Porsche Boxster with Siemens and DMOC

[arber333]

Ah! You use Melexis chips. Hm... i understand why Johannes mentioned you should turn them.
Definitely take them out first and leave them out of the box for quick test. ACIM brain will not protest if they will not move from centerline, but will definitely protest if it doesnt see them connected.
Try to run the motor and see if they are to blame. There could be a cold solder joint somewhere on the board also...
BTW what is voltage for sense lines when inverter is started? It needs to see 1.6xV.

[johu]

Yes you can try disconnecting them, but be aware that there will be no overload protection what so ever when you do this. Even a too steep increase/decrease in open loop frequency could blow up the power stage. So be very gentle.

I don't see any spikes on the plot, maybe they are so short that they are filtered out.
I have used these Melexis chips at a couple of 100A without trips, but they are the most likely culprit. Try swapping in your LEMs with a 3k3/4k7 voltage divider (just put 4k7 across C6 and C7).

[arber333]

Try swapping in your LEMs with a 3k3/4k7 voltage divider (just put 4k7 across C6 and C7).

Why 4K7? I use 6K8 with 3K3 and it works exactly to 1.6V.

[johu]

Yes thats correct, 6k8 is even better. Whatever you find in your parts bin. The inverter will remove the offset either way.

[Isaac96]

BTW what is voltage for sense lines when inverter is started? It needs to see 1.6xV.

Hmm... Sense voltage is 1.804v. Supply is 4.99v.

Isn't it supposed to be at 1.65v at idle?

Also overcurrent tripped even with sensors out of the inverter....

[joromy]

I had the same problems with my Ranger/Siemens inverter, when i upgraded to newest firmware.
Was ok again with ver 4.12

You could also try this one, from "nailgg" before FOC implementation.

viewtopic.php?f=2&t=464&p=8250#p8233

[Isaac96]

I just hooked up the LEM sensors. I have 6.6k of resistance across C5 and C7 (the caps to ground on the comparator inputs). Resulting voltage is 1.658v and 1.661v.

On the negative side of the comparator (pins 4 to 7) the PWM output from the STM32 is 1.007v.

On the positive side (pins 8-11) PWM output voltage is 1.562v.

Thus the overcurrent is tripping immediately.
This is with ocurlim set to 500; with ocurlim at 100 the positive voltage is 1.342v and negative is 1.254.

This does not seem right at all to me...

MAJOR EDIT

So I decided to see if the stm32 was doing okay. It is heating up quite a bit so I am going to call it dead....
This is the second time this has happened to me! Probably because I added a pullup resistor on UVLO (it was a 10k resistor to 5v, just like the DESAT pullup)...

I've already got an ST-LINK, or rather a STM32 Nucleo which has ST-LINK built into the board. I used this last time to reprogram and it worked nicely.

What's the fastest source for the STM32 board? Willing to pay a bit extra, I'm on Easter break at the moment.

Edit 2: Ordered STM32-H103 from Mouser. Should be here by friday with a bit of luck. I guess now I have time to work on the cooling system and properly install the current sensors.
What do y'all use for coolant reservoirs? Also should I combine the motor and inverter on one loop or keep them separate? I have enough pumps and radiators to separate them.

[johu]

Hmm, if you can still check can you scope the PWM that STM32 outputs before R16 and R20? Like does it reach 3V3?

I don't think a 10k pull-up causes that kind of damage, but a hot MCU is definitely alarming.

[Isaac96]

PWM outputs still look good. Chances are I fried the ADC (which is multiplexed, right?) and thus it's reading weird things on the current sensors, resulting in weird outputs.
My replacement board should be here tomorrow so I will get back to testing then.

[arber333]

What do y'all use for coolant reservoirs? Also should I combine the motor and inverter on one loop or keep them separate? I have enough pumps and radiators to separate them.

I look for scavenged reservoirs from various cars. I like VW bulbs, because they include level sensor that i can use to sense if there is some leak. It is also a nice round shape that you can put anywhere...

Well i would also use single loop with single pump that would also go through the charger since charger will usually work when motor will be off .
But my EVTV DUE will control fuel pump relay when CP signal will be sensed.
Main point is to keep inverter first in loop after the radiator and only then motor. Also you will need accurate sensor inside inverter so it will reduce power if your loop got too hot. In any case you can also keep a stainless steel tube container with 19mm hose fittings under the car as a buffer to have more thermal mass. For your range i dont think there will be much problems. What kind of radiator will you use?

