Resolver Excitation with V2 mainboard

[arber333]

Its alive!

I installed the latest firmware rev 4.12R and it meant worlds!
It turns out it is very important how the phases are connected. I went through couple variations. I even had to change wires on resolver cos to get rotation. But finally motor spun at 600rpm.
Well it spun also at 300rpm etc... it all depended on how i setup fweak. It seems fweak with udcnom will adapt max voltage and motor will spin up faster if i vary them. Tomorrow i will try with 360Vdc. But i have to make some 16mm2 cables because my test cables are in poor condition .
Attached is parameters file for the latest and best option.

params (7).json

(1.25 KiB) Downloaded 528 times

My motor seems to have an offset at 26000. I got it down to 1000. Johannes can you let me know how to pinpoint the offset more precisely?
Also attached some photos of signal from resolver.
Did you connect shield on all cables together and then to GND? I still havent, i think i will have to do that for 360Vdc test.
Next tests to be done:
1. testing of Forward - Reverse function
2. Testing of Cruise mode
3. Test at 360Vdc with same settings save boost
4. Test of max RPM with 360Vdc
5. Put inverter back in a car and drive

A

[zippy500]

nice work.

look forward to your results

[arber333]

Well i ve done it! Leaf is turning at 360Vdc and behaving at that. I had some problems with boost. Your setting of 400 is not enough. I used 1600 and this gets the motor turning. BUT it is not absolutely smooth at high speed. I can hear some rattle in there. How do you think i could better determine offset? I can get motor running from 24000 to 32000.
Should i measure phase currents?

I have tested:
1. Testing of Cruise mode works.
2. Test at 360Vdc with same settings save boost, works very good. I will keep settings.
3. Test of max RPM with 360Vdc. I got to 5000 RPM indicated at 360V nominal with freq 600Hz and weakening at 800Hz. Reading moves a lot at high speeds.
There is rough running beyond 3000rpm. In some cases inverter throws - error? This means unknown? I am not sure if this are EMI or is offset not quite right. What do you think?

I still have to test reverse function....

[johu]

Hi Arber,

thats absolutely remarkable, now you're quite far ahead of me
I think fweak is too high at 800, I was expecting more like 450. But then, what do I know... At least you should at least reach fweak in terms of rpm.
I don't think EMI is the problem, that would show at low rpm.
The instability in speed reading and strange noises are probably what is holding you back now.

Questions:
1. Are you using udcnom setting?
2. Which PWM frequency are you running?
3. Does the resolver feedback saturate, i.e. drop below 0V or above 3.3V? Can you plot it at various speeds?

[arber333]

Hm, if I use fweak 400 inverter accelerates so quickly it trips OClimit. Or is resolver skipping a beat?
1. I use 360V udcnom setting. Since motor is 8pole we should use fq up to 800hz to reach the speeds.
2. I use 8khz since it is Rev 2 board.
3. I am not sure about resolver I will measure tomorrow. My scope should be able to record functions. At 5k motor is generating some scary noises though and skipping at that.

I don't have resolver cable shields connected to gnd. Maybe I should do it first.

Edit: I found out I used ampmin setting 20 which I use with my Mazda inverter at much higher dead time. I think I would have to lower this to like 5 or 8. That should slow the voltage rise with throttle.

[johu]

Yes I think unconnected shield is like no shield at all.

1. 8-pole is 4 pole pairs. So the assumed base speed of 6500rpm would happen at 6500/(60/4) = 433Hz
2. You can use 17.6kHz on Rev2 board. You have to make R1 about half the value (first resistor of 3-pole filter)
3. Could the scary noises be caused by iacmax? It is quite low at 200A (the parameter file you send me)

[arber333]

Yes I think unconnected shield is like no shield at all.

1. 8-pole is 4 pole pairs. So the assumed base speed of 6500rpm would happen at 6500/(60/4) = 433Hz
2. You can use 17.6kHz on Rev2 board. You have to make R1 about half the value (first resistor of 3-pole filter)
3. Could the scary noises be caused by iacmax? It is quite low at 200A (the parameter file you send me)

1. I joined shields to common gnd and resolver is no better.
2. I played with ifltrise and ifltfall parameters and it seems response got better. Can you explain their function again please.
3. I set iacmax at 500A and response got even better. It seems motor wants to draw amps at higher speeds.
I set iacmax at 300A also. Fmax is 433Hz and fweak is 500Hz.

