NI-USB 6211 measurement

[miniMe]

Hello all,

I have an issue with a measurement made using a NI-USB6211 interface.  I have to measure ripple of a 15Hz saw-tooth shaped signal, of a 80mV (aprox) DC signal, using AI4 in differential mode connection. Measurement is made on 10000 samples taken a 10KHz. "Basic Averaged DC-RMS" VI is used to extract DC component of the signal, it's value is correct, and "Amplitude and Levels" VI is used to make peak to peak amplitude measurement of the AC component, I was expecting this value to be the ripple value of the 15Hz signal.
Graphic representation of samples indicate a noise like signal, with 1-2 mV peak to peak amplitude..

From this it seems that desired signal is buried in noise and probably it might not possible measure it without additional signal processing.

I looks like one part of the problem might be due to the fact that we haven't used an anti-alias filter between DAQ input and device under test. To somehow compensate this we have oversampled and than post-filter samples, with a digital filter with  fc=1kHz. Peak to peak amplitude is going down, however I'm still seeing noise rather than a clean sawtooth signal.

Question is, if it possible to extract the signal by further lowering filter low pass corner frequency and eventually increasing the filter order - I have used a digital FIR filter with 31 taps or still the best solution would be to  add the anti-alias filter before analog to digital conversion.

Thank you

Regards

[santo_13]

I guess your signal is weel under an mV and you're trying to work around the limited ADC resolution by oversampling but you're further blocked by the system noise which is higher than the actual signal.

What is the actual signal range?

What voltage range are you using for the instrument?

Can you share the wiring diagram and also the VIs you implemented?

Can you share some sample collected data?

[Kevin_Price]

Whatever high-freq noise is allowed to get through to the converter due to the lack of an anti-alias filter stage *CANNOT* generally be cleanly removed after the fact with software-based digital filters.  In my experience, noisy environments tend to contribute many frequency components worth of noise.  Sampling makes this noise get "mirrored" across the Nyquist freq, and the net result is that the noise tends to "spray" itself all across the bandwidth you intend to capture.

The first thing you need to do is add in an anti-alias filter.  Just be aware that many real-world measurement systems also struggle with AC line noise at 60 Hz or 50 Hz, depending on what part of the world you're in.   You're going to have some tradeoff there between retaining your ~15 Hz phenomenon while also attenuating above 50 Hz.

-Kevin P

[miniMe]

Hello Santhosh,

I was able to measure amplitude of 15Hz single tone from the samples and it's peak to peak amplitude is 150-160uV.

This is an indication, but it is not the correct value of the ripple, as the real signal resembles a triangle shape (not saw-tooth as previously indicated).

Channel I make measurement is configured to +/- 0.2V input range.

Thank you for your answer.

Best Regards.

[miniMe]

Hello Kevin,

I am able to filter out some of the noise however, not all.
I'm now trying to reconstitute the signal I have to measure, from spectral components by to running FFT on the samples. On FFT array, I will keep only those frequency bins that are present into the spectrum of a triangular signal shape (as my signal shape is), also frequency bin 0 to preserve DC component and discard all other. In the end i will run an IFFT to reconstitute the signal, than measure it.

I hope in this way it get closer to the real value than by filtering.

Thank you for your answer, I agree with your observations, anti-alias filter is a must, initially it was planned to use a DAQ with internal anti alias filter, however we ended up with this card an no time to design and make a good external filter 🙂

Best Regards.

[Kevin_Price]

FWIW, even a simple external anti-alias filter such as an appropriate RC combo would be better than none at all.

An FFT approach *does* sound promising, wish I'd thought to suggest it.   I'm no full-blown expert on FFT stuff, but attempts to reconstitute the time domain signal via inverse FFT is not something that'll always automatically come out correctly.  For example, you can't properly recreate the time domain signal if you do a PSD-style FFT.  I"m not certain, but I think this is also true for Mag/Phase.  You'd probably need to work with the Re/Im version, at which point you'll have a little more work to do to get the correct amplitude info.

All subject to correction by someone who knows better, just trying to help.

-Kevin P

[Henrik_Volkers]

If your samplerate is in sync with the 15Hz ... no big problems with the FFT IFFT approach, otherwise you have leakage and the bin sorting isn't that easy 😉

Check with simulated signals and try frequencies a little off the perfekt 15 Hz.

Again: Aliasing filter is strongly recommended

AND: The 6211 is a multiplexing card... so prone to crosstalk / settle time error !  Have a look at the spec. Fig 1

If you only use one channel, no problem.

[miniMe]

Hello Kevin, I'm not that lucky to get away with just a plain RC filter 🙂 , however I will add this anti-alias filter.
I've been able to simulate signal reconstitution, implementation was made with Re-Im FFT and inverse FFT, I'm still tweaking some parameters before integrating it with DAQ application.
Anyway thank you for your comments and good ideas !

Best Regards !

[Kevin_Price]

You might also consider trying the function "Extract Single Tone Information".  I haven't found occasion to use it myself over the years, but was reminded of it in a recent thread about FFTs.   

-Kevin P