T6 This is chosen because its lightweight and easier to process than heavier cheaper metals The bottom plate of the setup is made of gray Poly Vinyl Chloride PVC The gears pulleys and bearings used in the setup are given in table 4 1 The crucial dimensions and features are given as well Table 4 1 Bill of Materials non custom made Amount Type Size outer di Size hole di Modulus ameter ameter 8 Angular Contact 30 mm 10 mm Bearing 4 Shielded Bearing 8 mm 3mm x 2 Pulley 16 teeth 3mm 0 5 2 Pulley 48 teeth 8 mm 0 5 2 Gear 16 teeth 12 mm 3mm 0 5 2 Gear 48 teeth 24 mm 8 mm 0 5 2 Timing belt 0 5 2 Gear 13 teeth 3mm 0 5 The bolts and nuts used to assemble
51. gData finalSunData sortOnRing finalRingData finalSunData end if sortBy 2 finalSunData finalRingData sortOnRing finalSunData finalRingData end ringForward finalRingData 1 ringSpeedForward finalRingData 2 ringBackward finalRingData 3 ringSpeedBackward finalRingData 4 sunForward finalSunData 1 sunSpeedForward finalSunData 2 sunBackward finalSunData 3 sunSpeedBackward finalSunData 4 EDER Combine both forward and backward into variables sunAll 1 numel sunForward sunForward sunAll numel sunForward 1 numel sunForward numel sunBackward sunBackward sunSpeedAll 1 numel sunSpeedForward sunSpeedForward sunSpeedAll numel sunSpeedForward 1 numel sunSpeedForward numel sunSpeedBackwe ringAll 1 numel ringForward ringForward ringAll numel ringForward 1 numel ringForward numel ringBackward ringBackw ringSpeedAll 1 numel ringSpeedForward ringSpeedForward ringSpeedAll numel ringSpeedForward 1 numel ringSpeedForward numel ringSpeedBz WU Be a Dal 4 Then for all values 95 First declare the matrices ringTrans transpose ringSpeedAll sunTrans transpose sunSpeedAll y mzLeEccec cezclccllollllolccoclocloccoollooclcc 4 Create a ones list with ones for velocity forwards and ones for velocity backwards This is to
52. gears where used for transmission These measurements are shown in figure 5 9 From these graphs two conclusions can be made e The measurement with gears only has a much bigger amount of high frequent noise but the steady state value is approximately equal every waiting period It is not completely clear if there is position dependency because of the high amount of noise e The measurement with pulleys has a lot less high frequency noise but the steady state value is different every waiting period Also position dependency is more clearly visible now The pulleys clearly give less noise so one improvement could be to only use pulleys The steady state error is no problem since for every measurement that is taken with the mVSA UT connected an measurement can be done without the mVSA UT connected The average value of the empty measurements can then be subtracted from the average value of the mVSA UT measurements resulting in the wanted values To remove the position dependency a couple of things have been tried The first effort was to improve calibration by improving the connections between the measurement setup and the NET F T sensor This did not work It has 34 been tried to remove some angular contact bearings so that only one bearing at each side of the sensor remained This slightly decreased position dependency from 1 5m Nm to 1 0m Nm but made the setup more fragile to the bending moment applied by the timing belt
53. get the coulomb force of both the sun and 4 the ring gear for i 1 numel ringSpeedAll if ringSpeedAll i 0 ringDirection i 1 else ringDirection i 1 end end for i 1 numel sunSpeedAll if sunSpeedAll i gt 0 sunDirection i 1 else sunDirection i 1 end end ringAllTrans transpose ringAll sunAllTrans transpose sunAll ringDirTrans transpose ringDirection sunDirTrans transpose sunDirection Vall Tall ringTrans sunTrans ringDirTrans sunDirTrans ringAllTrans sunAllTrans RE ur EE EE 4 Then the least squares function 4Aallmatrix pinv Vall Tall Aallmatrix 0 4863 0 2630 0 3698 0 9649 0 0819 0 0482 0 0077 0 5752 TallStar Vall Aallmatrix 96 figure plot Tall 2 black hold on plot TallStar 2 red figure plot Tall 1 black hold on plot TallStar 1 red figure plot ringTrans k hold on plot sunTrans r 97 98 Appendix F F Measurements data F 1 Final measurement data Table F 1 Final measurement data for the sun gear input Ti OTI GQ revers avg 1 9458 0 4537 0 8193 0 4258 48 6258 1 4256 1 3289 2 5049 1 3306 47 9485 2 8140 1 7634 1 4718 1 7627 48 6591 1 7384 0 8415 1 6372 0 7843 48 0740 1 6010 1 6549 2 7663 1 6561 47 4862 1 7095 0 5096 0 8813 0 4247 47 6999 1 3331 0 8376 1 8240 0 8386 48 3041
54. hatronics Official pololu website page for motor driver 755 2011 2012 40 Appendix A Paper of the mVSA UT The mVSA UT a Miniaturized Differential Mechanism for a Continuous Rotational Variable Stiffness Actuator M Fumagalli E Barrett S Stramigioli and R Carloni Abstract In this paper we present the mechanical design of the mVSA UT a miniaturized variable stiffness actuator The apparent output stiffness of this innovative actuation system can be changed independently of the output position by varying the transmission ratio between the internal mechanical springs and the actuator output The output stiffness can be tuned from zero to almost infinite by moving a pivot point along a lever arm The mVSA UT is actuated by means of two motors connected in a differential configuration which both work together to change the output stiffness and the output position The output shaft can perform unbounded and continuous rotation The design ensures high output torque capability light weight and compact size to realize a multiple purpose actuation unit for a great variety of robotic and biomechatronic applications I INTRODUCTION Variable stiffness actuators VSAs realize a new class of actuation systems characterized by the property that the apparent output stiffness can be changed independently of the output position Such actuators are particularly significant when implemented on robots that have to interact safely with
55. he information to int data ans PINB amp 00010000 gt gt 4 data data 1 ans Set clock low PORTB B00000001 End of clocking set CS back to HIGH PORTB B00000101 68 Sample start sets CONVST LOW 69 PORTB B00000100 70 wait in total 1 microsecond for sampling 71 __asm__ nop n t Wait 62 5 ns 72 __asm__ nop n t Wait 62 5 ns 73 __asm__ nop n t Wait 62 5 ns 74 __asm__ nop n t Wait 62 5 ns 75 __asm__ nop n t Wait 62 5 ns 76 __asm__ nop n t Wait 62 5 ns 77 __asm__ nop n t Wait 62 5 ns 78 __asm__ nop n t Wait 62 5 ns 79 __asm__ nop n t Wait 62 5 ns 80 __asm__ nop n t Wait 62 5 ns 81 __asm__ nop n t Wait 62 5 ns 82 __asm__ nop n t Wait 62 5 ns 83 __asm__ nop n t Wait 62 5 ns 84 __asm__ nop n t Wait 62 5 ns 85 __asm__ nop n t Wait 62 5 ns 86 __asm__ nop n t Wait 62 5 ns 87 88 Sample end sets CONVST HIGH 89 PORTB 00000101 90 91 92 First set new channel longer time for channel to set value 93 Starts with initialized channel 1 twice then loops correctly 94 95 switch swtch 96 97 case 1 98 PORTD B00010000 Channel 1 99 break 100 101 case 2 102 PORTD B00110000 Channel 2 103 break 104 105 case 3 106 PORTD B01010000 Channel 3 107 break 108 109 case 4 110
56. he input shaft T hus a solution has to be found for the wires going to the sensor Also if the sensor is not well calibrated there could be noise from the rotation Therefore the calibration has to be done accurate Then there is also the option of creating a static torque sensor setup This setup measures the reaction torque the motor induces when applying a torque to the mVSA UT This torque sensor to which the motor is connected can as well be either bought or designed and built The advantage of this option is again adaptability however it is less adaptable than the previous option The precision will depend on the properties of the sensor The disadvantage of this option is that the sensor has to carry the motor which can lead to cross talk errors Also carrying the motor could give reduced sensitivity as the sensor has to be over sized to carry the extra load Table 2 1 List of advantages and disadvantages of three approach options Approach tion Advantages Disadvantages Characterize motors Once motors are charac terized properly measur ing is easy Requires better motors Low adaptability Preci sion depending on motor and characterization Dynamic torque sensor setup High adaptability Preci sion depending on sensor Rotating sensor could give problem with wires Static torque sensor setup Medium high adaptability Precision depending on sensor Possibility of c
57. hi VSA IE prototype of variable stiffness actuator for safe and performing r interacting with humans in Proceedings of the IEEE Internat Conference on Robotics and Automation 2008 S Wolf and G Hirzinger A new variable stiffness design Mat requirements of the next robot generation in Proceedings of the International Conference on Robotics and Automation 2008 K Koganezawa Y Shimizu H Inomata and T Nakazawa Act with non linear elastic system ANLES for controlling joint stif on antaonistic driving in Proceedings of the IEEE Internat Conference on Robotics and Biomimetics 2004 M Catalano G Grioli M Garabini F Bonomo M Mai N Tsagarakis and A Bicchi VSA CubeBot a modular variable ness platform for multiple degrees of freedom robots in Procee of the IEEE International Conference on Robotics and Autom 2011 L C Visser R Carloni and S Stramigioli Energy efficient vai stiffness actuators IEEE Transactions on Robotics vol 27 2011 S Groothuis G Rusticelli A Zucchelli S Stramigioli and R loni The vsaUT II a novel rotational variable stiffness actuato Proceedings of the IEEE International Conference on Robotic Automation 2012 A Jafari N Tsagarakis B Vanderborght and D Caldwell A actuator with adjustable stiffness AwAS in Proceedings o IEEE RSJ International Conference on Intelligent Robots and Sys 2010 AwAS IE A new actuator w
