Connecting Analog Signals Between Boards

Question

When you have a sensor with an analog output on one board, and an analog-to-digital converter (ADC) on another board, what's the best practice in connecting them together? For example, if you have a single-ended signal, should you twist that signal with a GND wire in the board-to-board cable? If it's a differential signal, should you twist the + and - signals together in the board-to-board cable? Or should the + and - signals each be twisted with a GND wire in the board-to-board cable? What are the best practices / rules-of-thumb / guidelines in this area for getting best results? If the answer is "it depends" then what does it depend on, and what is the decision making process for determining the right approach in a given context?

Obviously (?) first prize is integrating sensors and ADCs onto a single PCB, having a nice solid GND reference plane, and having routing discipline to separate your analog and digital domains, if that's possible. But I'm focusing on cases where it's not possible to do that.

Answer 1

I'm going to assume we're talking about signals that are less than a few kHz. In the single ended case, I would twist a ground wire with the signal. In the differential, I would twist the + and - alone with no ground, just like you said.

How you condition the signal is where some questions need to be asked. If:

• the application environment is noisy
• the boards are more than a few feet apart
• the sensor is something like a strain gauge or pressure cell

I would run the signal immediately into an instrumentation amplifier to half fill the span of the ADC, and send the signal over a differential pair, with a line driver IC. On the other end a difference amplifier can unbalance the signal and make it single ended for conversion, an inch from the ADC. The conversion back to single ended will then fill the span of the ADC.

If:

• the application environment is not noisy
• the boards are very close to each other
• the sensor is something like a strain gauge or pressure cell

I would still run the signal into an instrumentation amplifier, and fill the span of the ADC, but send the signal directly over to the ADC, single ended. If there isn't a lot of noise and the boards are close, there's no need for a high common mode rejection system.

If the signal was already greater in amplitude than a few hundred millivolts, there may not even be a need for the instrumentation amplifier. It was there to keep very small signals out of the noise floor.

Answer 2

I always have concern about power supply noise or current sharing a path with signals between two boards. I answered a recent similar question the same way but this was related to digital serial links.

Anyway, if one PCB supplies power to the other board and especially if there are switch mode power supplies used on either of the circuit boards you need to avoid PSU currents sharing the same wires as signals. To that end I'd like to see a differential amp at the ADC end and of course the previous answer mentioned an instrumentation amp and this is the best configuration for this job. This mainly prevents power currents trying to get down signal wires.

Twisted pair between the two boards for the signal too. But it all depends on how small the signals are and if small (or susceptible) then screened twisted pair is needed.