[swoozle]

Tried a few things today, let's see if I can tell the story.
In general the CE traces are 10V per division, everything else is 5. 0V is indicated by a little purple triangle on the right axis.

The main cap is a Kemet 380uF, .8 mOhm ESR at 10khz. Confirmed the capacitance using your time RC time constant measurement.

Here's the setup. 12V batt to the left, controller board center-ish (w VLA500 driver & one 300A 600V Fuji IGBT). Using 30R gate resistor to slow everything down. Motor is out of view to the right. Wires to and from the batt/motor are 10 or 12 AWG.
setup IMG_7414.jpg


Here's the gate drive signal. Looks clean, roughly +17V to -10V swing. The 15V power supply line is steady (not shown).
IMG_7393 gate drive signal.JPG

Tried a couple of things that are not shown. Applied -12V to the FWD gate to keep it off. No change. Tried reducing the gate resistor again (to 10R) and yes, it gets MUCH worse. Turn-off spikes to 100V on a 12v supply. Back to 30R.

Here's CE for a small resistive load on 12V supply.
IMG_7395 resistive load.JPG

Here's a low duty cycle CE trace, 12V supply, starter motor load. Bleh.
12V low duty IMG_7399.JPG

Here's a medium duty cycle CE trace, 12V, starter motor. Better. In general low duty cycle is a mess with the long tail on turnoff. That goes away as the duty cycle increases. The turnoff spike stays but is diminished. But everything goes to crap using a 26V supply. Turn-off spikes are outta sight, the tail is much worse and never goes away regardless of duty cycle.
12V med duty IMG_7398.JPG

Here's a low duty cycle B+/B- trace (still 12V). Not very stable.
B-B 12V IMG_7412.JPG

Medium duty cycle B+/B-.
B-B 12V IMG_7411.JPG

Now things start to get interesting. Baffling (to me), but interesting. Added a 2200uF electrolytic cap, but it has about 80mOhm ESR (all I have to play with at this point). This is in addition to the kemet.
Here's the B+/B-. Significantly more stable.
B-B 12V 2200uf IMG_7407.jpg

But here's the CE with the added cap (high duty cycle). Again, this is on 12V. On turnoff, the voltage spikes to ~40V and never gets even close to 12V. That yellow horizontal line is at 24V. This was NOT using a 24V supply.
When the switch turns back on the voltage has problems getting close to 0v. This is under no load stable running. At this high of a duty cycle and without the added cap, the CE looked fairly good (see above). The high ESR cap made it much worse.
CE 12V 2200uf IMG_7409.JPG

So how do I interpret this? B+/B- instability indicates insufficient main cap size? The added cap didn't help because the ESR was too high?

[PStechPaul]

Originally Posted by swoozle

Tried a few things today, let's see if I can tell the story.
In general the CE traces are 10V per division, everything else is 5. 0V is indicated by a little purple triangle on the right axis.

The main cap is a Kemet 380uF, .8 mOhm ESR at 10khz. Confirmed the capacitance using your time RC time constant measurement.

Here's the setup. 12V batt to the left, controller board center-ish (w VLA500 driver & one 300A 600V Fuji IGBT). Using 30R gate resistor to slow everything down. Motor is out of view to the right. Wires to and from the batt/motor are 10 or 12 AWG.
Attachment 13424

It's hard to tell if the setup has any major problems. It's best to keep all connections short, or at least twisted together to minimize inductance. The battery leads seem to go all over the place. But if you have a really big bus capacitor as it seems, that should take care of those current surges. And it seems like you're running on 7.8 kHz, so that's not terribly high frequency.

Here's the gate drive signal. Looks clean, roughly +17V to -10V swing. The 15V power supply line is steady (not shown).
Attachment 13431

I don't see any problems there.

Tried a couple of things that are not shown. Applied -12V to the FWD gate to keep it off. No change. Tried reducing the gate resistor again (to 10R) and yes, it gets MUCH worse. Turn-off spikes to 100V on a 12v supply. Back to 30R.

