Electronic – Avoiding ground loops with multiple coax cables

In fact, it is almost mandatory to run at least one ground wire per cable-- even when that ground wire is not needed for power supply reasons! Some cables have many ground wires in them, and it is not uncommon to see as many ground wires as signal wires in a single cable.

The reason for this is Signal Integrity (SI), which is a big area of electrical engineering that includes EMI, ESD, RF emissions, and others. It involves being able to send a signal from point A to point B with as little distortion and noise as possible without emitting radio energy, picking up radio energy, or picking up static electricity discharges (which is like radio energy).

The whole subject of SI is huge and way beyond what I can cover here, but let me briefly cover two topics of SI: AC signal return path and loop area.

When you send a signal from point A to point B, wither those are chips on the same board or different boards, that signal has to return back to point A. Normally this is done on the power or ground wires. The path that this return signal takes is called the AC Signal Return Path. Controlling this return path is super important for SI, and we do a lot of things to control it. We do things like control the PCB trace (or cable wire) impedance. We put decoupling caps on the PCB (also called signal bypass caps!). And, of course, we put ground wires in our cables.

The loop area is simply the area that our signal loop takes. From point A, to point B, and back to point A. The smaller this loop area is, the less our signal will get distorted and the less noise we will emit and/or pick up. To make this loop area smaller, we use multi-layer PCB's with power and ground planes. We use twisted-pair cables where a ground wire is twisted with a signal wire. We use multiple power and/or ground wires in a single cable.

Now, imagine your setup with two PCB's talking to each other, but no ground wire directly between those PCB's. A signal from PCB A goes to PCB B, then to the power supply via the GND wire, then back to PCB A. Depending on the exact configuration of cables, the loop area could be a square foot or more! But if the PCB's have a ground plane, and the cables have a ground wire, then the loop area could be just 1% of that (about 1.5 square inches).

Again, I overly simplified this but I'm sure you get the idea. SI is a huge and difficult subject that even "experts" who have been doing this for 10+ years have difficulty understanding. But stick with it and in time you'll get the hang of it!

Wow, I hope you have some slack in the timeline to redesign this after working with the prototypes. I would definately send a higher supply voltage and have an LDO for the 1.8V rail, perhaps even put in a isolated DC to DC converter (3-5V in, 1.8V out) this will eliminate one ground problem. If you only have 1.8V to drive your analogue and digital signals you are going to be in noise margin city. It sounds like a proton precession magnetometer projectc, see what others have done. I would run 80V DC, two fibres for clk and data and put the analogue to digital conversion in a shielded section at the instrument end.

If you have to go with the current sensor/cable setup then I would use bifilar wound chokes (with balun) on the inputs to eliminate common mode noise, try and use a transformer to isolate the ground. Hope the signal levels are large enough to get through ok.