averaging triangular waveform

[suethechemist]

Here's some data: both were acquired with a triangle frequency of 333mHz, and both show the drift. "data1" contains the signal, "data2" doesn't (thought this might be easier to work with initially - there are less data points)

[unclebump]

Here is a picture of the vi as it is running. I had to save the excel file as tab delimited text. I read this into labview and then used the Denoise signal vi from the signal processing toolkit. Then I used the Detrend vi from this toolkit. I made a waveform chart with all four signals, then put this inside a cluster. I couldn't get this cluster into an array, so I made a while loop to let you select which original signal to display.

[unclebump]

Here is the vi. There are a bunch of filter settings that can be applied to the denoise/detrend vi's.

[suethechemist]

Thanks uncle bump - can't open the VI though, and I don't have the toolkit! Denoise certainly looks good, but I'm not sure I understand what's happening with detrend - does it produce derivatives of the peaks/dips? What is the trended waveform?

[unclebump]

I don't think it uses the derivatives. Here are the labview help files for the denoise and detrend vi's.

[suethechemist]

unclebump (if you're still out there) - could you run these files through the denoise VI and send the data back to me? I want to know whether they contain any extra info buried in the noise. There should be tiny dips within the noisy line of data IN ADDITION to those we've been discussing. I'd love to upgrade to 7.1 and buy lots of toolkits but I have to be able to justify spending the money. If this works (it's quite a long shot, there are lots of experimental reasons why the dips might not be there) I might have a chance... Either post the data here or e-mail it to me (sbelshaw01qub.ac.uk), when you have a spare minute of course!

[DFGray]

Attached is a simple try at Fourier filtering that does about what the LabVIEW toolkit does in this case. Much as I would love you to spend lots of money on NI products, you probably don't need to. This is brain dead Fourier filtering that I suspect messes with the data a bit. You could get better results by subtracting the known signature of a triangle wave, but I have not had time to do that (yet). FYI, the coefficients are only the odd ones and proportional to ((-1)^((n-1)/2))/n^2.

The attached VI is LabVIEW 6.0, produced by a save from LabVIEW 6.1. I can not check it as I do not have 6.0 installed. Point the front panel path at one of your data files and run. Let me know if you have problems.

[unclebump]

Is that your whole email address?? Is there a part missing after an @ symbol??

[suethechemist]

Sorry - I put the 'at' inside pointy brackets (trying to be clever) and it has been removed in the final display! The 'at' should be between 01 and qub.ac.uk, sbelshaw01ATqub.ac.uk