58. humans and have to feature properties such as energy efficiency robustness and high dynamics In particular these actuators find their application in different fields of robotics and biomechatronics In prosthetics or rehabilitation robots for example the introduction of a VSA allows the device to adapt to the task and to increase not only the efficiency of the actuation but also the comfort of the patient 1 5 VSAs have the intrinsic capability to store and release energy during nominal tasks However their main drawback Is the inefficiency of transferring energy from the internal motors to the output due to the presence of internal me chanical elastic elements This limits their employment in precise positioning tasks which motivates the research effort on their mechanical design and control In the literature mechanical compliance has been imple mented in different ways in VSAs In the Jack Spring actuator 2 the apparent output stiffness is varied by changing the number of active coils of the internal spring Other actuators e g the MACCEPA 2 0 6 the VSA II 7 the VS Joint 8 the ANLES 9 and the VSA CubeBot 10 change the apparent output stiffness by varying the pretension of the internal nonlinear springs Other actuators including the vsaUT 11 the vsaUT II 12 the AwAS 13 the AwAS II 14 and the HDAU 15 change the apparent This work has been funded by the European Commission s Seventh Framew
59. in gauge half bridges using thin film resistors These resistors introduce less thermal noise into the system When a strain is applied to the strain gauges the bridges will not balanced anymore The voltage returned will now be different from 2 5 volts e g 2 2 volts The amplifier will compare this voltage to the reference and detect a difference of 0 3 volts This difference is then amplified by the 12 gain factor subtracted from the reference voltage and presented at the output In case of a difference of 0 3 volts and a gain factor of 5 the voltage at the output will now be 1 volts Other crucial features of the INA118 are the high slew rate 0 9 low gain er ror small package SOIC 8 and the possibility to set the gain using an external resistor The formula to set the gain is given by equation 3 1 5050 4 G 1 3 1 R being the external resistor This external resistor should not produce to much thermal noise and thus thin film resistors will be used for Rg The ADC chosen is the ADS7279 This low power 14 bit analog to digital converter has a sampling rate of 1 MHz It has been selected because of the high accuracy 14 bits max 1 LSB small package TSSOP 16 unipolar input and the fast sampling rate Since the footprint is the same as the ADS8329 the accuracy of 14 1 bits can even be increased to 16 1 bits but this is not necessary for this application The accuracy in volts when using a power supply
60. in the torque mea surement setup The 8 channel amplifier board designed within this project will be presented as well as a summary of the other electronics used in the measurement system 3 1 Strain gauges It was discussed in section 2 3 that semiconductor strain gauges were the best choice for this project The semiconductor strain gauges used were purchased at Micron Instruments The U shaped SS 037 022 500PU strain gauges were selected These are smaller and have better thermal coefficients than most others The in rest resistance of these strain gauges is 5400 with a maximal variation in resistance of 50 Ohms These strain gauges are selected because of the big linear region The strain gauges have a linearity better than 0 25 to 6004 22 which is approximately i of the full range and a linearity better than 1 5 to a strain of 15004 2 which is half of the full range They also have been selected because of the small size 10 3 2 8 channel amplifier board This section of the report describes the development process of the 8 channel amplifier board referred to as 8AMP from here on used for the measurement setup It gives the criteria and design choices for the 8AMP as well as a brief description of the design process For more information an users manual is included in appendix B This manual provides among others more specific information on the design of the 8AMP and a bill of materials It also includes an extensive
61. inal residue and is thus not completely accurate 13 The printed circuit board PCB design process of the 8AMP is also described in the manual A short summary on the design The 8AMP is a 4 layer PCB It has carefully be designed to have as low noise on signal wires as possible Digital and analog signals are separated Signal wires are as short and straight as possible and wiring at different layers is oriented in perpendicular directions where possible Capacitors ferrite beads and coils are used for filtering the power supply of different components Ground planes are used on each layer to decouple noise and cross talk between adjacent wires and to ensure the same reference potentials The final PCB is shown in figure 3 1 with the components soldered to the board except for the Serial to CAN transceiver The dimensions are x x mm and the total weight including components is X grams It contains five LEDs indicating power reset transmit receive and clock The clock indicates if the ADC is transferring data to the micro controller The SAMP can be connected to another PCB with as little as 4 wires power ground RXD or CANrow and TXD or CANg Gg There is however an additional connector located at the SAMP which can be used to program the ATmega328P using an In Serial programmer The 8AMP has 24 pads 8 x 3 pads which supply power and ground towards strain gauge half bridges and extract the signal Figure 3 1 A photograph o
62. ine is the velocity of the input of the ring gear the grey for the sun gear The x axis shows the number of different measurements To verify the integrity of the final results some additional measurements have been done To be exact 20 measurements are used to compare the actual measured values to the calculated torque values for both the sun and ring input The graphs showing the fit are shown below in figure 5 12 The used velocities are shown in figure 5 13 The table with results of all measurements can again be found in appendix F 37 Figure 5 12 The two graphs showing the fits of the actual measured torques of the verification measurements and using the Pseudo inverse calculations in mNm Left are the torques of the ring gear Right are the torques of the sun gear The black lines are the measured values the grey lines are the calculated values Figure 5 13 The combination of input velocities belonging to the graph in figure 5 12 The black line is the velocity of the input of the ring gear the grey for the sun gear It can be seen that the measurements do act a like the expected values It is however as expected not a perfect fit and the fit is not close enough to conclude that the acquired values from the Pseudo Inverse matrix are valid and not just random 38 Chapter 6 Conclusion and recommendations The final conclusion of this project is that the main goal
63. ins and the position dependent noise is still clearly visible However the displacement of the average value when moving from the no velocity value is a lot smaller than with the mVSA UT connected This is the case for all measurements that were done This implies some part of the measurement is the desired result However since the noise in most measurements is bigger than the desired result it is hard to determine the quality of the results Therefore the dissipation values of these measurements are only an indication of the actual values To achieve better results improvement analysis has to be done and the measurement setup has to be improved 33 5 3 Improvement analysis A couple of efforts where made to reveal what is causing the measurements to have this amount of position dependency and noise 100 7 wow 70 eor E 50 0 2000 4000 6000 8000 10000 12000 0 2000 4000 6000 8000 10000 12000 Figure 5 9 Attempt of improvement of measurements Left the torques are shown The grey line is a measurement taken on the sun gear with only gears as transmission The black line is a measurement taken on the ring gear with only pulleys as transmission Right the position vs time is shown for this mea surement The velocity of the gears is deliberately lower to control the size of the noise First of all the effort was made to verify whether the results would be better for one setup if only pulleys or only
