SigmaDelta (PDM) maximum frequency ESP-S3

[eriksl]

I want to have a PDM frequency as high as possible, to prevent ripple in the output signal.

The documentation says you can set the "frequency" as a divider to the 80 MHz APB clock (4.x) or just "the frequency and the clock source" (5.x). It does not say what's actually the maximum frequency that can be used. Also it doesn't say what's meant with "frequency".

So my questions:

- what does "frequency" mean here, is the frequency of a full PDM phase (which means the frequency of the PDM pulses are 256 times this frequency, which seems very high to me, 256 * 80 MHz...) or is the frequency of the PDM pulses themselves (meaning the frequency of the full PDM phase is max 80 / 256 MHz), which looks more logical to me.
- what is the highest frequency than can actually be obtained? Apparently 80 MHz is possible, but I think various factors limit the actual maximum frequency, like the IO mux. Would something like 20 MHz be a safe assumption (and also easily scalable from 80 MHz APB clock).

I need a ripple-free DC signal, so every Hz higher means smaller filtering capacitors.

[MicroController]

- what does "frequency" mean here, is the frequency of a full PDM phase (which means the frequency of the PDM pulses are 256 times this frequency, which seems very high to me, 256 * 80 MHz...) or is the frequency of the PDM pulses themselves

It's the latter:

After scaling, the clock cycle is equal to one pulse output cycle from the modulator.

I need a ripple-free DC signal, so every Hz higher means smaller filtering capacitors.

As I wrote in the other thread: Check if you need an analog voltage in the first place.
Most LED drivers support being controlled directly via PWM. If your driver is the 'switching' type, using an inductor, this will already kind-of low-pass filter the effect of the control signal.

[eriksl]

Where did you find this quote?

Also, the current sourcing buck converter I am using can be controlled with either PWM or an analog signal (even on the same pin). Voltages between 0 and 2.7 V are analog current control, voltages between 0 V [0] and 3.3 V [1] are digital PWM. The nice thing about the analog control is that is actually limits the sourced current to a lower value so lowering the actual power use. PWM will also achieve this, but only indirectly, by intermediately drawing full or no power. Given the use of capacitors this, in the end, will also lead to a reduced power use, but I trust the direct way more. Also it avoids any chance on flickering. Not only from the PWM frequency (which I can set quite high), but I am more worried about the PWM interfering with the buck converter's switching speed. And I don't want to use capacitors in the power path after the current source, it would defy the whole PWM advantage thing.

FYI it's about this IC: PT4115 (or the functionally completely equivalent PAM2861).

[MicroController]

Where did you find this quote?

ESP32-S3 TRM v1.5, section 6.5.4 "Sigma Delta Modulated Output".

PWM will also achieve this, but only indirectly, by intermediately drawing full or no power.

And this is exactly what the switching step-down converter does all the time.

I am more worried about the PWM interfering with the buck converter's switching speed.

The converter's datasheet says:
f[DIM] (Max. DIM frequency) = 50kHz, and apparently the shortest pulse-width it supports corresponds to f[OSC].
As long as you stay within those limits, I wouldn't expect any problems from using bare PWM. The driver is designed for it.

[eriksl]

I am not afraid it won't work. I am afraid it won't work optimally. The switching is probably more efficient when the converter does it, at it's optimal switching frequency, than when I have it chop at random times. Where talking about ~20 power LED's, at different locations, so they need their own current source, and they need to be on ~12 hours a day, 365 days a year. E.g. 100 mW unnecessary dissipation per LED will make a difference here.

[MicroController]

I understand.
And I suggest to try and measure if there's any difference in consumption between PWM and analog control. (I don't think there is.)

[eriksl]

I intend to do so, indeed. Knowing whether it all was worth the while

[MicroController]

Let us know if you see a difference - I might be a bit curious now