[johu]

PWM outputs still look good. Chances are I fried the ADC (which is multiplexed, right?) and thus it's reading weird things on the current sensors, resulting in weird outputs.
My replacement board should be here tomorrow so I will get back to testing then.

It's very unlikely that you fried a subsystem that selectively. I'd keep looking outside the chip. You could use a trimmer to generate a variable voltage 0-3.3, set it to 1.65, set ocurlim to its max and start the inverter. Then start adjusting the trimmer and see if the voltage translates to current as expected and does not saturate unexpectedly.

[Isaac96]

I look for scavenged reservoirs from various cars. I like VW bulbs, because they include level sensor that i can use to sense if there is some leak. It is also a nice round shape that you can put anywhere...
<snip>

What kind of radiator will you use?

I'm probably going to end up using Nalgene water bottles because there aren't any junkyards around here and most are closed anyways.
They are autoclave-rated, meaning they can survive 120 degrees celsius (made of polycarbonate).

The motor radiator is one of the three original Porsche ones.

It's about 3ft long so plenty of capacity.

I planned to use the OEM heater core for inverter cooling; it's about 12" by 8" and about 1.5" thick. I think it can dispose of the inverter heat effectively.

But having a single cooling loop would probably simplify the system quite a bit... I'm just afraid of the motor dumping heat into the inverter.

You could use a trimmer to generate a variable voltage 0-3.3, set it to 1.65, set ocurlim to its max and start the inverter. Then start adjusting the trimmer and see if the voltage translates to current as expected and does not saturate unexpectedly.

As it stands, with the 1.661v input from my LEM current sensors, the overcurrent trips as soon as the inverter is powered up. So I can't even get a current reading. The negative side PWM to comparators seems to work correctly, but the positive side PWM doesn't. I will test as you suggest and see how it behaves.

[arber333]

Hm... What does the PA0 input show? I had a lot of problems because of this pin draining back into the chip and it took the NAND gate low. I think pullup inside olimex was shot. When i cut that trace and pulled the NAND up by separate 10K resistor i could use the chip other pins.
I gather PA0 is connected to reset button pullup? NAND chip needs to see 5V for it to output low on PB12.

Also try to see if you get above 1.5V voltage on other pins like PA1, PA3, PC7, PC9.
Maybe the problem is in the LM2901 chip...

The motor and inverter will be fine if you keep enough water in the system. Really the problem comes when driving from a stop a lot, like in the city center on the crossroads etc... There inverter sees the most amps, so you need more mass to hold the heat of the motor and release it inside the radiator.
I used heater as radiator before and i dont have good experience. They have a wire spiral inside that can get blocked by the grease in glycol and they can seize up the flow.
I heared some guys used two intercooler (air) radiators one on each intake under front lights... Great flow, small dT.
I also used computer radiators (240cm x 120cm) but despite two fans it wasnt enough dT, i needed to add another one.
Now in Peugeot i am using cooling radiator from a Fiat Uno car, which has its own airing tank on the side. It is nicely proportional to the free space i got left.
I kept original AC radiator and fans in front so i will be able to experiment with AC later.

[Isaac96]

Hm... What does the PA0 input show? I had a lot of problems because of this pin draining back into the chip and it took the NAND gate low. I think pullup inside olimex was shot. When i cut that trace and pulled the NAND up by separate 10K resistor i could use the chip other pins.
I gather PA0 is connected to reset button pullup? NAND chip needs to see 5V for it to output low on PB12.

I've been looking at the schematic on the wiki and it shows nothing connected to PA0. I was getting overcurrent errors and so I thought that it might be going low due to random transients. Pulling it up solved nothing, and after that I noticed that the MCU was getting hot (too hot to touch).
I'd rather not test anymore with this MCU because I don't want to break other components.
The NAND chip was getting 2.8v from PA0 and from PC9 (the DESAT pin). PC9 had a 10k pullup to 5v, and PA0 had no pullup at all, probably an internal one on STM32.

LM2901 seems fine, it's outputting the right signals based on what it's being fed by current sensors and STM PWM signals. The PWM signals are just wrong.

