Inverter drive module overvolt fault

[capt bullshot]

Is there any feedback from the motor back to the inverter? Some motors use three Hall effect sensors to tell the inverter where the rotor is, needed for proper switching. If there are no sensors/feedback, it should run with a resistive load so long as the current is close to that of the compressor load.

No, the motor has only three wires, the three power phases from the inverter.

My conclusion from this and other observations is - there's some kind of synchronous motor inside the compressor, and the inverter uses "sensorless feedback" for its control loop. The inverter won't run with anything else than a similar synchronous (PMSM) motor.
Is 320V above the peak value of your normal mains AC voltate (e.g. 230V * 1.414 -> 325V)? If the PFC is active, the voltage from DC+ to DC- should be larger than this value. In practice, the rectified voltage from mains is somewhat smaller than this theoretical value, so 320V could mean the PFC isn't active. I'd expect 350V (ballpark value) while the PFC is active. If there's something wrong with the compressor, causing it to feed back energy into the DC-Link, voltage could easily rise up to 380V ... 400V before an error message is thrown. This is at least some general figures, I've got no experience with this particular appliance.

[Eamon]

My conclusion from this and other observations is - there's some kind of synchronous motor inside the compressor, and the inverter uses "sensorless feedback" for its control loop. The inverter won't run with anything else than a similar synchronous (PMSM) motor.
Is 320V above the peak value of your normal mains AC voltate (e.g. 230V * 1.414 -> 325V)? If the PFC is active, the voltage from DC+ to DC- should be larger than this value. In practice, the rectified voltage from mains is somewhat smaller than this theoretical value, so 320V could mean the PFC isn't active. I'd expect 350V (ballpark value) while the PFC is active. If there's something wrong with the compressor, causing it to feed back energy into the DC-Link, voltage could easily rise up to 380V ... 400V before an error message is thrown. This is at least some general figures, I've got no experience with this particular appliance.

You are probably right about the motor but for all intents and purposes the fact that this is a heat pump and that the load is a compressor is irrelevant to the problem at hand I believe; this is a variable frequency driver for a 3 phase motor that uses as you said "sensorless feedback".  I agree that I'm unlikely to have much luck using anything other than the actual compressor it is built for as a load.  In any case, there's a real chance that a different load might not exhibit the fault for various reasons and ultimately I need it to work with the compressor not some other load so I'm somewhat resolved to do my testing in place at the appliance even though that is outdoors and means a bit of setup and packing up each time I want to take some measurements.

My nominal AC voltage is "230 volts +6/-2%" according to regulations.  In practice I see it vary a bit throughout the day (there is a high penetration of roof top solar where I live, including my own house).  What I measured with the multimeter (BM796) this morning was nearly bang on 230v. 

I was playing around a little with the Analog Discovery 2 USB scope and the 100x probes at my desk. When I put them across active-neutral, the AC signal was not a pure sine wave; it appeared somewhat clipped.  Image attached.  I don't think this is likely to be a contributing factor to the fault, just that it might mean the peak DC voltage would be a little different than expected for pure sine wave 230v RMS signal. Over about 5 minutes I recorded a range 225V-233V, my multimeter recorded almost exactly 1V less. max and min are also recorded in the image.

[capt bullshot]

Your 230V waveform looks pretty normal. It's not an ideal sine wave, but rather a bit "clipped" as you see. This is the reason why I said, the actual peak voltage is lower than the theoretical value.

A "short" (up to maybe 5...10m) cable from the inverter to the compressor/motor most probably works, at least for testing purposes.

[florentbr]

though DB2 does not seem to have a reciprocal part in the diagram

Check this one:
https://www.ti.com/lit/ug/tidu249/tidu249.pdf

How high is "well over"? I double checked the +vDC plane several times with the multimeter in DC mode and the highest I have seen is about 325V.

I don't know how much higher it's supposed to be.
All the PFC I've tested were around 370v (small SMPS) and all the literature I came across is showing a significant difference between Vmax input voltage and output DC.
If you check the previous pdf, there's an example at the end showing Vrms=230V and Vbus=390V (DC-link).

This paper says that it must have an output voltage greater than 380v:
https://www.ti.com/seclit/ml/slup390/slup390.pdf

Is there something I can do with a multimeter to check if the PFC circuit is turning on?

