Torque ripple problem on the HV PFC Kit

Hi Danny

I am assuming that you dont use PFC section, but motor inverter only. Your problem might be current sense quality. Here are some suggestions to improve HVPFC kit performance:

- You can replace the 5mohm shunts with 10 (+-10A) or 20(+-5A) depending on you maximum peak current. You dont need to modify the code as long as you use sensor feedback. This will help to increase the resolution.

-Remove caps ( C16,17,18,19 ) around the current sense opamp ccts.

-Kelvin connections for the shunts, R10 & R11, will improve the current sensing accuracy. Unfortunately, the impedance of the PCB traces can cause up to 15 percent error on HVPFC kit. The trace impedances are comparible with the 5mohm shunts. Use blue wires to minimize the impedance between shunt resistor terminals to opamp cct. R11top terminal to R13, R11 bottom to R14. And the same modification for R10 terminals to by pass the traces on the PCB. Increasing the shunt values will directly contribute to get rid of this current sense linearity issue. Compare the current signal at the opamp output with a current probe output, and the current signal on your own hardware to make sure that they are all consistent. Are you using shunts, or current sensors on your design?

Thanks.  

Bill

Dear Bill,

Thank you for your reply, it makes a lot of sense, we will try all your suggestions.

This is not just an academic issue about the kit’s performance, it is very important for us to succeed because we redesigned our smooth torque drive to a new design based on TI’s kit schematics to reduce cost and we get the same problems.

Our product must have smooth torque.

Our older own design (that works well) uses shunt resistors in series with the motor phases with isolated linear current sensing ICs that convert and bring the expected voltage range to the DSP ADC inputs.

Regards,

Danny

Hi Bill,

We tried all your recommendations. The only thing that helped partially was increasing the value of the sampling resistors. 10 milli-ohms reduced the ripple by about 50% but for our application this is no solution because we need higher currents and torque ripple is still too high.

I tried few methods for compensating in software but with no success so far.

One more thing:  torque ripple is more noticeable at low speeds (used frequently in our application).

Can you think of anything else that may help solve this problem?

Regards,

Danny

Danny,

Your finding explains that the issue partially depends on the current measurement quality. I had similar experiences earlier where the current measurment resolution directly effects the performance. Unfortunately low side current sensing can cause these kinds of disadvantage when compared to LEM based apps or high side sensing.

Can you compare the current waveforms of both systems:

1- at open loop, should be level 2

2- only current loops are closed should be level 3

I just wondered whether the control loops have any contribution to your issue or not, and where exactly the problem starts.

Since the feedback ccts on two different hw are different from each other, the same controller gains wont work identical for both systems. Make sure that you tried different Kp, Ki coefficients for current loops.

How do you energize the kit, using a DC supply or from AC mains?

Regards

Bill

One more hw modification suggestion. Can you remove the dc bus return shunt resistor (R30), if you are not using it. Just short it with a piece of copper. This can cause common mode issues  for some applications.

Regards

Hi Danny,

as the torque ripple is an even harmonic and not a multiple of three I would guess that your current sensing is not equal for both channels. The easiest test would be if you put your drive in torque (or even better just in current control mode) supply a reference 0 for q axis and a couple of amps in d axis. Then just monitor DC-link current while manually rotating the drive for one electrical revolution. If my assumptions are correct DC-link current should have to maximums and minimums.

Regards, Mitja

Hi Bill,

1) The current waveforms in both systems seem to be identical.

2) We have played a lot with the Kp and Ki parameters of both q and d PIDs. We can't get rid of the ripple in any gains combination. The ripple is more noticeable with higher q gains.

3) The kit is energized from the AC mains.

4) Shorting R30 was one of the first things we tried, this had no effect on the torque ripple.

Regards,

Danny

Thank you Mitja for your suggestion.

We will try it next week.

But what do you think could cause the current difference between the 2 channels?  We experience this problem in TI's HV Kit and also in a couple of our own boards that were laid out and manufactured according to the HV kit schematics.
We verified all component values of both channels are identical.

Regards.

Danny

Danny,

Good to hear that your currents looks identical. In this case I suppose we dont have to worry about possible measurement   issues.

Since you are having ripples 2 per electrical cycle, this might be cause by the rectified bus voltage. Can you monitor bus voltage through the test point Vfb-Bus using a scope in AC mode. If the ripple on the bus voltage is significant, can run the system using DC power supply.

Are you using a low inductance motor? What is the value of stator inductance?

Regards

Bill

Danny,

from the email you sent me you verified proper SENSORED operation with

- your custom board design with low side current shunt

- modified version of TI HV board with phase current hall sensors

But very bad torque ripple with

- modified version of TI HV board with low side current shunts

And you verified that

your application does not use a speed loop. It only uses the iq, id loops.

you did try with speed loop and in position loops and it runs fine. Apparently in a common scenario where these loops can be used there is no apparent difference between the 2 current sensing methods (low side shunt sampling and phase sampling).

This last point is very strange to me. That simply by adding a slower speed loop to control the Iq reference of the current controller you are able to find stability?

What are you using for an Iq and Id reference for your torque controller?

Hi Chris,

1) I use Id.Ref = 0 (its a PM motor) and for the tests Iq.Ref is a constant starting with 0.1,    (changed from the debugger to 0.2,  up to  about 0.4) The higher the number the larger the unwanted ripple.

2) For the experiment I left speed and position + speed loops at 10 Khz (so they are not slower).

Danny