The motor and inverter will be fine if you keep enough water in the system. Really the problem comes when driving from a stop a lot, like in the city center on the crossroads etc... There inverter sees the most amps, so you need more mass to hold the heat of the motor and release it inside the radiator.
I used heater as radiator before and i dont have good experience. They have a wire spiral inside that can get blocked by the grease in glycol and they can seize up the flow.
I heared some guys used two intercooler (air) radiators one on each intake under front lights... Great flow, small dT.
I also used computer radiators (240cm x 120cm) but despite two fans it wasnt enough dT, i needed to add another one.
Now in Peugeot i am using cooling radiator from a Fiat Uno car, which has its own airing tank on the side. It is nicely proportional to the free space i got left.
I kept original AC radiator and fans in front so i will be able to experiment with AC later.

Okay I'll skip the heater and just use the single loop, with the inverter ahead of the motor as you suggested.

The radiator I have now is mounted on the back of the transaxle near the rear bumper, so it will be in the airflow when driving faster. I'll also add some ductwork to get good flow at lower speeds. Should it have fans on it? At high speeds the fans might just break, or at least block some air, and I plan to mostly drive this car on the highway at 65-70mph. I'll usually only drive 30 miles at a time.

[arber333]

Hm... You can use a larger radiator without fan, since all consumer fans other than automotive are useless under hood. Maybe plan ahead to mount small heater/radiator with couple 120mm fans in loop in some dry space if you should need one. I got good industrial 24V fans. High revs and vibration resistant. I have 12v to 24v converter to run them...

[johu]

PA0 is pulled up on the header board as it is connected to the button (the larger one, not reset).
It was once intended to detect an under voltage lockout fault from advanced gate drivers but was never used as such. Just don't touch it.

The NAND tree does not need 5V on all inputs to output 0 - its threshold voltage is more like 2V. This goes for the chip I provide with the kit, other models have a >3V threshold - maybe thats what happened to Arber.

The symptoms you describe still lead me to think there is a bad solder joint on the path current sensor -> stm32. You cold test for that by probing in close proximity to the MCU and see if your test voltage shows up there. Also when setting ocurlim very high (65535) you should see 3.3V on the comparator upper channels. If not: loose solder joint there or as last resort, indeed a sick MCU.

[Isaac96]

PA0 is pulled up on the header board as it is connected to the button (the larger one, not reset).
It was once intended to detect an under voltage lockout fault from advanced gate drivers but was never used as such. Just don't touch it.

The NAND tree does not need 5V on all inputs to output 0 - its threshold voltage is more like 2V. This goes for the chip I provide with the kit, other models have a >3V threshold - maybe thats what happened to Arber.

The symptoms you describe still lead me to think there is a bad solder joint on the path current sensor -> stm32. You cold test for that by probing in close proximity to the MCU and see if your test voltage shows up there. Also when setting ocurlim very high (65535) you should see 3.3V on the comparator upper channels. If not: loose solder joint there or as last resort, indeed a sick MCU.

I just checked the board some more. On PA5 and PB0 I'm getting the sensor voltage of 1.661v and 1.652v. These also show up at the comparator.

When ocurlim is set to 65535 then PB7 goes to 0v but PB8 only reaches 2.415v. This is measured directly on the Olimex breakout board. So I believe the MCU is partially dead.

My new one should arrive in a few hours (thanks to Mouser!) and I will test more with that board. I'll have to relearn how to flash with the ST-Link... I've got all the software and hardware already.

[Isaac96]

New STM32 board arrived today, the comparator is working great now! So I guess there was some sort of ADC issue.

In manual mode without encoder I can spin the motor beautifully. The current readings, however, are much higher than my clamp meter reports - 60A claimed by inverter, and 32 from clamp ammeter. I still suspect noise in the current sensor wires. I just switched the wires to twisted pair shielded wires which are grounded at the control board, total length is 9". They run perpendicular to 2 phase wires. Anyways manual mode seems okay, besides the sensor noise.

But when the encoder is connected strange things begin to happen. No matter the settings, as soon as I set ampnom high enough to move the motor it clunks and claims overcurrent. This occurs both in start mode 1 and start mode 2.

My encoder is AB type, open collector, and requiring a 5-20v supply.
When building the board I populated everything except J1. R7 has a jumper across it and is feeding the pedal and encoder.

I just scoped the outputs of the encoder itself and they look correct. Past R2 and R5 (which feed the signal to the STM32) they also look good.

In regular "Run" mode (slip control), the 'angle' variable begins to move around. When the pedal is in the regen area the angle goes backwards, and when the pedal is in the acceleration area the angle goes forwards.

If I rotate the wheels by hand I can counteract this and get the angle to remain steady.

