Sigma Delta Modulation - any disadvantages?

[kxie]

Given the advantages of sigma delta modulated square waves over PWM square waves, why has it not become more popular? For example, hardware PWM implementations can be found on virtually every microcontroller, but I dont find any implementing sigma-delta modulated wave. (is it called something else?) Am I missing some disadvantage?

Here is a good visualization of sigma delta modulated wave. It seems to me that it can essentially be dropped in anywhere in place of a PWM wave.

 

[dick_freebird]

PWM is one step away from real time. S-D takes a lot
of cycles to do its thing. That's OK if you've got a high
enough clock or a low enough must-have sample rate.
But at your low-end microcontroller clock rate a S-D
ADC would be hard pressed to keep up with audio.
PWM also needs only a counter and compare, or two
counters, no thinking whatsoever. S-D gotta mash,
mash, mash and then maybe you get a number.
Dithering a PWM signal (for reasons I guess) only
makes it all more demanding, more thrash / switching
losses, constraints on minimim on / off time that a
PWM would likely never encounter (if sensibly designed).

Ask yourself what the purported benefit of S-D has to
do with your project's specific care-abouts. Why other
people prefer what they prefer? Doesn't much matter
unless they're doing what you're doing.

 

[FvM]

You don't mention an application range for SD modulation. It's preferred for medium speed high resolution D/A, e.g. audio signals, but no option for many power converter applications. Switching frequency and respective switching losses would be to high.

 

[kxie]

PWM is one step away from real time. S-D takes a lot
of cycles to do its thing. That's OK if you've got a high
enough clock or a low enough must-have sample rate.
But at your low-end microcontroller clock rate a S-D
ADC would be hard pressed to keep up with audio.
PWM also needs only a counter and compare, or two
counters, no thinking whatsoever. S-D gotta mash,
mash, mash and then maybe you get a number.
Dithering a PWM signal (for reasons I guess) only
makes it all more demanding, more thrash / switching
losses, constraints on minimim on / off time that a
PWM would likely never encounter (if sensibly designed).

Ask yourself what the purported benefit of S-D has to
do with your project's specific care-abouts. Why other
people prefer what they prefer? Doesn't much matter
unless they're doing what you're doing.

It's not clear to me how SD modulation is much more complicated.
SDM has: a number representing error integral, a couple additions/subtractions with target "duty" as the only variable to update the error integral, and a comparison with a constant to decide the next output level (high/low).

float error_integral = 0.5;
bool state = false;

bool step(float duty) {
error_integral = error_integral + state - duty;
return error_integral < 0;
}

The algo can be modified to use integers to purge floats.

PWM has: a counter which can be incremented, and a comparison with a variable to decide the next output level. Usually it is implemented with features like counter load value which im not sure has an analog in SDM.

Please elaborate on "thrash" losses. I'm not familiar with it. I'm also not catching onto how PWM dithering would increase switching losses.

You don't mention an application range for SD modulation. It's preferred for medium speed high resolution D/A, e.g. audio signals, but no option for many power converter applications. Switching frequency and respective switching losses would be to high.

At first I'm thinking SDM would be preferrable to PWM in power electronic applications, since its higher harmonic frequencies would translate to better THD metrics. But then I realized that at higher frequencies (e.g. 0.5 duty) it would switch more often than its PWM counterpart, resulting in higher switching losses. This makes sense.

If this is true, we should start seeing SDM replacing PWM in soft switching power electronics

bingo

 

[KlausST]

Hi,

The PWM has a fixed frequency. Two edges per period time.
Every edge causes an error. (dead time, switching loss...) so the error becomes higher at higher frequency. If the frequency is not predictable, the error is not predictable either.
With PWM you have predictable error, and good linearity.

Klaus