Series PWM IGBT snubber problem

Hi All, I have designed a 240V 2KW inverter for a permanent magnet alternator. This is working okay. But, it uses 600V IGBTs and 600V driver ICs- basically I designed it for a (max) 600VDC bus. The three phase alternator puts out anything from 600VDC to around 1100VDC at a frequency between about 400 and 700Hz..

So, I thought, "no problem, I'll just design a PWM IGBT step down pre-regulator". I had a few problems with this but ended up with a semi-working unit using a 1200V IGBT switched at 10KHz by a TLP250. All drive signals look good and fast- which is part of the problem. Dissipation is in accord with calculations. It also has a soft start circuit so that there is not an almighty current inrush to the discharged capacitors at startup.

I am feeding the +ve from a three phase bridge rectifier to the collector of the IGBT. Then, the IGBT emitter goes to some series parallel capacitors. Effectively 400uF. So, I am variable PWMing the gate and have a constant output to the capacitors of 520VDC. That part works,. However, I am blowing IGBTs. After the first one, it dawned on me that I was basically switching off the alternator windings so the falling current was then inducing a big spike which of course passes through the bridge and hits the IGBT with (presumably) somewhat over the 1200 limit, destroying it. The bridge is a 1600V unit so the IGBT dies first.

No worries (I thought). I added an RC snubber across the IGBT (which is therefore also between the +ve bridge output and the +ve of the 400uF capacitor bank). Started with 550pF which is just over twice the IGBT output capacitance and a 47ohm resistor. That worked for a few minutes but then the IGBT died again (at least this time it lasted more than 1 second!) I tried 0.1uF with 47ohms and that worked fine for 10 minutes or more with nothing blowing up. I figured 10amps (max) from the alternator & 47ohms would limit me to under 500V and this seems to be the case. But the problem is that at 100V (genny 620V less 520V output on capacitors), my dissipation calculates at about 10W for the resistor and indeed this seems to be the case. I have not taken the genny to max revs, but that would bring the IGBT voltage drop to 400VDC which would increase the snubber resistor dissipation to 160W which is unsustainable.

So, I am looking for someone reading this who might have a clever suggestion on how to solve my inductive spike problem when commutating the alternator (via the bridge). I did look at various zero crossing solutions for pre-regulation using triacs and SCRs but it all got too big and complex largely because of the frequencies involved. It is frustrating to be so close with the PWMd IGBT but to be thwarted by the destructions (I've temporarily run out of suitable IGBTs so thought it time to ask for help).

I'm looking forward to any suggestions.

Thanks,
Dave,
Sydney, Australia

This may help: **broken link removed**. See especially part 4.7.1.

Hi KJ6EAD,

Not sure that helps. It is a useful article, but 4.7.1 is talking about shoot through. It really is not a problem if the IGBT falsely turns on. That would be a problem in a bridge, but this is a single series IGBT. What I am concerned about is snubbing the inductive spike so I do not exceed the Vceo or the dV/dT rate. because of the three phase bridge rectifier, and the fact the alternator windings are on the other "side" of it, the reverse voltage spike is "rectified" by the bridge so there is nothing for the anti-parallel diode to do. It is a large +ve voltage spike across the IGBT.

Regds,
Dave

Here are more hopefully useful articles: , **broken link removed**.

It's not a problem of snubber dimensioning, it's a problem of designing a suitable buck (step-down) converter.

At first sight, I don't think that it can work without a IGBT of sufficient voltage capability, an additional filter cap, an inductor and a diode. Otherwise, you have to dissipate the energy stored in the alternator's leak inductance quantitatively.

Hi FvM,
Yes, the problem is what to do with the energy from commutating the alternator windings and even calculating what the levels really are. I have not yet attempted to measure the alternator winding inductance. The problem with adding a filter cap on the bridge output is firstly the voltage capability needed at high engine revs and secondly the problem of dumping a lot of energy from that cap to the main 400uF bank when the series element switches on. At the moment, the current is limited by the windings. I suppose I could look at an additional filter cap on the output of the bridge then an inductor to the collector of the series element to limit the surge....but then THAT inductor will be generating a spike as well. Of course, the good news about that spike is that it is not rectified by the bridge.

Dave

but then THAT inductor will be generating a spike as well

Click to expand...

Thus I suggested a buck converter topology, adding a diode.

Hi Guys,

Thanks for the food for thought. KJ6EAD, thanks for the snubber docs. I previously had the first but not the second which fills in some gaps very nicely. FvM, you are quite right that this should be about the converter design and not snubbers.

