Overcurrent trips at low speed [SOLVED]

[manny]

As my E saxo project is nearing completion. I am doing some test drives. It drives pretty good.

Saxo Project:
viewtopic.php?t=4837

Volvo V60 motor:
viewtopic.php?t=3859

Volvo XC90 inverter:
viewtopic.php?t=3675

But when pulling away from a standstill sometime the inverter trips with a overcurrent error. If I get some speed than I can floor the pedal without a problem.

Have played with the iqkp and idkp values did not make a notable difference. With the handbrake on the Iq and Id are nice and stable.

With this being a OI board I made myself it could be loads of different problems.

Would like some comments on things to check/try before I remove the inverter to scope the current sensors to check the signal for noise or something. And check the reference voltages for the overcurrent protection circuit.

Any help is greatly appreciated.

[P.S.Mangelsdorf]

Is this occurring with low throttle input or high throttle input?

On the Tesla LDU, we had lots of issues with low speed, low throttle which were solved by the introduction of fweakstrt, which is a higher value of fweak for low speed, low throttle. But that's on the sine firmware, I have no idea how to apply that info to the FOC firmware.

[johu]

First I would lower throtramp. Between 3 and 10. Gives the PI loops more time to settle. Make regenramp < 1

You can also try lowering curki.

That's pretty much all you can play with.

Does it actually OVERCURRENT trip or is it some fault signal from the gate driver? The latter could be from over temperature. When the inverter is conducting pretty much DC without handing off current from one leg to the next, the "most active" IGBT at a given motor angle heats up faster

[manny]

Is this occurring with low throttle input or high throttle input?

Mosty with high throttle, but it also happened when carefully backing out of the garage.

First I would lower throtramp. Between 3 and 10. Gives the PI loops more time to settle. Make regenramp < 1

You can also try lowering curki.

Lowering throtramp did not do much. Lowering curki made it better.

Took the inverter apart to check the voltages at the ocurlim comparator.

• current sensors output 2.5V at 0A
• current sensors scaled 1.47V
• low threshold 0.587V
• high threshold 2.352V

• Delta low 0.883V
• Delta high 0.882V

So 0.882 /3.3 = 0.267 x 4095 / 2.4375(il1gain) = 449A
That is doing exactly what is should

When going through the schematic I found the reason why the udc whas stuck at 398V. thought is whas the limit from the inverter. Dumb in hindsight . With udcgain 10.28 @ 420V = 4317 digits, the input the adc input is overloaded. So I added a resistor to make a voltage divider with the 10k resistor in the input filter. Have seen overloaded inputs affect other adc inputs on microcontrollers. but it is not that much.

When taking the inverter apart I found out that my contactors are welded shut. So not able to plug that back in.
need to fix that first to continue testing. Maybe I can rig up a power supply to check the current sensor signal with the scope.

[manny]

So took the inverter apart. Luckily I did not need to drain the cooling system.

There seems to be quite a lot of switch ringing on the current waveform. 150mV equates to 45A

measurement setup

Red: current sensor signal Yellow: is ch3 PWM (for timing)

Overview

zoomed in

after voltage divider

Have tried adding a capacitor before the voltage divider/input filter (as on the first picture). That didn't do much.

Thinking about adding a chip bead ferrite inline. Need to find one with that works on the 26MHz most are 100MHz or higher

Don't think that I can do something about the origin of the noise.

[manny]

Finally got the car back together. After fixing the stuck contractors and a BMS issue.

Before that i got the chip bead ferrite installed on the inverter board. Went for the MMZ1608S202ATD25 from TDK.

After some initial testing it looked to good to be true so I soldered a small wire across the bead, So I can measure in one setup, without and with the bead ferrite active. By simple cutting the wire.

So it suppressed the noise a lot. So I was hopeful that this solved the issue but sadly it did not. Still happy I "fixed" this.

When I was looking at the wiki I saw this.

ocurlim: Hardware over current limit. RMS-current times sqrt(2) + some slack.

The throtcur sets the Amps per throttle % so my 3.5 will result in 350A max, Is this peak or RMS?

If this is RMS then 350 x sqrt(2) = 495A Then my setting of 450A for the over current is to low.
Then I need to go to 530A or so. The current sensors in the volvo inverter are 600A max so that's fine

[johu]

To be honest I'm not sure if throtcur shows as rms or peak
I run throtcur 5.5 and ocurlim 900 with leaf inverter

[manny]

Will check with the scope. Put the throtcur to .5 or so.