To determine if the PFC is ON, measure the voltage or frequency at the transistor's gate.
You should measure something even if the frequency is too high.

My thoughts are similar; the reported AC voltage when it jumps is almost exactly the same as the reported DC voltage, i'm just not sure where to measure to test that theory.

Measure the DC voltage at the output of the bridge rectifier and at the output of the PFC (DC-link).

if the problem is the +DC leaking back to the output of the bridge rectifier, it would seem that C53A could cause that?

If C53A was defective, it could have an impact on the measured voltage at the output of the bridge rectifier.
I have no explanation for the InverterPlateACVoltage switching between 225v, 320v and 255v.
If there was a path back to the rectifier, the InverterPlateACVoltage should be the same as InverterDCVoltage, on standby or PFC on. It's not the case and your curve shows 3 distinctive states.

[Eamon]

A "short" (up to maybe 5...10m) cable from the inverter to the compressor/motor most probably works, at least for testing purposes.

I hadn't even considered a longer cable . The run will be more like 30m but it's worth a shot.

[Eamon]

ok, so with a much longer cable, I have just over 2 ohms phase to phase now instead of 1.1 ohms but it runs and triggers the fault the same. I can use the scope on the desk now so will try to get some traces.  Just have to work out the safest way to go about this as the scope is a USB scope with earth ground and the inverter is mains earth connected with the common ground as the -DC

I've ordered a USB isolator but that will take a couple of weeks to arrive I think.  I am reading that I should probably be using an isolated transformer to power the inverter? The full load of the inverter would require a fairly expensive transformer, but as it faults at around 30Hz which is well below full load I will probably get away with something smaller.

Also the scope can be powered externally rather than USB powered.  The inverter has a 15V line for the fan (relative to -DC/common ground), If i were to convert that down to 5V the scope uses would it be the way to go as that would put the scope and the inverter on a common ground? (assuming I use the USB isolator)

[capt bullshot]

An USB isolator is fine. Some of them have capability to power up the connected USB device - don't know if this is enough for the AD. If not, I'd recommend a quality external power supply for the AD, these should have enough isolation to cope with -DC potential (don't use a lab supply). I wouldn't recommend powering the AD from the inverters internal supply, as 1. it might not provide enough power, and 2. the AD powers down when you disconnect the inverter mains supply, this might be really inconvenient. Third reason: You'd connect the AD GND to -DC anyway in this scenario, but you'd better ensure no power supply current flowing through this connection as this might induce "interesting things" to your measurements.

Alternatively, use an isolation transformer and connect -DC to protective earth. The AD should work then without any further isolation (in theory at least, in practice one could discover what kind of funny things EMI is able to kid on you in such kind of setup).

The isolation transformer must not be rated for the full power of the inverter, as at the lower motor frequency the load is lower and your typical isolation transformer can be short term overloaded by single digit multiples of their rated power. It just heats up faster, if you cut off the power before it overheats and wait for it to cool down again, it'll work.

[Eamon]

Ok, I think the isolation transformer is the way to go and I will then tie -DC to earth, along with the Analog discovery's ground.

I came across a couple of step up/down transformers for cheap and had the thought that I could step down through one and up with the other to get isolation but turns out they are both autotransformers so no luck there, however the input can be switched between 240v, 220v, 200v and 110v with both "110v" and "220v" outputs, net effect is that if I switch it to 240v input, I get 210v on the "220v" output rather than my normal 230v so I can try what you originally suggested of feeding it ~20v lower voltage, I need to source a US style plug first, but I'll give that a go while I wait for the isolation transformer.

[Eamon]

Ok, I haven't given up on this and I'm finally in a position to use the oscilloscope on the inverter (isolation transformer, USB isolator).
Also, while waiting for the USB isolator to arrive I've painstakingly mapped all the ADC pins on the MCU.

