DAC like MCP4726, but with more bits of resolution?

[TERRA Operative]

I'm looking for a DAC like the MCP4726 but with more bits of resolution.
I need to create an adjustable DC voltage from 15.800V down to 458.69uV and the 12 bits of the MCP4276 isn't quite enough.

Is there any DAC that I can control with an Arduino/etc that accepts an external vref with more than 12 bits of resolution?

[Kleinstein]

There are a few 14 and 16 bit string DACs from Ti and likely others. The DAC8531 would be such an example - not sure if currently available and if there is a more afordable alternative.

[dobsonr741]

You’ll find hundreds of DACs with >12bit resolution if you use the distributors or the manufacturers parametric search.

[Kanbus]

AD5662
Easy to use. A bit pricey though.
https://github.com/avandalen/avdweb_DAC16

[TERRA Operative]

Cool thanks!

Is there a 20-bit DAC that is common to use with Arduino etc? (As in I don't have to start from scratch with libraries etc)?
So far I'm looking at the MAX5719.

[PCB.Wiz]

There is also parts like DAC1220
How fast does this DAC need to be ?  'adjustable DC voltage' suggest trimming / set points ?
For slow DACs, another popular method is merged PWMs, where you create multiple PWM outputs of 8~12 bits and use 2^N weighted resistors to merge them.

[TERRA Operative]

It will be used for setting precise DC voltages, from a few hundred nV to 15.8V

[voltsandjolts]

another popular method is merged PWMs, where you create multiple PWM outputs of 8~12 bits and use 2^N weighted resistors to merge them.

https://www.edn.com/combine-two-8-bit-outputs-to-make-one-16-bit-dac/

[PCB.Wiz]

another popular method is merged PWMs, where you create multiple PWM outputs of 8~12 bits and use 2^N weighted resistors to merge them.

https://www.edn.com/combine-two-8-bit-outputs-to-make-one-16-bit-dac/

That's a good reference, I also found this update, around improved step response, tho here they target an 8-bit settle / ripple as the benchmark.

https://www.ednasia.com/cancel-pwm-dac-ripple-with-analog-subtraction-revisited/

[Kleinstein]

The PWM ripple compensation makes the filter a 2nd order filter. It can still be worth it as it is a relatively simple way to build a 2nd order filter, but no magic.
For combining 2 PWM channels for extra high resolution it helps if the fine part is really only a small fraction, as the resistor (and switch resistance) accuracy limits the useful resolution for the fine part to some 5-8 bits. So for 16 bit total one would not use 8 + 8 bits with a 1:258 resistor ratio, but more like 12+4 bits.

[PCB.Wiz]

The PWM ripple compensation makes the filter a 2nd order filter. It can still be worth it as it is a relatively simple way to build a 2nd order filter, but no magic.

I checked that on Spice, and you are correct in one part of the curve : the ultimate ripple attenuation is very similar - a few % different at ~30uV in my spice run.
However, the rise time is quite different, as the plot shows  (red trace is second order filter, brown trace is ripple cancel from 2023 edn )
951us improved to 260us