Whenever i try to run motor it ramps up to 4500rpm and there starts to wobble like it became unstable and finally trips OC. Or so it seems. I am not sure if it is OClimit. Interface indicates like it would jump to 5000rpm but mmediately fall to 4000RPM. Too fast to be other than indication error. See pic attached.

Is is possible those RPM drops are caused by voltage drop?

You think 17kHz wil be better for resolver? Why?

[johu]

I can't explain the wobble just yet. But is most certainly has something to do with syncofs.
1. Ok, so EMI does not seem to be the problem here.
2. Iftlrise and fall dampem the response of the over current limiting. The higher the value, the slower the reponse. ifltfall dampens how quickly the amplitude is decreased once it overshoots iacmax*80%, ifltrise dampens the recovery.
3. Ok weird. The plan is that the AC current stays constant over the whole rev range at a given throttle percentage.

So 17.6kHz would yield a doubled update rate of the resolver. Thats why I think it's worth a try.

[arber333]

I will try to replace resistor. with 1K2 and try 17KHz.
In meantime i will probably play with ifltrise/fall setting. It is all theoretical up to now since motor does not have any torque resistance.
There is another Leaf motor however which is bolted in a car . Today in the afternoon i will transfer my test boards to that car and we will try to get it moving. It does have only 200Vdc so i will have to compensate with 2nd gear.
I will notify you how it goes.

BTW i saw if i set fslipmin to 1 and fslipmax to 3 (or anything larger) motor jumps immediately after i press start button. Without cruise or anything else. If i have fslipmin to 0 nothing of the sort happens.

[arber333]

I have tested the other motor in a car with the same inverter. In fact i just replaced my board for his.
He has 200Vdc.
I performed offset calibration and it seems best setting is just 28000 wheras my motor offset is 27000.
Then i tried to turn motor with force.
I used one tire in the air and tried to turn it like a flywheel.
I applied parking brake and hold it to create friction.
I tried to move the car inside garage.

Facts:
1. We have to replace phase cables because car wants to move backwards
2. When torque is applied motor will turn up to some 2600RPM and not more. I tried to force the weakening and i got some 450A phase amps and OClimit, even though i had it set to 600A! Well just before it went out of scope i could hear motor straining and function wasnt sine anymore.
3. I took some shots of RPM plot and Hz. It seems motor consumes some 16A at 200Vdc max for 2500rpm. But at that rpm amps can be at 450A!!! What logic is that? I expected it would stabilise and raise voltage with RPM. I even expected for the speed to go towards 4000RPM at 200Vdc.

If you observe the first plot there is RPM and Hz. You can see when something tripped at 2300rpm.
Second plot observs max current and rpm. I had some success in getting more rpm from weakening. You can see when OClimit tripped and consumption went up.
hird plot is about Hz and max amps. Again observe where amps go up to 450A and inverter starts to whistle. Then OClimit.

[arber333]

Todays test went on incredibly good.
We changed phase wires and got a bang from the motor and then i also swaped the cos resolver wires.
We managed to spun the motor, but only after i set up boost of 12000 and went to offset 32000. I then tried multiple offset settings and it seemed that motor could spin faster the more we were closing to 38000. Finally i tried 40000, but motor would rumble when turning and used too much amps.
At 38000 however motor would spin to 7000rpm at 200Vdc! Even better at 39000 motor would go towards 8000rpm, but with pron ounced roughness.
See my settings in json file.

params (8).json

(1.25 KiB) Downloaded 426 times

Then i took shots of RPM and freq at different settings. Also i took some shots of current paramters

With 40000 offset it seemed some smoke came from the hood! We stopped and investigated. Nothing seemed broken so we went for a drive test.
We tried to start the car and drive it from the shop... the car went backwards!!!! We were turning transmission 8000rpm in reverse. Obviously bearing lubrication system wasnt made for fast rearwards turning. I am still surprised how the smoke lifted from under the hood...
How is it possible to turn backwards since we changed phase cables since last time car was going backwards? We turned motor at first with 27000 and it was very jumpy there. But it turned itself forwards with a car surging a bit then OClimit tripped. This was in the afternoon, before the nice frames.
We tried to change cables again, but this time it wouldnt start DCcontactor! We checked and found desat was tripping.
I hope we didnt damage IGBTs. I have OClimit at 600A and tonight it tripped many times....