Here's CE for a small resistive load on 12V supply.
Attachment 13430

There is some indication of inductance even there. And the voltage should bang solidly between 0 volts and 12V, within maybe 0.5V. I can't tell where the zero reference is. So, first sign of possible trouble...

Here's a low duty cycle CE trace, 12V supply, starter motor load. Bleh.
Attachment 13428

That looks really bad. Still, I can't tell where zero reference is. I assume it's about 2-1/2 divisions from the bottom. So it's jumping to about 42V on turn-off, and then after about 100uSec it settles to about 15V, so still above battery voltage. This indicates that the upper IGBT used as a FWD is not functioning.

Here's a medium duty cycle CE trace, 12V, starter motor. Better. In general low duty cycle is a mess with the long tail on turnoff. That goes away as the duty cycle increases. The turnoff spike stays but is diminished. But everything goes to crap using a 26V supply. Turn-off spikes are outta sight, the tail is much worse and never goes away regardless of duty cycle.
Attachment 13429

It still appears that the voltage is rising to about 30V, and then settling to about 20V. Seems like the starter motor is acting as a generator - and the FWD is open.

Here's a low duty cycle B+/B- trace (still 12V). Not very stable.
Attachment 13427

Hmmm. It looks like it's a triangle wave alternating between about 7V and 17V. Just what would be expected with a very large inductance, or a motor running as a generator.

Medium duty cycle B+/B-.
Attachment 13432

So your B+ starts rising at turn-off, and then falls during turn-on. Seems like the B+ has a huge amount of inductance or the battery is sagging greatly. Your battery connections might be bad. Read the voltage from the battery to the end of the cable where it's clipped onto the B+ and B-. If you get more than one volt, it's bad, and you need go no further until you get a solid connection. Or, if the battery voltage is jumping that much, you have a bad battery. (Unless you're drawing over 100 amps!)

Now things start to get interesting. Baffling (to me), but interesting. Added a 2200uF electrolytic cap, but it has about 80mOhm ESR (all I have to play with at this point). This is in addition to the kemet.
Here's the B+/B-. Significantly more stable.
Attachment 13426

OK, a large capacitor will compensate somewhat for bad connection or bad battery, but not completely.

But here's the CE with the added cap (high duty cycle). Again, this is on 12V. On turnoff, the voltage spikes to ~40V and never gets even close to 12V. That yellow horizontal line is at 24V. This was NOT using a 24V supply.
When the switch turns back on the voltage has problems getting close to 0v. This is under no load stable running. At this high of a duty cycle and without the added cap, the CE looked fairly good (see above). The high ESR cap made it much worse.
Attachment 13425

If the motor is acting as a generator, it will charge the capacitor to whatever it puts out, and then that charge will bleed back through its ESR. I think you have a multitude of problems.

So how do I interpret this? B+/B- instability indicates insufficient main cap size? The added cap didn't help because the ESR was too high?

It's been quite a challenge interpreting these scope shots, but I think if you reconnect your test setup with short lengths of cable and take care to take the scope measurements directly on the terminals of the IGBT, you will find that things make more sense and it should work fine. But I'm also not sure if your setup is good with a starter motor instead of a series wound traction motor. If you spin the starter motor by hand and get voltage out, then it's a PM motor and it might not like being connected to a FWD, and it may have produced enough power to have damaged the IGBT. These are all things that can be checked individually, so that when they are all connected together as a system you should have few surprises.

HTH. Good luck!

BTW, here is a presentation on the use of IGBT snubber and DC link capacitors: http://dkc1.digikey.com/us/en/tod/Ke...ower-apps.html

[swoozle]

Great, thank you. I checked quickly and it is a PM motor.
I think you mentioned that previously and I forgot to check.

We'll work on the other things as well but I'm hoping that's the big, basic mistake I was expecting to find.