However, the thought of effectively making a buck converter when I already pretty much have this already and adding an inductor when I already have one in the alternator (!) got me thinking a little more laterally. The down converter I have built (rough schematic at pre_reg1) works fine with a non-inductive source and also works with a hefty snubber (.1 and 47R) and the real alternator. However, it is not sustainable at higher current drains and alternator speeds [which I have confirmed by alternator winding measurements]. So, I decided that first I needed to know the size of my "problem" when the IGBT is turning off. So, I measured the alternator phase inductances and resistances. The phase inductance is 9.4mH which is higher than I would have guessed. Doing the calculations shows that my snubber will work at low current drains (which indeed it does in practice) but I would destroy a 1200V IGBT at several amps drain from the alternator.

So, spurred on by FvMs suggestion, I have two choices. Store the inductive energy in either a capacitor or an inductor. I've gone the capacitor route because of space. But I've also added a small inductor and a couple of diodes. I have also decided to restrict engine revs so that a total of 800VDC is the maximum from the alternator rather than the present 1,000VDC. This gives a bit more headroom.

When the IGBT switches off, there will be a maximum of 0.19joules stored in the alternator inductance. I have assumed the inductance will be approx 5.5mH (9.4/root 3 because of the three phase) but this might be an overestimate when the alternator is producing maximum current/voltage. This will dump via D1 in diagram pre_reg2 to capacitor C1 of 1uF when the IGBT turns off. This will increase the voltage of C1 by about 630V which will give around 900V maximum across the 1200V IGBT. Then when the IGBT turns on, initially C1 discharges via D2 and L1. L1 limits the current inrush between the capacitors and via the IGBT. I have calculated that 1uH (small air wound coil) should restrict the IGBT current to 50 or less amps before the voltage on C1 equalises with the alternator output. This will only take around 10uSecs. So, the IGBT will be way within specs.

A few other points. The inverter does not turn on unless there is >400VDC on the bus (ie across the 400uF). I also have a soft turnon circuit for the IGBT to prevent massive current flow when the alternator first starts. So, the IGBT sees the difference between the alternator output (with or without spikes) and the bus voltage. The IGBT driver is a TLP250 with associated logic plus the voltage regulator stuff- all shown as a couple of boxes here. The constant current source is for a small power supply driving the TLP250 and it is driven from a peak circuit which will benefit from spikes by giving plenty of volts for the constant current source even when the alternator output is draggeed down under heavy load. On top of this, the PWM can go between 100% off and 90% on [but not 100%]. I have a much smaller snubber now to cope with wiring inductance and local capacitance

So, I would be interested in peoples' thoughts about whether this modified design is likely to work and if you think I have missed something or made a logical mistake. IGBTs don't last long with 2KV spikes around! I am much happier to be able to (hopefully) store the energy and re-use it when the IGBT turns off then back on. Whereas using a snubber would have been brute force and hugely dissipative.

Best Regards,
Dave

I can't see, if the capacitance of C1 is sufficient to limit the maximum IGBT voltage according to your intentions. Generally I suggest to reorder the parts to form a real buck converter (saving 1 diode). You can then vary C1 and L1 respectively.

Hi FvM,

The problem (as I see it) with this topology is that the voltage across the IGBT is much higher again by the bus voltage. In my schematics, I am looking at voltages referenced to the bus voltage. When the IGBT turns off and the emitter goes to zero, the inductive spike from the alternator (630V with 1uF) added to the peak alternator voltage (800V with my revised thinking but 1,000V previously) goes way over the Vceo.

A standard buck regulator like this would work if there was a lot more capacitance on the bridge output but that becomes very messy because of the very high DC voltages.

Why do you think my schematic will not work? {My control logic turns everything off in the case the bus voltage is too low} I'm happy to admit I have it wrong but am struggling to see why?

Regds,
Dave

When the IGBT turns off and the emitter goes to zero, the inductive spike from the alternator (630V with 1uF) added to the peak alternator voltage (800V with my revised thinking but 1,000V previously) goes way over the Vceo.

Click to expand...

As far as I see, it's exactly the same with your circuit. The difference with the "real buck" circuit is, that you can increase C and L without problems, if necessary.

There's however a serious disadavantage of a small C circuit. If you turn it off in an overcurrent situation, you have no chance to limit the transistor voltage to a safe amount.

I also thought about the snubber idea, and found, that it's an uneffective method to limit the transistor voltage. A better and quite common method for IGBTs is active clamping. Unfortunately, it's not supported by simple integrated gate driver solutions, but it's limiting the transistor voltage precisely at a defined value and is causing no losses at a lower voltage, or on the falling edge.