[johu]

Hi Arber, that looks very promising! syncofs can also reverse direction, I think thats what happened.
Are you running coolant in your inverter? Modern car inverters have only very little thermal mass.

[arber333]

OK then. Can you privide insight how to rotate the other way? We almost burned oil inside transmission in reverse.

Just now i tried again on my inverter with 360Vdc and offset 38000. I noticed i got correct rotation, but only about 1000rpm. Then i went to 28000 and motor went wild! I got 8Krpm, but it should give more proportionaly to voltage increase. Leaf gearbox is very quiet one. You almost dont hear the gears.
I did change boost setting from 8000 to 12000 to 25000 in the end. Each time i got to more RPM, but i also tripped OClimit when i reduced throttle. I did set 100A for regen limit though. How does the boost work here. I thought boost is only present when starting?

[arber333]

Since i have proven to myself the simple crowsnest board works i will be ordering 10x resolver interface boards. I was thinking to use SMD construction, but really DIP version is more adaptable. Also i will change my design for Ampera driver interface to include a LED for signalling a Fault condition.

If you are interested and PM me i can order more resolver boards. Collecting orders now! Tomorrow in the afternoon i complete the order.

[D@F]

PM

[arber333]

DOH! I have a problem now. DC contactor wont stay on. It seems desat is fighting me. Maybe we damaged IGBT module somehow... its not fair !
I will try to swap Olimex chips in case they command all transistors ON! It happened before....

So it seems lesson says Ampera inverter is rather sensitive. I had OClimit set to 600A as well as AClimit to 500A. I am not sure if this is applicable.
But OClimit was tripped lots of times....

[johu]

OK then. Can you privide insight how to rotate the other way? We almost burned oil inside transmission in reverse.

Of course:
So "red" always wants to be exactly opposite of "blue". Therefor if you start out with a field offset of 270° or -90° the rotor will rotate anti-clockwise. If you start with 90° it will rotate clockwise. The inverter just follows the movement, it does not try to force a certain direction of rotation.
You can see how "red" overlapping "red" will weaken the magnetic field. But lets not discuss that here.

More importantly the challenge in finding a good syncofs if two variables:

1. if rotor and stator align perfectly, what is the value of syncofs? Ideally it would be zero.
2. Next, there is a fixed phase delay between rotor voltage and rotor current, how much is that? Ideally 90°

So we need to add up 3 offsets: resolver/stator alignment offset, desired magnetic offset, voltage/current offset.

Of course the drawing is a 2-pole rotor, Nissan Leaf has 8 "sectors".

Does that make it clearer?

[arber333]

Hm... so lets assume then my sweet spot is 38000 and the motor is turning in the unwanted direction. To get it turning in the RIGHT way i need to apply force at 180°. So from 65000 what does mean 180°? 90° from 65000 should be 1/4? That is some 16000?
So if i now make this arithmetic 38000+(2*16000) = 70000 so in a circle that means some 5000 from 0? Would that mean motor would just turn the other way? Is it really that easy?

tnx

[johu]

Yes that is how it should be. That said, in code I shift the field by 180° for reverse but it never quite runs in the same manner as in forward. But at least 5000 is in the right ball park.
I use 55000 as default but that is found using an idle motor at 60V. So your value is probably better.
EDIT: as opposed to my drawing above, IPM motors (magnets buried in the rotor rather than being glued to its surface) reach their best torque per amp at 45° field offset. https://www.controleng.com/articles/und ... et-motors/
So in that case the angle difference between forward and backward would only be 90°, right?

[arber333]

Ok will try again when i resolve the disobedient inverter issue.

Attached is the PDF file of resolver board i put to manufacture. I had to put it in zip. Johannes can you make the pdf file load directly to the forum? Its a little awkward zipping it first.

I will get 10 pieces that i value cca 6€ per bare board. If someone would like some boards populated i can offer this service for 12€.
Shipping of those small boards is some 2€ in EU i think. Payment through Paypal. PM me for details of payment and shipping. Also it would be good to know how many boards i assemble and test...

[arber333]

I have to write my observation before I forget.
Last time when I tested two leaf motors with Ampera inverter I noticed some consistent behavior.

1. Boost dependence. When I used low boost I got low rpm, when I used higher boost I jumped to higher rpm. By low boost I mean 4000 VS 12000 setting. I didnt get to see what happens at 15000 boost yet.

