Thursday November 27th

The box (mostly) survived an outing to Woodland Pattern with Will. Some bits – the amp & solenoid power supplies, attached to the side with screw-in standoffs – shook loose and ended up ripping out one of the amplifier board wires. Otherwise everything held together pretty well.

Things got somewhat stripped down and simplified during rehearsal. Now that I’m home I’m working on building the complexity back up again. ;-)

The koma field kit was initially an input mixer for the system, with two contact mics built-in to the top panel running into it. It seemed nice to have a couple free inputs for other mics (induction & others) because that’s worked well in the past. I like having the flexibility to route in new inputs as the mood and situation strikes.

I ended up only connecting one of the contact mics though. I’m going to see how far that goes before I go back to some kind of input mixer. With this new system the whole top board is the input surface. It might be useful to add one more instrument mic when I start incorporating brass again, but in the meantime the input board/surface as the only input to the system is a nice simplification.

Phantom xruns

The whole performance I was fighting these damn xruns that seemed to come out of nowhere. I thought maybe there was an overflow somewhere in the mixing stage, since basically once in a while the world stops – all audio just cuts out – for a few seconds, and then jolts back to life. Like a recording being muted.

Long story short, there was a bug in the normalization routine that would produce Inf or NaN with silent or nearly silent buffers. Of course, during rehearsal I added a normalization stage to some of the voices to make voice management easier… and some of those voices could produce very quiet buffers sometimes. So now & then they’d get fed into the system, and for a few seconds they’d be mixing in Inf or NaN and blow up the entire output stream… then they’d finish playback and everything would return to normal.

I think that’s been a bug in pippi for a few years now – it explains so much. I don’t often set up situations that produce silent buffers. I will sometimes do it when live coding during rehearsal or a performance by accident or to simplify something (I’m not optimizing my code during performance, that’s for sure) which is precisely when I’ve noticed this issue crop up. These are also the times I really don’t want audio to mysteriously blow up haha, so it’s very nice to have tracked this down.

💯

And my grandmother turned 100 this month.

no log for August