Noise print or denormal occur when ampeg_release not reached

[kinwie]

When the ampeg_release stage not yet reached and the region ends, sfizz produce some kind of noise print of denormal after it.

[paulfd]

There should be no denormals, so it might be noise or garbage. Do you have a test case?

[kinwie]

Yes, tempo 120 bpm

<region>
sample=*sine
ampeg_attack=1
ampeg_hold=1
ampeg_decay=1
ampeg_sustain=0
ampeg_release=1

Draw one bar midi note, so it reach the release stage, no noise :

Draw two bars note, it's not reach the release stage, there is noise after it (it should ends in the 2nd bar, 3rd beat, like first image) :

This seems only occur when the ampeg_sustain value is 0

[paulfd]

Was this one improved this summer?

[jpcima]

It's because we cut the envelope release at -60 dB.
You can lessen it by aiming at -80 dB.
If we rewrite the ADSR entirely in curves, there is no more need to take care of this asymptotic behavior.
(but then unnatural behavior on level/time dynamic updates. it's the ARIA behavior)

[paulfd]

I think it could make sense to have the ampeg as a FlexEG too, if only to have a single codepath to check and optimize for all this.

[paulfd]

@kinwie does this still happen? I have trouble seeing it on Reaper

[kinwie]

@paulfd Using same example case as above, the release is cut shorter (sfizz - DropZone - sforzando)

[jpcima]

A simple solution is to add a final phase that finishes the envelope in linear.