Hacker News Reader

Comment by eternityforest | Hacker News Reader

If you don't count your time whatsoever and don't go for the nice case design.

I mean at the extreme case, if you can tolerate 40ms of latency and stuttering, a Raspberry Pi and some USB soundcards can do it all. The actual power needed is reasonable and ADC/DACs aren't that insane.

A Teensy 4.1 connected to an ESP32 could probably do most of this, if you wanted to do $15k of software development in your spare time and were really good at DSP.

View on HN

Interestingly enough, it seems like the KORG wavestate is actually based on some Raspberry Pi, according to KORG's github page, so you can definitely get less than 40ms latency and no stutter. And yeah, Teensy or even FPGA evaluation boards can get you very far, since software development in your spare time is "free". If you search the web, you'll find many examples of such projects. For example I remember some FPGA based project with iirc a small ~50€ Spartan-6 eval board which implemented a fancy 32 voice 6op FM synthesizer.

I wonder how they're doing it? Do they have a custom kernel?

Or maybe they just have some good IO. It doesn't seem to be CPU limited exactly, it's just the built in sound device doesn't like it if you turn down the block size too much. USB cards are a bit happier.

Anyone who can tolerate 40ms of latency should stick to Bluetooth. If it's more than 3ms, it's already more latency than a good RF system gives you, if it's more than 10ms things are getting questionable, and at 20+ms you've lost the ability to monitor, because what your fingers/voice do and what your ears hear are now noticeably out of sync. Try singing while monitoring yourself at 20ms delay. It's hilarious.

I can kind of mostly sing up to 35ms or so, but at 40ms and up it's a real clown show.

My playing is already sloppy enough to not be affected till maybe 15 or so.... but it would clearly be different with a pro!

Reply on news.ycombinator.com