Electronics > Projects, Designs, and Technical Stuff

how to choose a precision ADC for a DIY power supply?

(1/2) > >>

exe:
Hello!

I'm trying to find a ADC for my project (a lab power supply). The problem I have is that there are simply too many ADCs to choose from. I'm lost :(. May someone could suggest a a good ADC or confirm my choice?  So far I settled on ADS131A04 (TI) . I only considered TI parts, but I don't really care about the manufacturer, it just stars alligned this way, any other vendor is fine.

To narrow-down a little bit, I defined the following requirements. They are somewhat arbitrary, but I had to provide at least some specifications, otherwise I'd need to choose from too many parts.

resolution: 16bit or more because why not :). I want to have 20000 count measurements, though not sure if I have skills for that (noise, offset, drift, etc problems)
architecture: sigma-delta because presumably it gives better linearity comparing to SAR, but I'm noob here
number of channels: 4+ differential channels
interface: SPI because it's simple and robust
extra features: external reference, I want to attach MAX6226 in ceramic package, this should presumably improve long-term drift.*
price: preferably below $10, but since it's a one-off build I'm fine going a bit above if needed, but I'd like to avoid too-advanced parts that are hard to use  (i.e. too many settings).
package: everything but BGA
input impedance: presumably doesn't matter, I'll have an input buffer
freshness: among many available designs, I picked the freshest one. There are 10-20y.o. parts from Burr-Brown, etc, may be they are as good as modern or better, but I hoped that modern parts should be superior or at least better spec'd. Obviously I cannot prove that, if I had so much experience with ADCs I wouldn't ask this question :).

* I know that friends don't let friends buy Maxim parts, but I don't see any other 2.5V refs in hermetic packages on Mouser for a reasonable price.

EDIT: typos

David Hess:

--- Quote from: exe on February 17, 2020, 08:10:07 pm ---resolution: 16bit or more because why not :). I want to have 20000 count measurements, though not sure if I have skills for that (noise, offset, drift, etc problems)
--- End quote ---

Pick a resolution commensurate with your linearity requirement and pick a linearity requirement commensurate with your display counts.  This means treat a 16 bit converter with only 14 bits of linearity as a 14 bit converter.  But see immediately below.


--- Quote ---architecture: sigma-delta because presumably it gives better linearity comparing to SAR, but I'm noob here
--- End quote ---

Either would be appropriate but an SAR will require extra processing to remove excess noise.  Delta-sigma will be a better choice if lower sample rates are acceptable which is likely the case for a visual readout.  Performance above 16-bits will require careful attention to noise and drift outside of the ADC.


--- Quote ---freshness: among many available designs, I picked the freshest one. There are 10-20y.o. parts from Burr-Brown, etc, may be they are as good as modern or better, but I hoped that modern parts should be superior or at least better spec'd. Obviously I cannot prove that, if I had so much experience with ADCs I wouldn't ask this question :).
--- End quote ---

It is often the case that more modern analog parts are not any better than older ones going back a couple decades.  I would be more interested in availability and packaging.

exe:
Thank you very much, I think I'll stick with my pick as it looks to be reasonable. I'd say it well the specs well exceed my needs, but it's good to have some headroom. I also don't fully trust datasheets, I think they are often written to give a better impression of the parts than they really are.

Kleinstein:
The choice looks Ok, though I don't know why one would need 4 channels and simultaneous sampling. Chances are this converter is originally made for something like power metering. This group of ADC is OK:  better DC performance and gain stability than the audio converters and often cheap.
What else except voltage and current would one want to measure ?
For a multiple channel supply the channels are usually isolated and thus separate ADCs.

exe:
What I plan to measure. Besides current and voltage, I wanted to measure temperature, and I might have a separate current shunt for current sink. If I manage to use the same shunt for bidirectional current sensing, then I can live with a two-channel ADC (temperature can be measured by the ADC in MCU as precision is not needed, but this means writing more code, so I'd prefer not to do that).

I want to build a two- or four-channel power supply, each channel is a separate and isolated board.

Navigation

[0] Message Index

[#] Next page

There was an error while thanking
Thanking...
Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod