1 MHz GDT problems / General Science and Electronics / Forums

Forums

4hv.org :: Forums :: General Science and Electronics

« Previous topic | Next topic »

1 MHz GDT problems

previous 1 2 3 4 next last

Move Thread

LAN_403

Tom540

Mon Oct 08 2007, 07:38PM

Banned on 3/17/2009.
Registered Member #487 Joined: Sun Jul 09 2006, 01:22AM
Location:
Posts: 617

That cap in series with your GDT seems too small. Even at that frequency. It might be limiting your gate charge.

uzzors2k

Mon Oct 08 2007, 08:55PM

Registered Member #95 Joined: Thu Feb 09 2006, 04:57PM
Location: Norway
Posts: 1308

I brought the DC block cap up to 400nF and the amplitude rose slightly. I'll put a 1ÂµF cap in tomorrow and see what happens.

I have also built another push-pull stage so each GDT drive mosfet has it's own gate drive signal. This gives each mosfet a near perfect square wave, if only it were that easy with the GDT as well. The improved signal for the gate drive mosfets evened out the signal (the excessive ringing was only on one half before.), at the expense of even more overall ringing. Meh.

Right now I have a 30V p-p signal on the test IRFP450, mostly sine, a little square. At least the gate drive signal is improving slowly.

Avalanche

Mon Oct 08 2007, 09:26PM

Registered Member #103 Joined: Thu Feb 09 2006, 08:16PM
Location: Derby, UK
Posts: 845

A sine wave drive on the base isn't as bad as you think it might be, especially when you have 30 volts between peaks (enough to push the fet way beyond its gate threshold on each cycle). Anything 'more square' than a sine is going to require higher frequency components in the drive which are difficult to achieve at higher frequencies. 2Mhz is probably a bit 'on the low side' to settle for sine wave drive though, but without proper gate drive ICs I'd probably settle for a sine drive. My high frequency gate drive attempts last year used no ICs whatsover at over 10Mhz - using an HF transformer with a center tapped primary, made to resonate with the gate capacitance of the fet producing a sine wave. It only just worked, the device was obviously switching because I lit a tube - but the gate drive wasn't really at a high enough amplitude to crank up the power (lots of dissipation in the active region). If I'd settled for a lower frequency it could have worked quite well.

Marko

Mon Oct 08 2007, 09:44PM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

Uzzors: for what you want, bigger really doesn't mean better.

I don't see why you used BD's for pre-drive, they are quite slow and have huge storage times.

Mosfets are large and slow too, although much less than those bjt's.
Least you could do is to replace the BD's with some smaller and faster transistors.

The trick I would use is to get some miniature extra fast mosfets like this

First, note their crazy delay times. They have tiny gate capacitance which you could actually drive directly from logic gates.

You may want to parallel few of them and use properly sized resistors to dissipate heat, but note that you really don't need as much current as you think.

With just 1amp peak you can charge 700pf gate in 7 nanoseconds, so you should really worry much more about the speed of your devices.

Still I don't know if messing with these is worth when you can just sample or buy UCC's.

uzzors2k

Wed Oct 10 2007, 06:08PM

Registered Member #95 Joined: Thu Feb 09 2006, 04:57PM
Location: Norway
Posts: 1308

How exactly do delays affect the signal? Do they simply add a delay/phase shift, or worse? I knew the driver would be pretty slow, but I thought I could compensate for the delay using Steve Conner's PLL circuit. I would use more ideal components, but this is sort of a junk box challenge, and the only signal transistors I have are BD's. And anyway, gate driver chips are so un1337, where's the fun in getting perfect gate drive first try, every time?

I'm starting to doubt whether I can actually get good enough waveforms at this frequency. This is the best gate waveform I have gotten so far. 5V div. Does it look good enough for running an IRFP450 in a 320V half-bridge without excessive switching losses? I fear it will only suffice for low voltages and not much else. Should I just run it and see, or can I tweak a better signal out of it?

Marko

Wed Oct 10 2007, 06:28PM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

You don't have any complementary pairs like 2N3904/06, BC237/308, ZTX450/ZTX500..? You must have.

As I think any of these without any mosfets would do far better than BD's, only dissipation may become a problem so you need to use properly sized resistors and maybe parallel devices.
Still it's questionable how fast can you go.

I think luckiest discrete driver would be of some of those neat small mosfets. (P channel's are hard to find).

Slowness of transistors is generally going to manifest itself both with phase shift (delay caused by charge storage time) and long transition times. Actually, I'm surprised how your BD's work that well at all.

I'd rather buy UCC's and not be sorry later when big mosfets die.

Marko

Sat Oct 13 2007, 01:33PM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

Uzzors:

Here is an example of 4Mhz waveform from a totem pole of small signal transistors. (2N 3906 and 2N3904) driving a 1nF load directly.
With good low-inductance design you should be able to use it with a GDT.

uzzors2k

Sun Oct 14 2007, 08:40PM

Registered Member #95 Joined: Thu Feb 09 2006, 04:57PM
Location: Norway
Posts: 1308

A 30V p-p sine wave when it should be 12V p-p square wave leads me to think that the gate capacitance is resonating with the leakage inductance. With a resistive load the waveform is 12V p-p and square.

Right now I'm working on lowering the fres of the coil instead.

Marko

Sun Oct 14 2007, 09:21PM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

12V p-p is only 6V peak which is insufficient for gate drive. I don't quite understand what you mean with that - with halfbridge you need 1:2 transformer or double supply voltage.

You can easily build a discrete driver capable for megahertz range.

I further improved my duty cycle and reduced initial ringing by use of a flip flop.

I used a single class A stage with BS170 and 100 ohm resistor and one emitter follower push-pull of bipolar transistors.

My 20Mhz oscilloscope now really seems to be limiting my resolution, I need a better scope to tell how fast this actually is.

Designing a GDT is more troublesome but you should not have problems for 1Mhz.
Use well twisted, or preferably shielded cable with shield for primary, with smallest possible number of turns. Don't use too large core, either.

2N2222/3906, BC308/337 and combinations all worked practically the same, so my scope is probably a big limiting factor.

You can also use a 2:1 transformer and double the supply voltage to full bridge of transistors. Or a halfbridge, double voltage and 1:1 transformer, as you want.
(make sure to also double the voltage to class A stage too in that case).

You can parallel transistors if you need more power, but you probably won't.

Two of these small transistors happily drove 2nF load at 4Mhz 12V, only got warm.

You should really avoid big slow things like BD's and large mosfets.

Pics are 4Mhz square wave into 1 and 2 nF load, just for you to get an idea of fall and rise times.

uzzors2k

Mon Oct 15 2007, 03:11PM

Registered Member #95 Joined: Thu Feb 09 2006, 04:57PM
Location: Norway
Posts: 1308

I realize that I need a 1:2 ratio, but the fact that I don't, and the fact that the signal is 3x larger is the problem. If I used the proper ratio and got rid of the resonant rise I think my problems would be over. If weak little 2N2222/3906 pairs work that well, how did the gate driver hype start? Sounds like a conspiracy to me.

I have a few 2N2222/3906s available so I'll try just using them alone. I'll also rewind the GDT and test things tonight.

One practical thing first; can I run a 2N2222/3906 pair directly from the VCO out on a 4046? I think the buffer makes the drive signal uneven.

previous 1 2 3 4 next last

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob