Jump to content


Photo

Differences in analog filter shapes different with oversampling

Renoise 3.1 Analog Filter DSP oversampling shapes

  • Please log in to reply
5 replies to this topic

#1 EatMe

EatMe

    Guruh Member

  • Normal Members
  • PipPipPipPipPipPipPipPipPipPipPipPip
  • 969 posts
  • Gender:Not Telling

Posted 15 September 2016 - 07:34

Hi,

 

with the internal DSP device Analog Filter

on more oversampling (2x 4x 8x)

the shape changes of the filter in the graphic pane

 

this happens on almost all Analog filter types.

 

win7 sp1 renoise 3.1 64 bit

 

RenoiseFilterBugShape3_1_0.jpg

 



#2 dblue

dblue

    Dodgy Geezer

  • Admins
  • PipPipPipPipPipPipPipPipPipPipPipPipPipPipPip
  • 5809 posts
  • Gender:Male
  • Location:Berlin
  • Interests:Code. Music. Graphics.

Posted 15 September 2016 - 14:02

Changing the oversampling will naturally have some effect over how the filter is processed, and by extension how its response graph looks.

Generally speaking, there is simply a bit of imprecision when plotting the lower frequencies in the graph, but this has always been the case and it's pretty much just a cosmetic issue anyway.

I honestly wouldn't worry about it too much. If it sounds ok then simply enjoy it and move on with the rest of your song.

#3 Bit_Arts

Bit_Arts

    Big Super GrandMasta Member

  • Normal Members
  • PipPipPipPipPipPipPipPipPipPipPip
  • 799 posts
  • Gender:Male

Posted 15 September 2016 - 15:53

Oversampling in some cases can help, in others it can cause signal degradation, depending on the entire signal processing. Don't use oversampling in each and every case you can, just because you can. It'd be a false conclusion to think "the more oversampling, the better it always sounds automatically". Use oversampling only, when it makes things sound better to your ears. If it doesn't, just leave it off. Your ears should decide, not some plain availability of a feature.


Herr, schmeiss Hirn vom Himmel! :mellow:


#4 EatMe

EatMe

    Guruh Member

  • Normal Members
  • PipPipPipPipPipPipPipPipPipPipPipPip
  • 969 posts
  • Gender:Not Telling

Posted 15 September 2016 - 21:08

Ok, so the way it is is intended because the lower end sampling becomes more precise at higher oversampling.

Thanks for explaining.



#5 Zer0 Fly

Zer0 Fly

    Guruh Motha Fakka is Levitating and Knows Everything About Renoise Member

  • Normal Members
  • PipPipPipPipPipPipPipPipPipPipPipPipPip
  • 1218 posts
  • Gender:Male
  • Location:Oops
  • Interests:Flying

Posted 15 September 2016 - 22:05

Oversampling does indeed change the sound, but rather watch (and listen to) the very top of the frequency responses at high cutoffs, no oversampling vs 2x or 4x. This is because of properties of putting filters into the digital domain, oversampling shifts these effects into the domain that isn't audible or represented in the final sample data anymore. This can also be seen with the digi filters.

 

Not sure about the low end display glitches. This should have nothing to do with oversampling, oversampling filters rather makes the high end response more acurate, not the low end. But the frequency transfer function of a filter changes when oversampling, and maybe the display routine is affected by it, I don't know. I already noticed the weird displays, too, and it can become even more apparent with narrow digital butter n8 bandpasses in the bass band (I think) but I tended to believe it was just display weirdness, and that I couldn't hear the strange responses visible but rather what one would expect at such cutoffs.

 

I only use oversampling when using high cutoffs, or drive that would generate harmonics higher than nyquist to avoid aliasing. Not sure about the necessity of the second point.

 

Also without or with only 2x oversampling I noticed bad artifacts at fixed frequencies when moving cutoff and strongly amplifying and overdriving bandpassed parts of the transition- and stopband of the output of a driven filter. Oversampling at or higher than 4x gives me clean sound for such techniques. Wasn't sure whether these glitches were worth reporting because the technique is rather obscure, it just drove me nuts with one instrument or another until I found out where the fucking crap noise came from and how to fix it by raising oversampling. I could try to craft an example to demonstrate it. Maybe those artifacts are still there in normal usage, but just very very quiet and thus are usually unnoticed.

 

More important for me would be dropping the lower cutoff limit of the analog filters to 20 or even 10 hz. A 40 Hz filtersubbass is badass, but compared to 35 hz it is mickey mouse in emotional impact of timbre.



#6 Zer0 Fly

Zer0 Fly

    Guruh Motha Fakka is Levitating and Knows Everything About Renoise Member

  • Normal Members
  • PipPipPipPipPipPipPipPipPipPipPipPipPip
  • 1218 posts
  • Gender:Male
  • Location:Oops
  • Interests:Flying

Posted 16 September 2016 - 00:36

I tested with the butterworth n8 bandpass max q filter 8x oversampling and a sine, both tuned to ~65 hz. Graph shows around 16 db reduction in its distortion, sine passes through without suffering any reduction, as one would expect from the filter without having seen the graph. Same with any kind of oversampling, just the graph showing different distortions. Just ignore the weird bass response graph.







Also tagged with one or more of these keywords: Renoise 3.1, Analog Filter DSP, oversampling, shapes