Some DDS SG will allow phase modulation with white noise.

Possibly the cheapest way of doing this is using Parallax Propeller II. It has excellent 8-bit DACs - literally one per GPIO pin. These DACs have a fixed low impedance and are purely resistive - a couple hundred ohms IIRC. They have a voltage output that is much easier to deal with compared to R-2R DACs that have variable impedance and thus need a virtual ground and a transimpedance stage.

The output of three DACs is combined into a higher resolution output using just three resistors, with some overlap between each DAC so that no values are missing.

With a slow and affordable external 24-bit ADC, the transfer function of the contribution of each of three DACs to the output is determined and stored in a lookup table in external RAM (HyperRAM on the PropII devkit).

When the ARB waveform is available, it is converted into 3-byte DAC values using the lookup table. These values fit comfortably in the internal hub RAM of PropII.

Finally, the smart pin functionality of PropII as well as hub-to-spoke streaming is used to stream the ARB waveform to the DACs repeatedly. This is friction-free, since PropII was designed with HDMI and VGA output in mind. It's super simple.

As everything warms up, periodic recalibration can be done using said 24-bit ADC. Assuming say 10 samples per second, it takes just over a minute to get the complete calibration table measured out. Instead, center/endpoint readings can be taken - all in under 1 second - to rescale the lookup table as resistances drift.

I've done this and it works surprisingly well for how cheap it is ($40 in parts sans UI).

There is a bit of a devil in the details when it comes to choosing which three pins to use - ultimately, several adjacent four-pin groups can be measured and the one that gives the lowest THD and IMD can be selected, using a bed of nails external multiplexer. But even a fixed pin assignment is pretty good. I can only measure THD down to 0.01% in audio band, and the 10kHz sine has distortion below my ability to measure. I have not measured the distortion at higher frequencies, so that's where things may get "interesting". But the building blocks are all quite docile, so it at least on paper has a chance of being pretty good - probably much better than any discrete R-2R solution that's not extraordinarily expensive.

Cheers, Kuba

On Mon, Dec 4, 2023 at 4:21?AM Gavin Watkins via <gavin_watkins=[email protected]> wrote:

Another challenge with R-2R DACs that I have encountered is slew rate. There will be stray capacitance on the board. For high slew rate, you want a large current to flow through the resistors, but are of course limited by the output current of the microcontroller etc to about 20mA. The resistors in the network limit this and the result is a restricted slew rate. I encountered this with a 12-bit R-2R DAC in a ramp generator where of course you want a fast drop after the ramp up. I had to add a transistor to switch in and help sink the current across any stray capacitance.

Another challenge with R-2R DACs that I have encountered is slew rate. There will be stray capacitance on the board. For high slew rate, you want a large current to flow through the resistors, but are of course limited by the output current of the microcontroller etc to about 20mA. The resistors in the network limit this and the result is a restricted slew rate. I encountered this with a 12-bit R-2R DAC in a ramp generator where of course you want a fast drop after the ramp up. I had to add a transistor to switch in and help sink the current across any stray capacitance.