Common ground for accurate measurement and analysis

[navneet]

Hello everyone,

I wanted your views on analog output and input configurations. Say we connect two resistors in series and provide an analog output sine signal across it of 10 V. Next we measure the potentials across the two resistors using differential input configurations, on AI channels 0 and 1. What I am curious about is:

1) Do we have to connect the analog output ground and analog input ground together?

2) If they are connected together, then am I correct in saying that both potentials measured on the differentialchannel are referenced to a common ground?

3) Does this help to improve SNR?

Will be grateful for your replies. (The two resistors here are an example simulating the loads across which am measuring the potentials).

cheers,

[Buechsenschuetz]

Of course it would be helpful to indicate which NI board/system you are using. Usually you find hints and diagrams for input signal and GND wiring in the manuals. 

> 1) Do we have to connect the analog output ground and analog input ground together?

Usually this is not necessary. Both terminals usually are connected internally, at a well chosen point. Connecting these terminals externally may even decrease S/N ratio. The AI GND terminal usually is used as a reference point for the input signals, the AO GND terminal as a reference point for the output load.  

> 2) If they are connected together, then am I correct in saying that both potentials measured on the differentialchannel are referenced to a common ground? 

I do not quite understand your question. Connecting AI GND and AO GND has nothing (or little) to do with the reference point of the input signal. Also, differential input signals are NOT (or not necessarily) referenced to AI GND. However, remember that many systems require an input bias current path between the signal input terminal(s) and AI GND, but you have to refer to the manual of the system you are using for details on this.  

> 3) Does this help to improve SNR? 

See above. There are several documents on this site covering the issues of proper input and GND wiring, such as

http://zone.ni.com/devzone/cda/tut/p/id/5362 

[Buechsenschuetz]

Maybe I got you wrong about the analog output. I was thinking you were referring to the AO of the measurement system you are using, many systems also have analog outputs. 

But probably your "analog output GND" rather refers to the GND output of the signal generator.

If you are going to use a differential input configuration the output GND of the signal source does not have to be connected with any GND terminal of the measurement system. With a differential input configuration, one input terminal can be regarded as the reference point and the other one as the signal input. Connecting the output GND of the signal source to the GND of the system is not necessarily and probably will not help anything.

It's somewhat different with connecting the shield of the signal wires. The shield always has to be connected to a GND terminal. For details see the NI document mentioned.  

[navneet]

Thanks Buechsenschuetz

i am using USB 6212. The reason I asked about this is, the instrumentation amplifier of the card is referenced to AI GND. So I was thinking is it better to connect both the AO GND (of the card) and AIGND together, such that there is a ideal common ground. Coz eventually, the differential inputs are referenced to the internal ground of the instrumentation amplifier (which in this case would be the quasi-universal ground). Thanks for the discussion.

[Buechsenschuetz · Accepted solution]

As mentioned, usually the GND terminals are connected internally so there is no actual need for an extra connection between the AI GND and the AO GND terminal. 

If you connect them externally there is risk to establish a "ground loop", i.e. a closed circuit covering a certain area which might pick up noise from the environment. "Double grounding" usually should be avoided. However, maybe in your setup you will not see any noticeable difference with or without external connection - but this is no evidence that the theory is wrong.