64. ion is given about the gain Figure 5 Part of the complete schematic The VOUT pin of the amplifier is connected to the input of the AD converter shown in figure 6 The AD converter has references of 0 and 5 volts The reference voltage is kept stable by a 10uF capacitor but this capacitor may also be bigger The AD converter communicates with the microcontroller unit using the MISO and MOSI wires The AD converter being the slave and the MCU the master The SCLK CS CONVST and EOC are used to convert the analog signal to digital This is explained more thorougly in chapter Programming The AD converter is the first component that is partly digital The digital supply VDD is therefore used for the digital voltage supply whereas the VCC is used for the analog voltage supply This is crucial since digital signals create alot of noise The digital and analog ground are also seperated The AD converter is commanded by the MCU of which the essential part of the schematic is shown in AD converter figure 7 An 16 MHz external oscillator is used instead of the 8 MHz internal oscillator to improve the speed capabilities Furthermore the AD converter is controller using the PORT B commands controlling all PBx pins The multiplexer is controlled using the PORT D commands This seperation optimizes the programming code since controlling the AD converter doesn t include controlling the current multiplexer channel The RESET pin PC6
65. ith adjustable stiffness base the novel principle of adaptable pivot point and variable lever r in Proceedings of the IEEE International Conference on Robotic Automation 2011 B S Kim and J B Song Hybrid dual actuator unit A d of a variable stiffness actuator based on an adjustable moment mechanism in Proceedings of the IEEE International Conferen Robotics and Automation 2010 K Galloway J Clark and D Koditschek Design of a tu stiffness composite leg for dynamic locomotion in Proceedin the ASME International Design Engineering Technical Confere 2009 R Carloni L C Visser and S Stramigioli Variable stif actuators A port based power flow analysis IEEE Transaction Robotics vol 28 no 1 2012 L C Visser R Carloni R nal and S Stramigioli Modelin design of energy efficient variable stiffness actuators in Procee of the IEEE International Conference on Robotics and Autom 2010 Appendix B B The 8 channel Amplifier board user s manual 49 KC Manne Mukjplexer boar 02 psepueig Developed by H W Wopereis M Fumagalli 9 Channel Multiplexer Board Overview The 8 channel multiplexer board 8AMP can be used to measure voltages between O and 5 volts depending on the supplied power It is capable of measuring approximately 10000 signals each second when communicating through serial at a baud rate of 115200 It contains 8 different input channels of which a s
66. ls and components 26 5 Measurements 27 5 1 Galabration v 25 has 38 sh en b t dr rn 27 5 1 1 Calibration principle 29 51 22 Calibration results 222225252 sea ee 1 29 5 2 Measurements and analysis 31 5 3 Improvement analysis 34 9 4 Data extraction ios save x og we pud eee p eR E 35 5 5 Results eheu ee VE 36 6 Conclusion and recommendations 39 A Paper of the mVSA UT 41 B The 8 channel Amplifier board user s manual 49 C C Data sheets 76 D D Solid works drawings 77 E E Matlab files 88 File used to extract single data values 88 E 2 File used to extract the final matrices from all measurement data 94 F F Measurements data 98 F 1 Final measurement data 99 F 2 Verification data 101 Chapter 1 Introduction One of the latest researches in the field of robotics is the research on variable impedance actuators These are actuators with adjustable compliance and damp ing which can store and release mechanical energy This can be used for soft and energy efficient interaction with the environment giving a certain degree of safety One of the research collaborations focused on this is research is the VIACTORS group 1 The viactors group developed multiple different variable impedance actuators 2 3 4 5 and one of the latest develo
67. m the NET F T sensor outputs its measurements in mNm The least squares fit is done using a pseudo inverse matrix The principle of this is shown in equation 5 1 a b 1 Ti az b2 1 Koridge b3 1 Koridges T3 5 1 a4 b4 1 Cof fset In this equation and b are the values of the two strain gauge bridges 14 bit signed integers is the measured torque by the NET F T sensor Koriage Koridges and Coffset have to be found These are determined by making the best fit 5 1 2 Calibration results Both sensors where calibrated in the same way but because both halves of the setups are not completely the same the results of the calibration are slightly different This difference is mainly caused by mechanical differences between both setups The position accuracy of the shaft holes and strain gauges play a part in this but also the extent to which the shafts are horizontally placed and aligned can make a difference One of the calibration results is discussed here The Matlab code used to find the pseudo inverse of the calibration can be found in appendix E In figure 5 3 the results after calibration can be found It is hard to see that both the measured torque with the NET F T sensor and the measured strain gauges multiplied with the pseudo inverse are plotted in the same graph To have a better look at the results calibration a more zoomed view of this plot is shown in figure 5 4 The grey line is the torque me
68. main goal of this project is to contribute to characterizing the mVSA by determining its internal dissipation The results of this project could then be used in modeling of the mVSA This has to be done with a low budget and thus expensive measurement systems cannot be bought The measurement equip ment therefore has to be developed and debugged within the project 1 2 Report outline This report will be split up into three main chapters since the assignment was as well mainly split into 4 parts analy sis of the problem designing and build ing the electronics designing and build ing a mechanical setup and performing measurements and data analysis First in Chapter 2 an analysis of the main requirements to achieve the Figure 1 2 One of the internal levels of the mVSA UT showing the differ ential connection to the pivot point project goal is given and the options on achieving this goal are discussed Then in Chapter 3 the development of the electronic data collecting board is de scribed as well as the other electronics used for the measurement setup In Chapter 4 an outline is given on the development and realization process of the mechanical part of the measurement setup Chapter 5 discusses the process for taking measurements and the results Finally in Chapter 6 the conclusions and recommendations are discussed Chapter 2 Requirement and approach analysis To achieve the main goal determining the dissi
69. nd filtering is placed here Also all input channels are located on the top layer denoted by numbers 1 to 8 Using ferrite beads coils and capacitors the noise is reduced as much as possible The top ground plane is used for analog ground Furthermore a number of LEDs are placed here indicating whether the board is powered transmitting or receiving data reset and whether the clock of the ADC is working The internal layers ground planes are analog voltage on the second layer top internal and digital ground on the third layer bottom internal The bottom layer shown in figure 11 contains the AD converter the MCU and also the serial to CAN converter The power inputs of these digital components are all filtered using ferrite beads and capacitors The communication method serial or CAN interface can be chosen using Ohm resistors When using serial communication The CAN transceiver does not have to be connected In the and is mentioned which components can be unsoldered in this case The complete wiring of the board is shown in the To reproduce the board and to guarantee quality measurements it is advised to use this wiring when redesigning the 8AMP The schematic is necessary to find which components have to be placed where A lt gt nos gt SA 6592 2 se When the shape of the 8AMP has changed and thus
70. ne brushed DC motor The motors are powered with 6 volts at least 5 5 V is necessary which has to be separately provided to the motor driver More information on this motor driver can be found on the website of Pololu 11 3 4 Arduino ATMEGA2560 For communication between the PC user the SAMP board and the two Pololu boards an Arduino Mega 2560 with an Atmel ATmega2560 processor is used It runs custom made firmware which takes care of communication with the PC sending PWM and direction signals to the motor drivers reading the encoder increment interrupts and receiving the torque measurement values from the The Arduino MEGA 2560 also provides power to the logic of the motor drivers and the SAMP board 15 Chapter 4 Mechanical part of the measurement setup This chapter of the report covers the development of the mechanical aspects of the measurement setup The analysis of the sensor and the design of the measurement setup are presented Also the realization of the setup is described 4 1 Sensor analysis The principle used for the torque sensor is based on beam flexing By connecting the strain gauges to a cantilever as displayed in figure 4 1 the impedance of these strain gauges varies when applying an force at the end of the beam This variation can then be measured Strain gauges Force N Figure 4 1 Principle used for torque sensor illustrated 16 Below first some options for the shape are presented