Hi FvM,

Not sure I agree with that. When the IGBT turns on, the emitter is taken to +525V whereas in the buck converter it is taken one diode drop below 0V. Of course, if there was a catastrophe and my 400uF caps were fully discharged, then it would be different. But my inverter controller turns off the inverter on low voltage (extremely fast) and high current (a bit slower) and also a combination of both. So, the difference between the circuits is that the IGBT voltages in mine are 525V better off than the Buck.

Anyway, I have built up my model with the changes and been doing some bench testing at low voltage but moderately high currents. This way is a bit cheaper than IGBTs destroyed at the rate of one per millisecond. I wound up a big toroidal inductor of about half the alternator inductance and have it between a hefty power supply with big output capacitors and the rest of my circuit. So far, it is performing per the calculations but I have a bit more to do before turning on the generator again. I also dropped the PWM frequency from 10KHz to 1.25kHz so I could use a cheaper IGBT if I fry another one and also because 10KHz was too fast with the high inductance of the alternator. This reduces the load on the diodes, inductor and 1.1uF capacitor. I will report how it goes....

I agree with your comments about the small C situation. It is a PM alternator so current will be limited (to some value) but the calculations of how much energy is stored in the flywheel in cases of short circuit etc are a little daunting at the moment.

Suffice to say, I am trying to replace an inverter module in an inverter generator. There are basic and serious design problems in that & the modules (two blown) are encapsulated in hard black epoxy.

Regds,
Dave

I see the difference for the fully charged DC bus. Unfortunately, it's not always the case. And your circuit is only working in a small parameter range.

Hi FvM, I finished my testing at low voltage and all went well & as expected. Then onto the real alternator. That went fine- on resistive loads.

Emboldened by the success, I (perhaps foolishly!) connected up an orbital sander. Ran for a second then a flash of light from the region of the inverter & deadness. I expect the sander is a fairly inductive load.

The pre-regulator, in particular the IGBT, is undamaged. So, I'm happy about that. However, the inverter is another story and I have a few questions that I am sure you will be able to assist with. It is a fairly standard H bridge design with the top IGBTs PWMd and the bottom ones commutated at 50Hz. The IGBTs are all Copacks. I had done a lot of testing on resistive loads. The output goes via a couple of toroids to a paralleled 4.4uF cap. with a limiting resistor (12ohms 2W carbon film) in series with the capacitor. What happened was that the resistor blew up. Then, I presume, the inverter was driving an almost entirely highly inductive load of two toroids in series with the sander. I lost both low side IGBTs and one high side IGBT. One of the low side drivers {IRF2101] was damaged probably when the collector, emitter and gate were all shorted together. That then took out a LM7812 [which drove the 2101s]. So, a bit of a mess.

I realised that I was originally going to run the inverter from 400VDC (IGBTs and drivers & caps and resistors are all sized for 600VDC). But, I increased it to 520VDC when pre-occupied by the pre-regulator. Now, a few simple questions:

1. I suspect that 520VDC bus voltage is probably too close to the limits (600), particularly when driving inductive loads. Correct?

2. I had not bothered with snubbers at all but perhaps that was stupid with something that is likely to drive a range of different load types- particularly at a high bus voltage. What do you think?

3. I am wondering why the limiting resistor might have blown up. It was well within ratings during earlier testing but that was with resistive loads and also when the pre-regulator was running at 10KHz. Now it is 1.25KHz but I really think that is not likely to be an issue as bus ripple is low. I am assuming that the inductive load somehow caused a big increase in voltage across the limiting resistor. I'm not sure why that might be so. Another possibility could be a big increase in high harmonics caused by the inductive load. However, I have NO real idea why this resistor might have gone up so spectacularly and undoubtedly you and others on this forum have probably seen such a thing many times before or will likely have a good idea. As it is in series with the 4uF, I am guessing there must have been some higher energy high frequencies to do this.

Thanks,
Dave

It is a fairly standard H bridge design with the top IGBTs PWMd and the bottom ones commutated at 50Hz.

Click to expand...

I only know this technique from hobby circuits presented at edaboard. It's e.g. unable to achieve a sine waveform with an inductive load, because it doesn't provide four-quadrant operation. But it shouldn't cause inverter damage, however. I generally prefer an unipolar pwm scheme, switching all four IGBTs.

The good thing with H-briges is, that it doesn't allow the voltage at any transistor exceeding the DC bus voltage. So snubber for a H-bridge means placing low inductance, low ESR capacitors directly at the module supply terminals, reducing DC bus spikes. A voltage margin of 80 V seems too low anyway.

The filter currents and possible overload conditions should be calculable.

Thanks,

Dave