Good Enough For The Pro?
Thoughts on MIDI's Next Decade
When it comes to synths, samplers, sequencers and effects processors, MIDI rules OK — but when it's suggested that MIDI is equally suitable for professional automation and sync, dissenting voices are raised. Paul D. Lehrman explains why they're just plain wrong, and why MIDI will be more rather than less important in music production as it enters its second decade.
As MIDI enters its second decade, there's a lot of talk around about whether it is still a viable technology, or whether it's time to pass it by in favour of faster, newer ways of pushing music and sound data around. The introduction of inexpensive digital audio recording and manipulation tools, utilising hard disk or consumer tape formats, promises an upsurge in 'real' recording in smaller studios that previously have relied very heavily on MIDI for sound production. Will MIDI continue to have a place in the studios of the mid and late '90s?
I think that it will, and in some ways it will play an even more important role than it does now. Even if a studio throws out all of its synthesizers and samplers and replaces them with 128 channels worth of linked multitrack digital tape decks, there's still going to be MIDI running around that studio. While we think of MIDI as being most successful at the 'lower end' of the music-production market — project and home studios, schools, dedicated low-rate 'MIDI rooms' in larger studios — it has quietly made significant inroads into the high end of post-production too, in areas like audio editing, processing, mixing, automation, and synchronisation, and it's in those areas that it will continue to be useful to studios at all levels.
Although the technology was created to be inexpensive, easy to use, and universal, MIDI has shown a remarkable resilience and ability to keep up with technology. Thanks to its open-ended structure, its widespread acceptance, and the development of advanced hardware and friendly user interfaces, MIDI is no longer just the poor man's way to accomplish many production goals, but is often considered the best way.
Beyond synthesis and sequencing, real-time control of sound via MIDI is today an important part of many professional installations, both in the studio and on the stage. This is particularly evident in the area of signal processing: any new reverb, delay, equaliser, or pitch shifter can be controlled in some way by MIDI, and most manufacturers are falling all over each other to provide more real-time control over more parameters. There are two crucial areas, however, in which MIDI has yet to reach its full potential, and that's where most of the innovations of the next few years are going to appear. Those areas are automation and synchronisation.
Part of the reason why so much work remains to be done is that despite MIDI's economic and ergonomic advantages, it is still looked down upon by many high-end manufacturers and users as a 'semi-pro' tool. This attitude can usually be traced to a combination of misunderstanding and superstition, and if it was ever justified, recent additions to the MIDI Specification certainly make it so no longer. It's a self-defeating attitude, because those who dismiss MIDI are missing out on a good thing, no matter what kind of music or sound they produce. As the prices of professional-quality audio gear continue to plummet, it is crucial to the health of the industry that MIDI, the most successful standard in the history of music technology, and arguably the single most important development in music and sound production since the introduction of magnetic tape, be recognised and utilised to its full potential.
Anyone with an automated console knows there is no such thing as an automation standard. Create a complex mix on one board, then bring the multitrack tape to another studio and chances are you're going to have to start creating the mix all over again, because the only thing the two systems may possibly have in common is that they both rely on a timecode track. But MIDI is a standard, and MIDI commands — even conventional ones like notes, continuous controllers, and program changes — lend themselves very well to automation purposes.
MIDI-based automation has actually been with us for some time. Yamaha paved the way with their DMP7 mixer, which is already over five years old, and its slightly-more-recent offshoots, the DMP11 and DMP7D. Program changes in these devices can be used to choose one of 99 complete setups, and I mean complete: not just levels and mutes, but also pans, parametric EQ settings, and send, return, and parameter settings for three independent effects loops. Besides program changes controlling 'scenes' or 'snapshots', MIDI notes or continuous controllers can adjust any individual parameter in real time, whether it's a channel fader, a channel's reverb send level, or the bandwidth of a particular section of a particular equaliser. Even the time it takes to go from one scene to the next is remotely controllable over MIDI.