71. nsists out of three subsystems denoted by the numbers 1 to 3 Figure 4 9 Figure of the complete measurement setup Note that used gears are not displayed Subsystems 1 and 2 are used twice one time for each motor and sensor Subsys tem 3 is used once to connect both sensors to the mVSA UT The separation of these three parts allows different types of transmissions and different sizes of pulleys and gears Only the pulleys gears that are attached to subsystem 3 are limited in size The maximum diameter for these is 15 mm Otherwise the pulleys gears would touch each other 22 Figure 4 10 Exploded view of subsystem 1 of the measurement setup The exploded view of subsystem 1 is shown in figure 4 10 It consist out of 4 parts Part 1 is the SolidWorks design of the contour of the motor and part 4 is the selected pulley Part 2 is designed in such a way that it fits motors up until 30 mm inside the slit Part 3 is the motor dependent part It contains the holes at the right positions to connect the motor but can easily and cheaply be changed in case other motors will be used Subsystem 3 figure 4 11 consists out of 7 parts Parts 1 2 6 and 7 are the bearings to let the shafts rotate smoothly Parts 3 and 4 are the shafts that connect to the mVSA UT with gears not shown on the right side Part 5 keeps all parts together 23 Figure 4 11 Exploded view of subsystem 3 of the measurement setup i igi Figure 4 12 E
72. nterface or the serial interface If the serial interface is used the following components have to be soldered 1 the top position in the white block in figure 14 1 at the bottom position TXD in the white block in figure 14 Remember this is wrongly denoted on the PCB If the CAN interface is used the following components have to be soldered 1 at the second CAN and third CAN position in the white block in figure 14 R1 CAN ON in 01 of figure 14 C4in D1 of figure 14 C6in D1 of figure 14 in D2 of figure 14 R2in D3 of figure 14 After soldering all small components the final components can be soldered These are Atmega328P on the backside of the PCB figure 14 in sector C1 ADS7278 also on the backside in sector B2 Warning On the PCB there is a circle at the wrong side of the component The correct orientation of this component is shown in figure 14 NA118 on the front side figure 13 in sector B2 DG408DY on the front side in sector B1 and C1 L1 onthe front side in sector D2 Both connectors on the front or backside whatever is preferred After soldering the connectors there is also the possibility to solder something to reset the board more easily The reset is located at sector A2 in figure 14 At this point the board is completely soldered Programming Programming connector This guide on how to program the ATmega using an AVR
73. ork Programme as part of the project VIACTORS under grant no 231554 m fumagalli e barrett s stramigioli r carloni utwente nl Faculty of Electrical Engineering Mathematics and Computer Science University of Twente 7500 AE Enschede The Netherlands Fig 1 The mVSA UT output stiffness by changing the transmission ratio betv the internal linear springs and the actuator output In this paper we present the novel design of the mV UT which realizes a compact rotational variable stiff actuator As other VSAs this system consist of a numbe internal springs and a number of internal actuated deg of freedom which determine how the elastic elements perceived at the actuator output The mechanical struc of the mVSA UT is such that the apparent output stiff can be varied by changing the transmission ratio betw the internal elastic elements and the actuator output nar by implementing a lever arm of variable effective len The length can be changed by moving a pivot point al the lever arm by means of a planetary gear system w realizes a linear motion along the lever By satisfying kinematic requirement the actuator s output stiffness ca changed without changing the potential energy stored in internal elastic elements The extremely compact design of the mVSA UT cai achieved by implementing the two internal degrees of 1 dom i e the two internal motors in a differential config tion This implies that by combining two small moto
74. oth RXD and TXD showing data communication The connector shown in the right upper corner of figure 9 also has the input connections for the 5 volts input voltage and ground This connector is connected to the master board or computer Programming LED 6 3 Figure 8 Part of the complete schematic This concludes the schematic of the 8AMP Schematic design choices that were not explained above are directly derived from the recommendations in the datasheets of the different components or from basic electronics design LED2603 xs CAN SERIAL choice Figure 9 Part of the complete schematic PCB Design Part of the design of the 8AMP is shown in figures 10 and 11 The 8AMP contains 4 layers a top layer two internal layers and a bottom layer The top layer is used for analog signals the bottom layer for digital signals Both layers are shielded by using ground planes on all four layers This separation and shielding is done to reduce noise Other noise reducing factors have also been taken in account The wiring on board does not contain sharp edges e g a 90 degree turn and the amount of EMF antenna s are reduced on critical wires by designing them as straight as possible Also analog and digital power and ground sources are separated through the board only connected by one filtered wire The top layer shown in figure 10 is used for all analog parts The multiplexer amplifier a
75. oth the rotational output as the movement of the pivot point will depend on both input velocities It is assumed that the dissipation can be approximated as linear dependent on the input rotational velocities If this is not a correct approximation the measurements will show this and the assumption has to be adjusted Relation 2 2 can be found between input velocities and measured torques in case of linearity Constants and in this equation are positive if the directions of and respectively wa are positive and negative when the directions are negative E IA 03 One possibility to gather the values of the two matrices is by taking a lot of measurements and finding the values that best fit to those measurements These measurements can be done by building a setup that contains two torque sensors Since torque sensors are usually quite expensive to buy these have to be designed and build within the project The torque sensors can be designed using strain gauges The most commonly used strain gauges are foil gauges and semiconductor gauges 8 Since the strains are expected to be very small the semiconductor strain gauges seem to be the best choice These have higher resistance values and bigger sensitivity compared to foil gauges but also have greater sensitivity to variations in temperature and the tendency to drift when aging The advantages outweigh the disadvantages since the temperature is going to be approximately cons
76. pation of the mVSA UT an measurement setup has to be build Therefore requirements on such setup have to be analyzed In this chapter the requirements from the goal s point of view are given More specific requirements on specific aspects of the to be build measurement setup are given in sub sections of the relevant chapters 2 1 Approach options analysis Determining the internal dissipation of the mVSA UT can be done by measuring the power put into the system in steady state and compare it to the power that is taken at the output The difference is the power loss inside the system The easiest way to measure this loss is by setting the power withdrawn from the output to zero In that case all power supplied to the system is used to overcome internal energy losses Thus a way to determine the input power of both inputs has to be found Since the input power of each input is given by 2 1 with 7 being the delivered torque N m and w being angular velocity in rad s7 the input power can be determined when measuring both the velocity and torque applied to each input shaft Since measuring angular velocity is fairly easy using encoders the main approach issue is how to measure the input torque Pin Win 2 1 One method to determine the input torque can be to characterize the servos used to actuate the mVSA UT To do this a setup has to be created that can determine the output torque of the servos at certain applied c
