here is a VI which implements the derivative w/ histogram method. The method only works after heavy filtering of the data, since otherwise the noise leads to extremely broad histogram peaks.
Probably another method (user selection of nearly linearly sloped segments, fitting of linear curve int that region) gives better results. That method would also allow to quantify the quality of the linear fit and therby produce an estimate of the non-linearity which could give valuable hints on whether the observed process behaves as expected.
The code comments are brief, so don't hesitate to ask if you have questions
Sorry, I forgot to include it, here it is (it just implements a sliding average filter where each value is recalculated as the average of itself and N neighboring points to the right and left):
I'm not sure if I'm following this right. The first square says that it extracts the first part of the 2nd channel? Is that the -1 on the Max/Min? Then the dominant slopes are calculated from the histogram correct?
I'm not sure if I'm following this right. The first square says that it extracts the first part of the 2nd channel? Is that the -1 on the Max/Min? Then the dominant slopes are calculated from the histogram correct?
Dave
ok, in the leftmost rectangle first I extract channel #2 (with the 'split signal' function). Then I look at which index the max of the Y-data array occurs, subtract 1 from ths index value 8to avoid the max itself) and take the subarray from 0 to maxindex-1. This is what I figured your data was, as it was presented in your startup post.
In rectangle 2 I filter the data, take the derivatives and plot them.
The 'dominant' slopes are correct (I believe) but you have to divide them by the correct dt of your measurement. In your original dynamic data set dt equals 1, and is therefore probably not correct.