The DMP7 and its brood had their problems, but they are still available and have very devoted users. They're particularly popular in audio rooms at busy post-production houses, where client turnover is fast, and 'total recall' of a mix means that a console must be able to do exactly that, not just make a printout of a bunch of dial settings, which a second engineer will then spend half an hour trying to duplicate. But despite their relative success, Yamaha's mixers still stand in a class by themselves — no other manufacturer has yet to implement anything approaching their comprehensiveness. (Yamaha's own DMC1000 digital console has an equally sophisticated MIDI implementation, but it is designed for a very specialised market, and has a price tag — an order of magnitude higher than the DMP7's — to match.) Quite a few consoles now offer MIDI control of mutes, a few offer control of channel levels, and there are several 'aftermarket' MIDI-based mute and level systems that can be added onto existing consoles, but until very recently no one has taken the comprehensive, no-parameter-is-safe approach of Yamaha.
Why haven't others jumped on this bandwagon? Why don't Sony, Otari, SSL, Amek, and other major console manufacturers adopt MIDI as an automation standard? If you talk to most of them about this subject, they'll dismiss MIDI control as being impractical, claiming that MIDI isn't fast enough to handle 'real' automation. But besides the obvious evidence to the contrary provided by Yamaha, an analysis of this attitude shows that it is not based on any real problems within MIDI itself.
Let's look at some numbers. Assume we have a 64-input console, and we want to fade all of the inputs from maximum to minimum at the same time. If we have a different MIDI controller corresponding to each fader, we theoretically need to send 128 different controller values for each of those 64 controllers. That's going to take a while: 0.625 milliseconds per command, or about five seconds, if we're using Running Status. If the whole board is not on one MIDI channel, or if you want to use double-precision controllers (for 16,384 level values per fader), that time increases to eight or ten seconds. Certainly much too long.
But this is assuming a very old-fashioned design for our console. What if it were designed so that instead of needing continuous instantaneous values, each fader could simply be given a target value and a delta-time in which to move to that value, from which the console itself would work out the intermediate levels? Now we're down to 128 commands, or 80 milliseconds. Not bad. How about if we create automation groups — which is, after all, the way most dedicated console automation systems work? With 64 faders in eight groups, we're down to 10 milliseconds. And if we take the program-change approach, and tell our console to go from the current status to a full fade in a specified number of seconds, we can do that in two commands: 1.25 milliseconds. This is more like it.
So, if your idea of a 64-input console is one that can only be operated by people with 64 hands, yes, MIDI is too slow. But in a real console, designed for use in a modern environment, MIDI is plenty fast.
What about hardware? Is there anything intrinsically inferior in using MIDI to control the hardware in an automated console? Absolutely not: VCAs are VCAs, no matter what you send them, and with increasing use of on-board digital processing in consoles, the issue becomes even more moot. If 128 discrete values for fader levels is too few, 14-byte double-precision controllers can be called into use when needed, and/or the console itself can do some of the work of interpolating.
Making things even more interesting are recently-adopted or under-discussion additions to the MIDI Specification such as master volume and balance controls, which have the potential to greatly simplify the data structure (and lower the bandwidth) of automation systems, and MIDI Machine Control, which is already beginning its evolution into a common language for any conceivable studio event, including automation.
So why hasn't MIDI become the automation standard already? Why don't automation systems simply consist of a £500 computer with £200 worth of software and a £50 interface? Part of the problem is that we are an industry in which image plays at least as important a role as reality. When high-end studios are constantly involved in, as a writer friend of mine likes to call it, an "arms race" for the most elaborate and impressive equipment, it's difficult to convince either studios or clients that a MIDI-based automation system costing less than a good pair of monitors just might be better than the system down the street that costs a quarter of a million pounds.
