Erratic Speed Measurements [SOLVED]

[nailgg]

Can you send a photo of your power stage? What cap(s) are you using for the DC-link? The voltage spikes generally caused by the stray inductance in the system but the spikes generally occur for very short moments, in less than 100 nanoseconds or so.

[GrapeNuts]

Capacitors are Electronicon PK 16 Electrolytic Capacitor, 420uF, 1100VDC
Also a Metallized Polypropylene Snubber Cap 2.5uF 850V 10%
IGBT is POWEREX - CM50TF-24H - IGBT MODULE, 50A 1200V SIX PACK

These components were all purchased used on ebay.

[arber333]

Hi

From your pics it seems your cap mounting point is connected to the case GND. Usually that bolt is caps negaitve pole and therefore DC negaitve!!! Can you check the cap mounting bolt against DC negative pole? This can cause strange behaviour and current flow also within logic section.
You say your transformer went up in smoke? Did you check if you still have isolation?

Now check your drivers for voltage differences while at load. Driver chips can be damaged but still functioning.
Also do the IGBT lamp test or gate desat test if you have drivers with desat available.

Maybe you need to use more snubber cap value?
Also DC link rails side by side is not a good idea at higher loads. If you want inductance to cancel itself out redesign DC link so that rails will overlap as much as possible. Or use two flat sheets of copper and spacers over IGBTs.
After i did it like that it seems all my previous inverter problems are history.

[nailgg]

I am not putting the blame on the busbars 100% but it seemed to me that they are way too long than they should be. It's generally desired that the caps are placed almost directly on the IGBTs to minimize the stray inductance. You're using electrolytic capacitors and their ESL is higher than that of the film caps, this can be another factor. So it would be better if you could redesign your DC-link, as Arber also pointed.

Even 50 nH of stray inductance between the capacitor and the IGBTs might result in a voltage spike of more than +1kV or even +2kV while switching off an IGBT where hundreds of (300+, maybe less or more) Amps is flowing. The spike value depends on many factors (the main factor is the fall time of the IGBT), and even the gate resistor value has effect in this.

[johu]

The caps are film I use the same and fortunately the case is isolated unlike on most elcaps. But I agree that the bus layout has too much inductance, there is only a snubber on one IGBT and most of all the wires between gate driver and IGBT are waaay to long and no twisted.

Not sure what caused the spike, had the motor been turning it would be regen (which by the way you really have to make sure does not happen)

Sorry about your scope, I've fried 3 digital ones, only my analog Philips has survived all mistreatment

Nail, thanks for bringing up gate resistors. Maybe for a 50A IGBT increase them towards 15 or 20Ohm (check data sheet)

EDIT: I just did, it says 6.3Ohm. But general increasing it makes the whole circuit more forgiving.

[nailgg]

I just did, it says 6.3Ohm. But general increasing it makes the whole circuit more forgiving.

From what I have experienced so far, although increasing the gate resistor makes the circuit forgiving, it also increases the switching losses, duration of the tail current, possibility of desaturation and so on. So you'll say goodbye to the IGBTs sooner.

Since @GrapeNuts is going to design a new power stage, the best practice I could humbly suggest is to keep the DC inductance as low as possible using the well known laminar busbar system where two bus rails are placed on top of each other with a thin dielectric material in between, and to choose the gate resistor value as what's written on the datasheet. With the current high inductance system, the only way to make them survive is to increase the gate resistors though as Johannes suggested.

[arber333]

I just did, it says 6.3Ohm. But general increasing it makes the whole circuit more forgiving.

From what I have experienced so far, although increasing the gate resistor makes the circuit forgiving, it also increases the switching losses, duration of the tail current, possibility of desaturation and so on. So you'll say goodbye to the IGBTs sooner.

