MIDI 2.0 - Do You Care?

[Theo Verelst]

Accuracy isn't completely solved for a long time to come.. Analogue when you integrate the phase of a VCO under control of a pitch wheel, you'll need a lot of bits and a high sampling frequency of the wheel in order to get really accurate. Hard target to get at.

 

Also, I think it's fine Midi upgrades, in fact there's something to be said for the "Jack" midi approach mostly present in modern Linux-es where midi messages can be timed to-the-sample accurate (if programmed so by the plugin or audio program, which is a whole other matter). That can also a thing I like to have: fixed latency, which is if the incoming midi messages is time stamped in the keyboard unit with a clock coupled to the audio or some other computer - internal (maybe extrapolated) clock, there's a close to completely accurate time difference between playing a key (or drum pad) and the sounding of a note. I'd even like sub-sample accuracy with that, to make the phasing effects in the higher frequencies when combining notes (like a high hat) statistically properly (de-) correlated.

 

As it is, usually the interpretation of midi signals is implicitly coded, i.e. there's hidden meaning in the way midi messages are interpreted in terms of timing, and even in terms of chords and velocity. Not great, but can work.

 

T.

[The Real MC]

Who mentioned anything about old equipment?

 

Craig was addressing this:

 

So the mechanism is there for a more accurate representation of the data but, so far, all of the manufacturers I've tested don't even bother to use the MIDI 1.0 spec to it's fullest.

Of course, my tests are limited to just a few, but take in some of the 'big names'.

 

So, what hope that manufacturers will make any use of the MIDI 2.0 extended ranges, when they don't even use the MIDI 1.0 features currently available?

 

Craig was addressing the shortcomings of processors and the design decisions of new products.

 

Back when MIDI was introduced, keyboards was a very competitive market. Everything was judged by price. To be frank, for a long time after MIDI became reality after 1983 the microprocessors sucked. With precious few exceptions, during the 80s and 90s microprocessors selected for new products were 1970s era 8 bit crunchers like Z80s and 68xx. That was by design in order to keep the price down in the very competitive market.

 

That meant those new products were limited how much they could process dense MIDI streams. I've seen boards that implement aftertouch, but only generated very discrete (not continuous) MIDI AT messages. While the ad copy may claim they implement aftertouch, monitoring the MIDI stream may tell a different story. That's just one example of many.

 

This still remains true today - in an attempt to keep the price down, designers don't use state of the art microcontrollers or peripherals. The trend I have noticed is that by the time a music product is introduced, it is using computer peripherals a generation behind the current technology. I love my Alesis HD24 hard disk multitrack recorder but the TCP/IP protocol they used (10BaseT) was a generation behind the current computers back then and is agonizingly slow (10MB/s) for transferring audio files.

 

In both cases - old and modern - the end user is going to find that not all the MIDI features - 1.0 or 2.0 - are implemented due to design decisions.

 

So is there hope that manufacturers will make any use of MIDI 2.0? That depends on if the customers are willing to pay for the better components to handle the new features.

 

[Anderton]

The processors that were in use in the early days of MIDI struggled to keep up with the serial data stream. With scanning keyboards, if you hit a chord the spread between the first and last note could be as much as 20 milliseconds, and more in a multi-timbral keyboard. Some were around 50 ms. Computers choked on the data stream as well. Remember all those "thinning" filters for controllers, and admonitions to make sure aftertouch was turned off if you weren't using it?

 

This became part of the code that got passed down from generation to generation. Now we have products like the LinnStrument that aren't afraid to send out lots of controller data, and computers that can cope with it.

 

The point of MIDI 2.0 isn't to convert everything out in the world tomorrow, it's to make sure that the spec remains relevant in the future. The companies involved don't want to have to think about MIDI 3.0 in another 10 years.

Who mentioned anything about old equipment?

 

I was writing about modern kit.

 

That's what I was addressing with the comment "This became part of the code that got passed down from generation to generation."

 

But of course, The Real MC's points are highly valid. An i7 8-core Skylake processor is going to cost you almost $700. You won't see that in a keyboard that costs $1,000. And that underlines my comment "The point of MIDI 2.0 isn't to convert everything out in the world tomorrow, it's to make sure that the spec remains relevant in the future." That processor is going to go for $99 someday, and manufacturers will be able to release instruments that can handle data-intensive streams at prices ordinary humans can afford. And when that happens, there will be a spec that delineates what to do with that additional power.

 

It's just a matter of time. At present, the only reason we can get virtual instruments for a few hundred dollars is because they're riding on the back of a $1,500-$4,000 computer.