Large jitter on angle in resolver and sincos mode

[bexander]

My angle data have spikes when motor at standstill, see picture.

I have a very long (4m) resolver cable as per original from the Lexus it came from. Shielded cable and shield connected to ground in one end.
I have tried moving the cable around without any changes.

Anyone else seen this or have any ideas where to start?
Can it be anything on the logic board or internally in the inverter? I am running the DC/DC in the inverter.

[Jack Bauer]

You should have at most a degree of jitter. I'd start by disconnecting the screen then tie all 3 phase lines as tight together as possible.

[bexander]

I'm running the inverter alone, with only the resolver connected to the motor. All HV wireing removed. No phase connection to the motor. Inverter powered from a bench supply.
I've tried and tied the motor chassie to the inverter chassie. No improvement.
No matter what I do, I still have the same angle signal with spikes.
When looking at the resolver feedback signals with a oscilloscope (picoscope 2204A) I can't see any strange things? Probed at C27 and C28 on the logic board.
I would have guessed this was some sort of grounding issue but I can't find it.
Starting to lose my mind over this...

[bexander]

I still have not found the solution to the problem with spikes in the angle...

In my hunt for a cause I looked at the resolver feedbacks (sin and cos) on the logic board. What concerns me is the 5V pullup (R30 and R29). to my knowledge the STM32 inputs should stay below 3,3V and when the resolver is not connected I read 3,75V at C27 and C28?
When looking at the resolver feedback (C27 and C28) when the motor is stationary, depending on the postion, I can get peak voltages of 3,35V? If my oscilloscope is correct, woulden't that result in an overflow of the ADC in the STM32? The signal is centered around 2V due to the 5V pullup. If it where centerad around 1,65V it would leave more margin.
I'm I misunderstanding something here or should I try and do some adjustments to the logic board? I'm thinking of increasing the resistance of R30 and R29 to decrease the pullup effect.
Have anyone else measured this?

Never the less I don't think this is the cause of my problem but I'm starting do get desperate. I have tested everything else I think could case this but without any positive results.

[arber333]

I still have not found the solution to the problem with spikes in the angle...

In my hunt for a cause I looked at the resolver feedbacks (sin and cos) on the logic board. What concerns me is the 5V pullup (R30 and R29). to my knowledge the STM32 inputs should stay below 3,3V and when the resolver is not connected I read 3,75V at C27 and C28?
When looking at the resolver feedback (C27 and C28) when the motor is stationary, depending on the postion, I can get peak voltages of 3,35V? If my oscilloscope is correct, woulden't that result in an overflow of the ADC in the STM32? The signal is centered around 2V due to the 5V pullup. If it where centerad around 1,65V it would leave more margin.
I'm I misunderstanding something here or should I try and do some adjustments to the logic board? I'm thinking of increasing the resistance of R30 and R29 to decrease the pullup effect.
Have anyone else measured this?

Never the less I don't think this is the cause of my problem but I'm starting do get desperate. I have tested everything else I think could case this but without any positive results.

I am seeing similar problem with ACIM motor ann resolver control. For some reason i have spiked signal in resolver return. I tried to supply 12V and 5V to exciter circuit and i allways get back the same almost garbled signal. ACIM runs. but very rough.
Whilst if i switch the controler front to back on Volt inverter i get smooth rotation on the front resolver interface that is exact copy of the one in the back. Go figure!

EDIT: I managed to scope exciter signal on three exciter resistors. And immediately after first resistor signal becomes triangular instead of square.
On my first resolver signal translates nicely to a square weave directly on EX pins. I am thinking three 22n caps are suspect. I will replace them and see what happens.

[Jack Bauer]

The 5v pullups are not used in resolver mode.

[bexander]

The 5v pullups are not used in resolver mode.

Are you saying that the resistor R29 and R30 should be removed when using a resolver? It seems to be the only way to disable the 5V pullups.

[Jack Bauer]

Yes thats correct. The next rev will have a solder jumper for this purpose.

[bexander]

In what case are the 5V pullups needed? I'm thinking that they might be controlled by SW instead by using a analog switch, FET or SSR. The resolver exciter signal stm32 pin coulde be used to controll this if no other pin is available? I can look into this if it would be usefull?

[Jack Bauer]

No need for that. They are used for induction motor encoders.

[bexander]

Removed R29 and R30. Sadly, it had no effect on the spikes in angle plot. Anyone have any more ideas?

[bexander]

Anything else that needs to be changed or removed on the logic board? I have the block two board.

[Jack Bauer]

No thats the only mod for resolver use.

[johu]

Maybe time for a resolver debugging video
Content:
- Is the excitation signal good i.e. 7-9V peak to peak and sinusoidal?
- Are both return signals good i.e. 1.5-3V peak to peak and sinusoidal?
- sin/cos swapped/incorrect polarity?

