if it does not get any better ,try to tie one of the speaker wires to ground with a 330 ohm resistor or something like that and see if it starts automatically
Interesting. I didn't think speed was that important for the bootstrap diode, but I think I misunderstood it's function. It's really part of a charge pump, sort of. Right? I can always whack in a Schottky diode instead. I'll try that when I get a chance to play with it. Also, I've seen people put a small resistor (4.7 ohm or something like that) in series with the bootstrap diode. I guess that charges the cap a bit slower, but I'm not sure I understand how that affects startup behavior.
Another thing is that I've been able to tune RV2 so it at least gets into oscillation every time, although it takes a few seconds of scratching. If I keep the self oscillating frequency around 200kHz it seems to work best. There's clearly a pole that's independent of the integrator parameters somewhere around that frequency, because that's where the gain maxes out. At the end of the day, the control loop is so full of non-linearities that I doubt you can accurately calculate a Bode plot from it. I've seen some people do SPICE simulations of it, but it's always full of a bunch of assumptions. So I guess it's mostly experimentation. But as I said, I've gotten from it sounding like scratchy crap to sounding decent, so it's getting there!
Another thing is that I've been able to tune RV2 so it at least gets into oscillation every time, although it takes a few seconds of scratching. If I keep the self oscillating frequency around 200kHz it seems to work best. There's clearly a pole that's independent of the integrator parameters somewhere around that frequency, because that's where the gain maxes out. At the end of the day, the control loop is so full of non-linearities that I doubt you can accurately calculate a Bode plot from it. I've seen some people do SPICE simulations of it, but it's always full of a bunch of assumptions. So I guess it's mostly experimentation. But as I said, I've gotten from it sounding like scratchy crap to sounding decent, so it's getting there!
The charge pump works at the speed of the switching, so it needs to be pretty fast,I Used a resistor to limit current peaks, something like 1 to 10 ohms,maybe remove or reduce c8 and c9 since they may be slowing things down too much, d4k5 does not have it,and while it isn;t the pinnacle of amplifier engineering,it works, so it must do some things right,lemme see if I can find a schematic of my own hunk of junk,it's been a while
Thanks! It’s getting there as you can see!
I noticed that your output filter has a pretty high cutoff (something on the order of 100kHz) and that you’re not using a snubber to deal with the LC resonance. I desperately want to get rid of my snubber, because the resistors get really hot and waste a lot of power. I assume you didn’t have issues with runaway oscillations? If I remove the snubber in SPICE, the whole thing oscillates at 20-ish kHz.
Anyway, this is what I have so far when dumping power into a 8 ohm resistor. The noise on the waveform is all switching frequency and should be way outside the audible range.
I noticed that your output filter has a pretty high cutoff (something on the order of 100kHz) and that you’re not using a snubber to deal with the LC resonance. I desperately want to get rid of my snubber, because the resistors get really hot and waste a lot of power. I assume you didn’t have issues with runaway oscillations? If I remove the snubber in SPICE, the whole thing oscillates at 20-ish kHz.
Anyway, this is what I have so far when dumping power into a 8 ohm resistor. The noise on the waveform is all switching frequency and should be way outside the audible range.
Oh yeah, the snubber’s gotta go. It’s wasting 30% of my power. Or at least R13 and R14 have to go WAY up.
On a positive note, the THD looks fairly low based on what I can see on my very basic scope. I’ll try to slap together a simple instrumentation amplifier tonight so I can feed the differential output into REW for a better analysis. Noise is still a problem. I plan to put an LP at the input and after the first amplifier stage. Also, my PCB is kind of messed up. I think I’ll get rid of some noise if I rev the board so I can get rid of all my bodges.
On a positive note, the THD looks fairly low based on what I can see on my very basic scope. I’ll try to slap together a simple instrumentation amplifier tonight so I can feed the differential output into REW for a better analysis. Noise is still a problem. I plan to put an LP at the input and after the first amplifier stage. Also, my PCB is kind of messed up. I think I’ll get rid of some noise if I rev the board so I can get rid of all my bodges.
a lot of the values on the schematic do not represent reality,the inductor was slapped together from a core from a pc power supply. I don;t have an oscilloscope,but the output ripple measured a half of the supply on the multimeter,speeding up the switching through the compensation network lowered it,but also increased distorsion for a reason I cannot remember.In an ucd type the output filter is part of the oscillator,and its values affect the speeds, as opposed to your type, with the snubbers you should only aim to damp oscillations way above the running frequency. like above 1 mhz
Ready for rev 2 of the PCB. Got rid of some annoying DC on the output by tweaking R7 a bit (not shown in the schematic). It's starting to sound really, REALLY good! This is fun!
Glad to hear, In ucd types, the lower the gain the better the soundthis may or may not apply to your type,but it is generally better to have lesser gain in the later parts of the signal chain, and seeing as you have some op amps already onboard, you could do like an times 5 or at least 2x to get a stronger signal into the amplifier, and lower it's gain so it has more loop gain
