How to extract periodic signal from raw XY data?

[avater]

The signal is periodic,so the acquired data(XY Graph) contains many of it.see pic.

I have to analyse the signal part,but don't have a way to extract it from XY graph data.

Any suggestions .Thanks.

[GerdW]

Hi avater,

use ArraySubset to get the interesting part from your XY data…

[Bob_Schor]

You call it XY data, but are the X values all equally-space, as though they were sample times?  If so, you can consider it a 1-D array plotted as a 1D Waveform (with the "dt" component equal to the time difference between the points).

With a periodic signal, the first (non-zero) peak in the autocorrelation function should correspond to the period of your signal.  Once you know the period, you can do things such as chop the signal into intervals corresponding to the period and averaging, which should enhance the signal-to-noise ratio.  You can then look for maxima and minima and use these values (and some visual inspection) to isolate the signal.

Bob Schor

[avater]

Actually I don't know the period of the signal.All I get is this kinf of periodic signal,raw data.

And I believe the intervals between signals are not presice the same.

By our eyes,it's easy to know where the signal is and get it apart.Is there any function in Labview that can do simila r thing?

[GerdW]

Hi avater,

Is there any function in Labview that can do simila r thing?

You can detect pulses using the PulseDetection…

(You can do it "manually" with a compare function and an array search operation.)

[Bob_Schor]

OK, it sounds like you need to learn a little Signal Theory.  Find an Electrical Engineer, show her/him the problem, and get suggestions.  Note the key question, are these uniformly sampled data, wasn't answered, and will be a very important part of how you proceed.

Bob Schor

[altenbach]

---

@avater wrote:

The signal is periodic,so the acquired data(XY Graph) contains many of it.see pic.

I have to analyse the signal part,but don't have a way to extract it from XY graph data.

Any suggestions .Thanks.

---

Let's quickly catch up what you mean by "extract", because it might not be clear to some of use. Obviously, your signal contains two distinct components:

1. a rolling wavy part
2. sharp spikes at regular intervals.

So, which one is the interesting part you want to further analyze and which one do you want to ignore? (does "extract" mean "remove" or "keep only"). Which part is interesting and which part is an artifact? Since the two components differ vastly is their spectral composition, it seems almost trivial to separate them and you already got some good ideas.

To give more detailed advice, it would help if you could answer a few questions and even attach some raw data. Are the x-values spaced equally or arbitrarily? Even if they are not exactly equally sampled, I would bring them on an equally sampled trace using interpolation. Is the distance between the sharp peeks an exact multiple of the time between points? What are you trying to get out of the data? (e.g. frequency and decay constant of the sharp signal averaged over many occurrences). Is the signal recorded in quadrature or plain?