Home brew polyphase amp

[guzzibreva1]

Here is a picture of the amp I just finished designing. It's a three phase amp designed with nothing but quad op amps (TL084) an transistors (2N2222 and 2N2907) and IRF510 outputs. No special components and all are through hole.
I used air core inductors to reduce distortion plus they were easy to make. The the overall feedback come from a point prior to the output inductors.
The only small problem is there is a bit more hiss than I would like and there is a bit of heterodyning that is only noticeable with pure sine waves. It doesn't seem to effect the sound when listening to music. I think it's because I used quad amps and the same chip that is making the triangles is the audio feedback amp.
Output is 200 watts RMS per channel into four ohms. The switching frequency is 100kHz so the summed outputs looks like a 300kHz amp.


[IMG] http://i118.photobucket.com/albums/o...0/SDIM7800.jpg [/IMG]

[guzzibreva1]

Here is a picture of the small signal step response at 2.5kHz
[IMG] http://i118.photobucket.com/albums/o...SDIM7306-1.jpg [/IMG]

[newvirus2008]

Hi guzzi,
Is it pre filter feedback that you re using? what is the aircore inductor value and what do the losses look like? How is the audio quality?

I think you should be able to have a nice amplifier (only ofcourse at extra cost)

[guzzibreva1]

The feedback is pre filter. The air core inductors are 100uh. The greatest losses in the circuit appear to be the inductors. I wound them with solid wire so that might be causing some losses.
I have a Carver M400 amp to compare it to and I can't hear any difference between them.
I can post the schematic in PDF format if you are interested.

[darkfenriz]

I'm interested in seeing schematic!
That is something unusual and inspiring

[guzzibreva1]

MEGAUPLOAD - The leading online storage and file delivery service

Changes needed would be to use a separate op amp with better noise specs in the audio feedback section.
I welcome any suggestions for improvements. The output filters do not show on this drawing. They are just 3 inductors 100uh each with a .1 uf to ground and a series resonant tank to ground tuned to 300kHz

[newvirus2008]

Pre-filter f/b is ok if inductor is air core. But the inductor is very lossy.

See guzzi, I think you have a workaround here.
For 3-phase buck, the filter values should be 3L & C but you can always have L & 3C instead. This will reduce turns in the air core inductor and hence lesser losses. However, there shall be peaking in the response. (For correct damping of L & 3C filter, the load impedance should be 4/3 = 1.33 ohms)

But see what you gain. when N turns => 100uH, 33uH = only N*0.57 turns

[darkfenriz]

Looks like you've done a good job getting completely rid of output ripple. Now think about a proper gate driver, a complementary class B emitter follower in place of your 2 diode-transistor connection is a good starting point.
Another thing is getting a lower Rdc inductors while keeping the current sharing.
A last thing to do is tweek the feedback to get as much as you can, at some point a lot of feedback will probably try to desynchronize the modulators by excessive ripple aliasing or even switch to self-oscillate, whenever you achieve that much, go back just a little

[guzzibreva1]

I wanted some losses in the inductor wire resistance to help prevent current imbalance in the output stage. I didn't have any precision resistors or caps so there is a very slight difference in the triangle amplitudes. The inductors are the hottest part of the system.

Has anyone ever tried using compound inductors? A power supply that I worked on used two different cores of different materials. One was powdered iron and the other ferrite. One core made the inductor a high inductance until it saturated then the inductor changed to a low inductance during the rest of the cycle. I am thinking this would be a way to reduce the idle reactive current and still have fast response to the audio signal. Maybe it would just cause distortion if there wasn't enough overall feedback.