2. Offset behavior. When I found offset where motor would move at first I got 800rpm no matter what my frequency was. That was at 30kpt. At 34kpt I got some 4krpm from motor. While at 38kpt I got 8krpm. It seems there is a constant for this 1krpm improvement for 1kpt offset. However with 39kpt I got to 9krpm but motor was turning rough and immediately tripped OClimit.

I happened to had a moment when in my head I actually saw the field moving back and forth when I changed offset. Johannes what I am saying I saw the result and it was 8krpm regardless of voltage. So some weakening happened there. I think software would have to vary the offset when voltage comes to fweak. That way rpm would extend beyond voltage range but torque would have to be limited somehow. Would that be amplimit? Hm. Is there a way to encode varying offset in the correct direction. Like setting the sweetspot offset and after that setting a range to extent offset the way it would increase rpm.
Maybe it would be good to just set PMSM motor in parameters to avoid confusion.

[arber333]

Well today i had ups and downs.
First last time on idle my Leaf motor obviously broke one side of Volt inverter!? I had to connect the other side for testing to continue.
Well once i had boards working again i put them in the car. First i checked direction of rotation. Of course it was reverse.... so this time i didnt switch phase cables. I just added/substracted 16000 to offset and i got it turning in the correct way. BUT i noticed motor was running rougher than before. I had to make some corrections and we got in a car and went for a RIDE!!!!

It was a slow ride. I had to correct offset to get a good driveoff and eventually it went without OClimit. In 2nd gear i have seen on my phone amp draw of some 50A from battery and spikes of 300A+ of phase amps. Max rpm was some 4200rpm at 400Hz setting. We tried uphill and downhill.
There was very uncomfortable surge after 30Hz when car was comming to a stop. Regen was also uncontrolable. It just happened without conforming to some pot or setting. That would be something to work on. I guess inverter should provide some excitation for regen to lessen direct force on magnets. Now it looks like its connected to diode bridge directly.

After inverter came to 65°C we returned home. No worries, we stopped and experimented some to lighten the roughness of the motor. Then I heared some noise like motor would be turning on idle very slowly like 1Hz throttle was at 0 travel so i stopped inverter. We went to inspect and didnt find nothing. Next time i tried to start inverter threw off DCcontactor right away. It let the precharge happen and after we would triger DCswitch it signaled Fault! NOT AGAIN!
I have inverter on my worktable now and it seems like Fault trigger transistor on Volt driver board is burned. It is a small SMD thing even smaller than SOT23 parts. I think it is SSMini3 package. I will try to replace both of them and then see what happens.

How would you expect they are polarized? Signal is grounded all the time inverter is in operation except when fault happens, then signal floats. Would that be NPN transistor properties?

Also Johannes would it be possible to have a button on web interface with next revision? That would save the graph frame and store it as a pic. I tried to capture frames with my phone and couldnt. I could capture interesting drive/regen data...

[doobedoobedo]

Can you get a magnifier on the package and see what's etched on there? Same size as SC-70/SOT363? Google might help with package and markings.

SC70 (below) sot23 (above) resistors are 0603

[arber333]

Hm... i found one burned Ampera driver board and changed transistors from fault position. Seems they are working. I get good fault response when i connect 60Vdc to inverter.
However when i connected mainboard there was still desat error. I inspected the main board and it seems that desat pin is allways pulled to GND. I saw on my interface board that i cnnected both desat and UVLO pins together. But now UVLO pin is pulling them both to GND. Johannes how is UVLO connected? Why would it draw it to GND?

EDIT: It seems my Olimex board developed UVLO pin towards GND. When i replaced it with another 2nd power stage works with 50Vdc battery. I still have to try it with 380Vdc.
1st power stage is relentlessly tripping desat no matter what i do. It seems it is damaged. I even tested without power stage connected and desat didnt trip then obviously.

[arber333]

To anyone who is interested...
I got 10 resolver interface boards from china . I will use 2 for my inverters. If someone is interested in bare board price would be 6€ with 5€ shipping inside EU. Outside EU i will let you know in PM.

I have started to populate one board. When i test it i can also make them populated. That will bite a bit more. I expect somewhere cca 15€ final price + whatever shipping. If you are interested PM me your adress and how many boards you are interested in.