My very first Class D pwm (switching) amplifier.

[fr0st]

Has anyone seen a simple (ish) PWM amp for the beginer?
I'd be interested to build one that I could later build on to make a monster car amp.

RE: quality... if PWM amps are used for low frequencies won't it mean that you won't exactly notice the added distortion? Bass is mainly distortion anyway.

[phase_accurate]

Bass is mainly distortion anyway.

Me as bass-player can't share this opinion of course . Although the ear is less susceptible to THD at low frequencies it is always nice to have low distortion bass reproduction

One problem with switching amplifiers is getting enough feedback
specially with the necessarily low unity-gain point involved. It is even desirable to take feedback off from the amp's output to correct for the nonlinear behaviour of output inductors, though it makes things even more complicated.

OTOH feedback-takeoff from the switching stage makes it easy to achieve a phase- marging that is above 65 degrees (one goal that sometimes gives linear-amp designers headaches).

It is quite simple to achive large NFB factors at low frequencies but more difficult at mid/high frequencies, but still not impossible.

The output filter has very weak influence at low frequencies in terms of amplitude response. But FR is influenced by the interaction of filter and load at upper-mid- to high- frequencies.

Since the output devices don't operate linearly, no SOA protection is needed (only max current limiting, if any at all).
This, combined with the low filter/load interaction makes them very good performers for bass-reproduction but sometimes less good above.

At very low frequencies there is some drawback: The output filter inductor makes the amp act as an inverter, leading to supply voltage runaway (also called supply pumping).

Regards

Charles

[Circlotron]

Well, things went a bit better than I had hoped. After I has sorted out a few pcb tracks that I had cut that I shouldn't have, messing up the earthing of the driver and output stages, I powered up the thing at it just went! No flames, no fuss. At first I only used +/- 10v rails on the mosfets just in case something was dreadfully wrong, and it sounded a bit dissapointing - very reminiscent of crossover distortion although it wasn't of course. Later on I got bolder (or more reckless) and in stages finally got the rails to +/- 35 volts, the limit of the supply I was using. By this time it was sounding heaps better. Switching frequency was 50Khz at first but I can crank it to something past 500kHz although it didn't seem to make a whole lot of difference to the sound.

The output inductors I used were 800uH each, 55 turns of 1.25mm wound on a Microlite MP4010PDGC toroid core. This core is made of a long strip of metal wound up like in a normal mains toroid, but is actually a distributed gap or something because it is meant specifically to handle a dc flux. Weird stuff. Output cap is just some 1uF 250v block thing; not really suitable actually. With these, the frequency response is -3dB @ 890 Hz (!) A bit more work to be done there. It sounds pretty much like an AM radio at this point, pleasant to listen to but by no means hi-fi. Would probaby be just fine for voice only PA use though. Winding the volume up, this amplifier is not afraid to get down and get with it; it keep saying "more, more" but it was 11:45 pm by this time so I had to keep it down a bit.

Power output is a bit limited at the moment because the current sense resistors in the fet source leds that go to the drivers are way too big so it cuts in too early. I can see right here and now, to make a top end amp using this aproach would be a huge job, but that was never my intention anyway. I *am* pleased.

P.S. Will try and post some pics tonight, about 8 hours or so from now - lunchtime 12:55 pm.

[phase_accurate]

If you have already implemented NFB then there's a simple explanation why it sounds better with the higher supply voltage:

The gain of a switching amplifier output stage is basically proportional to the ratio of supply-voltage/peak-voltage of the triangle used for the PWM modulation.

With increasing gain the feedback factor raises proportionally.

Regards

Charles

[Circlotron]

Originally posted by Circlotron
the current sense resistors in the fet source leds that go to the drivers

That should be "in the fet source leads"

Originally posted by phase_accurate
If you have already implemented NFB then there's a simple explanation why it sounds better with the higher supply voltage:

Nuh, no feedback yet. Much fiddling remains to be done.

[LukasLouw]

Which modulator configuration are you using, Circlotron - the basic one from teh BCA white paper?

Lukas

[Circlotron]

Hi Lukas. Yes, that is basically the one I am using, only instead of two opamps making two out of phase signals that get fed to the comaprators, I am using a centre tapped transformer, the ends of which get fed to the comparators. This is just for the experimental setup so I can drive the transformer primary with whatever I like and also have isolation. I am using a UC3637 motor control pwm IC to do the job. The crummy sound I had at first turns out to be a crummy source, not the amp itself. Also I have changed the output filter cap from 1uF to 0.1 uF so it sounds a bit better now too.

I was going to post some pics, wasn't I?

[LukasLouw]

Gotcha,

Aren't you missing some of the possible switch combinations with the single transformer drive, as you will only be able to get the two combinations where one FET is one and the other off.

If you use their modulator, with 2 comparators, and separate drives to the FETs, you will get two more combinations where both are off and both are on, which should improve performance quite considerably.

Lukas

[LukasLouw]

Aaah scratch the previous, the caffeine hasn't kicked in yet. I just reread your reply, and see you ARE using two comparators.

Lukas

[djk]

"Since the output devices don't operate linearly, no SOA protection is needed (only max current limiting, if any at all). "

You need a duty cycle limiter to prevent shoot through on clipping.

Motorola has an app note with a discrete design, Linear Technology has a driver IC with it built in.