77. plifies the data of one of the 8 different input signals and passes this amplified value towards an ADC The ADC is controlled by a micro controller as well as the multiplexer After conversion the ADC passes the digital value to the micro controller which on its turn presents the value on either the Serial line or to the Serial to CAN transceiver The multiplexer chosen for this application is the Maxim DG408DY This mul tiplexer is capable of multiplexing between 8 different channels by setting three pins of logic to either ON or OFF The advantages of this multiplexer are the small package 16 narrow SO and the fast transition time 250ns The tran sition time is the time needed to switch between channels Although the ON resistance of this multiplexer is quite high 100Q this will not provide any problems since the input impedance of the selected amplifier described below is extremely high 10 90 The INA118 UB is selected as amplifier for this application This amplifier is a general purpose amplifier offering excellent accuracy The amplifier can amplify signals using only a single power supply comparing the input signal to a reference signal In this application the voltage supplied to the strain gauge half bridges will be 5 volts Assuming the half bridges are balanced a strain of 0 mm mm will return a voltage of 2 5 volts The reference of the INA118 UB will be set to 2 5 volts while matching the impedance of the stra
78. pments is the development of the minia turized Variable Stiffness Actuator by the viactors division at the university of Twente 6 which is shown in figure 1 1 From here on the miniaturized Variable Stiffness Actuator is referred to as mVSA UT The mVSA UT is a miniaturized version of the VSA UT 7 developed by the same research group The mVSA UT is actuated using two motors con nected to a differential stage to both rotate the out put shaft and change the stiffness of this output The stiffness of the mVSA UT can be changed by un Baus torn of the newly developed moving a pivot point along a lever This is illus mVSA UT with a two trated in figure 1 2 The outer frame can be rotated and the inner gear can be moved around o with radius l These are the two movements can be ac tuated by the two input motors euro coin for size refer ence The novelty of the mVSA UT and of the VSA UT lies in the possibility to vary the stiffness of the output shaft from zero to almost infinite while retaining a wide 60 maximum deflection for zero stiffness settings The novelty of the mVSA UT is its small size which creates great possibilities for implementation in smaller and more precise compliance requiring devices If the reader is unfamiliar with the conceptual design of the mVSA UT it is recommended to at least read chapter III of paper 6 This paper is added in appendix A 1 1 Project goal The
79. r in the code 4 unsigned int data 0 5 int shiftData 0 6 bool ans 0 7 8 void setup 9 10 11 Initializes the serial and BAUD rate to 115200 12 Serial begin 115200 13 14 Sets CS B2 MOSI B3 and SCLK B5 as output and sets MISO B4 as input 15 and CONVST BO as output and EOC B1 as input 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 DDRB B00101101 Sets EN D4 AO D5 A1 D6 and A2 D7 as output DDRD B11110000 Sets the initial values for the AD converter and sets the channel of the multiplexer to channel 0 PORTB B00000101 PORTD B00110000 This code keeps looping void loop Loop between the 8 multiplexer channels channel can be skipped using an f loop or by setting the for boundary for char swtch 1 swtch lt 8 swtch data 0 Wait till EOC is LOW while PINB amp B00000010 1 __asm__ nop n t Wait 62 5 ns Start signal CS set to low PORTB B00000001 Function for reading clocking the data from ADC to MCU for int i 0 i lt 16 i Set clock high PORTB B00100001 only record 14 bits since it is a 14 bit ADC if i lt 14 Read pin MISO into the bool ans and bitwise add t
80. ract data the following effort has been done First the average torque of both sensors is taken when actuating in the desired way as well as the average velocity The measurements are composed of two periods so the average of both periods is taken The starting points and end points of the actuation are determined and all torque values are summed up and then divided by the number of data points Second the average value of all the waiting periods is determined Since it takes some time to achieve the steady state value the average is taken by determining start points and then taking the average of a few values before these start points Then all average steady state values are again averaged giving one final value for the steady state torque Third the steady state value is subtracted from the measurement averages and the final torques and velocities are put into one table This is done for all measurements And last the final table is used to find the Pseudo Inverse of all the torques and velocities However since Coulomb s friction is also considered there should 35 also be an column which gives the direction of the velocities of both the sun and ring gear The final result is made by fitting the directions and velocities to the torque and finding the optimal values for matrices A and B in equation 5 3 shown below e d lan 7 62 A dir B vel 5 3 5 5 Results In total 64 measurements were don
81. re some safety factor has to be considered since the sensor has to be attached to the setup by hand which could cause some unwanted torques as well as misalignment This safety factor is set to 0 5N m Therefore the cantilevers should be designed such that they won t permanently deform when this torque is applied The material that is going to be used for the sensor is Aluminium alloy 6082 T6 It has a Proof Stress 0 296 of 310M Pa This is the alternative for the yield strength if this value is hard to determine The strain gauges used are given in section 3 1 These strain gauges have a maximum allowed strain of 430004777 and a suggested maximum operating strain of 3 20001 777 The linear region 0 25 linearity is between 60077 Since maximum use of the linear region gives maximum precision the aim is to let the expected strain with the torques lower than 100m Nm be within this region while taking in account the safety factor 19 0 015 0 03 m 0 0075 0 022 Figure 4 6 Von Mises Stress diagram of the sensor design when 1 Nm is applied to the outer ring while the inner ring is fixed The part is colored grey with shades The only important values are the black colored values of the cantilevers The optimal cantilever thickness has been found using FEM analysis Its cross sectional area is determined to be 2 mm x 1 mm The results of this analysis are shown in figures 4 6 4 7 and 4 8 0 0 015 0 03 m
82. rings and the fr This results in the springs being elongated to several ti their rest length Rubber springs can meet these requirem however care has to be taken in the selection of the mate The rubber should be strong enough resistant to the gr used to lubricate the actuator and keep its elastic prope V ROBUST DESIGN Even though small dimensions were a main design ob jective the actuator still needs to be mechanically robust to prevent failure The possibility of compliant actuation already implies some passive protection Impact loads can be absorbed by the internal elastic elements thus preventing damage to the actuator However the output shaft lever arm and pivot point are still in the force flow path Furthermore the output shaft hits a stop at a passive deflection of 0 45 and the mechanism is not protected through compliant actuation if it is in a stiff setting A challenge in miniaturisation is the fact that stresses in components e g bending or torsion stresses of a shaft do not scale linearly but with the third power of the diameter A balanced design was achieved by analyzing how big the expected loads on certain components are and laying out their dimensions accordingly In this way it was possible to design a robust system with small dimensions The design was guided and validated by strength calculations and finite element analyses for critical parts especially in the first level of the actuator The result
83. ross talk Possibility of reduced sen sitivity Comparing all advantages and disadvantages in table 2 1 the choice is made for the second option the dynamic torque sensor setup The accuracy of the measurement has to be precise which excludes the option to characterize the servos Although this option can be precise there are a lot of uncertainties which could give problems When comparing the second and third options the static torque sensor setup possibly generates more noise in the measurements and could give a reduced sensitivity These disadvantages outweigh the possible problem with the wires of the dynamic torque sensor setup 2 2 General requirements The general requirements towards achieving the main goal are given in section 1 1 The internal dissipation has to be determined by creating a low budget measurement system The dimension of torques to measure is unknown and therefore it is hard to give an exact requirement on the precision of the mea surement system The precision therefore has to be as high as possible with the resources available and precise enough to achieve the main goal The frictional torques to measure are expected to be within the range of 1 100Nmm 2 3 Detailed approach analysis To determine the internal dissipation of the mVSA UT it is necessary to know what kind of results to expect Since the mVSA UT is actuated by two servos connected through a differential stage the dissipation of b
