MIDI Automation Systems (Part 2)
Last month we explained the development and requirements of VCA-based mixer automation systems. This month, Graham Hinton scrutinises three retrofittable automation systems - the Twister PAC, MegaMix, and J.L. Cooper MidiMation - and asks 'How good are they?'
Over the past year a rash of products have appeared all vying for your hard-earned (or yet to be earned) cash and offering in return the universal panacea of mix automation - and even more appear as I write. Yes, your basement/home recording studio/country mansion can be magically transformed into the equal of the best recording studios in the land, merely by buying a box containing some VCAs and connecting it to the insert points of your mixing desk. You may only have a half-working, beaten up, vintage MM 12-2 mixer, but you will be transformed to true professional standards with facilities you never knew you needed. After all, if famous pop stars can waste a fortune working in SSL studios watching pictures of faders going up and down on screens, why can't you waste a fortune too and do the same thing in your basement/home studio, etc?
Enough!! We've all heard the hype, but why would we want mix automation anyway - and what precisely is MIDI mix automation?
Mix automation has been around for over a dozen years now, first appearing in the mid-Seventies when multitrack recording and mixdown started to become complex. Sixteen tracks could be coped with, but 24, 32 or 48 tracks became just a bit too much to remember all the fader movements. Various automation systems came out, mainly retrofittable to existing mixing consoles, that would record the fader movements and allow them to be replayed and edited. As microcomputers were practically non-existent then, these systems were either based on cumbersome hard logic circuits or a minicomputer, and as floppy disks were only just coming in, the fader movements were often attempted to be recorded on a track of the audio tape.
All of the early systems were far from ideal and have been mercifully forgotten. Only two systems have survived from the Seventies: Neve and SSL, with the latter now strongly dominating the field. It is important to note that both are disk storage based and both are integrated systems. If either mixing desk were not computer-assisted it would still be a complex and powerful recording tool, and in both cases the computer does quite a bit more than just remember the fader levels.
Mix automation started getting big about six years ago when it was beginning to be used more to achieve artistic effects than merely remembering a manual mix. Many complex mixes heard today could not have been done without an automation system.
The job of the automation computer is to control the audio level of all channels of the console as transparently as possible, to control all the tape machines in the studio by autolocating them and synchronising them together, to control other devices in the studio, to eliminate mundane tasks like keeping notepads of track lists and cue points, and to present the whole operation as one coherent system without being overbearingly 'computery', the way computer systems are wont to be.
The system is locked to SMPTE timecode recorded on tape, but as the computer has control of the tape machines it is possible to cycle between two cue points until the mix is right and then move on. This is especially important for TV and film work, where it is impractical to keep running through programmes that are often over half an hour long. Once random drop-ins are possible the computer has the responsibility of reconstructing the fader levels from a known point, which it has to do by following every movement from a snapshot reference that it periodically takes. The effect of dropping in from a quiet passage with the faders fully up to a bit of fortissimo heavy metal is quite disconcerting, even if it only lasts one second.
As with any programmable system, the controls get out of step with their true settings. The Neve automation system introduced motorised faders so that the fader knob could always be brought to indicate the current level, whereas the SSL system uses a high quality VCA (voltage controlled amplifier) with a fader that has various modes (the most obvious being absolute, relative and off) and a means of indicating that it is matched to the remembered level.
Both systems are a compromise and have their strong and weak points. Motorised faders never quite feel like a conventional fader or allow really fine tweaking, but are always in the right position, even if they knock cups of coffee over to get there.
The main argument for using motorised faders is that they do not alter the sound quality, which is not in fact true. They are subject to all the same quantisation and control slewing and lag effects as VCAs. Non-motorised faders have to be manually nulled to get them in the right position when remixing and all the different modes necessary can become confusing. Faders on a desk are like a steering wheel on a car - no other form of control is acceptable to the user despite whatever is technically possible.
With the huge success of the SSL mixer automation system in recent years - there are now over 400 installations worldwide it is not surprising that more economical imitators would appear. There are now no less than six retrofittable VCA systems being marketed, of which three have been scrutinised over the last couple of months for this article: the Twister PAC, the IMD Megamix, and the J.L. Cooper MidiMation.
All of these systems provide a set of Voltage Controlled Amplifiers (VCAs) that are intended to be connected to the insert points of a downmarket mixing console which normally would use a stereo jack socket for the insert point. Consequentially, all have multiple stereo jack connectors with the input and output on the ring and tip. I'm not in sympathy with the reasoning behind this approach. The insert point on such consoles is often the poor relation in the channel, and has been known to cause hum problems. When the mixer manufacturers finally notice that we're living in the Eighties and start fitting VCAs to their products themselves, they will be able to optimise both the audio chain and the cost leaving these products out in the cold.
In all three systems looked at, the audio path seems added almost as an afterthought, being no more than the VCA manufacturers' application note, sometimes less. None of the systems are balanced and only the Twister uses an output amplifier capable of driving a decent length cable run. The MidiMation doesn't even have an input amplifier, just a resistor straight into the VCA.
The reason these systems all claim low noise and distortion specifications is that the audio path in these devices is minimal. When connected across a room, possibly through a jackfield if the insert point were already in use, the actual noise figures (let alone the crosstalk) will be nothing like what the manufacturers' specs would like to imply. However good the VCAs are they will not improve the audio performance of the mixing console, they can only add the functionality to whatever exists already.