Since @GrapeNuts is going to design a new power stage, the best practice I could humbly suggest is to keep the DC inductance as low as possible using the well known laminar busbar system where two bus rails are placed on top of each other with a thin dielectric material in between, and to choose the gate resistor value as what's written on the datasheet. With the current high inductance system, the only way to make them survive is to increase the gate resistors though as Johannes suggested.

I think if you use a DIY paralel busbars you need to use at least 2x more Rg than specified in IGBT datasheet. Otherwise you can get away with 1.5x.
If you have high inductance DC bus you must have high capacitance elcaps AND fast reaction snubber caps. The more your DC bus is laminated the less snuber caps you need. But my experience shows that having only film caps can produce DC resonance and snubbers are still neccessary for large currents.

I think the best way would be to construct DC bus from two 1mm copper sheets one over each other with layer of pressed silk or other isolator in between. That way you can keep your vertical cap configuration. I see you use some IPM module. Do you have option to connect fault line to Johannes master board?

[GrapeNuts]

Thanks for the feedback. I will make modifications to the power stage to reduce inductance. I tried to mount the capacitors as close to the IGBT terminals as I physically could. I was hoping this vertical arrangement and the thick copper bars would achieve that. The vertical section of the bus bar is 3.75 inches. Unfortunately reducing the vertical clearance of the bus bars crowds the IGBT gate terminals. The IGBT module I have has power terminals on either edge of the device. I decided to shape my bus bar in this "C" shape so I could have my bus connected to both sets of terminals with the caps in the center. I may have to move the capacitors out to the side rather than above the IGBT. I will also add a second snubber cap, and reduce the length of the wires from the gate drivers to the IGBT gate terminals and twist them.

Would the inductance on the power stage cause the voltage I measured on the gate driver output and the general symptoms I'm seeing over all? Is there some testing I can perform while I await my new parts to arrive? The gate drivers seem to put out clean signals when the IGBT doesn't have high voltage on it, which makes me wonder if voltage is leaking from the IGBT emitter to the gate?

[arber333]

Thanks for the feedback. I will make modifications to the power stage to reduce inductance. I tried to mount the capacitors as close to the IGBT terminals as I physically could. I was hoping this vertical arrangement and the thick copper bars would achieve that. The vertical section of the bus bar is 3.75 inches. Unfortunately reducing the vertical clearance of the bus bars crowds the IGBT gate terminals. The IGBT module I have has power terminals on either edge of the device. I decided to shape my bus bar in this "C" shape so I could have my bus connected to both sets of terminals with the caps in the center. I may have to move the capacitors out to the side rather than above the IGBT. I will also add a second snubber cap, and reduce the length of the wires from the gate drivers to the IGBT gate terminals and twist them.

Would the inductance on the power stage cause the voltage I measured on the gate driver output and the general symptoms I'm seeing over all? Is there some testing I can perform while I await my new parts to arrive? The gate drivers seem to put out clean signals when the IGBT doesn't have high voltage on it, which makes me wonder if voltage is leaking from the IGBT emitter to the gate?

You can use the same setup for DC bus. You just have to use copper sheet for both +/-. You have to drill out correct holes under holes so that it clears the adjacent contact. Observe Pauls design here: https://www.instructables.com/id/200kW- ... ctric-Car/
You can even use some copper material as reinforcement where caps connect.

[nailgg]

I may have to move the capacitors out to the side rather than above the IGBT. I will also add a second snubber cap, and reduce the length of the wires from the gate drivers to the IGBT gate terminals and twist them.

Would the inductance on the power stage cause the voltage I measured on the gate driver output and the general symptoms I'm seeing over all? Is there some testing I can perform while I await my new parts to arrive? The gate drivers seem to put out clean signals when the IGBT doesn't have high voltage on it, which makes me wonder if voltage is leaking from the IGBT emitter to the gate?

