MIDI is so last century!

[MushMusic]

Fiber optics were used in an organ I worked on as a builder back in the mid 90s. The key actions are optical and had fiber optic runs, some over 200 feet. All that stuff was custom made for us by Solid State Logic. They make the most beautiful electronic components I've ever seen, but it was really really expensive. That instrument just by layout design had crazy latency issues. There was about 150 ranks 50 feet in front of you and 50 feet off the ground. Then there was another 150 ranks that were over 200 feet behind you. And another 50 ranks split in half, 75 feet to each side of you. Yes your math is right, this is a 350 rank organ.

[JeffLearman]

As I understand it, what separates dedicated keyboards, with their own custom sound chips, from computer generated sources, is the latency.

That's one thing that tends to separate them.

 

I assume a muse receptor is best case scenario.While I would expect it to be well-tuned, it's not necessarily the fastest/bestest/latest hardware. It would be interesting to see actual test results.

 

So assuming a SSD

What's SSD? Solid-state device? That generally has no bearing because audio data is not buffered on disk. (Sample players have to deal with pulling samples off disk, but to avoid latency they keep enough of each sample in memory so that they can start playing the sample while waiting for the rest of it to come from disk. When that fails, you don't get latency, you get nasty stuff like zippering, clicks & pops, or dropouts.)

 

is 2-3 ms the lowest latency that's possible? Does that increase dramatically as you put a heavy load on it? Would any of these faster pipeline technologies improve on that?

The primary cause of latency is buffering. The more buffering, the longer the latency. With good drivers, the right kind of internal software architecture (e.g, ASIO interfaces), and a lack of other software that hogs the processor with interrupts disabled ("interrupt latency") or hogs the I/O channel (e.g., "PCI latency"), buffering can be quite low.

 

Some technologies (e.g., convolution filters often used for reverbs) require internal buffering. That is, the 1st sound sample can't be calculated until the Nth sound sample is available for processing. For live use, we avoid these if N is big enough to notice.

 

As processing speeds increase, the amount of buffering required can be reduced, assuming we also keep the interrupt and PCI latencies low. One problem is that as hardware gets faster, programmers get lazier (or more accurately, more functionality is expected from the software) and the software gets less efficient. So, we don't get the full benefit of the speed increase.

 

The ultimate way to get super low latencies would be to return to programming practices of 20 years ago, where the lower processing speeds demanded extremely careful coding on the whole path. But ... you'd have to give up things like VST plugins.

 

Another cause of latency is transmission delay. With today's bandwidths, these are insignificant "in the box". They can become significant over the internet, and if the distance is long enough (say over 500 or 1000 miles) the speed of light becomes the limiting factor.

 

With MIDI (on MIDI cables, disregarding MIDI/USB), there is a 1ms per note transmission delay. That is, if you play 8 notes simultaneously, the first note has a minimum 1ms transmission delay, and the last note plays 7 ms later.

[JeffLearman]

I can tolerate it, but then I'm a hack.

 

I call BULL!

 

You're no hack, Jeff. I've heard ya play.

Flattererer! OK, I'll amend it to "passionate hack".

 

Thanks for adding your wisdom to my semi-krazi thread. (It was BAIT for ya, don'tchaknow?)

Well, you're a master at that.

[Mark Schmieder]

I was reminded recently of OSC's tragic decision to place identifiers at the end rather than the beginning of strings/names/fields of indeterminate length. This and a few other choices related to text-based messaging will probably limit OSC's ability to challenge MIDI as the reigning king.

 

I spoke with the head of the MIDI consortium at AES a few years back, and he was confident at the time that MIDI 2.0 would take off. There's been dead silence since then.

[Theo Verelst]

Midi isn't boring enough compared to most plugins which share a common sound feel that sends most musicians probably bored away to look for new stuff before anything.

 

Midi isn't incorporated in a predictable scheme and can still drive stage lights and synthesizers, computers and modules, even though a 10x speed increase would be nice for heavy piano playing and accurate drumming, probably. (maybe even 100x for accurate musicians).

 

Because of software organization issues it seems that the software replacement of midi is possible here and there (even to another box over ethernet) but it isn't easy to get what's there to act nice with all those timed messages.

 

A timestamped USB midi like at least I know should be possible in new Linux systems is of course an improvement, but is that still midi?

 

[JeffLearman]

A timestamped USB midi like at least I know should be possible in new Linux systems is of course an improvement, but is that still midi?

That solves one or two of the issues, but leaves a bunch more open.

 

I'll be interested to learn more about MIDI 2.