Poweramp with IRFP4668

[moschfet]

I want to build a simple amp for an active Subwoofer and I try to use the IRFp4668 for it.
The circuit is based on the Elektor amplifier called Buzamp. It was not easy to find a stable setting with LTspice and I hope the circuit works.
Unfortunately I have no matching transformer at the power limit of the MOSFET. The only thing available is a 500VA toroid with 2 x 30V.

The first idea:


There is the question of how much heat can dissipate the Mosfets.

[sawreyrw]

You are going to have a hard time with this circuit, because the transistors Q6, Q7 and M3 and M6 do not "track". The power rating and derating of the MOSFET should be in the data sheet.

[Tekko]

It would work if you used lateral mosfets, with vertical ones you need a VBE multiplier mounted on the heatsink to keep thermal stability in check.

[moschfet]

Yes, Q10 is provided for the thermal coupling.

@sawreyrw: What do you mean with track? The IRFP4668 have a max. power dissipation of 520W. But can the chip size dissipate the heat?

[woody]

Yes that device is rated at 530W BUT that is if you keep the case at 25C or less.
But to keep the case at 25c or below is damn near inpossable you would need to mount it on a flat heatsink with thermal compound and no insulator and have the
area under the device at about 100C below zero and anyway that device will be hard to drive with a ciss of 10,720 pf.

[sreten]

Hi,

Seems bridging (BTL) would use that transformer more effectively.
Sub amps just have to be effective, not flash, bandwidth is so low.

rgds, sreten.

[moschfet]

I have changed the circuit a bit. The small driver for the Mosfets works in the simulation very well.

I think, it's time for a prototype.

[moschfet]

Here some simulations:

Bandwidth with compensation:


Bandwidth without compensation:


open loop gain:


FFT 50Hz 1Watt:

THD: 0.011%

Square 100Hz:


The amp is stable up to 100 nF in parallel to the exit. I am thinking of a coil in the output to ...
Let's see what the reality shows.

[EL36]

Hi,
that's probably going to give some nice firework. With R14 connected to GND and no source resistors, biasing is highly dependent on supply voltage. I'd use some kind of current source in the second LTP's tail and also put Q14's collector on GND and the whole front end on some higher supply voltage (around 12-15 V above the outputs) to be able to drive the OPS rail-to-rail for some better efficiency. There should be also some resistance between the driver's emitters to stabilise their Iq. For compensating the OPS' tempco, the controlling transistor (assuming a two transistor CCS) at the 2nd LTP's CCS should be thermally connected to the outputs/heatsink.

I've been thinking to build something like this for some time, here's some schematic (not been built yet) with the suggenstions above applied:

[moschfet]

Hi EL36

With R14 connected to GND and no source resistors, biasing is highly dependent on supply voltage

That's right. Well observed I try to build it simple as possible. Maybe I put a simple fet current source in it.

... put Q14's collector on GND ...

It's probably better, and its works. At the beginning I had a circuit that ground Q14 and made problems in the simulation. There the dog was buried somewhere else (german proverb)

the whole front end on some higher supply voltage (around 12-15 V above the outputs)

That's also right, but it need a second little power supply.

...some resistance between the driver's emitters to stabilise their Iq...

The current is so low that I hope there is no need.

For compensating the OPS' tempco, the controlling transistor (assuming a two transistor CCS) at the 2nd LTP's CCS should be thermally connected to the outputs/heatsink.

I had already mentioned above

Thank you for your advices