9

Context

In trying to make modular linear power supply with digital control, I figured I'd need a DAC to output the correct voltages into the feedback circuitry (two - one for voltage and one for max current).

I wanted these to be 16-bit so I could have ~1mV precision in a 60V supply.

Question

When looking on mouser in the DAC section, the 16-bit (and above) ones seem to start at almost $6.
When looking in the audio DAC section, they start at 16-bit and the cheapest is a little over $1 and it's a dual (stereo) 24-bit converter.

The sample rates on the regular DACs for entry prices are not impressive either when compared to the seemingly ubiquitous 192kS/s of the audio ones.

So, is there any inherent disadvantage to the audio DACs? And if so, can it be avoided by picking a 24-bit audio ADC for a 16-bit application?

TrisT
  • 193
  • 4
  • 2
    One obvious problem is that audio DACs typically are AC coupled but you presumably want a DC level. I've never tried running one DC coupled, but I suspect a lot of them are not going to perform as well outputting a constant voltage. – user1850479 Aug 24 '23 at 03:17
  • @user1850479 the one I'm currently looking at (PCM1754) doesn't seem to be internally AC coupled, they instead specify a 10μF cap on the outputs for decoupling. There are filtering/buffering circuits that show direct connection to an opamp (with the decoupling cap after). That would (I think) disprove also the idea that they're engineered to drive the capacitance in a way that would be detrimental to it's use without it. – TrisT Aug 24 '23 at 03:37
  • What I meant is that I'm not sure how accurate the (in this case) delta sigma modulator is at producing a constant voltage. The output driver probably doesn't care about the load, but that doesn't necessarily mean it works very well. – user1850479 Aug 24 '23 at 03:47
  • @user1850479 why would they advertise it at x-bit if it's not accurate to within that? or do you mean in terms of it's output DC offset? – TrisT Aug 24 '23 at 03:53
  • As a side note, take into consideration that a 16-bit converter is not going to really provide 16 bits of resolution - look up ENOB. If you really need 1 mV accuracy with 60 V full scale, 16 bits is not going to be enough - 18 might barely make it, 20 or more should be okay. – Vladimir Cravero Aug 24 '23 at 11:50

2 Answers2

20

Audio DAC are optimized for low distortion specs (THD, SINAD, etc.) at the cost of worse DC specifications. Compare the datasheets, the PMC1780 for example (just the first one that came up in the mouser search link) claims a decent 106dB SNR, but its gain error is 6% of Full Scale Range, and its Bipolar Zero Error (i.e. offset) is +/-80mV. Not important for AC coupled audio, since the offset error goes away when it's AC coupled, and the human listening to the audio will just adjust the volume control, the absolute gain doesn't matter.

Precision DAC are optimized for DC performance specifications which are important when using the DAC to control an actuator or close a servo loop. Basic DC accuracy specifications like offset error, gain error, (sometimes you see TUE Total Unadjusted Error); non-linearity specifications like INL/DNL, and in some cases also distortion specs THD/SNR/SINAD.

Similar to op-amps, there are clusters of application areas that benefit from optimizing certain parameters at the cost of others. (Price is a parameter not shown on the datasheet but prominent in engineer's mind.) The optimizations that favor precision, low-offset, low-drift, low-power op-amps are contrary to the optimizations that favor high bandwidth, fast settling op-amps. Same idea with DACs.

MarkU
  • 14,413
  • 1
  • 34
  • 53
  • It makes perfect sense, thank you. It seems the datasheet doesn't mention how these (gain error, offset) change with temperature. If they don't (or don't by more than the desired precision) it could be calibrated out no? (either in software or hardware) – TrisT Aug 24 '23 at 05:48
  • 3
    @TrisT You can only 'calibrate out' things if you have repeatability. That's what is not spcified in a 24 bit audio DAC, DC repeatability. Your ear doesn't care if the gain or the DC offset drifts 1% between the start and end of a song. It might not, but it's not specified not to. If you had the time and inclination, you might buy and test several DACs, to see what their actual performance against time and temperature is. – Neil_UK Aug 24 '23 at 07:29
  • @Neil_UK that does seem like a fun project indeed. if it works perhaps the same can be done for the audio ADCs as well. I have been wanting to make one of those fancy digital multimeters with the integrating adc, this would be just the excuse. Another approach though might be having an extra 2 ADCs and implement a feedback loop, it would be more work but still comes out cheaper. – TrisT Aug 24 '23 at 21:07
4

From an audio perspective, the notion of "dynamic range" can be used to describe two concepts:

  1. The difference in amplitude between the loudest sounds, and the amount of noise that would be present even with the quietest sounds.

  2. The difference in amplitude between the loudest sounds, and the amount of noise that will be present when the loud sounds are present.

For many audio tasks, the first figure would be more important, but for many other tasks where a DAC output is used to control something, the latter figure may be more important.

If one were to feed the output of a 12-bit linear DAC into a 12-bit linear volume control, then at low signal levels the system may behave as a DAC with 24-bit resolution while the volume-control DAC is at its lowest setting, but as the signal to be reproduced gets louder and the overall gain increases, the amount of noise on the signal would likewise increase. For many audio-related tasks, such a device might yield better-sounding audio than would a 16-bit linear DAC, but for many control-related tasks the performance would be inferior.

supercat
  • 45,939
  • 2
  • 84
  • 143