myRIO Analog Inputs GND Reference: How to mitigate noise?

[FilipeTaveiros]

Hello fellows,

I have an intriguing issue to discuss: Noise.

I am using the all the twelve analog inputs available in myRIO. Ten of them are single-ended, whereas two of them are differential. The figure below, which was extracted from myRIO datasheet, illustrates these channels.

The datasheet also informs that all analog inputs/outputs are referenced to the analog ground reference (AGND) pin. There exists continuity between this pin and the digital ground reference (DGND) pin. I don't know if internally there is another components decoupling these pins, all I found out is that, measuring with a multimeter, the resistance between these pins is very low and then the MM beeps.

The majority of the inputs are being used to read currents of a three-phase electrical machine, by means of a current transducer.

I am using the differential inputs as single-ended inputs, that is, the negative input is grounded along with the AGND pin. This way, once the differential inputs are now configured like single-ended, regarding noise, I should have similar signals when reading from these inputs, right? In theory yes, In practice wrong.

I'm aware the differential inputs neglect noise much more than single-ended. However, as I pointed out, the negative input of the differential inputs is grounded along with the AGND pin. The figure below shows signals obtained from two single-ended inputs and one differential input (configured as mentioned):

 

As you can see, the signal read by the differential input has less noise than the others read by the single-ended inputs. As the negative input of the differential input is grounded, I thought that there should be no difference in the signals regarding noise. The three should be "dirty", or the three should be "clean".

Therefore, I ask if there is anything the community would advise me to do to improve the others inputs signal-to-noise-ratio (SNR), in other to get good-looking signals just like the one achieved with the differential input.

Thanks in advance.

[nanocyte]

How fast are the signals. Maybe you can get rid of some noise by applying a digital filter (averaging?)

As for the hardware, maybe your wiring is introducing noise. 

[FilipeTaveiros]

Hi nanocyte,

Thanks for the answer.

The signals are sampled at 20 kHz and the main frequency of the signals is 60 Hz. However, as I am conducting performance tests, during transients, there are expected frequencies up to 3 kHz.

I would rather prefer to not use filters, once they introduce phase lag.

About the paper you pointed out, it seems very rich. I will read to see if I can improve my system in any way.

In the mean time, have you read this paper? Any thoughts from it that could help my experiment?

Thank you.

[nanocyte]

Here's an experiment you can do. Try getting a "pseudo differential signal". That is, read the GND channel using one of the single ended inputs. That will give you and idea of how much noise there is on GND.

I'd have to see your wiring and your transducers to give you more specific advice.

Is 20kHz your max sampling rate? If you could sample faster, say at 40kHz, you could decimate your 40kHz to 20kHz by averaging consecutive samples. That wouldn't introduce extra lag and should provide a less noisy signal.