skylogger wrote: I think that once I replaced the fuse, and 1000pf caps, I tried testing in car, and it seemed to detect AC input for a few minutes but then stopped charging. There is probably still some shorted component that is causing more than 16 amps to be drawn, which has now damaged the 4.7R resistor and the Relay during that test.
Initially, I thought that the resistor or relay failure might have been the original cause of the charrger failure. But with a little thought, I agree with most of your assessment. I think that the 4.7 Ω resistors must have been intact for the test, and failed after the few minutes. Hence, no more AC detection. But I think that the relay contacts may have been burned before the test. The pre-charge resistors allow the 3 220 μF capacitors to charge and the charrger to actually charge at very low power, but of course they will drop way too much voltage as the current ramps up. In fact, maybe the relay is the only problem now, apart from the extra 4.7 Ω resistor, and when these are replaced, it might all just work. But of course, it would be good to gain confidence about this. Replacing the relay means the full power of the mains is available for blowing up IGBTs
Once I get the replacement relay and resistor replaced, I'll still need to see what other fault is causing the input current to be too high.
As indicated above, I think there's a moderate chance that there isn't such an other fault. The normal mains current of over ten amps is too much for the resistors, if the relay isn't shorting them.
Earlier, Coulomb suggested using a current limited power supply with an AC output of around 50v
Actually, I use around 50 V DC. The charrger doesn't care; it will rectify whatever you give it. I was going to say that a ~37 VAC source (remember that 50 V will rectify to some 71 VDC before diode drops) isn't nearly as safe as a current limited DC source, but due to the 3 220 μF capacitors, the current limiting will have little effect. The capacitors with enough voltage will have enough energy to destroy the IGBTs before the current limiting can come into effect. Perhaps you could use a 12 VAC or even DC plug pack to connect to the mains. There isn't a lower limit, because the power supply doesn't come from the mains, but from the auxiliary battery. In fact, you might be able to use your 13.8 VDC power supply to power the mains input and also the 12 V input, wiring them in parallel. There should be isolation to prevent problems with that. Unfortunately, without intervention, the PFC front end will probably work really hard and turn that 13.8 VDC into ~400V for the IGBTs to switch. The Elcon/TC charrgers had mains detection, and disabled the PFC section if the mains was not detected. We don't know if these Nichicon charrgers have this detection and disabling.
I guess that since I can go across the whole bridge array and see a 2x diode drop, that would look like something else if any of the 4 inside diodes were shorted.
Yes, the fact that you can't separate the two paralleled bridges doesn't really matter at this point. As you say, any single diode shorting should be detectable.
I was worried that one of these diodes would get damaged by the same inductive kick that killed the 2x 1000pf caps since the output filter feeds back to them. So far that all looks ok on the meter.
Excellent point. But I think the IGBTs have a reasonable surge current rating, so they might
survive where the capacitors literally explode.
I can see a normal diode drop across the protection diodes that are across the IGBTs, but that's not much of a full test.
True, but considering the usual failure mode of IGBTs (total meltdown and short circuit, or alternatively fused bond wire and hence open circuit), seeing these normal voltage drops is quite a good sign. I'd say very roughly a 75% confidence that they are OK.
Since the waffle is separated from the PCB now, I was wondering if I used a 12v supply with a 10k series resistor connected to the gate of the IGBT, would a ohm meter show a difference across the output of the IGBT pins comming out of the waffle, or would I need to setup some other pullups or pulldowns to get a bias?
That sounds reasonable; I think it should work. You might have to come down a bit with the 10k series resistor; there may be something like a 10k pulldown from gate to emitter already. Though I've seen values as high as 47k. 1k from 12 V (12 mA) isn't going to blow up these large devices. I'd use the diode range on the multimeter, rather than ohms; often ohms range are designed to be lower voltage so as not to turn on semiconductors, and diode ranges are higher. My second hand Fluke can even slightly light up blue LEDs (near 3 V). I just used another multimeter to check it: 0.15 V on the ohms range (auto ranging, so that's the 40 megohm range), and 3.76 V on the diode range. Other multimeters may not have as large a difference, but still.[ Edit: "merely work really hard" -> "probably work really hard" ]