Author Topic: audible artifacts in hilite trace?  (Read 2003 times)

p8rpp

  • Sr. Member
  • ****
  • Posts: 58
    • View Profile
audible artifacts in hilite trace?
« on: September 18, 2019, 03:12:15 PM »
Hi,
using 7.1 on Linux I do get audible artifacts with hilite trace, which do sound a bit like bit noise or aliasing.
I am attaching a file via
https://framadrop.org/r/lNSvFSYs3t#FzODUss6BYJ9NZHQ4xCXSl50bEL1Gl47oyqv6fViYkM=

which I created with
  hilite trace 3 infile.ana outfile.ana 18 3000
thus leaving only the 18 loudest partials below 3kHz.

This has puzzled me since a long time and happens in all sort of different calls to hilite trace.

Would be great to get any idea and finally make this beautiful algorithm usable again!

joehuergo

  • Full Member
  • ***
  • Posts: 36
  • very personal text
    • View Profile
    • SoundCloud
Re: audible artifacts in hilite trace?
« Reply #1 on: September 18, 2019, 03:38:15 PM »

Seems to me that it's working properly, all the partials are indeed contained under 3khz... I suppose it does sound slightly alias-y, but to know if there were artifacts for sure I'd have to hear the original sndfile, and what analysis settings you used.

p8rpp

  • Sr. Member
  • ****
  • Posts: 58
    • View Profile
Re: audible artifacts in hilite trace?
« Reply #2 on: September 18, 2019, 03:53:34 PM »
Thanks!

Now here
https://framadrop.org/r/YhMF5xUfNZ#OmoQKr3R06Zrd0P0oEb6nz1IJmkQfFhZKVoAY73aUww=
is a sine sweep in its original form and then processed with
 hilite trace 1 sweep.ana sweepTrace.ana
after having been pvoc analyzed using the default parameters of 1024 window size.

What is interesting is that if I set the window size to 2048 the artifacts happen at different times during the sweep.

How did you create that spectrogram by the way?

joehuergo

  • Full Member
  • ***
  • Posts: 36
  • very personal text
    • View Profile
    • SoundCloud
Re: audible artifacts in hilite trace?
« Reply #3 on: September 18, 2019, 05:17:54 PM »
Yeah I can kinda hear/see what you're talking about now... not sure what the deal is exactly. I've definitely experienced a good amount of weird aliasing artifacts in the spectral processes before... part of me wonders if this could be fixed if maybe we had control over the window shape (PVOC only uses a Hamming window) though that may have absolutely nothing to do with anything, lol. Maybe setting a higher value for the window overlap parameter may do something? I think it's already 3 by default though, and the highest is 4, so idk. Very curious indeed.

And that is the spectrogram in Renoise, I have it open pretty much all the time now. On my laptop I use VB Cable (https://www.vb-audio.com/Cable/index.htm) which you can set the "Cable Input" as your primary output in system settings, then set "Cable Output" as a Line Input within Renoise, and this way you can monitor the spectrogram of whatever you're listening to, including any cdp stuff. I highly recommend! (on my home desktop I forego the VB Cable stuff because I have an 8 i/o interface which can do the same thing with less latency.)

(Oh yes, I just remembered you use Linux, so VB Cable prob isnt an option. You can prob do the same thing using JACK audio server though!)
« Last Edit: September 18, 2019, 05:19:44 PM by joehuergo »

p8rpp

  • Sr. Member
  • ****
  • Posts: 58
    • View Profile
Re: audible artifacts in hilite trace?
« Reply #4 on: September 18, 2019, 10:44:38 PM »
Thanks!

Does anyone else have a take at these artifacts?
Much appreciated!

P