Is it true that the ESP32 S3 will remove the DAC?

[MicroController]

Apparently, when using a first order RC filter, using a capacitor of 470 nF and the intrinsic impedance of the control pin (200 kOhm), you'll get a -3 dB cutoff point of around 2 Hz.

You still need an external resistor between the ESP and the capacitor; the 200k pull-up on the control pin doesn't work as part of the filter.

[eriksl]

It's the final impedance that counts, not whether there is a resistor (or any other component with some impedance) is fitted.

Having said that, it's better to add an additional resistor anyway, because the internal resistor to the control pin is a pull-up and not a pull-down as it should be for a RC filter. Also it probably has an internal current source for a few micro-amps connected too.

As soon as I have received the 100 μH coils I can start experimenting!

[MicroController]

Example: https://tinyurl.com/2y675qtq

[eriksl]

Nice

And now with 470 nF instead of 4700 nF. I prefer to use solid state components, the 470 nF can be obtained in a ceramic version and is still quite small.

Also note that the 10 kOhm is a stand-in for the output impedance of the ESP32's GPIO pin. I think the output impedance will be quite high, what happens if you use 100 kOhm instead (and thereby removing the physical resistor)?

And also note that the PDM frequency seems to be able to reach higher frequencies, e.g. 1 MHz (but I'd still like to have that confirmed), can't find specs on that.

[MicroController]

Nice

And now with 470 nF instead of 4700 nF. I prefer to use solid state components, the 470 nF can be obtained in a ceramic version and is still quite small.

Also note that the 10 kOhm is a stand-in for the output impedance of the ESP32's GPIO pin. I think the output impedance will be quite high, what happens if you use 100 kOhm instead (and thereby removing the physical resistor)?

The ESP's output impedance will be on the order of 10-100 Ohm.
Notice that the R in the RC filter creates a voltage divider with the 200k pull-up. This causes an offset between output voltage and duty cycle, and prevents the control pin from seeing a voltage lower than Vcc*(R/(200k+R)) even at 0% duty cycle. Lower C -> higher R -> higher "0%" voltage into the control pin.

[eriksl]

Yes you're absolutely right. I was confusing an GPIO pin in open drain mode. In push-pull mode the output impedance is quite low, especially when led-grade currents can be delivered, like the ESP32 does.

So indeed an extra resistor is required to reduce the impact of the output impedance on the RC filter, duly noted

The LED driver needs a very specified range of voltages for dimming, between 0.3 V and 2.7 V from the top of my head. So that is definitely something to take into account. It will probably need some pulling up or down. I am not an expert on calculating these, so I'll probably just try.

BTW don't forget we're talking PDM here. There is no duty cycle. There is a pulse density, it's not really the same.