To be honest, all of the problems with MIDI automation are far from solved. Even though the hardware is up to the task, there are still plenty of user interface issues facing MIDI automation designers — just like the makers of more traditional systems. A computer mouse, no matter how useful it is for sequence editing, is a lousy tool for running a multi-channel mixing console. Smaller companies like JL Cooper and Russ Jones Marketing (Steinberg's US distributor) have introduced dedicated control surfaces — fader boxes — for MIDI automation, but they are just the beginning. Yamaha was able to take advantage of its corporate enormity to include moving faders on the DMP7 (they borrowed them, a company spokesman told me, from their organ division), but other companies do not have such resources available, and if they were to try to include moving faders it would make their consoles prohibitively expensive. Therefore, other solutions have to be devised.
Another challenge to manufacturers is how to avoid 'zipper noise' — audible discrete jumps in level caused by adjacent steps being too far apart — while at the same time making sure the MIDI data stream doesn't choke and cause delays due to an overload of data.
The software side needs work too. Music-oriented sequencers certainly have enough power to handle automation, but only a few of them present user interfaces even remotely appropriate for the task. Companies like Steinberg and Digidesign have created computer-based automation front ends, but they still don't provide a familiar enough interface, or fit easily enough into existing facilities, to gain wide acceptance. Again, it's going to be up to the larger console makers in America, England, and Japan to recognize the advantages of MIDI, and come up with systems that engineers can learn quickly without having to acquire lots of new skills, before it will become the automation standard it deserves to be.
But we're getting there. The good news is that there are companies working on MIDI-based automation systems that approach Yamaha's in completeness, but are much cheaper. First among these is Fostex, who last year introduced their DCM100 mixer and Mixtab control surface. In addition to 99 snapshot memories, recallable with MIDI program changes, the system provides MIDI control over the levels of eight stereo input channels, plus 2-band fixed-frequency EQ and two auxiliary sends on each channel. In addition, the auxiliary returns (and their own EQs) are controllable. All of these functions use standard MIDI continuous controller commands, so they can easily be edited with any decent sequencer.
The whole Fostex system retails for less than £750. For that kind of money, moving faders are out of the question, so 'nulling' the control surface — that is, getting the physical faders' positions to correspond with their 'virtual' positions as controlled by MIDI — has to be done manually. Fostex has included a 'preview' mode, not unlike that in some far more expensive dedicated automation systems, that uses multi-coloured blinking LEDs to help you do this quickly and relatively painlessly. Also included is a 'Smoothing' control that determines the density of the MIDI data being sent by the Mixtab. For large, quick fader movements the control is supposed to be set so that fewer MIDI commands are sent, and the values of adjacent commands can be relatively far apart. For slower, more subtle moves, the control should be set so that the data density is greater, and the values of adjacent commands closer together, thereby preventing any possible zipper noise from being created.
Another champion of inexpensive MIDI-controlled automation is Mackie Designs, the American company whose CR1604 16-channel console has proven immensely popular in project studios. Mackie has announced (and should be delivering by the time you read this) OTTO-1604, a MIDI-based automation package for their mixer, which will handle muting and levels of every fader, the main and alternate stereo outputs, and the four stereo returns, all for $800 (US). The hardware part of the package will be on a user-installable circuit board that mounts inside the mixer, attached to a MIDI 'breakout' box with a ribbon cable. It will use standard controller commands, so it will be compatible with most sequencers, and the company is also developing a dedicated Macintosh-based front end. Since the physical controls on the board are not used to address the automation, nulling is not a crucial issue: when you are in automation mode, you put all the controls at their '0' position (the board is designed around a 'unity gain' concept, so this is the controls' normal position anyway), and let the sequencer or computer do the work. A hardware front end is also in the works, although details are not available as of yet.
With these two companies paving the way, it hopefully won't be long before other manufacturers get the idea, and we begin to see comprehensive MIDI automation packages at many different levels of the industry.
