Class D Design Issues

It's tough. I've been working on one based off a PWM design for a year now and it's kicking my butt. I've got a prototype in the other room that plays, but doesn't sound too good and has a problem with ultra-low frequencies on the output. One of the few amp designs you'll come across that pass the signal through nand gates and optocouplers. On the upside, I think I have designed a successfull output stage for it. (Gotta take pride in the small victories.)
 
Well, ok. Since you asked nicely.

Basically it is PWM based off a triangle wave. I generate a triangle wave, and then bias it with an offset voltage to create two waves, one biased positively and one biased negatively. The seperation between them is my dead time between the pull-up fets being energized and the pull-down fets being energized so both fets arn't on at the same time. I then compare the input to these two waves, one wave for pull-up and the other for pull-down. My output stage is an H-bridge so any diffence in turn on/off times for my high side and low side switches is nullified by the fact that (ideally) the pull-up and pull-down current paths are (ideally) identical (ideally). I've also got a minimum pulse width circuit on there, to keep the bootstrap capacitors charged when it is driven into clipping.

That's it in a nutshell. I don't want to post really specific details or schematics yet because I haven't hit a wall yet (I like to solve my own problems, it's the fun of it).

* I know I've roasted my output stage when I turned the dead time control down too low, dissipated around 250W in fets with no heatsyncs. It isn't extra crispy, but damage was done. I think the subsonics are from this because they started after I zapped them
* I know something was odd in the output stage even before that. One fet just wasn't acting right. It doesn't look like a design problem, so I need to look into that.
* I was looking at a class-D design on this board and saw some ideas in the feedback loop I want to steal. I may rip them off.

Electrone said:
Check the power supply for the high side switch. It may be dropping too low if it is normally refreshed by bootstrapping.

I'll try that next time. I don't think it is the problem as I took steps to prevent it, but it is an easy test.


Also, I have to start moving so in a few days it is going in a box and won't come out for a month. So I don't want to get really into it now.


And believe me, I'll post details when I get it working. And if I can't, I'll ask when I've run out of ideas. I had this prototype sounding good under slightly different conditions. I just need to test more and be more scientific about it. ;)

My first comment was just that: a comment. I was new and felt like commenting. Pardon me.
 
Major design issues!!

Hi all,

I've been building amps for some time now and i have two very nice sounding prototype classD designs. One is a carrier based, and the other self osc. The problem is i would love to take one of the designs further and complete into a nice stereo amp for my daily amp needs ( the commercial amp in my living room is on all the time i'm in and is getting a bit tired ). The issue i have is this, EVERY CLASS D I BUILD HOWEVER GOOD IT IS ALWAYS ENDS UP DEAD EITHER BY STARTUP PROBLEMS OR BAD LOADS! I know that if used properly they are prob very reliable, but i NEED SOME KIND OF EFFECTIVE PROTECTION against overdrive, overcurrent, bad loads, startup and stall conditions. No matter how much i try i cannot impliment an effective protection system, this is the reason why regretibly i still cannot enjoy these wonderfull sounding amps in the real world, only on the test bench or at occasional parties.

Have any of you DIYers have similer issueswith your designs??

If so and you found an effective way of protection i'd love to hear about your experience.


Regards
Mad.P
 
Multiphase PWM

Has anyone endeavored to come up with a multiphase PWM amp, self oscillating or clocked? This seems to be the solution to the band limiting of moderate switching speeds and the loss of pushing it. There is an obvious difficulty in tuning each phase leg gain to the average of the rest without servo, but it seems possible. With eight phase legs, each one could run at a cool 500kHz, with smaller output transistors, and the total switching frequency would be 4Mhz. Available bandwidth would rival the best linear amplifiers.