How to filter an irregular waveform?

[oliver1023]

Hi guys,

i have a graph from 2D array data.

May i know a way to filter it to make it a smooth line at the red box area and remain the spike?

[kartiknattar]

You can use filters to remove the noise. High Pass, Low pass or band pass. These vi's are availble in filter pallete.

[Henrik_Volkers]

Detect the pulse , apply a filter (FIR with half the detected pulse length) replace the pulse part with the unfiterd part ?

Or detect the pulse , from start to 'left'  apply a mean filter (backward) with n(TBD) points  and init the meanfilter with the pulse start value.  Do the same on the other side ...

[oliver1023]

Hi, appreciated if you could give me some example. Attachment below is my file for x and y axis.

[oliver1023]

Attachment is the VI of my XY graph from array

[jwscs]

i can not solve your filtering problem,

but i can critique your code.

1. please don't choose file extensions, that let you expect a format which it is not.

use either '.txt' or '.csv', NOT '.xls', this is a excel document, which your file is not.

2. the read-delimited-spreadsheet function, converts your strings into numbers,

there is no need to be doing it by hand

3. i never used the "index+bundle" function, but simply the 'bundle' function,
i don't think that the execution time matters in your case, but it might be if you have larger datasets (bundle only takes 60% of the time index+bundle takes (+- variation).

(this is a snippet, download the .png file from below and drag+drop it onto a blockdiagram)

regarding your problem, take the already given suggestions, and play a little with all the filtering VIs.

e.g. search for "butterworth filter" and then try lowpass and change the sampling frequency.

find out how to display multiple plots in one xy-graph (search for "xy graph" in labview's example finder) so you can compare your filtered result with the original.

have fun

[McQuillan · Accepted solution]

Hi oliver1023,

I'm not going to implement the fix for you. However, I will point out the functions that I would use to solve this challenge. If we break it down in to sections:

1. We are interested in the peak - that's our focal point, so lets isolate that that. To to this, we can use the Peak Detector function.
2.  
   1. This function will tell us what the peak is and where it is. Enter a threshold level and this function will return an array of values.
3. You want to smooth out everything except for the peak.
   1. Considering we have all the information about the peak, we can be cheeky and actually filter everything (including the peak), then add the peak back in later.
   2. To filter your signal use the Butterworth Filter - this can be set to bandpass
   3. . 

I hope this has given you some food for thought!

[Henrik_Volkers]

To apply a meaningful signal processing, it is important to know why you need to make that areas flat?

What is the goal?   What is the physics behind?

If you 'just' want to calculate the pulshight to the biased level, there are faster/easyer/better ways to do it:   

• detect peak value, 
• fit a line with weightningfilter leaving out the peak ,(smoth to the ends, use prior knowledge..)
• calc distance of peak to line

[Henrik_Volkers]

Lazy sunday ...  don't know if that's the task you need, however

here is something to play

[oliver1023]

Hi jwscs,

Noted of all the error I've done that you pointed out.

I will update my main code.

Thanks with appreciation.