Using MIDI to synchronise different elements in a production session actually means several different things. Although most engineers tend to think of synchronisation in terms of timecode, vertical drive, and tach pulses, there's another kind of synchronisation that MIDI excels at, which is seeing increasing use in music and post-production studios. With the growth of high-fidelity, high-capacity, fast (ie., SCSI)-loading samplers, these beasts are getting very popular at tasks that the designers of a few years ago couldn't even dream of. They are doing grooves and backbeats for dance records, Foley and ambience tracks for films, dialogue and effects for video and broadcasting, and even entire music tracks for a wealth of audio and multimedia applications.
Samplers work with ordinary MIDI commands, and their clocks are phenomenally accurate: if you record a one-minute piece of dialogue into an Akai S1100 or a Roland S770, when you play it back, it will still be one minute long, give or take absolutely nothing.
As with automation, however, a lot of people think that MIDI isn't powerful or accurate enough to trigger samplers reliably in the real world of post production. They're wrong — and they're wrong when they put down other aspects of MIDI-based synchronisation as well.
First of all, let's look at a typical modern computer-based sequencer. It should be able to divide a quarter note into at least 200 parts. If that sequencer is playing at 240 beats per minute, it has an accuracy of 1.25 milliseconds, or 1/27th of a frame, or three SMPTE bits. Some sequencers give far higher resolution, and let you run them at higher tempos, so you can spot cues accurately down to a single bit, or even better.
But doesn't MIDI's limited bandwidth (31,25kHz) get in the way of this accuracy? No — the limitation on the number of commands a MIDI line can handle in a given time period has nothing to do with how accurately it can send individual events. The bandwidth limitation only becomes a problem when you are sending lots and lots of MIDI data down a line, such as when you are controlling 16 synthesizers, each with its own pitch bend, modulation wheel, aftertouch, and other data-intensive parameters. When you exceed the capacity of 1000 or so commands in one second, some of the commands will be delayed. Musicians who work with many MIDI instruments have learned to overcome this problem by using special interfaces and sequencers that support multiple MIDI cables, and distributing the load among them.
However, the amount of MIDI data that is used in a typical post session is far, far less than what is used in music composition. A half-dozen 16-voice multi-timbral samplers will give you the effective capacity of four linked 24-track tape decks, and yet will require only a small fraction of MIDI's bandwidth to control: you could fire every sound on every sampler in the space of three frames, and not run into bandwidth problems. And when the amount of data is relatively low, as it usually is, MIDI can be stunningly accurate.
The accuracy is due to the fact that MIDI is an asynchronous protocol, meaning that it doesn't follow any kind of master clock. A MIDI command can be transmitted and received at any time, not just, as would be the case if it were a synchronous protocol, on arbitrary boundaries of 1/31,250th of a second (Figure 1). The limiting factor is not MIDI itself, but the transmitter and receiver.
If the device generating the MIDI command has a processor clock running at 1MHz, a common speed for MIDI interfaces, then the accuracy of an individual MIDI byte can be predicted to 1 microsecond. Good enough for you? Actually, it's probably a lot better than you need, because the receivers themselves aren't nearly as accurate: recent research published in Keyboard magazine (in the US) shows that most MIDI devices not only have delayed responses to MIDI commands, but those delays often vary randomly. However, modern samplers fared well on the tests, with the deviations ranging from less than ten to a few hundred microseconds. Still plenty good enough for post work.
Of course, none of this means anything if the MIDI sequencer/generator is not itself hooked up to some kind of video-based synchronisation. Here we get to what might be the crux of the confusion about MIDI synchronisation: how good is MIDI Time Code? Since most Macintosh sequencers and an increasing number of those on other platforms lock to MIDI Time Code, this is an important issue indeed.