The inductance might have killed one of your IGBTs in your six-pack. Reducing the wire lengths and making them twisted-pair would be better than the current one, but ideally you must mount the gate drivers directly on your IGBTs. See the image below. In this one the gate drivers pins are soldered at the IGBT pins. Johannes' gate drivers cannot be mounted like this one so you have to use wires, but I would not exceed 1 inch. Assuming you have +15/-9V DC/DC converters and assuming you're using 6 Ohm gate resistors, you're pushing and pulling 4 Amps and those long wires could damage both the gate driver and the IGBT gate because of the inductance in the gate drive circuitry.

The stray inductance we were talking about is the inductance of the DC-link which causes a voltage overshoot at the plus and minus terminals. But the gate driver cables are also stray inductance (another one) and they cause a voltage overshoot at the gate terminal of the IGBT and the output of the gate driver. If the gate voltages exceed 20V, the gate is very likely to get damaged. To avoid that bidirectional TVS diodes are placed at the gate terminals, but since the diode is on the PCB and there are very long cables between the PCB and the IGBT, the TVS becomes useless. Most importantly, because of the stray inductance of the gate wire and the gate capacitance, you will have a LC resonant circuit and you'll have deadly voltage oscillations at the gate.

If you have high inductance DC bus you must have high capacitance elcaps AND fast reaction snubber caps. The more your DC bus is laminated the less snuber caps you need. But my experience shows that having only film caps can produce DC resonance and snubbers are still neccessary for large currents.

@Arber, I actually have to object to this, partially Film caps are superior in many ways, especially on current ripple durability, ESL and ESR values over long frequency ranges. With a good-enough DC-link design created with film caps, you don't even have to use any elcaps or snubbers. Most importantly, what I have seen from the experiments I have made, more capacitance doesn't make things better if the stray inductance is high. Even if you have "infinite" capacitance, the capacitor becomes ideal voltage source and the circuit becomes like this with the stray inductance (kept it very simple).

In my current design I didn't use any snubbers. Only a 1000uF/700V film cap (power ring). I used 16 Amps gate drivers directly mounted on the IGBTs, and my gate resistance is 1.8 Ohms (datasheet suggests 1.5). Not a single desaturation error has occured so far and I am taking off with full throttle drawing around 500-600 Amps. All the IGBT faults are history like yours

[GrapeNuts]

Thanks for the help everyone. I'm going to mark this thread resolved and after I make the suggested changes I'll open a new thread if needed.

[arber333]

Yes Power ring would do away with elcaps and snubbers for sure. I am still not sure about DC link resonance at large amp swings though. This thought is like a sword over me, but for now i dont have any inverter problems.
I just got a desat error last week when my transmission coupler lost its splines! Motor went into overdrive, but nothing bad happened, well besides towing the car home...

[GrapeNuts]

Update: I have finally received all my replacement parts from China and re-assembled my power stage. I'm happy to report so far it seems to be operating as expected. At this point I'm fairly sure the failure was the IGBT module. I figured it would be cheaper and easier to just try a new IGBT module than to keep buying new scopes or other failed bench equipment while trying to troubleshoot the old power stage.

I purchased a new Mitsubishi CM50TF-24 to replace my old used Powerex CM50TF-24 and added an additional 2.5uF snubber cap. I did not modify the copper bus bars. I also replaced the driver board to IGBT wiring with short twisted pairs.

Here is the new power stage:

[arber333]

I stil recommend you fit/solder a TVS diode directly to Igbt GE pins the same you use in driver. It will reduce the spikes because of long wires. It can help with high amps.

[nailgg]

Please don't get me wrong, I don't know what power are you expecting out of your inverter but at powers around 50kW+ you'll probably have serious issues with those bus bars

[GrapeNuts]

Please don't get me wrong, I don't know what power are you expecting out of your inverter but at powers around 50kW+ you'll probably have serious issues with those bus bars

Understood. Right now the goal is just a low power bench top test unit. I wanted something to play with as a proof of concept. When the time is right I'll scale up to something which can handle more serious power.

At this time I'm just happy I can finish fine tuning the parameters and start playing with a throttle.