[Done 2.8 B2] Master Spectrum Peak Fall Speed

Loving the new spectrum!

One of my main gripes with the old one is still present though… which is that there should be a peak fall speed option for ‘instant’. So it might read:

  • Still
  • Slow
  • Normal
  • Fast
  • Off

This will result in all types of flickering when drawing the spectrum curve…

I’m not that of an experienced spectrum reader, but what would the benefit be of instant fall speed over fast?

I find that having the curve smoothed out over time slightly detaches it from what you’re hearing, because it’s useful to be able to see the dynamics, the pops, the spikes and voids - it’s useful to see how short percussive sounds actually fall away, and how their frequencies skew as they fall.

With too slow a peak fall speed, you’re mostly watching the bands fall at fixed speeds from (less meaningful, imo) peak values while the actual data wildly dances hidden beneath. It’s more like watching a series of high-scores lingering and being superseded than watching a stream of information. I reckon spectrums are at their most useful when they attempt to display something as close to the raw data as possible - maybe it’s just me who feels strongly about this? Would be nice to have the option though.

The frequency curve in Adobe Audition doesn’t appear to have much, if any interpolation - it looks real-time, or at least the peaks fall faster than most ‘common decay rates’ in recorded music?

Is the flickering you mention just a side effect of FFT?


I’d also be in favour of a faster peak fall setting. Even the fastest one does feel quite slow.

I’ve made it faster, that is i apply no fallback at all (on big block size it is a bit funny but anyway it’s ok to use it).

Glad this was implemented :)

Tested a sine kick drum with a block size of 8192. Its informative to see the quick pitch modulation then slow down at its designated pitch.

Before, even with a peak fall at fast, I could still see the remnants of the pitch modulation across several frequencies.

So, for example, seeing the actual speed of the pitch modulation has given me insight on whether to cut the mid freq during the modulation or resynthesize it depending on other sounds in the mix.

Glad you like how it is now.

I have to say it’s pretty usable now even with longer block sizes of processing by using variable overlap averaging, that means we have

  • 12.5% overlap (8 fft per block) @ 16384 and 8192
  • 25 % overlap (4 fft per block) @ 4096 and 2048
  • 50% overlap (2 fft per block) @ 1024 and 512

Awesome news! :walkman:

Just tried it out, with a window of 4096 with instant peak fall speed it’s the tits! With a better framrate than Adobe Audition, even.

Glad you requested for an instant fall speed. I personally wouldn’t have been able to fully describe why its useful as you did in post #4