[manny]

Did some testing in the garage because the weather is not compatible with driving with the hood up and a scope in there.

Set the throtcur to 0.5 and the I can go full throttle and slowly accelerate and then break before smashing in to the work bench.

So throtcur at 0.5 at 100% throttle 50A produces 107 peak to peak(probably, the current clamp is a cheap one. Hantek CC-650)
After looking at this I remembered that I have the MTPA parameters set so this may change things.

Took a plot on the web interface.

So at 50A throttle Iq is 50A and there is some Id current resulting in the 107Ap/p on the scope.

To answer my own question:

Is this peak or RMS?

~Peak, but with some Id current. peak = sqrt(Iq²+Id²)

This all seems fine. Hopefully the weather will cooperate this weekend to do some testing on the driveway and get a scope shot when the overcurrent occurs. To see if there is something strange going on or if it's just noise/trigger happy overcurrent comparator.

[manny]

♫The sun is shining and scope is out♫

Went out for testing today. Tested every phase with the current clamp.

no trip

trip

So this is definitely not a problem with the OI software it seems to have perfect control over the phase currents. This also means that the current sensor signal is also good.

Also bump up the ocurlim to 600A(sensor max). The response stayed the same. So it must be something else. Maybe signal coupling in on the track between the comparator and the nand gate(will check the pullup on this track). Maybe I will disconnect the current sensor signal from the comparator and replace it with a 1.65V from a voltage divider. To test this.

It is difficult to measure the current sensor signal on the controller. Because the PCB is on the backside of the inverter near the firewall

[manny]

Good news slightly bad news.

IT IS FIXED but I am not 100% on the solution.

I have been a bad scientist and changed 2 variables between tests

replaced the 10k pullup with 1.2k to get ~max 4mA as the datasheet from the comparator suggests(4mA only when overcurrent is active).
and added a 100nf capacitor near the VCC pin for the comparator.

so which one ¯\_(ツ)_/¯

[manny]

Could not leave it like that.

Took the PCB out again and put the 10k resistor back. Got 1 trip in 5 launches.

put the 1.2k back in and got 1 trip in 10 launches.

So it not completely fixed but this is the problem area.

[manny]

Did some more testing today.

Measured the current sensor signals in the car. Did not do this before because, its difficult to measure the signals because the PCB is in a awkward place. Attached two small wires to the scope probe so that I can solder the in place.

This signal confirms the current clamp measurements. There is no overcurrent.

To check the output of the comparator added a second probe.

So there is no change on the comparator output. To verify this I used the comparator output as the trigger for the scope. floored the car, it tripped and the scope did not.

So what else can cause a overcurrent fault? Have been locking at the code on github.

Code: Select all

extern "C" void tim1_brk_isr(void)
{
   if (!DigIo::desat_in.Get() && hwRev != HW_REV1 && hwRev != HW_BLUEPILL)
      ErrorMessage::Post(ERR_DESAT);
   else if (!DigIo::emcystop_in.Get() && hwRev != HW_REV3)
      ErrorMessage::Post(ERR_EMCYSTOP);
   else if (!DigIo::mprot_in.Get() && hwRev != HW_BLUEPILL)
      ErrorMessage::Post(ERR_MPROT);
   else //if (ocur || hwRev == HW_REV1)
      ErrorMessage::Post(ERR_OVERCURRENT);

   timer_disable_irq(PWM_TIMER, TIM_DIER_BIE);
   Param::SetInt(Param::opmode, MOD_OFF);
   DigIo::err_out.Set();
   tripped = true;
}

So if the interrupt is triggered and no input is active than the overcurrent is set.

PS:
Found a small error in the OI mainboard. The pull up on the comparator output is to 5V but this pin on the STM32(pa1) is not 5V tolerant. the Desat pin is also pulled up to 5V but it is 5V tolerant. Changed the pull up on the comparator output to 3v3 did not change my problem.

[manny]

After finding that the overcurrent error could be one of the 3 inputs to the NAND gate if the pulse is short enough.

I soldered the scope probe to the track between the NAND gate output and the TIM1_BKIN input on the STM32.

red=current sensor yellow=NAND

Zoomed

So the interrupt pin is triggered by the NAND gate. Next I soldered the scope to capacitor on the desat track.

Zoomed

The datasheet for the NC7SZ10 gives a LOW Level Input Voltage of 0.3 VCC. That is 0.99V at VCC 3.3V
the scope got to 1.4V so in the inverter it whas even lower.

Added one of the chip bead ferrite to the track.

This seems to have fixed it