Well thats about it, rather short video that would be.

[Jack Bauer]

Yeah thats a good idea

[bexander]

Information and explanation videos is always a good idea!!

I did some more digging and I don't think I have a resolver issue but rather a HW issue inside the logic board or the inverter itself.
I used two potentiometers and an external 3,3V source to produce some adjustable DC voltage resolver feedback and with this I still get the spikes in the angle signal. That excludes the source is the exciter signal or the resolver.
Next step is to unplug all connections between the inverter and the logic board and check.

[johu]

Yes well traced. 3V3 supply of processor dodgy?

[arber333]

Information and explanation videos is always a good idea!!

I did some more digging and I don't think I have a resolver issue but rather a HW issue inside the logic board or the inverter itself.
I used two potentiometers and an external 3,3V source to produce some adjustable DC voltage resolver feedback and with this I still get the spikes in the angle signal. That excludes the source is the exciter signal or the resolver.
Next step is to unplug all connections between the inverter and the logic board and check.

Well I have dual board on my inverter. I have same problems with 2nd brain. When I swap olimex chips I have the same result so that eliminates chip. I truly think resolver circuit has some fault in it. I have to inspect it.
1st brain works good with both motors.

A

[bexander]

I have disconnected everything between the logic board and the inverter. With a 5V connection to R38 I can start the logic board but I still have the spikes in the angle plot.
I have measured the 3,3V to the STM32 on C44, C50, C51, C53 and most important C43 with a DMM and also checked with the oscilloscope without any findings. However my scope is slow with few samples so not the best for finding voltage spikes.
I'm suspecting that the ADC might have gotten damaged when the 5V pullups (R29 and R30) where in place. The pins should handle 4V as per datasheet and I meausred 3,85V with nothing connected to the input so not alot of margin.
I don't know what I should do next...?

[johu]

The data sheet specifies 10mA drain capability for the ESD diodes on non 5V-tolerant pins. Since the resolver/encoder inputs are protected with a 10k resistors that value isn't nearly hit. One thing to note when feeding constant voltage into the resolver inputs: at startup the processor does an offset calibration. At that point you must supply 1.65V to both pins. Alternatively you can switch to SinCos mode, feed some constant voltage like 1V and 3V and show us the angle plot

[bexander]

I've switched to SinCos mode and feed some constant voltage of 1V and 3V to the inputs. The resulting angle plot in the picture.

And a picture of the setup.

I noticed that when both inputs are low I get a good angle plot!! so started to do some more measurements and can conclude that when both input voltages are below 0,5V the angle plot is good. Angle jitter is about 0,3 degrees.
When increasing one of the two input voltages to 1V and the other still below 0,5V I get spikes with pk-pk at 3 degrees in angle plot.
When increasing one of the two input voltages to 3V and the other still below 0,5V I get spikes with pk-pk at 9-10 degrees in angle plot.
So there is a voltage dependence on the angle spikes.

Can the flter capacitors C27 and C28 cause this? If they somehow are defect? Seems unlikely.
Otherwise this is most likely caused by the STM32. Or possibly the SW but then more people should experience this.

[Jack Bauer]

Its definitly not a software or hardware design isue as I personally have used this board in resolver mode with no problems. That leaves either hardware build or hardware damage. As you are seeing increasing noise with increasing signal this would make me suspect the filter. I'd look for solder joint problems on the filter caps or even wrong value filter caps. JLC are really good but I make mistakes as I'm sure can happen with them. Keeping in mind they typically populate one of those boards in about 1.2 seconds:)

[bexander]

Sounds resonable and easy to check.
First visual inspection on more time.
I will then desolder the filter caps and measure them as I couldn't measure them inplace.
This also allows me to test again with constant voltages with the caps removed.
If this doesn't work I will replace the 10k filter resistors and check.
If problem persists I can try to feed the signal directly to the STM32 input pins, to exclude any defects in the PCB.
Still a problem, the STM32 needs replacing.

[bexander]

I have tested removed the filter caps and they are both 1,2nF when measured with my DMM.
I have also tried feeding the signal closer to the STM32 (at the capacitor pads) using an external resistor. I don't dare to probe directly at the STM32 pins.
No luck, the fault still remains so I'm down to replacing the STM32. Will look into whats needed to see if I or someone around can do it.

[SRFirefox]

The easiest way I've found to replace leaded components like the STM32 is to use a razor blade or exacto knife to cut all of the leads next to the body, then use a soldering iron with a generous blob of solder to pick the legs off the board. Use solder wick to remove solder from the pads, and then tack down your new part by a couple corner leads after lining it up. This gives you lots of time to adjust. Once you're happy, get lots of flux on the leads and then wipe the leads with lots of solder on the tip of the iron. Once you have a nice shiny joint on all the leads, use your solder wick to take off all the extra solder and voila, you have a nicely soldered QFP micro.