[cpate]

Today we hooked up the controller to our big motor hoping to make some progress. This didn't work though.

We tried measuring the voltage across the freewheel diode - from M+ to M- there were positive spikes but only very short negative ones every once in a while. So I guess the freewheel diodes work alright.


Freewheel Diode Collector-Emitter with 12V in-out


12V main IGBT C-E graph with 30 ohm resistor


26V voltage testing graph


12v B+ to B- Graph


26V B+ to B- Graph

What's interesting is that the voltage across the battery cables in both directions looks a lot like this graph. It actually oscillates between positive and negative which makes me think that there's a lot of stray inductance in the cables.

Anyways do you guys think that more capacitors with lower resistance will solve our woes? Is there anything else we can try? This frustrating occurence is holding up our project

Here is a picture of our setup in case there is something obviously wrong.

[Stiive]

Hi,
Haven't read your whole thread, just the last post where you've implemented some changes.

Don't know if you have before, but you should show a close-up of your hardware... From afar the leads looks long and cross over high EMI parts of the controller. This could be causing problems.

How long is the motor lead? Is it looped on the floor? Maybe you have high inductance here causing these voltage spikes...


What I think has happened here is somewhere along the way you have damaged your FWDs in the IGBT. Its easy to check with a multimeter with diode setting. Check between collector and emitter of each gate, probe leads one way should read open circuit, the other way around should read diode drop. Check this value against spec sheet of IGBT to see if you have damaged the FWD.
Both diode drops should be very close (within a few mV), otherwise there's something wrong.

[Stiive]

Well, your other post arrived while i was writing mine.

Can you provide more pics of the hardware? How many IGBT bricks are you using? From the lack of screws it looks like its set up for 3 but your only using 1 at the moment

From a very brief look i would suggest place a EMI shield between IGBT and drive board and look at turning the IGBT around so the gate leads can be minimised and don't have to cross over the DC bus/motor leads.

[JRoque]

Hi. That's strange, I can't see the bank of bulk capacitors or the heavy wires from the pack/to the motor. Need to look closer I guess.

JR

[Stiive]

Originally Posted by JRoque

Hi. That's strange, I can't see the bank of bulk capacitors or the heavy wires from the pack/to the motor. Need to look closer I guess.

JR

He's got a film cap there, no need for bulk caps - that will actually make things worse. Reducing inductance is the name of the game here

I also looked for the motor leads... The very thin power leads I can see are perhaps being used as PTC current limiters :P Maybe we have low voltage/current testing here.

[PStechPaul]

If you are getting any significant voltage across the length of any of the cables, then the resistance and/or inductance is too high. Only one of the cables from the battery appears to be heavy gauge. The green wire looks like it's only about 12 AWG. Even if you are expecting only 20-30 amps RMS, and the wire may handle it without overheating, it will affect the high current surges of the PWM. A much larger capacitor may help a little bit, but it needs to be able to handle surge currents as high as a couple hundred amps, even if the average (or RMS) current is much less.

It's still difficult to determine where the zero volt reference is on your scope traces. But it's pretty obvious that there is a great deal of inductance in the circuit. Remember that inductance increases with the enclosed area of a conductor loop. A big circle is worst. If you twist the leads together or at least bundle them in pirs with ty-wraps, you should see some improvement.

I know this can be frustrating but there are definite steps you can take to determine where things are going bad, and there are ways to make improvements. Good luck!

[cpate]

Originally Posted by JRoque

Hi. That's strange, I can't see the bank of bulk capacitors or the heavy wires from the pack/to the motor. Need to look closer I guess.

JR

We used 8 gauge wires and the motor only pulls 75 amps max, are you saying we need to go bigger?
I would think that 8 gauge would be big enough for that current!

And with the large 2200uf cap wired up, it smoothed the battery signal but seemed to distort the C-E signal as seen previously.

The diodes were checked earlier today and they look fine.
Here is our setup when we were at 12v