problem with cg igbt switch
Reaching, Sun Nov 05 2006, 03:16PMokay, ive been playing around with mosfet and igbt switches for coilguns but i havent found the right way to switch the mosfets/igbts.
i have a simple trigger circuit (555) to adjust the on time with a driver ic and then the igbt. works good so far but i had some mysterious failures.. so, is there any way to completely isolate the igbt from the triggering circuit, ive heard someone uses a gdt but i had no success trying this. and i dont want it to be too complicated cause i want it to use for a 10 or more stage cg. is there anyone out there who can give me some tipps?
Re: problem with cg igbt switch
Marko, Sun Nov 05 2006, 03:28PM
Have any schematics? Pics?
You want the driver IC to be fast enoughand peak current within ratings of mosfet/IGBT.
You can also 'overvolt' IGBT gate up to 30V to keep saturation voltage down and allow some more peak current than maximum rating.
Driver IC needs to be powerful enough (very powerful for lots of paralel devices) and gate leakage inductance must be low.
No failures are 'mysterious' actually...
Marko, Sun Nov 05 2006, 03:28PM
Have any schematics? Pics?
You want the driver IC to be fast enoughand peak current within ratings of mosfet/IGBT.
You can also 'overvolt' IGBT gate up to 30V to keep saturation voltage down and allow some more peak current than maximum rating.
Driver IC needs to be powerful enough (very powerful for lots of paralel devices) and gate leakage inductance must be low.
No failures are 'mysterious' actually...
Re: problem with cg igbt switch
Reaching, Sun Nov 05 2006, 03:44PM
i use irg4bc20ud igbts, 10 in parallel to get a very good igbt brick, with a 600volt rating, 130A continious, 520A pulsed, power dissipation of 600watts and a turn on time of only 39nS , gate capacitance 5,2nF.
when i connect the igbt directly to the mosfet driver ic, everything is fine, but the source of the igbt is the same as for the triggering circuit and the other electronics, and thats my problem, . i have enough knowledge about driver ics etc but how can i switch the igbt without connecting it electrically to the triggering circuit,.. yeah but i see no one knows again where my problem is
Reaching, Sun Nov 05 2006, 03:44PM
i use irg4bc20ud igbts, 10 in parallel to get a very good igbt brick, with a 600volt rating, 130A continious, 520A pulsed, power dissipation of 600watts and a turn on time of only 39nS , gate capacitance 5,2nF.
when i connect the igbt directly to the mosfet driver ic, everything is fine, but the source of the igbt is the same as for the triggering circuit and the other electronics, and thats my problem, . i have enough knowledge about driver ics etc but how can i switch the igbt without connecting it electrically to the triggering circuit,.. yeah but i see no one knows again where my problem is
Re: problem with cg igbt switch
..., Sun Nov 05 2006, 03:45PM
Well if you want to completely insulate the gate...
Get an optocoupler. Any one will work. Hook up your logic to the led input. Now you have an isolated input for your gate driver. Ideally you would put in some type of gate driver chip, as you want to have a circuit that actively pulls the gate to + 20v and at least -10v, preferably like -20v. The most common way to do this is to have 2 gate driver chips, set up to that when the output on one is grounded the other is high, and vice versa, so that the gate swings +/-.
But the power to that circuit needs to be isolated. You could put a small 12v battery ('lighter battery') in each stage, with a couple of nice fat caps (say a 100ufd electrolytic, a 1ufd tant, and 100nf ceramic, with the ceramic as close to the ucc as possible). Or you could build a nice big smps for the stages to use. I would recommend using a mazzilli circuit (due to extremely simple nature), run it of 12v using whatever moseftes you have arround, and then have a 10+10 turn primary, then 10 secondaries (wind one and see how many turns you need, aim for 20v p-p). Then on each gate driver board use some fast diodes and zeners to rectify the output and viola.
But if you want dirt simple.... You could just take a resistor sized to draw ~1ma across the high voltage side of the fet, to a 20v zener to the ground of the fet. Then put your big cap across that. I don't think that you will be able to power a ucc off that, but you could just put a pulldown on the fet, and use the internal transistor in the ucc to turn on the gate. It won't be perfect, but if it works it works
..., Sun Nov 05 2006, 03:45PM
Well if you want to completely insulate the gate...
Get an optocoupler. Any one will work. Hook up your logic to the led input. Now you have an isolated input for your gate driver. Ideally you would put in some type of gate driver chip, as you want to have a circuit that actively pulls the gate to + 20v and at least -10v, preferably like -20v. The most common way to do this is to have 2 gate driver chips, set up to that when the output on one is grounded the other is high, and vice versa, so that the gate swings +/-.
But the power to that circuit needs to be isolated. You could put a small 12v battery ('lighter battery') in each stage, with a couple of nice fat caps (say a 100ufd electrolytic, a 1ufd tant, and 100nf ceramic, with the ceramic as close to the ucc as possible). Or you could build a nice big smps for the stages to use. I would recommend using a mazzilli circuit (due to extremely simple nature), run it of 12v using whatever moseftes you have arround, and then have a 10+10 turn primary, then 10 secondaries (wind one and see how many turns you need, aim for 20v p-p). Then on each gate driver board use some fast diodes and zeners to rectify the output and viola.
But if you want dirt simple.... You could just take a resistor sized to draw ~1ma across the high voltage side of the fet, to a 20v zener to the ground of the fet. Then put your big cap across that. I don't think that you will be able to power a ucc off that, but you could just put a pulldown on the fet, and use the internal transistor in the ucc to turn on the gate. It won't be perfect, but if it works it works
Re: problem with cg igbt switch
Marko, Sun Nov 05 2006, 03:59PM
What driver IC are you using, again? you need something like at least one UCC to drive those IGBT's properly.
A big problem wiht IGBT's is that they simply don't tend to share current like mosfets their negative tempco. This is especially a problem with very large IGBT stacks, as it's hard to equalize resistances and inductances between them. Once one random die starts conducting a bit more current it gets hotter, and as it is hotter it's reisstance drops and makes it conduct even more current (thermal runaway) until it fries.
I would really recommend using smaller number of stronger IGBT's, it would be perfect if you used just 1 per stage.
You could also split the cap bank or make multifilar coil, although I don't know if it would work well. (I know it works on OLTC's)
What diodes are you using to clamp the coil?
If you want optoisolation you *could* use a simple resistor-zener-cap arrangement to provide single-shot power to ucc, and use optocoupler on it.
(I don't see any advantage on isolation if you are using just 1 switch and you don't need level shifting).
Marko, Sun Nov 05 2006, 03:59PM
What driver IC are you using, again? you need something like at least one UCC to drive those IGBT's properly.
A big problem wiht IGBT's is that they simply don't tend to share current like mosfets their negative tempco. This is especially a problem with very large IGBT stacks, as it's hard to equalize resistances and inductances between them. Once one random die starts conducting a bit more current it gets hotter, and as it is hotter it's reisstance drops and makes it conduct even more current (thermal runaway) until it fries.
I would really recommend using smaller number of stronger IGBT's, it would be perfect if you used just 1 per stage.
You could also split the cap bank or make multifilar coil, although I don't know if it would work well. (I know it works on OLTC's)
What diodes are you using to clamp the coil?
If you want optoisolation you *could* use a simple resistor-zener-cap arrangement to provide single-shot power to ucc, and use optocoupler on it.
(I don't see any advantage on isolation if you are using just 1 switch and you don't need level shifting).
Print this page