TMS320F28027F: PMSM motor identification for low inductance motor using 28027F microcontroller

Note that the table of identified parameters posted by James M. is for the same project and board we're testing. The inductances are consistency ~1e-13 using lab2c. The motor appears to be spinning well with a reasonable current draw, and the identified flux and resistances appear to be correct.

We updated our values as suggested and ran lab2c 3 times with the following results:

The Lsd and Lsq seem very low. I would suspect we would get an error if we tried to use those values. What is involved in measuring those more accurately? 

The current is not enough yet to run motor smoothly, you should increase continuously to run motor again.

Used the following motor params:

Got the following results from Lab2c:

Lsd and Lsq seem way too low. How do we increase the test current to get better results? Thanks for your help.

I have attached some pictures of the phase A current from the data logger while running Lab2c. The current seems choppier than I would expect. Does this look correct?

Good afternoon Yanming,

Jim has attached three graphs showing the measured IA current, which has been recorded by the 28027F processor, for the ramp up and flux estimation states. I see there are some cases where the current jumps. Can you verify this is the expected behavior for these two states?

Best regards,

Chris

Check out the inductance measurement channel A current waveform.

Also check out the waveform during the Rampup phase

Do these look right?

Any suggestions from the TI community? Yanming? We've tried increasing the Ls estimation current but have not had any success. What are you recommended next steps for troubleshooting? Do the graphs of Ia look correct for the different estimation states?

Thanks,

Chris

Good evening Chris,

Good to hear from you again. You've been a huge help for us in the past with our previous motor controller project. Here's a copy of the user.h file. We've tried experimenting with several different USER_MOTOR_IND_EST_CURRENT and USER_MOTOR_FLUX_EST_FREQ_Hz settings, but have not been able to achieve a successful inductance estimate. Do the Ia graphs in the previous posts look correct to you? What do you recommend for the next debugging steps?

Best regards,

Chris

75774.user.h

6327.user.hI assume this is a completely custom board.

Have you checked it out to insure proper pin mapping, current polarity, etc?

Have you ID'd and run any other "standard motors"

MotorWare v18 has proj_lab01 to help with hardware verification.

Assuming all HW/SW settings are good...

I don't see a bunch of changes that need to be made, but you can use this user.h

Try motor ID again using proj_lab02c

report the Rhf, Lhf, and Rs

does the motor start spinning and ramp to 80 Hz = 2400 RPM?

if it doesn't start up increase

USER_MOTOR_RES_EST_CURRENT      (5.0)

report the flux value

does the motor keep spinning at 2400 RPM for the duration of the ID process?

if it doesn't, increase

USER_MOTOR_RES_EST_CURRENT      (5.0)

You seem to be be graphing 50 samples wide blocks at 1 Hz into a 400 sample graph. This will give you roughly 8 segments of unrelated data in your graph, I'm betting this is not what you intended. Check the configuration.

Also a flux of 0.003 VpHz x 100 Hz gives ~ 0.3V. On an ADC_FULL_SCALE_VOLTAGE of 22.53V, that is just 1.3% of your ADC scale, 54 counts in 4095. That is not impossible to measure, but it is quite low and I would not expect a noisy board to be able to measure this.

At the full 14 volts, this motor would run up to (roughly) 140.000 RPM (14 / 0.003 / 2 * 60) which seems rather high. It seems to me this motor and this board are not a very good fit.

Good afternoon Chris,

Thank you for the suggestions. I took a step back and connected a more standard inductance motor, an Anaheim Automation BLY17 Series motor, but am still not achieving consistent identification. It seems to work perhaps 50% of the time. This motor is an 8-pole,  1.5Ohm, 2.1mH, and 2.7V/kRPM motor. The values I'm finding after a successful identification are 1.3 Ohms, 1.3mH, and 0.025V/Hz, which seem reasonable. But there seem to be two problems

1) After solving for these parameters and re-setting Flag_RunIdentify = 1, the motor is not controlled well. There is lots of vibration and it becomes unstable at roughly 200RPM.

2) It only solves for these parameters roughly 50% of the time. In other cases, the identified Ls is 1e-9H.

I've modified the datalog to store 150 values x 3 phases (this seemed to be the limit for the 28027F chipset), and recorded a few sample images of the current during the Rs, Flux, and Ls identification stages. It doesn't appear to match the sine wave shown in the lab documentation, but perhaps I'm not recording enough samples. Any suggestions for adjusting the graph settings to help troubleshoot?

Any additional thoughts on why it may be failing for this new motor? I did go back and run labs 1b and 1c and I can spin the motor in both cases. I've attached my latest user.h for review.

Rs identification

Flux identification

Ls identification

7360.user.h

Thanks again for your support!

Chris

Good afternoon Chris,

Thanks again for reviewing our case. After a thorough hardware review, we found that we were using a 2.048 Vref signal, but the DRV8323 requires Vref > 3 volts. Once we made this change, we're obtaining much more consistent current and voltage offsets. Despite this improvement in the offset biases, I'm still seeing the same issue with inductances being measured as 1e-9H perhaps 50% of the time. Note we are using Kelvin ground connections. What are you next suggested steps for troubleshooting? I've attached my most recent user.h for the Anaheim motor.5270.user.h

//! \brief ADC current offsets for A, B, and C phases

//! \brief One-time hardware dependent, though the calibration can be done at run-time as well
//! \brief After initial board calibration these values should be updated for your specific hardware so they are available after compile in the binary to be loaded to the controller
#define I_A_offset (0.973623335)
#define I_B_offset (0.975451648)
#define I_C_offset (0.989299715)

//! \brief ADC voltage offsets for A, B, and C phases
//! \brief One-time hardware dependent, though the calibration can be done at run-time as well
//! \brief After initial board calibration these values should be updated for your specific hardware so they are available after compile in the binary to be loaded to the controller
#define V_A_offset (0.318499684)
#define V_B_offset (0.318760693)
#define V_C_offset (0.316394270)

All the best,

Chris

The motor spins in lab1c but I'm not sure what to expect from Ia current measurements. I've set up my data logger using

datalog.iptr[0] = &gAdcData.I.value[0];   // datalogBuff[0]

datalog.iptr[1] = &gAdcData.I.value[1];   // datalogBuff[1]

datalog.iptr[2] = &gAdcData.I.value[2];   // datalogBuff[2]

and set

gMotorVars.IqSet_A = _IQ(0.5*USER_MOTOR_MAX_CURRENT);

but I'm only seeing the current in my three graphs reach perhaps 0.1 Amps (where it should be half of max current or 0.9 Amps). Also note that the current from my power supply is measuring 0.5 Amps, but this is not the individual phase current.

Would you expect &gAdcData.I.value[0] to reach IqSet_A value of 0.9 Amps?

Also I found in the DRV8323 datasheet that the Isense / SOx outputs appear to be inverted. Where can I configured my current sense in hal.c to be inverted?

The values of inductance indicated are nearly accurate from my experience. Try defining the motor parameters if have any doubts.

Did you have any current waveform on oscilloscope for identification to show the issue? It seems that the motor didn't run smoothly. What's the rated current of the motor? You can set the USER_MOTOR_RES_EST_CURRENT and USER_MOTOR_IND_EST_CURRENT to 20%~40% of rated current to ensure the motor run smoothly.