Hi,

Following this document I built the attached circuit on a breadboard, though I'm using an LM393 and a TL072 and some different component values in the triangle generator. It's not the greatest looking PWM but it works.

What I don't really understand is why unfiltered V_PWM is fed back through R4 to the error amplifier. Wouldn't you much rather first run V_PWM through a filter and feed that filtered signal back to the error amplifier?

 

C1 in the error amps feedback generates a low-pass filter with a corner frequency of 159kHz (equation 4 on page 5 of the reference).
That will filter the PWM frequency which is around 500 kHz with the component values shown.
But a larger C1 with a corner closer to 20kHz would seem more appropriate.

 

Oh, right. So feedback signal is filtered. My triangle generator is running at around 100kHz so I need something like 1 nF or more for C1. The breadboard is starting to look like a battle field so I'll start over and try again.

 

My triangle generator is running at around 100kHz

Why so slow?
It should be around 500kHz.
100kHz is likely too low for good high frequency audio fidelity.

 

Why so slow?
It should be around 500kHz.
100kHz is too low for good high frequency audio fidelity.

It's the generator you helped me with here:
https://forum.allaboutcircuits.com/...en-multivibrator-and-opamp-integrator.137603/

My current version uses an LM393 for the square wave and a TL072 for the triangle. This gives very nice crisp triangle. But I noticed I have a 100 pF cap on the integrator, not 50 pF as you suggested. Maybe because I don't have 50 pF, I need to check. All in all I just haven't had the time yet to see how fast I can make it go.

One of the things I might try is make a buck converter out of this in which case 100 to 200 kHz would probably be enough(?). But maybe a class-D amplifier too if I can push it to 500 kHz.

P.S. My claim in post #1 that "it works" can not be true because the pull-up resistor on the PWM output comparator was missing.