I've put probes on all the ADC inputs and none appear to exhibit the pattern reported by the "AC Plate voltage" value reported by MODBUS when I run the compressor

That said, there is another characteristic that might provide some clues; when powering off the inverter there's enough power in the capacitors that the 3.3v rail holds for another 15 seconds or and continues to power the MCU and I can continue to read the values for this period.  The DC voltage drops in a fairly linear fashion until about 160v at which point the MCU cuts out BUT the reported "AC Plate voltage" drops immediately to 190v and stays exactly at that until the MCU cuts out.  In fact, regardless of what input voltage I feed the inverter (using a step up/down transformer I have been able to generate 250v, 217v and 210v in addition to my current line voltage of 227v) and consequently where the "AC Plate voltage" starts when the power is cut (it always seems to read 14v lower than I measure at the terminals with a DMM), it always drops abruptly to exactly 190v for the remaining duration before the MCU cuts out.

One reason this behavior is useful is that I have probed every pin of the MCU while powering off the inverter from idle and the only two pins that show any real change between power off and the MCU cutting out are pins 6 and 9. 

Pin 6 (ADCINA7) drops from about 1.6v in a linear fashion to about 0.8v before the MCU cuts out (pin 6 is connected to DC+ through 4 x 470kΩ resistors and I'm fairly confident this is where the DC voltage measure is taken)
pin 9 (ADCINA2) drops from about 2v much faster, more of an exponential decay. Maybe not directly converted to the "AC Plate voltage" but certainly suggests its it involved.

My guess is that because the "AC Plate voltage" does not directly track any input it is some sort of calculated value (with a floor of 190v) but I'm at a bit of a loss as to where next? 

I was thinking that maybe I should I try manipulating the voltage on pin 9 with an external source and see what it reports as the "AC Plate voltage"?  What sort of supply would be suitable for this?

[capt bullshot]

The supply must be capable to override the voltage at pin 9. So it needs to have a lower output impedance than the circuitry that brings the signal to pin 9. I'd recommend to investigate / reverse engineer the circuit driving this pin at least partially to see if it's driven from a resistive divider or an OpAmp output.
In the latter case, be careful when applying external voltage to that node, as it could (thermally) overload the driving OpAmp.

For a supply, keep in mind it has to be able to source and sink current to overdrive the voltage, and should be of low output impedance. In a former job, I've used a purpose designed amplifier to achieve that, but there was low risk, as I easily could get a fresh or repaired target board if I blew the device by mistake. For whatever reason (most possibly some kind of reversed murphys law) I didn't blow up a target.
https://www.eevblog.com/forum/testgear/show-the-homemade-equipment-you-are-using-now/msg1983056/#msg1983056

[Eamon]

This is what I can determine the circuitry feeding pin 9 looks like:


Not sure exactly what "D5" is, it has "HA7" written on it (no luck find the part on google) and it looks like this:

I've drawn it as a regular diode and assumed the polarity because the opposite makes less sense but not sure of the purpose of it, is it just to prevent pin 9 going -ve or is it actually a zenner diode to prevent too high a voltage on pin 9?

Anyway, I put this into LTSpice and it corresponds with the measured voltage at pin 9 and it shows 706.40177µA through R44 (not sure of the input impedance of pin 9).

The analog discovery 2 has a power supply capability between 0.5v and 5v in 0.001v increments so I'm hoping I can use that?  Not sure how to determine what its output impedance is though?

[shakalnokturn]

Maybe worth checking C53A.

[Eamon]

Maybe worth checking C53A.

For short? (Pretty confident it's not shorted) or do you mean to check that it's value matches what the label says?  I was hoping to go as far as I could without needing to de-solder anything first.

[shakalnokturn]

I'm expecting a low value. This capacitor gets some harsh current spikes when the PFC is active. MKP capacitors tend to degrade by slowly loosing capacity. A value too low could cause more switching noise to reach the MCU's "AC" sense.
If you must desolder use an iron with sufficient power.

Edit: If desoldering is a problem, for testing you could add a 2.2μF parallel to the existing one see if it makes any difference.

[Eamon]

I can confirm that Pin 9 is the direct input for the "AC Plate voltage" value.  I connected my analog discovery 2 power supply to the end of the 150k resistor chain and varied the voltage observing the effect on the "AC Plate voltage", sure enough as I varied the voltage, the reported AC voltage varied between a floor value of 190v (at 1.83v or lower) up to a 255v ceiling (2.45v or higher)

I'm expecting a low value. This capacitor gets some harsh current spikes when the PFC is active. MKP capacitors tend to degrade by slowly loosing capacity. A value too low could cause more switching noise to reach the MCU's "AC" sense.
If you must desolder use an iron with sufficient power.

