At work we have a DAQ system that is measuring a 0-10vdc signal from a fully isolated signal conditioner, i.e. the signal should be floating. The analog input card has differential inputs, with a reported 50Mohm input impedance. The manufacturer, along with other resources, recommend that a 10K or 20k resistor be used to provide a return path from the low side of the signal to an AGND on the input board. As you might guess, previous designers on this system did not bother to provide the return path. However, everyone says that everything is reading fine, with no problems ever observed or reported. I'm pretty sure I have a solid understanding of why the return path is needed, which makes me wonder how this setup is working despite the apparent design flaw. The designer DID connect the low side of the signal conditioner's power supply to AGND. But, I don't see how this accomplishes anything, as the power supply is supposed to be isolated from the signal output by a ridiculous 1500Vrms. I guess there could be an unintended return path somewhere, or maybe the circuit has never run long enough for a problem to show up. Anyway, I'd appreciate any thoughts on what could be going on.

 

Without seeing the actual circuit its difficult to say what is happening but if the signal conditioning input is a differential input then earthing one side (even through a high resistance) is likely to introduce noise into the system. The whole point of using differential inputs is to allow you to isolate the input circuit from earth so that you can use twisted pair cables to cancel any induced noise.

 

If there is a galvanic path from the signal conditioner output via the signal conditioner power source common to AGND, then wouldn't one have the required bias current return path? If there was a capacitor in series with the conditioner output then there might be problems for lack of a DC conduction path.

There are several recognised methods for providing bias return for instrumentation amps with isolated sources.

If the source is a balanced output the usual recommendation is to ground both lines through two equal resistors.

If the source is a low impedance device like a thermocouple then a single resistor to ground suffices.

There seems to be no concern on the various manufacturers' side concerning an possible increase in noise - provided the grounding is done correctly.

 

One of my philosopies is that no matter what you believe or calculate (or have a firm understanding of), the electrons are always right. That is, electrons obey the laws of physics...with no exceptions. You can lay in your hospital bed and swear that couldn't have happened, but the electrons are always right, even if you never figure it out.

You might believe that there is no ground return, but the electrons say there is at least enough ground return to allow them to register properly at the analog inputs. The analog board is measuring voltage difference, not current. I've seen instrumentation amps with DC needs in the picoamp range. (50 Megs times 100 picoamps = 5 millivolts of error.) Whatever the required current is, it is getting where it needs to be. You can tell because the output is correct, stable, and has been for quite some time.

I do have one idea: The 50 megs you mention is the differential resistance. May I suggest that you find a time when the power is shut off and measure the resistance of each input of the analog board to the ground that the input amp is using (without the input wires connected). Your DC signal is finding enough path to ground somewhere.

 

The most likely reasons are:

The differential input on the analog card would typicaly be fed through a resistor network to an opamp or instrumentation amp. This type of circuit has a finite common mode range, beyond which it may not measure correctly or may be damaged.

Or, if the analog input is actually fully floating with an isolation circuit, there will still be a safe voltage range before the on-board insulation breaks down.

Either way, the added resistor from input low to ground would prevent the input signal from exceeding the safe input range.

 

Thanks everyone for the replies. Here is a link to the signal conditioner, although it doesn't appear to offer any extra insight, it just shows everything being isolated.
http://www.dataforth.com/model.view.aspx?modelid=20

t_n_k, I measured resistance between between the low sides of the power input and signal output. I initially measured a finite number, but the DMM reading quickly rose to an "open circuit" reading. I'm guessing the DMM charged up some capacitance.

Bychon, I did try to measure resistance between a single input line and AGND. But, it only indicated an "open circuit" reading. In all fairness, I only have access to a hand held Fluke DMM. I don't know what the top of its range is.

Thanks again for the replies. I obviously don't have a real problem here, but I would like to make sure we have the return path accounted for in future designs. It adds a little bit of complexity, which is usually met with some "ain't broke - don't fix it" mentality.