Shielding and Ground loops

Question

I have various pieces of equipment that all need to be connected to GND (Chassis GND). The issue I'm having is that I'm connecting an ADC card inside a computer through a metal box with custom electronics that connects into a cryostat (metal shield).

The problem is that if I just use my standard cables then the computer GND is connected to the metal box with the electronics in which is in turn connected to the metal of the cryostat. Now the computer is connected to GND via the plug and the cryostat is connected to GND via a big metal strap to the building GND. This to me smells like a bad case of a ground loop.

So I'm thinking that I need to break the shield somewhere on one of the cables. The questions is where? The electronics are amplifying a rather small signal from the cryostat so I'm guessing that I want to try and keep that shield connection continuous. I was going with breaking the shield at the box on the cable that runs to the computers ADC. Is this a good idea? Should I not worry if the Computer GND and cryostat GND are pretty much connected to the same power strip?

Note that the electronics GND should be floating from the Chassis/building GND.

Answer 1

Yes, it sounds like (a little confusing) you have a ground loop problem, and yes they can matter, especially when trying to measure small analog signals. If all grounds tie back to the same outlet strip via relatively short line cords, then it would probably be OK. However, you say that this cryostat thing (whatever that is) is connected separately to building ground, so that is obviously not the case and it's confusing therefore why you brought it up.

In general, it's good to convert analog signals to digital as close as possible to the source, then ship around digital signals. Those are much easier to isolate, like via opto-couplers, pulse transformers, radio, etc. In other words, a old fashioned A/D card in the computer is not the best overall architecture from a system level point of view.

However, look at the A/D card carefully. Most likely it can be configured for single ended and differential operation. This is a case where you want differential inputs. The cryostat thingy may produce a ground referenced signal, but take its ground and output signal as being differential. This will essentially subtract the ground offset from the signal before converting it.

This trick will only work up to some frequency, probably a few kHz or low 10s of kHz. It should work pretty well in subtracting off any ground signal due to 60 Hz or 50 Hz power line return currents accross ground paths in the loop. Sharp common mode spikes can still confuse the diff amp in the A/D and show up as noise in the final output. It's worth a try though. If it's not good enough, go back and convert to digital at the sensor, then opto-isolate the digital telemetry signal.

 

Answer 2

There is a tutorial on ground loops and other forms of electrical interference at the Loop Slooth web site.

This tutorial shows that ground loops behave differently at low and high frequencies. At low frequencies what counts is the resistance of the conductors forming the ground loop, but not the physical layout whereas at high frequencies it is the other way around because at high frequencies what counts is the inductance not the resistance. The cross-over frequency is given by R/L where R is the cable resistance and L is the loop inductance.

These issues are also discussed in a Review of Scientific Instruments paper (also at a link on the website).

The tutorial explains how ground loops result from the distinction between Faraday's law and Kirchoff's law (electronics is based on Kirchoff's law which is valid for DC only, whereas the real world of time-dependent currents involves Faraday's law). It also discusses the relation of ground loops to other forms of interference such as inductive coupling, electrostatic coupling, and radiative coupling.

Answer 3

1. It is often believed that the key threat of a ground loop is that there is a substantial DC difference between grounds (e.g. 0V and 0.3V). In which case connecting the two with a low resistance wire may cause high current and damage one (or both) of the devices. It can be the case if current consumption of the devices is orders of magnitude different between them (say, a rotary engine connected to a mobile phone). However, this is unlikely your case. So, DC drop should not be a problem.

2. Ground loops are usually a problem because they act like a transformers converting changing magnetic field (caused by operating nearby devices like power grid wires, bulbs, switches, engines, etc.) into induced voltage.

In an ideal case, if there would be only one pair of wires (connected to a power source) closely located to one another everywhere, the current loops would be the same and induced EMF would be the same. Thus, measuring voltage difference between the wires at any point would produce the same result whether with a magnetic field or not.

However, in practice this is rarely the case, wires don't go in pairs and they don't go always go close to one another. This way, noise appears (EMFs induced in different loops are different).

From practical standpoint you can do the following: a) ensure all wires go close to one another as much as possible (no big loops of isolated wires are formed). b) ensure there are no time-varying magnetic field sources (noise aggressors) nearby.

In your case, the two devices have different connections to ground (mains gnd and building gnd), so, likely, a giant loop is formed, which isn't good.

The actual noise level will depend on parasitic magnetic fields surrounding your setup. If there are no strong sources of time-varying magnetic field in your laboratory, the ground loop may not be a problem.

Breaking a shield may not be a good choice because:

1) it will change impedance of the cable which can be troublesome at high frequencies.

2) The receiver's gnd will float relative to the driver's gnd. Any stray current appearing at the receiver gnd will affect the signal.