84. rque 0 11 sec 60 1 5 rot sec 2 0 kg cm 0 2 Nm 9 10 11 12 13 14 15 16 17 18 R Van Ham T Sugar B Vanderborght K Hollander and D Let Compliant actuator designs IEEE Robotics and Automation M zine vol 16 no 3 pp 81 94 2009 K Hollander T Sugar and D Herring Adjustable robotics te using a Jack Spring gt in Proceedings of the IEEE Internat Conference on Rehabilitation Robotics 2005 G Carpino D Accoto M Di Palo N Tagliamonte F Sergi E Guglielmelli Design of a rotary passive viscoelastic join wearable robots in Proceedings of the IEEE International Confe on Rehabilitation Robotics 2011 Vallery J Veneman E van Asseldonk R Ekkelenkamp M van der Kooij Compliant actuation of rehabilitation rol IEEE Robotics and Automation Magazine vol 15 3 pp 6 2008 P Beyl M Van Damme Van Ham R Versluys B Vanderb and D Lefeber An exoskeleton for gait rehabilitation Prot design and control principle in Proceedings of the IEEE Internat Conference on Robotics and Automation 2008 B Vanderborght N Tsagarakis C Semini R Van Ham and D well Maccepa 2 0 Adjustable compliant actuator with stiff characteristic for energy efficient hopping in Proceedings of the International Conference on Robotics and Automation 2009 Schiavi Grioli S Sen and A Bicc
85. rque force sensor which was bought for another project This sensor the NET F T sensor with sen sor Mini40 with calibration SI 80 4 build by the ATI Industrial Automation company has a maximal measured torque of 4 Nm and has a resolution of approximately 1 4000 Nm The link to the manual is included in appendix C 27 Figure 5 1 A photograph of the complete setup that was built Figure 5 2 A figure showing how the NET F T sensor is connected to the setup for calibration The NET F T sensor is the most right part on the figure The other shaft is temporarily removed for this process Both sensors are calibrated this way 28 5 1 1 Calibration principle The NET F T sensor is read at a frequency of 200 Hertz which is done using the UDP port of the computer The torsional moment 77 is then saved to the workspace of Matlab The NET F T sensor is connected to the measurement setup as shown in figure 5 2 When applying a torque to the end of the NET F T sensor this torque is also transferred to the self built torque sensor This will give both measurements of the NET F T sensor and of the two strain gauge half bridges After applying a lot of different torques while the sensor is rotated around by the motors an least squares fit can be made between the strain gauge values and the NET F T sensor values This way position dependency can be filtered from the measurements and the readings of the strain gauge bridges can be shown in mN
86. rs possible to have high torque speed capability on the actu output and an independent control of the apparent ou stiffness which can be varied from almost zero to infinite An additional feature of the proposed mechar of the output shaft 10 which guarantees a wide range applications of the system Figure 1 shows a picture of the mVSA UT prototype The paper is organized as follows Section II presents the fundamental requirements for a miniaturized rotational variable stiffness actuator In Section IIL we describe the in novative mechanical design of the mVSA UT In Section IV we evaluate the output stiffness profile produced by the variable stiffness mechanism In section V the mechanical robustness is elucidated The internal actuators sensors and electronics are presented in Section VI Finally concluding remarks are drawn in Section VII II REQUIREMENTS The main goal of our work is to design a multipurpose compact and mechanically efficient VSA Many of the VSAs present in the literature have a limited range of output position due to their mechanical structure In order to be multipurpose the variable stiffness actuator should be capable of performing unbounded and continuous rotation This feature increases the possibilities of application of VSAs on both robotic and biomechatronic fields Moreover a compact design i e lightweight and small is an extremely important property of actuation systems for applic
87. s many different torque measurements have to be done with different combinations of velocities of both inputs of the mVSA UT Since the wires roll up slowly only a certain amount of rotations in one direction is allowed For this reason the measurements are performed in the following way First the mVSA UT is actuated constantly in the desired manner for a certain amount of time Then the actuation is stopped for the same amount of time After this the mVSA UT is actuated in exactly the opposite manner and again is stopped This cycle is done two times This has two advantages This way two measurements are done in one time and the wires won t roll up One of the measurements taken is shown in figure 5 5 as an example for all measurements The black line is the torque measured by the sensor attached to the ring gear of the mVSA UT and the gray line by the sensor of the sun gear The measured positions are shown in figure 5 6 Again the black line is for the ring and grey for the sun First in region a the desired actuation is initiated Then in region b there is a waiting period In region c the opposite actuation is performed and in region d there is again a waiting period A couple of conclusions can be made from these plots The steady state value in the waiting period is not constant There is a lot of noise especially on the sun gear e The noise has the same shape in the first movement and the third move ment and the second
88. s of the FEM analysis of the carrier output and pivot are depicted in Figure 9 The material of the components was chosen according to the required strength and light weight among other considerations like low friction or machinability In any case it was made sure that the nominal stresses in the parts do not exceed the yield strengths of the material Components with high stresses are made out of tool steel type 1 2510 with a yield strength of 400 N mm The output shaft the internal lever and all the gears are made of this steel as well as the shafts of the planet gears Because of their complex but planar geometry the gears for the prototype were spark eroded The pivot pin is made from the end of a hardened steel drill The carrier which experiences smaller stresses is made of stainless steel type 1 4305 for easier machining The lower parts of the internal rotating frame and the housing that have contact with the planet gears and the lower ring gear are made of bronze to minimize friction and to achieve good machinability Bronze offers favorable friction properties however it is also rather heavy The remaining parts of the frame and the housing are made of aluminium alloy 3 1645 for low weight high strength and good machinability All rotating parts are mounted with ball bearings apart from the ring gear of the differential stage and the planets which are mounted directly onto shafts The bearings were selected because their dim
89. sults oen ne nee ren ne ee 20 E eer 21 Schematic To be able to understand the design of the 8AMP the schematic is given below To keep it clear it has been split up in different parts with the same purpose It is a complete schematic of the board First part of the schematic is the input voltage see figure 1 This is divided in three different specific voltages Vss Vdd and Vcc Vss is the voltage that is redirected to the external Wheatstone half bridges Vcc is the voltage used for analog essential parts also referred to as analog voltage Vdd is the digital voltage used for digital components like the MCU Ground is Power and grounc LEDPOHER1 LED 6 3 Figure 1 Part of the complete schematic split in analog and digital ground As can be seen in the schematic there is alot of filtering Although it looks like alot of filtering close to eachother in the schematic this is not the case on the PCB The capacitors are tactically spread over the board to assure optimal filtering In figure 3 the connections for the holes are shown which are grounded In figure 2 all the wiring for the pads is displayed This is pretty straightforward The Sig1 to Sig8 notations are connected to the multiplexer in figure 4 8 input channels Figure 2 Part of the complete schematic The multiplexer selects a channel using 3 logic inpu
90. surements it has to be connected between the mVSA UT and the motor Therefore a setup has to be designed Important for this design is that it is designed simple but adaptable In case the motors are not strong enough it should be possible to use different motors with different sizes Transmissions between different parts of the setup can either use gears or pulleys but the design should be such that both can be used For optimal torque measurements it is desired that the transfer ratio between the sensor and the shaft going into the mVSA is big This means more torque is applied to the sensor and its rotating more slowly The transmission of the motors to the sensor should be small This allows the motors to run faster and more smoothly 21 At this point it is also important to know how the wires coming from and going to the sensor which is rotating are connected to the fixed world The three main options are using long enough wires that wrap up slowly using magnetic power transmission and using slip rings The slip rings are too expensive for this assignment and the magnetic power transmission could introduce extra noise into the measurements Therefore it is chosen to use wires that are long enough to wrap up slowly and that allow a certain amount of rotations in one direction before breaking The designed setup is shown in figure 4 9 It does not include the gears as the SolidWorks drawings for these were not available It mainly co
