Let’s suppose I have a few home-made instruments I’m keeping in a rack, each in a 2U chassis. They all plug into a big analog supply at the base of the rack. How do I set up the grounding in the instruments to avoid creating ground loops through the BNC cable shields? Everything must be DC coupled.
The problem is that the BNC shields, and every other "ground", which should really be called "common", are connected to one of the DC power wires, usually the negative one, and the current drawn through that wire by each instrument creates a different voltage offset from where they are all connected together. So the answer to "How do I set up the grounding in the instruments to avoid creating ground loops through the BNC cable shields?" is that you do not, which leaves the following options:
1. Isolate the instruments at their power inputs. This is how AC line powered instruments do it, albeit usually poorly because of large amounts of capacitive coupling. The Tektronix TM500 series instruments do this by providing a pair of floating 25VAC transformer secondary windings to each instrument. I sometimes include an isolated DC-to-DC converter at the 12 volt power input inside the instrument to do exactly this. If regulation is not required, then an inverter, which has some advantages for low noise, could be used to convert 12 volts to isolated 12 volts at each power input.
2. Use differential signalling. Audio instruments commonly do it this way with XLR connectors which provide the two differential signal pins and a common pin. Higher performance test instrumentation may do this with a pair of coaxial connectors. Oscilloscope differential probes which connect through a single coaxial connection to the instrument do this.
3. Remote sense the common mode signal through a single ended connection. This is essentially the differential signaling method in 2 above but using a single coaxial connection and is also common with audio instruments. The BNC shield connection is used to sense the remote common mode voltage. Note that sensing could work either way with the source sensing the common mode voltage at the receiver or the reverse. Small Signal Audio Design by Douglas Self discusses how this is done in audio applications. Sometimes this is done with triax cable or shielded differential cable.
4. Galvanically isolate the input or output or both. This means generating a floating power supply for each input or output and transferring the signal across the isolation barrier. The old 10Base-2 Ethernet standard which uses BNC connectors did this. Oscilloscopes with isolated inputs do this for each input.
Cleverscope does this for its function generator output.
Next to the rack is a device which can only be driven in a single-ended manner. In the end, there must be multiple connections to the ground of the DUT. The technique espoused in Ott and other books on noise is to drive the shield and center conductor on the source and then sense using a diff amp the potential between center and shield on the load, avoiding shorting the shield to the chassis. At the end of the signal chain, I can’t do this, as the device is too simplistic to sense using a diff amp. How do I work around this?
That is method 3 above and it can work in reverse. Sense the remote BNC shield voltage from the transmitter and adjust the center signal accordingly, so the difference amplifier is at the transmitter side rather than the receiver side. I have done it by driving the center signal with a current instead of a voltage and then providing a resistive load at the receiver but this requires a special cable with the resistive load built in. Sensing the remote BNC shield voltage with a difference or instrumentation amplifier at the source is more universal and works with a standard BNC input.
Method 4 above works even better but is more complex because it requires an isolated power supply and it requires the signal to be transferred across the isolation barrier somehow. Method 1 above with an isolated DC-to-DC converter at the DC power input is simpler.