Changing a Class D LF amp FET driver??

[Chris Wilson]

I am replacing the FET's in a high power (1kW) Class D 136kHz amp circuit for higher power devices. They have increased gate capacitance and I feel the current driver IC may not be up to the task. I attach the current schematic which shows what devices are in use. I intend to use a pair of  TC4452  FET drivers and replace the IRFP450 FET's with some 600V ones, namely STW42N65M5 which have considerably more gate capacitance. My question is, can I use the two higher current capability drivers as direct replacements for the TC4452? I have the TO-247 cased devices.

Secondly, is it better to capacitively couple the driver chips to the gates?

Thanks

[T3sl4co1l]

So, you're scaling this beast up to about 7kW?

I hope everything else can take it...

Note that you'll need to triple the supply voltage, and about triple its current capacity as well.  (You still need an isolation transformer; line-op is not a good idea here.)

If you're keeping it at 1kW, then use a more efficient (read: low conduction AND switching loss, not just one or the other) part.  Something 42N65 does not sound efficient, by either measure, here.

1kW at 136kHz could be done with a pair of TO-220, though I wouldn't really recommend it.  You should have no problem using any cheap TO-247 part with modest ratings.

There is no need to ever use IRFPxxx parts.  Today's MOSFETs are approximately 10 times better, in terms of all performance metrics (Rds(on), Rg, Qg, avalanche, etc.), and many are even cheaper now.

Tim

[Chris Wilson]

Hi all, thanks for the replies. No, I am not scaling it up, I had a couple of FET failures and became more and more annoyed that I hadn't built it with ease of repairs in mind. I am building a MKIII version with 3 LED's to hold on to show what tripped it, assuming the trip protections save the day, and while I am at it was hoping to upgrade the rather marginally rated FETs with something newer and better rated. These thoughts led to updating the driver chip too.


I have some some TC4452 drivers, in TO-220 case configuration, but I have just realized they are none inverting, and the current driver chip, a TC4426 is an inverting one. So I need to buy TC4451 devices I guess...

The info on newer devices is very welcome, I think as blueskull says, the packaging of the GaN devices precludes their usage by myself, so will look to the C3M0065090D
devices. Other suggestions welcome!

Thanks again everyone!

[PA0PBZ]

... but I have just realized they are none inverting, and the current driver chip, a TC4426 is an inverting one. So I need to buy TC4451 devices I guess...

I don't think it makes a difference

[Chris Wilson]

I am not qualified to comment, Googling inverting and none inverting drivers left me more confused! Can you or anyone else explain how this would affect operation in my circuit please? I did go through a moment where I thought it couldn't matter, then self doubt overcame me

[PA0PBZ]

Think about what happens when you reverse the fet inputs (TC4426 pins 5 and 7)

[T3sl4co1l]

The worst problem is that there's no current mode protection, so a mismatch or short at the output will lead to burned transistors, no problem.

Or a poorly designed output transformer causing voltage spikes.  There's no voltage clamping, just some RC dampers.  A low impedance transformer, suitable for P-P operation at this voltage and current, is not trivial to make!

Tim

[T3sl4co1l]

I've seen lots of avalanche ratings, but I've never seen a single one that likes being exposed to switching transients on a repetitive basis.

MOSFETs and minority carriers just don't belong together... avalanche, even reverse bias (body diode), they're just not very good at it...

Tim

[T3sl4co1l]

2nd breakdown -- sure they have it, or most do.

Though SuperJunction types seem to buck that trend, which is... amazing.

Have you seen any GaN devices with avalanche or linear ratings?  (It occurs to me I haven't noticed any GaN parts with unusual SOA. Not that you'd use them for almost any linear applications -- too expensive!)

Well... I suppose as GaN gets cheaper, it will become worthwhile to make RF amps -- high power, high voltage ones -- out of them.  That's probably how the microwave oven magnetron will finally be displaced.

Tim

[Andy Watson]

There's no voltage clamping, just some RC dampers.

Agreed. Also, the action of CH1 is only checked by the two 4n7 capacitors - what value is CH1? How much energy does it store?

[Chris Wilson]

There's no voltage clamping, just some RC dampers.

Agreed. Also, the action of CH1 is only checked by the two 4n7 capacitors - what value is CH1? How much energy does it store?

Choke CH1 is made from a two-inch (50mm) piece of receiver type ferrite rod wound with about 20 turns of 1.5mm enamelled copper wire

[Chris Wilson]

The worst problem is that there's no current mode protection, so a mismatch or short at the output will lead to burned transistors, no problem.

Or a poorly designed output transformer causing voltage spikes.  There's no voltage clamping, just some RC dampers.  A low impedance transformer, suitable for P-P operation at this voltage and current, is not trivial to make!

Tim

Hi Tim, if you are referring to over current protection in the circuit, there is something. The Hall effect device is mounted b=very close to the correct pole of the choke's ferrite core. It's distance is adjusted to trip the driver signal if current exceeds about 12 Amps. It's fiddly to adjust and I am bread boarding an ACS712 current measuring device hooked up to a comparator to trip at a very accurately defined over current.

[Chris Wilson]

Does the 6R8 value of the gate drive resistors look OK? I have one resistor to each of the 4 gates (two FET's in parallel each side)

I am curious as to why capacative linking the driver to the gates is frowned upon, would that method not add protection for a driver issue leaving one pair of FETS permanently conducting?

Thanks for all the replies, even if some rather go over my head