How can I use the Diadem FFT Analysis GUI to give me multiple outputs given multiple inputs?

[aribrich]

I have a few hundred channels I would like to perform FFT analysis on. They are mostly of different lengths, but I know I will have groups of 7 that will have the same number of data points. The default GUI will accept those 7 sets since they are the same size, but it will only give me one output as a combination of the sets. Is there a way to get Diadem to give me an FFT output for each set without having to select each data set individually?

[HeroOfHyrule]

Hello aribrich,

I'm trying to understand your question, but got a little confused. You have a few hundred data channels for analysis, lets say 700 of them. You have those data channels broken down into groups of 7, so we have 700 channels - 100 groups of 7 channels each (each group has signals of the same size).  Here is where I got a little confused. You input your group with the 7 channels and get back just one result, for the group, but want to get individual results for the channels (ie 7 results) without having to individually select them?

[aribrich]

That is correct. If I use the available GUI for this analysis, selecting multiple channels as my input will only give one output. I assume that is simply an FFT of the combination of the channels which would be useful under some circumstances, but in this case is not what I am looking for. I want on input output per input, even if multiple inputs are selected. There may be a way to build a loop for this programatically, but I don't have the experience coding in Diadem that I would need and I can't find enough information online to do that. Do you understand what I am trying to do now?

Thanks,

Ari

[HeroOfHyrule]

Ari,

Which FFT Analysis are you using for this? There are some of the functions that will give you multiple outputs for multiple inputs, but others that will give you just one output. Could you answer back with the one you are using? If it is one of the multiple output functions, it may be as simple as selecting all the channels correctly on the correct part of the GUI. 

[aribrich]

I was using the first FFT option. The one labeled FFT (One Time Signal). There are other options, but they don't seem to do what I'm looking for either.

[HeroOfHyrule]

When running the FFT One Time Signal, you should be able to select multiple channels and get a new channel for each of them as a result.

[aribrich]

The way it worked with the default channels, I had to make sure I wasn't going crazy because the way you demonstrated worked very well for me. That being said, for some reason what I am doing does not work. I've recently found that it is in changing the details in the GUI that makes my output different than yours. The differences are as follows:

1) I do not use a time channel. That is always left blank.

2) in the 'Window Functions' tab, I change the window type to Hanning

3) In the 'FFT Functions' tab, only amplitude is checked with the properties 'Peak amplitude', No Third/Octave, and Average amplitude is changed to arithmatic.

4) In the 'Time Intervals' tab 'Length/Start/Overlap' is checked, Interval length is changed to 1024, and overlap (%) is changed to 50.

I've also found that changing any number of these details from the default will cause my proplem to crop up. I am fairly sure that most of those options are just as feasibly done to one channel as they are to multiple. Is this a bug, or am I misunderstanding a property of FFT that has gone over my head? Please try changing a few of those properties since it seems like it should work just as well for all the channels as it does for just one, but that doesn't seem to be the case.

Another point of confusion is that while the window previewing the inputs in the GUI will show all the inputs I have selected, the window previewing outputs only appears to preview a single unknown output. That is just a side note though.

[HeroOfHyrule]

You can simulate this with your data or the example data by choosing no time data and using the Noise data (they're waveform data types) like the ones you are using I imagine. That being said, it seems like step 3 is your problem. The average is what is killing your multiple outputs. Check out this document so you can select the right one for the job. Notice that switching from Arithmetic to another will give you the expected number of outputs. The example data worked just fine because there was no averaging being done. Hope this helps. Happy New Year!

[aribrich]

Happy New Year Rob!

Thanks for that tip. That information did help, and I don't think it should be an issue using no averaging. I still don't really see why it messes up the output when using multiple channels, but that should work for now. That being said, please try doing the same thing again with no input for the time channel. I am not collecting time data, so getting it as an input would be tricky. Using no time channel will not let me preview data, and when I try calculating with inputs it gives a strange output - generally giving me way too many outputs. I'm not really sure what the outputs represent either. Even still, if I do use the time channel, I am unable to set a "overlap (%)". Is this an issue that you see as well? I think I should be able to use the tool without a time in and just assume my sampling rate as my step size. I see your screen captures do not show this to be an issue. Is this something you have seen before? Is there a reason for that as well?

Regards,

Ari

[HeroOfHyrule]

Ari,

Happy New Year! As you can see from the newer screen shots I posted, I am not using the Time data as an input. It would, however, make the issue a little more clear if you could post a picture of what you are seeing so that I can understand it a little better. Thanks!