Autocorrelation help!!!

[~MS~]

Hi everyone,

I am having some issues with this autocorrelation VI that I am building for my summer research opportunity. I need some help with my autocorrelation VI. I am suppose to autocorrelate collected noise. I am having issues with my autocorrelation(AC) VI, I am not getting the correct results. The AC function of labview is one part of my function. I implemented the remaining part of the function in labview and when I autocorrelate the data I don't get the expected results. We have a more sophisticated software in the lab that calculates the AC. I am using the data set used with that software and when I compare the results from my program and the software, my results don't match the results from the software.

I have attached a document file with my function and the results from my VI and what I am looking for. I am also attaching my labview vi and sample set of data that I am working with.

Any help would be great, Thanks in advance.

[~MS~]

Hi everyone,

I am having some issues with this autocorrelation VI that I am building for my summer research opportunity. I need some help with my autocorrelation VI. I am suppose to autocorrelate collected noise. I am having issues with my autocorrelation(AC) VI, I am not getting the correct results. The AC function of labview is one part of my function. I implemented the remaining part of the function in labview and when I autocorrelate the data I don't get the expected results. We have a more sophisticated software in the lab that calculates the AC. I am using the data set used with that software and when I compare the results from my program and the software, my results don't match the results from the software.

I have attached a document file with my function and the results from my VI and what I am looking for. I am also attaching my labview vi and sample set of data that I am working with.

Any help would be great, Thanks in advance.

[PedroMunoz]

How did you get those expected results? How does that other software works?

[johnsold]

Your formula did not make it through translation to a non-Word program:

I looked at the data and am not seeing anything which looks like it should produce the kind of autocorrelation you expect. When I run LV autocorrelation on the raw data get a very sharp peak at exactly the mid point and a straight line (log Y, linear X) to zero at both ends.  There is a slight downward trend in the data.

What produces this data and why do you expect the result you report in your document? What is the underlying physical model?

Lynn

[~MS~]

@ Pedro,

We have another setup in the lab for long sometime. My senior has been working with that setup and the software she uses gives the autocorrelation. I am using her dataset to produce similar results in labview. The way the other software works is it takes the data and just autocorrelates it.

Thanks

@ Lynn

Sorry about that issue I have attached the function again.

I was getting similar results as you have posted. But when I was going through forums I read in a discussion on how to get rid of useless data points. I am looking for exponential decay.  So, if you can open my VI you will see I am getting rid of the first half of the collected data set. And the results have to be plotted on log scale to see the decay. Now considering the zeroing part as per what I know about autocorrelation function it correlates a data from a selected period with data collected in pervious period and sees how much it differs.  So eventually, the Autocorrelation function is suppose to go to zero.  This data has been produced by my senior she has been using her setup for long time now. She ran her data in the software she uses and gave me her results. I am using here dataset to produce similar results in labview. I will post my VI saved in older version of labview.

Thanks

[johnsold]

That does not look like any autocorrelation function I have seen.  Where is the time shift?

I am not convinced there is much in the data you posted. There is a slight downward trend.  The spectrum is nearly flat. Smoothing the signal with a Savitsky-Golay filter shows the trend.  Fitting a straight line or an exponential results in almost identical curves: The difference between the two fitted curves is much less than the noise on the smoothed signal. The unsmoothed data gives the same fit within less than 1% change in any of the coefficients.

Lynn