Page 1 of 1

Windowing and Time Based Epoching

Posted: Fri Jun 29, 2018 9:56 am
by dragouns
Hi, everybody,

I have a question about two plugins: windowing and time based epoching.
As I understand it, but it doesn't look like it, hanning windowing for example allows to work on a finished signal in time and epoching allows to cut the signal in time.
So I don't understand why some examples use time based epoching or the hanning window but not both? Is it the same function?

Thank you very much and good end of the day.

Re: Windowing and Time Based Epoching

Posted: Mon Jul 02, 2018 7:46 am
by jtlindgren
Hi,

Epoching slices the signal in different ways (you can think of it as implementing a sliding window with different step sizes), whereas Windowing applies a coefficient mask on each signal chunk. Both names are due to historical reasons.

If you have single channel signal chunk [x(1),...,x(k)] for chunk size k, then the output of windowing is

[w(1)x(1) w(2)x(2) ... w(k)x(k)],

where the coefficients of w depend on which window function you choose. Windowing is usually used to avoid/mitigate border artifacts by dampening the signal towards zero at both ends. You can google for it.


Hope this helps,
Jussi

Re: Windowing and Time Based Epoching

Posted: Mon Jul 02, 2018 8:41 am
by dragouns
Hi and thanks for your answer,

Actually I've already read a lot of doc about windowing and epoching, what I don't understand is if the two functions are independent or complementary. Because I understand that windowing is used for edge artifacts. So if I want to do a good treatment before doing a spectral analysis, I have to use the windowing on each epoch to smooth the edges and have a nice fft?

Merci beaucoup et bonne fin de journée ;)

Re: Windowing and Time Based Epoching

Posted: Mon Jul 02, 2018 8:49 am
by jtlindgren
In OV, signal is always implicitly epoched even if you're not aware of it: the data is always passed on from box to another in small matrix chunks. In OV, the 'epoch size' and 'chunk size' are the same thing, and mean the number of columns in small data matrices that are passed around. For example, data coming from the Acquisition Server by default has chunk size of 32 samples ('sample count sent per block').If you run Windowing without epoching, you're doing weighting of 32 sample buffers. Different epoching boxes can be used to change this chunk size for your data.

So the methods are independent in a way that you can use neither, either, or both, depending on what you want.


Good luck!
Jussi

Re: Windowing and Time Based Epoching

Posted: Mon Jul 02, 2018 9:14 am
by dragouns
Thanks a lot,

I try to make a graph of the power of the different types of waves as a function of time. I thought I was averaging the fft powers so I was wondering if I should place a hanning window between epoching and spectral analysis.
Anyway, thank you very much for your very quick answers.