Whats wrong with this circuit?

 

hi,
Do you have a specific question regarding the circuit.?

E

 

HI,
Do you have a specific question regarding the circuit.?

I ran it at full output powering resistive load for over 1hr- no issues no overheating, I was using a varac to provide 120vac, as I reduced the supply the 22000uf cap exploded and the igbt blew - not necessarly in that order, what did I miss?

 

I was testing the circuit, load was a 90W incandescent bulb, relatively low current test output voltage was 53vdc. I am working on a power supply to drive an induction heating unit, max input 60v @ 50Amp. I don't want to use a transformer - too heavy so I'm shooting for a standard 120v input, to bridge rectifier, clean it up a bit -hit the igbt (rated at 60A/650v)

 

I am think maybe because I reduced the ac input (without thinking) and did not shut it off (as would happen with a standard power off switch) there was a surge or backflow that exceeded the igbt threshold and it self destructed or is there an error in the circuit? If it was my mistake and the circuit is viable that's good, if not I only have 2 transistors left and don't want to trash another because I missed something so looking for another set of eyes.

 

hi hr,
It is possible that by reducing the Vsupply, the Gate voltage became too low to fully switch ON the FET, which would cause overheating.??
E

 

hi hr,
It is possible that by reducing the Vsupply, the Gate voltage became too low to fully switch ON the FET, which would cause overheating.??
E

Not that explanation - the circuit is a linear regulator, so it is always working in the linear region.
I suspect something on the lines of the gate being supplied from the output via the reverse CE diode taking the gate voltage more than 20V below the emitter.
Before you turn it on next time, install 15V back-to-back zeners between gate and emitter.

 

Simulation shows a possible culprit is the 22000uF cap across the output. Dissipation in that averages ~2W but peaks at ~4kW when it initially charges at power-up. Dissipation in the IGBT averages ~95W but peaks at ~8kW at power-up while the cap initially charges.
The sim shows no Vge reversal when the supply voltage drops.

 

Ya, I assumed a fairly high inrush on power up, I don't have a simulator and didn't run the numbers, I did however scope the junctions initially to monitor for excessive differentials and did not see anything outside mfg parameters. In the final circuit I have a relay that must be switched on before the power is applied to the load to allow for voltage stabilization. I used the 22000uf because I had it in my parts bin and wanted the excess capacity due to the nature of the load. Adding the zener protection to the circuit had also been considered but not in place at this time as I was still in the testing phase. What is throwing me off is the fact that it crashed on power down (accidental input reduction). Something that should not happen normally in the real world situation - unless there is a brownout or something. I had not considered this possibility but inadvertently caused it myself. however this would be an issue in the real world and I need to prevent it as it can happen and apparently the circuit doesn't like it, at all...

 

You want 120 VAC rectified and smoothed giving you about 170 VDC. You then drop it to 50 V. The voltage across the IGBT would be 120 V, and the current is 50A, so it would have to dissipate 6 KW. Good luck with that.

 

I reduced the input (not quickly by the way)from @120v and was around 70v on the way down when the cap blew fire out the top (quite impressively I might say lol) but unexpected. so there was no quick drop and return surge that would cause a quick inrush.

 

You want 120 VAC rectified and smoothed giving you about 170 VDC. You then drop it to 50 V. The voltage across the IGBT would be 120 V, and the current is 50A, so it would have to dissipate 6 KW. Good luck with that.

IGBT is rated at 650v @60amp, load was a 90w light bulb so output was no where near 50amp, you think the IGBT was pushed beyond the limit that far?

 

It really doesn’t matter exactly what happened in your experimentation, my point is, that even if the circuit is operating correctly, you are producing more heat in the IGBT than in your induction element. It is not a feasible way to power it. You need either a transformer or a DC to DC converter, or an element that works at 170V.

 

IGBT is rated at 650v @60amp

Do you think this means that it can pass 60A while dropping 650V, thus dissipating 39KW? If so, think again.

 

Yup...that's a very common mistake.

 

If you had the output set to 53V (gate at about 56V) , and the input voltage was 170V, then the tap would be set to 33% of the input voltage.
Remove the input voltage, and the source remains at 56V, the drain goes to 55V, the gate is at 33% of drain voltage which 18V. Voltage between gate and source = -37V. . . . and there goes your gate oxide.

 

looked at the dc/dc converter but haven't found anything that fit the bill for a reasonable cost, I am open for suggestions, I am trying to keep the cost low since I am basically giving this to a friend. I have the hardware myself to power the induction system but yes it consists of a 350lb power transformer. is there any reasonable design that you know of to take 120vac down - that's really what I am looking for. The load is a classic 3500w zvs design he will be using to heat bar stock to hammer out for knife making.

 

Remove the input voltage, and the source remains at 56V, the drain goes to 55V, the gate is at 33% of drain voltage which 18V. Voltage between gate and source = -37V.

Shouldn't source and drain be emitter and collector, respectively?
Regardless, the gate voltage is a set fraction of the voltage across the set of 6.8uF caps and remains more positive than the emittter (load) voltage as the supply voltage drops.

 

Shouldn't source and drain be emitter and collector, respectively?

Of course it is! But it's a MOSFET-y sort of a thing.

Regardless, the gate voltage is a set fraction of the voltage across the set of 6.8uF caps and remains more positive than the emittter (load) voltage as the supply voltage drops.

The input side capacitance is ony 54uF and the output side capacitance is 22000uF. The input side is discharged by the potential divider (27k) giving a time constant of 1.4 seconds.
The output capacitance is discharged by a "90W incandsecent lamp" - assuming it is a 120V lamp. Impedance is 160Ω. Time constant 3.5 seconds.
The input side discharges first, so the Emitter-Collector diode starts to conduct. At that point Gate voltage is 0.33*Vc and Vd=Ve. Depending on how far the output side has discharged, it's a good chance that the gate voltage limits are exceeded.
Because it is an emitter follower, the gate-emitter voltage is not the same as the gate voltage.

 

The varac was fused at 5amps, the first bank of caps would have some charge but not a huge amount, the load was still attached, that was quite a current dump for what it had to draw from. I haven't played with igbt's much but have built a few high current regulators, mostly adapted from what I have found online and in text, added a few extra stages just to push the boundaries. 8 transistors instead of 4 or 6 I really am quite impressed with them. How was it able to pull that much power that quickly - the fuse on the varac did blow, it's a small 500va I use to prototype, usually gives me a chance to catch things like this before it goes nuclear I guess.... It had been running for over an hour , the heat sinks we fan cooled and maybe a little above room temp. I'm not disputing the fact that the design very well may crash at higher power, just the fact that it was able to blow fire out the top of the 22000 cap the way it did, I have seen it before but usually it has a good power supply and plenty of available current to draw from. In comparison what it had available was like taking leak in the ocean .