Filtering steep jumps in number flow

Reply to Filtering steep jumps in number flow on Sun, 12 Jul 2015 14:05:30 GMT

852585 — Sun, 12 Jul 2015 14:05:30 GMT

Hi pure daters,

I have a flux of numbers which is NOT a signal flux.
just a regular number box, which outputs a more or less constant value.
The problem is that this value jumps down to zero due to some parasites.
Is there an object that filters steep jumps in a float flux ?
Something that outputs the variation rate (derivative) of the flux,
and eventually displays the filtered data.

Should exist as it is basic !
Thanks in advance

Reply to Filtering steep jumps in number flow on Sun, 12 Jul 2015 14:34:35 GMT

whale-av — Sun, 12 Jul 2015 14:34:35 GMT

@852585 Hello there...
If it always jumps to zero (which is an error?) then you can filter zero with [select 0]
If you need to smooth the float stream you can use [smooth] or [maxlib/average] which calculates the average of the last "n" floats. Maybe you should remove the zero's first.
David.

Reply to Filtering steep jumps in number flow on Sun, 12 Jul 2015 14:55:01 GMT

852585 — Sun, 12 Jul 2015 14:55:01 GMT

It is not an error. I am just using cheap rusty sensors.
It actually jumps to -64, sometimes -63.6 .... around these values.
and these parasites are transposed into different ranges when the sensor is used.
So using [select] is out of purpose.

Smoothing is not a solution. Average neither.
I tried, the response delay is too high for both
and the parasite peak is happening too often to be averaged.

A filter would be great though, bypassing steep jumps (high derivative values).

Reply to Filtering steep jumps in number flow on Sun, 12 Jul 2015 18:22:15 GMT

whale-av — Sun, 12 Jul 2015 18:22:15 GMT

@852585 Hello again.... I have made you a little filter.
You can set on the right inlet the maximum deviation acceptable between successive floats, and you feed your data to the left inlet.
The outlet will only give floats that were within the acceptable range of (immediate) deviation.
If you have successive bad floats then they will not be caught. Anything can be solved though!
This probably does exist somewhere else, but anyway... here it is:
filter_data.pd
filter_data-help.pd
David.

Reply to Filtering steep jumps in number flow on Sun, 12 Jul 2015 18:43:14 GMT

852585 — Sun, 12 Jul 2015 18:43:14 GMT

[delta] is a very useful object I did not know !
The filter looks good, but the deviation itself varies a lot in my case (bigger problem).
I'll make something out of this. Thanks for help ! Problem solved (soon).

Reply to Filtering steep jumps in number flow on Fri, 15 Jul 2022 11:07:14 GMT

granular — Fri, 15 Jul 2022 11:07:14 GMT

Hey there, I'm coming up against this same issue now and wondering how it resolved?
I'm running Pd Vanilla 0.52.2 and have no [delta] object. What library is it in?
Thanks

Reply to Filtering steep jumps in number flow on Fri, 15 Jul 2022 18:36:26 GMT

whale-av — Fri, 15 Jul 2022 18:36:26 GMT

@granular Try this vanilla replacement....... delta.zip
The external was, and probably still is in the maxlib library.
David.

Reply to Filtering steep jumps in number flow on Fri, 15 Jul 2022 20:17:15 GMT

oid — Fri, 15 Jul 2022 20:17:15 GMT

@granular deken can be searched.

Edit: just remembered that searching for objects got added too deken in pd itself, just menu over to Help->Find Externals and type in the name of the object you are looking for and it will return a list of libraries which contain an object with that name for you to easily install.