And lastly, in slip control mode, the motor will not do anything until I reach potnom of 60%. Then it hits overcurrent and dies. The 'amp' variable also stays zero until 60% potnom. I think this is linked to the angle weirdness somehow.

Any ideas? I've been about tearing my hair out the last 2 weeks with this...

-Isaac

[arber333]

Is the output open collector or push pull? If it is open collector you can supply encoder with 12v power. Maybe the chip on encoder is on the limit of brownout at less than 5V.

Also where exactly do you power encoder from maybe you drain power from somewhere you shouldn't.

[Isaac96]

I'm pretty sure the encoder is open collector, just because of its wide voltage range. I'll throw 12v at it and see what happens.

The encoder is now being powered from the ENC_IN_5V pin on the control board; R7 (which is intended to supply power for an IR diode there) has been bridged as advised in the Rev2 schematic.

[johu]

And lastly, in slip control mode, the motor will not do anything until I reach potnom of 60%. Then it hits overcurrent and dies. The 'amp' variable also stays zero until 60% potnom. I think this is linked to the angle weirdness somehow.

I'll try to get to the bottom of this, there is no parameter that should cause this.

[arber333]

I'm pretty sure the encoder is open collector, just because of its wide voltage range. I'll throw 12v at it and see what happens.

The encoder is now being powered from the ENC_IN_5V pin on the control board; R7 (which is intended to supply power for an IR diode there) has been bridged as advised in the Rev2 schematic.

Can you source power for the throttle pedal somewhere else. Maybe pedal is causing some voltage fluctuations.
I source throttle Vcc and GND at JP4 which also supplies sensor board.
I actually provide encoder power from 15V IGBT driver power supply which is super decoupled and strengthened by some tough caps.

[Isaac96]

Can you source power for the throttle pedal somewhere else. Maybe pedal is causing some voltage fluctuations.
I source throttle Vcc and GND at JP4 which also supplies sensor board.
I actually provide encoder power from 15V IGBT driver power supply which is super decoupled and strengthened by some tough caps.

I just switched the encoder power to main 12v. No difference, the encoder is still doing weird things. But when the encoder is disconnected the 'angle' doesn't change by itself.

And lastly, in slip control mode, the motor will not do anything until I reach potnom of 60%. Then it hits overcurrent and dies. The 'amp' variable also stays zero until 60% potnom. I think this is linked to the angle weirdness somehow.

I'll try to get to the bottom of this, there is no parameter that should cause this.

Thanks Johannes! I think it might be tied to the 'slipstart' and the weird angle changes I'm getting. Slipstart is at 50 right now, when I set it to 10 then the motor begins to respond at 30% potnom.

[johu]

Thanks Johannes! I think it might be tied to the 'slipstart' and the weird angle changes I'm getting. Slipstart is at 50 right now, when I set it to 10 then the motor begins to respond at 30% potnom.

Please post your parameter file. I assume fmin=1, fslipmin=1. That means until you reach 50% slip stays one and is therefor not > fmin. Try setting fslipmin=1.5 and/or fmin=0.5

EDIT: just checked the parameters, fmin=1, fslipmin=1 is actually defaults - silly me.

[Isaac96]

Please post your parameter file. I assume fmin=1, fslipmin=1. That means until you reach 50% slip stays one and is therefor not > fmin. Try setting fslipmin=1.5 and/or fmin=0.5

EDIT: just checked the parameters, fmin=1, fslipmin=1 is actually defaults - silly me.

YES! That enabled the motor to spin under pedal control! (although it was super jumpy and soon hit OC). It also spun up nice and fast and then shut off from the OC. It now makes whining noises just as a motor should

But the encoder is still behaving strangely. I dove pretty deep into its behavior and figured out exactly when it happens.

Whether the encoder is connected or not, when I turn the key or send 'start 1' then the angle variable begins increasing steadily. I have posted screenshots in this thread already.

If I just turn on 12v and then send 'start 2' the angle doesn't change. But if I turn on 12v and then turn the key for start (turning on the main contactor) the angle starts ticking upwards. Even if I send 'start 2' afterwards the angle keeps changing. I think this is causing the jumpiness of the motor.

The encoder, however, does get read correctly; I can counteract the changing of the angle by turning the wheel backwards.
Also the 'turns' variable doesn't change just from the ticking of the angle variable.

I'll scope the MCU inputs and see if there's some noise on the line.

Progress!

EDIT: Scoped the MCU inputs (or rather the MCU ends of R2 and R5) and they are clean.