The answer is, it's very good. Again, it's the asynchronous nature of MIDI that's the key: if it were synchronous, an MTC byte would have an accuracy of 1/31,250th of a second. But since it's asynchronous, as long as there's not a whole lot of other MIDI data clogging up the line (and the MIDI Time Code specification suggests — and most manufacturers go along with this — that MTC should be on a cable of its own), the resolution is theoretically infinite. If we use our friend the 1 MHz processor, MIDI Time Code can be generated with an accuracy of 1μs (microsecond). How does this compare to SMPTE? The SMPTE spec allows 4.2μs deviation in the length of time between SMPTE bits, so in this case MTC is four times more accurate than SMPTE.
But what about the fact that MTC only can resolve down to quarter-frames, while SMPTE can resolve to 1/80th of a frame (one bit)? Irrelevant, because of interpolation.
Once upon a time, when MIDI Clocks were the only way to get sequencers to lock to tape, a few pundits decided that because Clocks only come down the line 24 times per quarter note, any sequencer locked to them could only play with a precision of 24 parts per quarter note. All the sequencer makers, therefore, that claimed higher precision (96, 240, 480, 1024, or whatever) were lying.
What these deluded souls failed to take into account was that a sequencer receiving a string of Clocks could interpolate much shorter time intervals between those Clocks. In other words, if a note were supposed to happen 1/96th of a beat after a quarter note, the sequencer listening to the clocks could make a reasonable guess when the next Clock (1/24th) was supposed to arrive, and fire the note when one quarter of that time interval had passed. The maximum resolution was therefore not dependent on the speed of MIDI Clocks, but on how finely and accurately the sequencer could divide up the times between Clocks — in other words, on the speed of the sequencer's internal processor, and the precision of the software (Figure 2).
It's the same with MIDI Time Code: although only four MTC messages arrive per frame, a sequencer can interpolate times between those messages to whatever degree its designers wish. That's why there are MIDI sequencers that let you specify SMPTE times of events in 80ths or 100ths of a frame: these programs are not just showing off, they really can divide time up that finely. An added bonus of MIDI Time Code is that, unlike MIDI Clocks which are tempo dependent, MTC is sent at a steady rate. If there is a tempo change in a MIDI Clock setup, the receiver will for a brief moment interpolate time incorrectly. But that doesn't happen with MTC, so MTC-based timings are extremely reliable.
The most exciting development in MIDI synchronisation since MTC is the new MIDI Machine Control (MMC) protocol. MMC is designed to make any mechanical or computerised device in a studio completely operable from one central control surface, using a common digital language for all. How seriously are manufacturers taking it? Tascam and Fostex are already using MMC to link computer sequencers with multitrack tape decks, and in a potentially revolutionary move, Alesis is using MMC as the 'native' control format for their Big Remote Control — the soon-to-be-available box for linking multiple ADAT decks to each other and the rest of the world. If more forward-thinking manufacturers take a good hard look at MMC, they'll find a comprehensive standard, easy to implement in both hardware and software, that they will be able to use for a huge number of tasks in the years to come. MMC may do more than anything else to keep MIDI alive and well in its second decade.
The conclusion you have to draw from all of this is that MIDI does indeed have a future. While musicians at all levels of accomplishment will continue to get great use out of MIDI's original capabilities, the professional production community will be doing more and more to keep it active and growing. MIDI is not 'for musicians only', or 'low-end', or 'not good enough', or 'just not pro'. It's a standard that has proven itself over and over again in the world of music production, and will become increasingly important in every aspect of audio production. It's still a vital, useful tool, and the prospects for its second decade are just as exciting as the great strides it made in its first.
Paul D. Lehrman is a composer, consultant, teacher, author, and long-time contributor to Sound On Sound. Based in Boston, USA, he is on the faculty of the Sound Recording Technology program of the University of Massachusetts-lowell, and sits on the Executive Board of the MIDI Manufacturers Association. His book, MIDI For The Professional co-written with Tim Tully, is scheduled for publication early in 1993.
Feature by Paul D. Lehrman
mu:zines is the result of thousands of hours of effort, and will require many thousands more going forward to reach our goals of getting all this content online.
If you value this resource, you can support this project - it really helps!