GDT - Gate Drive Transformer, UH duty cycle variations (Which Ferrite Core?)

[rentner]

As the title propably describes recently well...

I designed a little circuit for driving the gate of a mosfet. It uses just 2 high speed BJT's with high gain, one NPN and one PNP.

Forming an emmitter follower from the power NPN into the PNP, it drives an ideal Transformer with 500uH/500uH. The isolated side uses 2 diodes and 2 caps and drives a gate with a gate charge of 32 at over 300kHz almost perfectly smooth with amplitude of 15V.

That part is working perfect...

Or is it not? I mean, I did say "ideal transformer", meaning, that it actually just was a simulation...

The real Problem is the fact, that I don't know, what ferrite core should be used. I was googling for a long time and came up only with the FT37-77. I did not find anything about ETD cores, which I would have lying around... I would have to order the FT37-77 and pay shipping, which is not, what I want. The question is: How do I make the "ideal" GDT?

Is it about inductance, about windings, about saturation current or what?

For example: The FT37-77 has AL of 880mH/1000n^2. ETD34 without air gap has 2450mH/1000n^2. Obviously the ETD34 will saturate quicker with same inductance, than the FT37-77... But is it even about saturation for GDT's?


If I have to, I will of course order a different core. But if the windings don't even matter and it only is about inductance, than the ETD34 would be ok.


(Just have to say, that I am amazed, that there are so many ways, to drive a MOSFET with a GDT. What I will use is propably the most simple solution and it supports a huge variety of Duty-Cycles. 1% works fine, if the components are fast enough abd 99% still is simple... Just absolutely BASIC components. Awesome, that some Solutions are best to realize without expensive IC's. I love electronics for live!  )



Oh man, I always write to much. Just kicking you BACK to the real stuff: I just need to get a good Transformer capable of 500kHz or higher, because of the ultrafast voltage rise with square waves...

Thanks for help!

[johansen]

forget the inductance you need, it comes down to the question of is the magnetizing current too high or can you run the ferrite at .3T field strength for minimal turns and fast rise times.

i've wound a lot of GDTs, even with a 50 ohm impedance signal generator i can still get reasonable rise times. since you're driving it with a much lower impedance network you can probably go with just 50 turns on a 10mm^2 core. (for 50 khz nominal square wave at 20 volts rms)
generally the resistance of the winding isn't critical unless you're trying to wind it with 30 gauge wire to fit it inside a watch to send a pulse to a trigatron for one of those loose nukes i keep hearing about...

anyhow, a core the size of a dime is plenty big enough, what is the gate capacitance of the fets you're driving?
emi beads are great, just use whatever you have on hand?

if voltage isolate is a problem you'll need insulation between the wires and the ferrite, those sharp edges cut into the insulation.
if not, then just dip the first layer in epoxy and wind the next layer on..

[rentner]

I am working with 240V DC. Will a double insulated Wire be enought? If so, I would like to wind it bifilar.

I only know Gate Charge Qg: 32 nC. I simulated it in LT Spice and it is quite a clean ramp. The driving current won't be a problem! It can go to up to 8A and the simulation showed me peaks of 800mA which is not much at all. What do you think of ferrite Cores like T94, T106 with the Material 26? Often used in a computer SMPS. I have a quite a few.

I need to forget about the inductance? But there must me an absolute minimum. 100uH switches the Mosfets just as fast as 500uH does. Maybe That saves a lot of work and materials. I hope, someone will borrow an oscilloscope, to test that out. (You can simulate absolutely everything, but if it comes to Inductors, it is hell of a work with millions of core types)


Since it is 140V, I hope I will find a good (almost perfect) solution. Thanks for the first answer so far! 

[johansen]

category 5 ethernet insulation is good to 6 kilovolts.. i've run 8 kv between the two wires of a twisted pair.
i suppose it depends on the code requirements, and its not very high temperature wire.

anyhow, yeah 240vdc and 500Khz system frequency will only require like.. 5 turns of wire around a dime sized ferrite bead of anything higher than like 2000 permiability.

you could use an iron powder core of permiability 75, like mix 26, but its going to get really warm at 1 volt ac per turn.. those cores are meant to operate at like 60KHZ and on the order of .1 vac per turn or less.

you really need a ferrite bead.
you can also use the common mode choke from any power supply as a 1:1 GDT. it might have 10 times as much leakage inductance as what it could have, if you cut the number of turns in half. (since your operation frequency is so high, you don't need many turns)
Most of them have a bobbin that you can unwind half the turns without taking the core apart. (heck, some of them have a gear sprocket on one end, and they actually spin the bobbin to wind the core.

32nC isn't that much, it should not be a problem.

I need to forget about the inductance? But there must me an absolute minimum. 100uH switches the Mosfets just as fast as 500uH does. Maybe That saves a lot of work and materials. I hope, someone will borrow an oscilloscope, to test that out. (You can simulate absolutely everything, but if it comes to Inductors, it is hell of a work with millions of core types)

well no, but the inductance doesn't really matter.
100uH at 500Khz and 15 vac is 314 ohms.
so if you disconnect the mosfet, the transformer itself is going to draw a triangle wave with about 50mA rms current just by itself.
but it won't get warm, all of that is reactive energy. however, with a 3 ohm resistor in there you're going to waste 6 milliwatts of power lol.

here's a practical example:
a ferrite core with a cross section of 4mm by 4mm, and a magnetic path length of 44mm (meaning its about 10mm inside diameter, 18mm outside diameter, 4 mm thick) and a permeability of 2000, means you've got 90-100uH from just 10 turns of wire.
in practice, emi beads will have a lot more permeability, even an order of magnitude higher.