Edit: If desoldering is a problem, for testing you could add a 2.2μF parallel to the existing one see if it makes any difference.

Thanks, I think I'll try that first as my de-soldering skills leave a lot to be desired .  Now to source a suitably rated capacitor.
This is the existing one:
Not sure where I previously determined that it was 2.2uF but will need to verify that before I can try another in parallel.  But that part number seems hard to find any datasheet for best I've found is this link https://world.taobao.com/item/585823361132.htm?spm=a21wu.24122187-tw.recommend-goods.6 which has a picture with the same markings on it and says it's 2.2uF.  Does that seem like a reasonable value for such a capacitor?

Edit: the "G>" symbol on the capacitor suggests its from "Xiamen Faratronic Co. Ltd" and that took me to http://files.faratronic.com/book/2022/C35-1.pdf which I think means that it is in fact a 2.2uF 630v Metallized polypropylene film capacitor

[capt bullshot]

The capacitance code 225 means 22 and five zeros pF -> 2200000pF -> 2.2uF

[Eamon]

Ok, so it's going to be a little while before I can get a replacement capacitor for C53A, but I have a much larger electrolytic capacitor (labeled 820uF but tests at 730Uf with my DMM) here that has a suitable voltage rating so I was wondering if I could wire it in parallel to test. 

I assume there is a good reason that they used a metalised-polypropylene capacitor with a relatively small value here instead of a larger electrolytic capacitor, but can I add an electrolytic capacitor in parallel in this circuit with a much larger capacitance and see if that prevents the fault?

Also, if I do just end up de-soldering C53A to test it, is the capacitance test function on a DMM sufficient?

[Eamon]

Well this is frustrating; I've ended up de-soldering C53A and testing it and it looks good so I'm back to the drawing board for the root cause of the problem

I've tested the C53A capacitor with both the DMM (which reports 2.18uF) and as it turns out the analog discovery has an Impedance Analyzer function that I used to measure the capacitance from 10Hz up to 100kHz and it was within 1%-2% of the 2.2uF nominal value between 10Hz and about 50kHz and within 10% from there up to 100kHz so I am reasonably confident that the capacitor is good, at least at low voltage (AD2 test was done at 5v).  Is it possible/likely that the capacitor might still be bad but only at AC line voltage?

Is there any other test I should do before I put the capacitor back in circuit?

I can manipulate the reported AC voltage by varying the voltage at pin 9 between about 1.8v to 2.5v and this corresponds to linear change in reported voltage between 190v to 255v, but everything from 2.5v to 3.3v remains at 255v like it was a ceiling value (3.3v was as high as I was prepared to go as that is MCU datasheet say to keep the ADC inputs below 3.3v), but when running the compressor, the reported voltage jumps higher than this (usually up to about the DC voltage) and I observed the voltage at pin 9 actually dropping (but getting noisier) when the compressor kicks in and the "AC voltage" jumps up.

Where to next?

[shakalnokturn]

I doubt the capacitor would behave badly at higher voltages, nothing else I'd check before re-fitting.
Sorry for giving you hope, at least it's checked.

So if in "bench test" conditions the AC voltage measurement is behaving correctly and I would suppose that 255V is a sensible high limit anyway, then the problem must be noise getting to the wrong place.
Others have previously mentioned some of these points:

-MCU supply voltage(s) decoupling (ADC Vref. voltage if externally generated or decoupled.)
-Snubber capacitors/networks on the motor output.
-"Y" capacitor on the small SMPS.
-Wrong grounding.

[florentbr]

as I varied the voltage, the reported AC voltage varied between a floor value of 190v (at 1.83v or lower) up to a 255v ceiling (2.45v or higher)

The output of the bridge rectifier DB1 should be 320v DC (flat) when the PFC is off (no load) and 230V DC (rectified sine, 320v peak) when the PFC is ON .
Maybe it's simply the software reporting the voltage as Vavg/1.41 when the PFC is turned off and just Vavg when ON.

Assuming the PFC is failing to pull the coils L1/L2 to ground with Q1/Q2, that would explain the ACPlateVoltage rising from 220v to 320v and why the InvertDCVoltage doesn't rise.

