Class-D GDT DC Restoration / General Science and Electronics / Forums

Forums

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

« Previous topic | Next topic »

Class-D GDT DC Restoration

Move Thread

LAN_403

dude_500

Thu Dec 09 2010, 04:14AM

Registered Member #2288 Joined: Wed Aug 12 2009, 10:42PM
Location: Cambridge, MA
Posts: 179

I'm working on a high power adjustable class-d inverter to make sine waves for powering large (10kw+) transformers.

I am generating the class-D waveform from a microcontroller which is working fine, however the issue comes with the gate driver.

The output of the amplifier (effective the output of the comparator of a class-d topology) - Looks fine
Class D 1

The output of the quad UCC37322/21 driver (sorry for bad pic) - Looks fine
Class D 2

The output of the GDT - BAD!
Class D 3

This is problematic since the signal goes below ground sinusoidally, so half the time the gates of the IGBT brick are never even triggered.

So, is there any way to use a GDT in class-d configuration with full duty cycle ranging? I was thinking about perhaps an LR high-pass on the outputs but this could introduce phase shifts...

klugesmith

Thu Dec 09 2010, 05:55AM

Registered Member #2099 Joined: Wed Apr 29 2009, 12:22AM
Location: Los Altos, California
Posts: 1716

Well the first couple pictures look fine. Good work.
As for the second two pictures:

1) How are the 'scope signal and ground connections made (all channels)?
Maybe you are just seeing an artifact of the probing.

2) What load is on the GDT?

3) is the GDT design good from far above 7 kHz (the switching frequency)
down to below 400 Hz (the modulation frequency)? It may have to
handle the full range (depending on the circuits around it).
It might be informative to look at the GDT output waveform when
driven by a square wave at your modulation frequency.
Note risetime, droop, and any ringing.

[edit] Your thread title mentions DC restoration. Good thinking. I bet there
are diode clamp solutions that can help when the modulation frequency
is low for the transformer, esp. if the low frequency voltage does not
saturate the core. Can you dial the UCC output voltage swing up
or down, to see if the GDT is already nonlinear?

dude_500

Thu Dec 09 2010, 07:08AM

Registered Member #2288 Joined: Wed Aug 12 2009, 10:42PM
Location: Cambridge, MA
Posts: 179

1. It's just the GDT secondary straight to an IGBT, and I have the probe grounded to the Emitter, and the probe on the Gate.

2. It does this either with no load and the probe just on the secondary, or with a 50nF brick gate on it. Waveform looks nearly identical (with the few random expected rings and overshoots changing).

3. The GDT design is good down to about 2khz and up through 15khz+.

With just a standard square wave input it looks fine on the output of course, it seems that the core is magnetically following the sine wave that I'm trying to make (Except I don't want it yet in the circuit!)

I've varied the voltage on the UCC37322 bridge and it behaves just like this from 5V up to the 15V max, it's totally linear.

I don't think this has to do with the fact that the core saturates at 2khz, I found a reference to this problem online in a class-d amplifier information page, but all it said was "some DC restoration circuits are required when using a gate drive transformer due to the AC coupling", however it did not mention what circuit to use and I can't seem to find one anywhere online.

Also as more reinforcement that I don't think it's the GDT, here is a spice simulation of the circuit, light blue is the op-amp output, dark blue is the probe on top of the 50nF Gate capacitance
Classd

HazzWold 1993

Thu Dec 09 2010, 09:32AM

Registered Member #2563 Joined: Mon Dec 21 2009, 10:17AM
Location: Australia
Posts: 280

Woh mega pics

Steve Conner

Thu Dec 09 2010, 10:22AM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

Fixed your oversized pics. Remember, 400 pixels max width! If you have to make them bigger, provide a clickable link.

You simply don't use GDTs in Class-D amp circuits, VFDs etc. for this reason. The gate drive necessarily has a component at the modulating frequency, and if you design the GDT to pass it without saturating, you need so many turns and such a big core that it will have a lousy leakage inductance, and won't pass the drive pulses properly.

There are lots of isolated driver ICs available that pass any frequency right down to DC: use one of those instead. I like the TLP250 optoisolated gate driver just now, I see it used a lot in student projects here.

The only way I know of to do it with a GDT is to modulate the signal: You put a flip-flop over on the secondary side, and send a positive pulse to turn the power switch ON, and negative pulse to turn it OFF. I think this is a form of Manchester encoding: it shifts the low-frequency information up to a higher frequency where it'll pass through a small transformer.

But I hate that method: the power section now has state, so Murphy's Law says that both switches in the bridge will end up on simultaneously somehow.

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