91. t position another sensor on the output shaft is necessary VII CONCLUSIONS In this paper we presented the novel design of the mVSA UT which realizes a compact variable stiffness actuator that can perform continuous rotation at the output shaft The mechanical structure of the mVSA UT is such that the apparent output stiffness can be varied from zero to almost infinite by changing the transmission ratio between the internal elastic elements and the actuator output namely by implementing a lever arm of variable effective length by means of a set of planetary gears The working principle of the variable stiffness mechanism has been successfully realized in a compact design thanks to the implementation of a differential configuration of the two motors The mechanical design has been presented and analysed Table I reports the preliminary specification of the mVSA UT together with the specification of the servo motors In the attached video the behavior of the system is shown TABLEI SPECIFICATIONS OF THE MVSA UT PROTOTYPE AND INTERNAL ACTUATORS mVSA UT Stall torque Nm No load speed 4 5 rad s Stiffness change duration 0 54 sec Actuated range continuous Passive range 7 4 rad 45 Weight 100 g Dimensions excl shaft 30 x 32 x 49 mm DS 001MG 4 8V No load Speed Stall Torque 0 14 sec 60 1 2 rot sec 2 5 kg cm 0 25 Nm DS 002MG 4 8V No load Speed Stall To
92. tant while measuring and the measurements are only done in a relatively short period of time The applied input torques can be found by creating a sensor that connects one side of the setup where the motor is located to the other side of the mVSA UT using small cantilevers The strain of these cantilevers can be measured by using a Wheatstone full bridge configuration of strain gauges A full bridge is used since this will give the most sensitivity To be able to measure the voltages coming from the two full bridges with high precision additional electronics are necessary These electronics have their own requirements which are described in section 3 2 2 To determine the optimal shape of the torque sensor and to fit the cantilevers to the expected range of torques to be measured finite element method FEM analysis has to be done After the shape and dimensions of the sensor are determined it is necessary to create a setup that connects the motor sensor and mVSA UT This setup should influence the measurements as little as possible and has to guarantee a high adaptability When the setup is designed and build the sensors have to be calibrated This can be done with a torque sensor which was bought for another project thus saving expenses After calibrating the setup it can be used to do the measure ments Chapter 3 Electronic part of the measurement setup This chapter will cover the strain gauges and electronics used
93. th of the I This means that for these configurations half the le of the variable stiffness mechanism cannot be used to the stiffness significantly but also that the stiffness is r sensitive to 05 in the other half Assuming a given preci of 0s then the stiffness can be more finely tuned in a n monotonous and continuous way when using configurat 6 a and 6 e Note that configuration 6 e does not reach zero stiffr because a deflection of the output always results in a flection of the lever The other configurations achieve stiffness by attaching the springs to the lever in a way they are not elongated for a certain pivot position Configuration 6 a thus shows the best output stiff characteristics Furthermore it turns out that it is not compact than the other configurations because the sj used to place the springs is not lost but used for the mo of the lever and for constructing the frame of the ring In Figure 8 the output torque and stiffness are show functions of both the output deflection 6 and of the pl carrier angle 05 Note the peak close to the singularity at 6 0 and also that not the whole plain is reachable especially area of negative stiffness is inaccessible due to mechar stops constraining the lever and output To be able to realize configuration 6 a linear spr are needed that can be stretched extremely far Because space for the springs is limited the stroke of the lever ne reaches the connection points of the sp
94. the pivot point along the lever arm realizes a more favorable design regarding the minimization of mechanical work during stiffness changes In the next Section we describe the mechanical design of the mVSA UT which fulfills the properties described above III MECHANICAL DESIGN Figure 2 shows a sectioned CAD view of the mVSA UT Its innovative mechanical design can be described in Fig 2 mechanism are 1 the output 2 the lever 3 the springs 4 the pin 5 the pivot gear 6 the planet carrier 7 the first ring gear 8 first sun gear 9 three planet gears 10 the second ring gear 11 second sun gear the motors 12 and 13 Sectioned CAD view of the mVSA UT The main parts c three main levels In the first level the variable stiff mechanism is realized It includes the output crankshaft that is connected to a lever 2 to which two linear spr 3 and a pivot pin 4 are also connected The motion ol pivot pin is realized by a set of planetary gears 5 and The second level is made of the differential mechar that actuates the output and pivot pin It includes a sec ring gear 10 planet gears 9 and the sun gear 8 w is connected to the shaft of the planet carrier 6 of the level In the third level the internal motors 12 and 13 act the differential mechanism by engaging the ring gear and
95. the ring gear is 06 and influences the apparent output stiffness The position l of the pivot pin along the diameter of the ring gear is given by d l COs 1 As the pin slides in a slot in the lever arm the motion of the lever arm is constrained to follow the relative motions of the pivot pin and the output crank at P A deflection 0 of the output shaft produces a displacement of the point p cos0 sin 97 and a rotation 05 of the lever arm The distance a between the pivot position and the output position P and the angle 05 are given by dz V d 3 12 NF cos 04 a d arcsin sin 01 2a The two pretensioned springs act on the lever like a single virtual linear spring i e the dotted spring in Figure 3 that Is connected to the reference frame at the equilibrium point Poq P BP 22 amp of the springs and to the lever arm at point P Assuming a linear zero length spring with elastic constant k the spring force on the lever with respect to 0 is given by 0 Level 2 The Differential Mechanism The second level Fig 4 mVSA UT is a set of planetary gears This mechanism is used to obt differential motion of the first ring gear 7 and the pivot gear 6 where P dz de sin 05 Let Fig be the rotation matrix of the lever arm with res to the reference frame 0 as defined by 05 Then it foll that the spring forces along the lever arm
96. the setup are commonly used and widely available Therefore they are not discussed in this report 26 Chapter 5 Measurements This chapter will cover the calibration and measurement analysis The measure ments taken up until the point of writing this report contain a lot of noise For that reason there is a section covering the analysis on what causes this noise 5 3 Although there is a lot of noise it is not necessarily only noise that is measured For that reason an algorithm to extract data from the measurements is still developed and the results are still taken as an indication for the real values Beforehand there were certain expectations considering the measurements The biggest uncertainty was the dimension of the torques that had to be measured This was estimated to be something between 1 mNm and 100 mNm which is a quite big range Furthermore the dissipation model was assumed to be linear and thus the measurements as well If actuating in one direction with a certain velocity needs a certain torque then actuating in the opposite direction should need the same torque It was also assumed that moving with a constant velocity would not necessarily give a constant torque due to vibrations in the system but ignoring high frequency peeks the torque should be approximately constant 5 1 Calibration The setup which was built is shown in figure 5 1 To do proper measurements it has to be calibrated This is done using another to