I don't think it's a sensing issue since the reported values are stable and they match either Vrms 230v or Vpeak 320v.
Check the voltage on the gate/drain of Q1/Q2 with and without load.

[Eamon]

Check the voltage on the gate/drain of Q1/Q2 with and without load.

Eureka!
The voltage on the gates did not move one bit when turning on.  I had assumed that the PFC was turning on as the outputs from the MCU suggested that was so but I should have measured at the final target!

The MCU is driving pins 1 and 8 (Channel A and B enable pins) of the UCC27524 gate driver to 3.3v about two seconds after I send the compressor on signal, then just over a second later the MCU sends a ~48kHz 90% duty cycle 3.3v square wave signal to pins 2 and 4 (channel A and B Input) BUT output pins 7 and 5 remain at 0v. pin 6 (VDD) has ~13v and pin 3(Gnd) is 0v.




Output pins 7 and 5 are each connected through a 100Ω resistor to the gates of the two 34NM60N Power MOSFETs (Q1 and Q2 in the schematic).

13v VDD is within spec (-0.3v to 20v) and outputs should be VDD +-0.3v so I think it's safe to conclude that the gate driver is bad?

Assuming the PFC is failing to pull the coils L1/L2 to ground with Q1/Q2, that would explain the ACPlateVoltage rising from 220v to 320v and why the InvertDCVoltage doesn't rise.

So a bad gate driver here fits with this theory too then I guess?

I'll source a replacement and get it changed over and provide an update once done.

In the meantime, a big thank-you to those who persevered with me through to this point, I've learnt heaps and it's been fun (and frustrating).

Special thanks to capt bullshot and florentbr

[florentbr]

so I think it's safe to conclude that the gate driver is bad?

I would change the driver and the mosfets since a leaky gate could have killed it.
It's usually the power switching components failing first followed by the driver.

So a bad gate driver here fits with this theory too then I guess?

Yes it does since the voltage at the output of the bridge rectifier is supposed to decrease when the PFC is turned ON.
Since it doesn't, I suppose it's the reason why the software is reporting an over-voltage.
Check this simulation and pay attention to the output of the bridge rectifier with PFC ON and OFF:

https://www.falstad.com/circuit/circuitjs.html?ctz=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

[capt bullshot]

so I think it's safe to conclude that the gate driver is bad?

I would change the driver and the mosfets since a leaky gate could have killed it.
It's usually the power switching components failing first followed by the driver.

I can confirm that from others (colleagues) experience with power electronics. I'd at least check whether the gates are shorted to drain or source - if not, one cannot be totally sure there's no hidden damage that'll strike again somewhen in the future. At that particular job, it was an unspoken rule to replace the whole driver / power board in case of such kind of failure.
The other way 'round: In case the MOSFETs failed, they were shorted Drain-Source which obviously would blow fuses or release magic smoke in this configuration. So i'd bet they're still OK and prepared to lose my stake.

So a bad gate driver here fits with this theory too then I guess?

Yes it does since the voltage at the output of the bridge rectifier is supposed to decrease when the PFC is turned ON.
Since it doesn't, I suppose it's the reason why the software is reporting an over-voltage.
Check this simulation and pay attention to the output of the bridge rectifier with PFC ON and OFF:

[Eamon]

I would change the driver and the mosfets since a leaky gate could have killed it.
It's usually the power switching components failing first followed by the driver.

Both power mosfets (Q1 and Q2) read exactly the same resistances in circuit:
source-drain=16.5k
gate-drain=27.7k,
gate-source=9.75k

DB2 = 12.8k

What about the power diodes (D1 and D2, EPU3006)? is it a case of better safe than sorry? They're about $10 each but a relatively small amount given the time invested so far, just more soldering .  Using my DMM Diode test function in circuit they each read 0.378v, but I assume the inductors are just a resistor at DC so they're effectively in parallel so that's about right? and in reverse the voltage just keeps climbing which I assume is because the diodes are good so not passing any current and that's just the effect of the capacitors charging.

[florentbr]

You'll have to remove the mosfets to check them.

Since D1, D2 and DB2 are in parallel (coil is negligible), it won't show on a multimeter if one failed open. If either D1 or D2 was open, it would expose the mosfet to high voltage spikes well over the specified limit. I would just check them off circuit.