Draining high voltage capacitors, MOSFET or IGBT ?

[LinuxHata]

Hello.
I have to drain a bank of HV capacitors, with capacity of 3300uf, charged up to 400V, to different voltages, down to 100V.
This might happen as fast as once per 3-4 seconds, but in most cases, that would happen once-twice per minute.
To do so, I plan to use thick wire wound resistor of 40 ohms, which will be connected to that bank via transistor, either mosfet or igbt.
The transistor will be controlled via dedicated optocoupler, to isolate hot and cold sides.
Since there's 40 ohm resistor, worst case current will be 10 amps, for less than a second. So what kind of transistor is preferable in this situation?
Mosfet or IGBT ?

[Andy Chee]

This site gives some advice on capacitor bleeder circuits:

https://sound-au.com/articles/bleeders.htm

[magic]

Switching speed is not an issue I think, either could work.
Pick the cheapest part which will meet peak current, peak voltage and power dissipation ratings.

Exact calculation of power dissipation would be a little tedious: current is 10A only initially, it then drops, which is particularly important for MOSFET. One could surely solve it with some integrals and stuff, but I'm lazy so I would do SPICE simulation. Use a resistor to represent MOSFET RDS(on) or a voltage source to represent IGBT Vce(sat). Try a few values copied from datasheets of parts you could buy and you will have an idea which type works better for this application.

edit
Look at "transient thermal impedance" plots which show permissible pulse power versus duty cycle and repetition rate.

[Psi]

It's going to be down to under 5A in 100ms
and under 40V @ 500mA in 400ms.
So even with 3-4 second cycle time you still have like 80% off-time for cooldown/recovery.

Lots of options for a cheap mosfet that can handle that.

[Terry Bites]

Dumping capacitor charge can be efficently achieved with an SCR (Thyristor). An SCR will have far higher peak current handling than an IGBT or a MOSFET.  Typically 10x the continuous current rating. They're easier to drive than the transistors . SCRs are very robust and switch on a couple of microseconds. Low cost too.

[wraper]

IGBT is not the right part here. They shine at higher voltages but for voltage like this it's just paying more for a worse part.

Dumping capacitor charge can be efficently achieved with an SCR (Thyristor). An SCR will have far higher peak current handling than an IGBT or a MOSFET.  Typically 10x the continuous current rating. They're easier to drive than the transistors . SCRs are very robust and switch on a couple of microseconds. Low cost too.

OP wrote in ambiguous way but as I understand he probably wants to discharge down to certain voltage, not completely. In which case SCR is not suitable. Also not easier to drive than a MOSFET unless you need isolation (add an optotriac).

[LinuxHata]

Yes, I need to discharge to specific voltage, so SCR can't be used there (but in some ancient circuits they do so - they had large SCR, which took main power, and there was a small "shunt" transistor, which was in parallel with SCR and turned if off when needed.

[LinuxHata]

And question was mainly asked because as I know, a lot of mosfets does not "like" the linear operation under high voltages - here, 400V will be constantly applied to MOSFET.

[magic]

The 40Ω limits current to 10A and drops most of the voltage.
That's not linear operation, provided that the MOSFET is driven either hard ON or hard OFF.
OFF state current is zero.
ON state voltage is close to zero, up to 10A·RDS(on).

Most IGBT aren't made for linear operation either, by the way.

[LinuxHata]

No I meant a different thing, like that if high voltage is applied to mosfets, even if they are in off state, high voltage causes some migration effects, ruining the mosfet. I've read about that in an old article, maybe 20 years ago.

[langwadt]

No I meant a different thing, like that if high voltage is applied to mosfets, even if they are in off state, high voltage causes some migration effects, ruining the mosfet. I've read about that in an old article, maybe 20 years ago.

I very much doubt that, it would make mosfets usesless 

[wraper]

And question was mainly asked because as I know, a lot of mosfets does not "like" the linear operation under high voltages - here, 400V will be constantly applied to MOSFET.

You said you want to use resistor, in which case MOSFET don't need to be in linear mode. IGBT does not like linear mode even more than MOSFET.

[wraper]

No I meant a different thing, like that if high voltage is applied to mosfets, even if they are in off state, high voltage causes some migration effects, ruining the mosfet. I've read about that in an old article, maybe 20 years ago.

And how do you think mains powered SMPS do work? The vast majority are MOSFET based.

[CaptDon]

Yes Terry, SCR....But he only wants to drain down to 100 volts. The SCR won't turn off without the use of a G.T.O. device and an elaborate turn off circuit. The O.P. will need some sort of circuit to supply either a safe gate drive voltage for the MOSFET/IGBT or a drive current for a normal transistor. Perhaps derived from the capacitor bank being discharged.

[wraper]

As it needs to be discharged to particular voltage, I see no point making isolated control circuit. Adding isolation layers between voltage sensing (ADC or analog) and switch devise driving is counterproductive.

[Andy Chee]

As it needs to be discharged to particular voltage, I see no point making isolated control circuit. Adding isolation layers between voltage sensing (ADC or analog) and switch devise driving is counterproductive.

Isolated control circuit could also imply they are attempting to use the MOSFET/IGBT as a high side switch.

One interesting example of a high voltage, high side switch, is the supply for this vacuum tube tracer (which incidentally also has a variable capacitor voltage):

https://www.dos4ever.com/uTracerlog6/tubetester6.html

Basically the anode voltage is pulsed into the vacuum tube, the transfer characteristics being measured during the pulse. The low duty cycle pulsed anode voltage limits the tube dissipation.

[LinuxHata]

Thanks everyone for your input, I really appreciate it. For the curious folks, no, this is not a tube tracer, ray gun or particle accelerator of any type

[PSR B1257]

Instead it is what? Now we are extra curious 

[LinuxHata]

Guess

There's APFC circuit, which, via MOSFET, which is controlled by user (MCU), charges that bank of capacitor to a certain voltage, and keeps it at that voltage.
However, at certain moment user might want to reduce the voltage on that capacitor bank, so for that task, this circuit will be used.

[Andy Chee]

As a beginner, you’re not making it easy for experts to help you if you’re telling them to guess.

One possible SCR solution would be to completely capacitor discharge into a dummy load to 0 volts, then recharge to your new voltage.

Unfortunately without more information on your application, it’s hard to give proper advice.

[LinuxHata]

I do not have issues with other parts or circuitry, nor I need any improvements - everything is working as it should.
I just want to add that "smart" discharge part, to save the time by avoiding the method you're describing and which is already implemented

[PSR B1257]

However, at certain moment user might want to reduce the voltage on that capacitor bank, so for that task, this circuit will be used.

Interesting application.

[LinuxHata]

Voltage * Capacitance = Energy
So it is that simple and managing capacitors that way is an industry standard for last 50 years at least

[PartialDischarge]

Voltage * Capacitance = Energy

V*C= Charge