plotting derivation of waveform data

[xombox@seznam.cz]

Hey Coq rouge,

I have tried the filter you designed and played with those interval for comparism but didn`t reached the suitable value. If I take a look at the graph it seems that linearity ends approx. somewhere around x = 600. What I need is to be able to detect this linearity breaking point.

I tried the values 1 and -1 and it seemed that filter just detected maximum in the graph and cutted the rest part behind this detected peak. Take a look at the first screenshot.

When I set lower values it was too sensitive, e.g. constants = 0,24; - 0,24 then it cut the graph up to x=80. Take a look at the second screenshot.

I tried to set some values between but it didnt work, the filter is too sturdy.

What does actually the filter do? It takes 2 consecutive values, makes it difference and then it compares whether it fits in the specified interval, right? Which means that it only detects major changes in two consecutive values and doesn`t involve any linearity.

I`m sorry, if I got it wrong.

Hello Vince,

I have seen these filters but except gaussian filter I don`t have a clue what it does or how it works. I`m quite blind in this case and I though someone skilled might suggest a specific filter.

All in all, thank you guys for your interest in solving my issue.

[OliviaC]

Hi,

You seem to be right about Coq rouge's example only looking at the difference but it does have some relation to linearity. If your data is linear and dt is equal between each point, you'd expect the step between consecutive values to be constant. For example, let's say your data fits the linear equation y=2x + 1 and dt=0.01. Then you would expect that there is a 0.02 difference between each point of your data. If the difference between 2 consecutive values is not 0.02, then you know a value doesn't fit the linearity.

Of course, there is going to be some noise in your data so you need some tolerance. Therefore, instead of expecting a 0.02 difference between points, make it between 0.195 and 0.025. If the difference between 2 consecutive points falls outside of this, then you know linearity is broken.

[xombox@seznam.cz]

Hi Olivia,

thank you a lot for perfect explanation. Now, I have got the idea how to adjust  Coq rouge's example. Altough I have tried many filters it seems impossible to set any limits from this very chaotic derivative data. Maybe some averaging method would help. Don`t have much time for it. So I will rather focus on major part of my project. When I come back to this, I will keep this post updated.

Thanks to all for help.

Martin