Analog Supply Over "Long" Wire

[dknguyen]

Hey all y'all:

Suppose I have a bunch of analog sensor boards with differential analog outputs that plug into a central 16-bit ADC board via twisted pair. The twisted pair will be "long" simply in that it will be off board and probably won't be more than 30cm/12".

Is it a losing proposition to feed out the supply from the ADC board (a low noise RF linear regulator) to all the sensor boards over twisted pair and just LC filter it on the sensor board? The sensor boards outputs are absolute and not ratiometric so the voltage that arrives at the sensor board isn't too important as long as it is stable and quiet. The linear regulator is relatively expensive so it would be nice not to need to sprinkle one on each sensor board but it feels really icky to not have one on each board.

 

[JonSea]

Twisted pairs should be relatively immune to noise. Unless you are in an extremely noisy (electrically) environment, you shouldn't have a problem.

I have used two twisted pairs with an overall shield to power remote preamps and to receive analog data (audio frequency range) in the tens of mV range with no problem...over 100 feet or more.

 

[dknguyen]

Twisted pairs should be relatively immune to noise. Unless you are in an extremely noisy (electrically) environment, you shouldn't have a problem.

I have used two twisted pairs with an overall shield to power remote preamps and to receive analog data (audio frequency range) in the tens of mV range with no problem...over 100 feet or more.

I'm not so much worried about "make or break" problems as much as I am wondering how much I would lose of the 16-bit resolution the ADC is capable of due to noise in the signal lines as well as the power lines to the sensors.

You said your overall signals were 10s of mV though but a 16-bit ADC at 5V has a resolution of 76uV. What kind of resolution were you measuring them with? I kinda wonder how many bits I might lose because if it knocks it all the way to 12 bits then I might as well go with the point-of-load regulators.

I guess the wisest thing to do would be to just make provisions on the sensor boards for such a regulator as well as a jumper to bypass it if it's not populated and test it out.

 

[crutschow]

What is an "RF linear regulator"?

 

[dknguyen]

What is an "RF linear regulator"?

I've nary a clue lol. That's what the datasheet calls it.
https://www.analog.com/media/en/technical-documentation/data-sheets/ADM7150.pdf

 

[AnalogKid]

For a 16 bit A/D you need a 17 18 bit accurate voltage reference. Without that plus some multi-sample averaging in software, you will not get 16 bit accurate results no matter what you do to the front ends.

Separate from that, what are the peak and continuous current draws of the sensor boards?

ak

 

[dknguyen]

For a 16 bit A/D you need a 17 18 bit accurate voltage reference. Without that plus some multi-sample averaging in software, you will not get 16 bit accurate results no matter what you do to the front ends.

Separate from that, what are the peak and continuous current draws of the sensor boards?

ak

The boards aren't built yet but the current draws will real low. Just a thermopile amp and a couple of photodiode amps. I'd be surprised if each sensor board will draw more than 50mA. No peakiness is expected.

These are sigma delta converters. The ADS1178.

 

[crutschow]

I've nary a clue lol. That's what the datasheet calls it.
https://www.analog.com/media/en/technical-documentation/data-sheets/ADM7150.pdf

Okay.
It means it's a low-noise output regulator, suitable for powering low-noise RF circuits.

 

[AnalogKid]

In a 16 bit system, the amount of power supply filtering and decoupling around the circuits is pretty much independent of where the nearest regulator is. Other then a possible issue with ground loops, no regulator and lotsa filters probably will work.

ak

 

[dknguyen]

In a 16 bit system, the amount of power supply filtering and decoupling around the circuits is pretty much independent of where the nearest regulator is. Other then a possible issue with ground loops, no regulator and lotsa filters probably will work.

ak

Okay. I'll give it a shot then.