97. ts OUI IC MOU Il an enable input These are connected to the microcontroller in figure 7 The V_amp output of the multiplexer is connected to the amplifier in figure 5 To be specific to pin 3 This is then filtered by a low pass filter using the internal resistance of approximately 100 O and the capacitor of 82pF The cutoff frequency is quite high to still allow high frequency multiplexing signals 1MHz Figure 3 Part of the complete schematic Multiplexer without notably influencing the results This can be the case when the board is used to sent data to another microcontroller e g that can handle high BAUD rates A more effective low pass filter could be created when increasing the value of the capacitor but risk is this will influence the shape of the output signal of the multiplexer and thus the measurements Check the optimal solution for the application you are using The reference of the amplifier is set by the wheatstone halfbridge onboard This has to be a low noise source and for that reason it Figure 4 Part of the complete schematic Amplifier is important that that the resistors are made of thin film material This gives less Johnson Nyquist noise or also called thermal noise and thus a more stable reference voltage The gain of the amplifier is set by the resistor connected to pin 1 and pin 8 of the amplifier In the section Materials more informat
98. ts the kind of clock which is used The default setting is the internal clock of 8 MHz This manual uses a 16 MHz clock so the fuse has to be changed The oscillator given in the BOM in section materials is a full swing oscillator To set the correct value for this oscillator the SUT CKSEL should be set to one of the EXTFSTAL options It doesn t matter which one in this case For other types of oscillators check the datasheet of the ATMEGA328P Now the next step is to check whether the Device ID can still be read If so this means the clock is correctly set and soldered If not check the soldering and or capacitors connected to the external clock If the clock doesn t work you can t set or read the fuses anymore You also can t check if the clock works using an scope since this disturbs the clock signal If the device ID can still be read the fuses are all set correctly and the program can be uploaded Boot loader and lock bits To program the application on the ATmega328 there are multiple options For instance the program can be coded in AVR studio and directly uploaded but also there is the option to program the ArduinoO boot loader on the ATmega328 and use the Arduino environment to code the program The latter is explained in this manual The first step is to program the Arduino bootloader on the MCU This bootloader can be found in the software pack of Arduino which is located here To upload the Arduino bootloader go back to AVR
99. types with the same footprint Part of the BOM Notation Description Requirements Value Package Material Quantity one board optional C1 Capacitor Voltage rating 5V 18 pF SMD 0402 Ceramic 2 C2 Capacitor Voltage rating 5V 82 pF SMD 0402 Ceramic 2 Capacitor Voltage rating 5V 0 01 uF SMD 0402 Ceramic 3 C4 Capacitor Voltage rating 5V 0 1 uF SMD 0402 Ceramic 8 1 cs Capacitor Voltage rating gt SV 4 7 uF SMD 0603 Ceramic 2 Non polar preferred C6 C7 Capacitor Voltage rating 5V 10 uF SMD 0603 Ceramic 7 1 Non polar preferred CFB Capacitor Voltage rating gt SV 1uF SMD 0603 Ceramic 1 R1 Resistor Voltage rating 5V 00 SMD 0201 Thick film 4 1 R2 Resistor Voltage rating 5V 1200 SMD 0603 Thick film 0 1 Power rating 0 2 W R3 Resistor Voltage rating 5V 499 O SMD 0402 Thin film 2 R4 Resistor Voltage rating gt 5V 4 7 KQ SMD 0402 Thin film 1 R5 Resistor Voltage rating gt 5V 12 5 kQ SMD 0402 Thin film 1 L1 Coil DC current gt 0 15 A 10 uH SMD 1210 Ferrite 1 DC resistance lt 1 Q L2 Ferrite bead Free choice SMD 0402 Ferrite 4 These resistors form the half bridge to which the input voltage is matched Try to match the value of these resistors to the value of the resistors to that of your strain gages dui This resistor determines the amplification factor of the INA118 12 5 kO corresponds to an ampli
100. ulator will eventually be attached to the a Quadrotor Unmanned Aerial Vehicle The manipulator will use strain gauges to take joint torque or 6 dimensional D force torque measurements This extra application brings forth some extra requirements listed below 4 The board should have at least 6 half bridges inputs for 6D force torque measurements 5 The weight should be as low as possible 11 6 The information processing should be done on board 7 Communication protocol should be either CAN or Serial Since the board is going to be used on a flying robot the weight should not be too high Because the processing power of the processor of the manipulator has to be used for doing other calculations the data collecting and transmitting of the 8AMP has to be done by an on board processor This processor then communicates with the other processor either by CAN or Serial communication 3 2 2 Analysis To build an electronic board that can achieve the given requirements 3 2 1 the correct components have to be found The choice has been made to use a single supply voltage for the SAMP of 5 volts This voltage provides the most options in selecting components and is commonly used for this kind of electronics All the links to the data sheets of the main components can be found in appendix C To meet the requirements the choice has been made to use a multiplexer to pass up to 8 different input signals to an amplifier The amplifier am
101. urrents Using this setup the relation between the current through the motor and the output torque of the motor can be determined Then by measuring the current that passes through both the servos while attached to the mVSA UT the torques applied to the inputs can be found While this can be done the question is how accurate the measurements will be It is highly probable that the servos are not robust and better motors are needed Furthermore it is expected that a setup built to characterize motors supplying 100 Nmm of torque will probably give an inaccurate characterization for torques below 10 Nmm In other words the setup will be less adaptable In order to create the best possible setup it is therefore necessary to have a good estimation of the torques that will be applied to the mVSA Another method is to create a dynamic torque sensor setup This setup is placed in between a motor and one of the inputs of the mVSA UT Using the torque sensor as the name states the input torque applied to the mVSA UT can directly be measured The torque sensor could either be bought or designed and built The advantages of this option are that the measurement setup is highly adapt able and it is straight forward to build Depending on the properties of the torque sensor fairly great ranges of torques could be measured with high preci sion with only minor adaptations of the setup The disadvantage of this option is that the sensor rotates along with t
102. waiting again The startEndPointsRising positionArray function determines all start and end points of the rising edges assuming the rising edge takes more than 500 pts Small rising edges are ignored since the PID can create extra small rising edges For the startEndPointsFalling positionArray function the same counts only the function takes all falling edges 89 sun rising startEndPointsRising sunpos sun falling startEndPointsFalling sunpos ring rising startEndPointsRising ringpos ring falling startEndPointsFalling ringpos ER ne Ek Functionality to check whether the velocity of either the ringgear motor or the sungear motor isn t 0 since then the startEndPoints functions 4 cannot find any values for when to measure and when not In this case the 4 measurepoints are taken from the other motor if numel sun rising lt numel ring rising sun rising ring rising sun falling ring falling end if numel ring rising lt numel sun rising ring rising sun rising ring falling sun falling end Split 2 dimensional arrays obtained from the startEndPointsX functions 4 into seperate more recognizable parameters Just cosmetic adaption ring risingStart ring rising 1 ring risingEnd ring rising 2 ring fallingStart ring falling 1 ring fallingEnd ring falling 2 sun risingStart sun rising 1 sun risingEnd sun rising 2 sun fallingStart sun falling
103. xploded view of subsystem 2 of the measurement setup 24 Subsystem 2 is the part that contains the sensor Its exploded view is shown in figure 4 12 All parts shown are listed below e Parts 3 11 13 These are angular contact bearings which allow better damping of non torsion bending moments e Parts 2 12 These rings allow some space between the angular contact bearings which increases the damping of non torsion bending moments even more e Part 4 This shaft is connected to the inner circle of the sensor It is hollow such that it can guide the wires of the strain gauges e Part 5 The bottom plate of the sensor holder This bottom plate is con nected to parts 6 and 9 allowing the sensor and shafts to stay assembled while making it move able from the complete bottom plate which holds all parts shown in figure 4 9 This way when changing gears or pulleys the sensor does not necessarily have to be calibrated again e Parts 6 9 Bar holders These parts contain holes for the bearings that carry the shafts of the sensor These holes have to be as aligned as possible It does not have to be perfect since the misalignment can be filtered from the measurements but if the alignment is too much off it will give big perpendicular forces on the sensor in combination with the angular contact bearings e Part 7 This is the sensor of which the design is discussed in section 4 1 2 Its inner ring is connected to part 4 the outer rin

PDF - University of Twente Student Theses

Contents

Download Pdf Manuals

Related Search

Related Contents