[sdiy] MIDI 2.0

[rsdio at audiobanshee.com]

On Jan 28, 2020, at 10:15 AM, sleepy_dog at gmx.de wrote:
> paula at synth.net wrote:
>> My worry is that it needs USB currently (they've not agreed any other
>> transport layers yet), and that a lot of smaller manufacturers will
>> struggle to implement this. Alongside that, you'll need to be
>> registered with the MMA to get an ID so the CI portion will work
>> properly.
> 
> Not sure what the market is for that.
> But couldn't it be possible that some manufacturer makes an IC that
> handles the USB for you and you deal with a simpler interface?
> Similar to how FTDI does USB<-->UART.
> 
> (maybe someone with connections could pitch this idea to those who could
> do it)

I don’t think it would quite work the same.

A UART is only 2 pins. It’s easy to make a chip handle everything you might need for that.

A synth has a lot more than 2 signals inside. A MIDI processor is going to need to read a velocity-sensitive keyboard, each model with a different number of octaves. You’ll also need to read pots, switches and generate CV while driving LED indicators and displays.

Granted, you could treat the MIDI part separately from the rest, and just make a chip that converts classic MIDI to USB-MIDI. You’d then need a second processor to handle the rest of the features. That’s already been done in products like the monome, with a couple of drawbacks.

One drawback is that the USB-UART chips don’t show up as USB-MIDI devices, so you need a custom driver. This drawback could be solved by simply changing the USB Descriptors. The ezUSB supported custom USB Descriptors, but I don’t know if anybody ever got USB-MIDI working on that old chip.

The other drawback is that a simple classic MIDI to USB-MIDI chip would not be able to take advantage of the higher speed of USB-MIDI. The first USB-MIDI firmware that I wrote delivered data about three times faster than classic MIDI. That’s not always needed, but considering the number of times I’ve heard people complain about classic MIDI bandwidth … it would sure seem like a shame to develop a chip that is stuck with the old limitations.

Brian Willoughby