LB302 sound output bugs

[diizy]

So, in the most recent development version (compiled from stable-0.4) at this time, I'm encountering some buggy behaviour in the LB302 instrument.

• Sometimes noteoffs get ignored, and a note that should be very short plays up until the next note. Weirdly, this mainly happens on the first note of the pattern.
• Sometimes when playing back a pattern, the sound glitches entirely and mutes everything. This persists until closing and/or reloading the LB302 instrument.

Both of these never happen (at least I've never seen it) when playing live, eg. on the keyboard or the virtual keyboard attached on the instrument window, only when playing back patterns - either individual patterns in the piano roll or playing the entire song in the song editor.

There are additionally some older bugs in LB302 behaviour that might still be unsolved (but it's hard to confirm due to these newer bugs and their random nature): sometimes, some notes play the wrong pitch - the pitch of the previous note gets "stuck", eg. if you have a sequence C - E - D, instead C - C - D gets played. This always only happened with overlapping notes, if the notes were separated it didn't happen.

[pgiblock]

Could @diizy attach a project file that demonstrates this problem?

[diizy]

I think these are mostly fixed now? The LB302 still does some occasional weird random stuff - this is hard to demonstrate with a project file because its occurence is random. Sometimes, if you put in two notes so that one starts exactly where the other ends, the second note plays the wrong pitch - it plays the same pitch as the first note. So you have to either overlap the notes or leave a gap between them, to get LB302 to behave nice.

All the other bugs seem to have been fixed though.

[diizy]

Seems like this issue still exists. It only seems to happen when slide is enabled, now.

Also, there's a new issue: sometimes if you enable the 24db setting on LB302, it crashes LMMS, but I'd better file a new issue for that.

[pgiblock]

@diizy: The C-E-D becomes C-C-D could be because you have slide enabled and a really huge ramp time. The slide (portamento) is reasonably stupid, but so is the TB303. We ramp from the previous note to the note that was just played. The slide is implemented as an exponential decay. So if the coefficient is very high, then the destination note could take a very long time to reach.

This behavior might not be of much use classically, but it can be useful to create frequency sweep effects.

[pgiblock]

@dizzy: can you verify things are working as of fc1475b ?

[diizy]

On 02/20/2014 04:41 AM, Paul Giblock wrote:

@diizy https://github.com/diizy: The C-E-D becomes C-C-D could be
because you have slide enabled an a really huge ramp time. The slide
(portamento) is reasonably stupid, but so is the TB303. We ramp from
the previous note to the note that was just played. The slide is
implemented as an exponential decay. So if the coefficient is very
high, then the destination note could take a very long time to reach.

This behavior might not be of much use classically, but it can be
useful to create frequency sweep effects.

Paul I'm sorry but no, that's not it. This thing is very noticeable,
it's not about the slide time, you can set the slide time to minimum and
it still happens. Also it doesn't happen consistently, if you loop the
same pattern, you'll notice it happens on notes randomly. Seems a bit
more common at the beginning of the pattern. It only seems to happen
when one note starts where another ends, if you overlap them slightly it
doesn't usually happen.

[diizy]

On 02/20/2014 04:46 AM, Paul Giblock wrote:

@dizzy https://github.com/dizzy: can you verify things are working
as of fc1475b
fc1475b
?

The crash/hang problem seems to be fixed, yes.

The weird issue with slide hasn't.

[diizy]

Since the critical parts of this bug have been fixed and the remaining part isn't a release-critical bug, I'm marking this for 1.1.0.

[diizy]

I have some ideas (hunches really) now why this is happening and I might be able to cook up a fix for this, can't promise anything though...

[diizy]

Should be fixed now.