When manufacturers start claiming better than -90dB specs I immediately get suspicious. After all, we are talking micro-volts. I get suspicious for two reasons: I know how difficult it is to achieve that performance in practice, and I know how difficult it is to measure it reliably. Most common pieces of test gear, like the Amber 3501 I used, do not go down below -85dB. I was not allowed to take a Megamix or MidiMation system away to make my own measurements, hence the gaps in the table. The Twister, which claimed better than -90dB noise, was actually measured at -81.5dB noise level (ref: 0dBm), which was predominantly mains hum, and that was without being connected to anything!
All of these units had mains transformers fitted inside the same enclosure as the VCAs and without any screening. Only the Twister used a toroidal transformer, the other two looked more like a car battery charger.
I would like to have seen a balanced line input amplifier and an electronically balanced line output amplifier, just like you would find on a thousand pound reverb unit, and I would like to have seen regulated DC supplies taken to the audio circuitry enclosure, just like most consoles do. Even a single-ended line output would be acceptable, if it used a decent op. amp and could drive into a low impedance, but a normal TL070 series type op. amp (as used in Megamix and MidiMation) is not suitable for such use. In short, these omissions mean that they do not conform to good, sensible and proven audio engineering practice.
It would also be far more practical to have one or two multipin connectors on the rear panel to ease the wiring of a cable loom to the patchbay or desk. The use of stereo jacks has actually precluded the possibility of balanced operation on the grounds of economy, yet multipin connectors would cost less and the simplest balanced input stage would only have cost two resistors more per channel.
All three systems use MIDI as the control data medium, but in different ways. The Twister transmits MIDI data from its front panel controls for recording by an external sequencer and obeys the same data on playback. It transmits fader and mute, subgroup and master information as Controller Changes and transmits a snapshot every two seconds as a System Exclusive block. It can also recognise Program Changes. The Megamix VCA unit only receives MIDI data from its own software, running on a Macintosh or IBM PC, which serves as a soft control panel and sequencer. The MidiMation uses MIDI to communicate between the MAGI unit, which controls the VCAs and remote control unit, and either SAM (a dedicated automation manager) or another MIDI sequencer. This system can transmit the VCA level as either Note On codes or Non-Registered Controller Changes and uses System Exclusive blocks for housekeeping between the SAM and MAGI units.
None of the VCA units have any knowledge of the passage of time, they obey only the MIDI or manual commands presented to them. One of the shortcomings of using MIDI for all fader movements is that it won't be busy a lot of the time - after all, faders don't go up and down like yo-yos (despite what computer programmers think), it's the music that's really doing the work - but when it does get busy, like a scene change, nearly every fader could change at once and it would take a minimum of 18 milliseconds to transmit 56 fader changes in MIDI running status, which is a bit too slow. None of these systems are conceived as sharing a MIDI data stream with a keyboard or sequencer, but rather to be used as a separate piece of MIDI plumbing.
The Twister, by passing the responsibility of control to an external device, is only as good as that device and the average MIDI sequencer is not a good starting point. MIDI sequencers are not automation computers. They are oriented towards controlling musical note-type events where it is safe to assume that the default state is Note Off and even Continuous Controllers have off positions, eg. Pitch Bend can be assumed to return to centre. This type of action does not work for fader levels, and equating faders and mutes to Note On events and/or Controller Changes, to trick a keyboard sequencer into recording it, does not conform to the official MIDI Specification and you can expect problems. For example, when synchronised to tape and that tape is wound to a different point, the automation system thinks it is still at the previous location. Therefore the faders will be out of step with the music at the drop-in point, until the system is refreshed. In Twister's case it takes two seconds for it to be updated.
The fader levels were transmitted as absolute Controller positions, but subgroup and master information was transmitted as relative Controller positions, eg. increment subgroup is sent as controller 'n', position 1 on each increment, and decrement subgroup didn't actually work on my sample so could only be used for heavy metal mixes! This is definitely non-kosher MIDI, as all MIDI Controller data is defined in the official MIDI Spec as being absolute, and some modern sequencers would condense the repeating codes interpreting them as superfluous and thus loose all the subgroup and master changes.
Everything on the Twister has been optimised for speed, there is actually an individual 8-bit DAC for each VCA and the response time is less than one millisecond from receiving MIDI data to actual level change. This, in fact, is its undoing - every change produces a click, as there is no control voltage slewing applied whatsoever. Zipper noise is noticeable on all but the fine control setting and worse on external MIDI control. At normal studio monitoring levels (loud), the click from a mute or unmute would "blow your teeth out", as they say in the trade.
As the model presented for review was supposed to be a production unit rather than a pre-production preview unit, I don't find this state of affairs acceptable and wonder what people that have shelled out good money think. I have discussed these faults with the manufacturer and have now heard that they have been rectified. Overall, the Twister is an otherwise very well engineered unit and is capable of better performance with some minor modifications to the hardware and software.
Whereas the Twister has a thousand pound entry price for eight channels, the Megamix system will set you back two and a half thousand for 16 channels, and then you'll need a computer (Macintosh or IBM PC) to drive it on top of that. The Megamix is controlled by a mouse clicking on computer graphic simulations of faders on-screen and moving them up and down. This, of course, means that you can only adjust one at a time which takes a lot of immediacy out of a balance. As for mouse-controlled crossfading, this displays a singular lack of understanding of what happens to the levels in a crossfade - it is never the reciprocal of the knob position.
Mutes and unmutes had a definite attack and decay characteristic of over 15ms, which is generally a bit sluggish and the software very vanilla. The Megamix was the worst engineering quality of all the equipment looked at. What was supposed to be a production model looked like a prototype knocked up in a garage, which is not acceptable for this sort of price. For the money, you don't even get a microprocessor to control the VCAs, the MIDI data being decoded in hard logic. I asked the distributor for details of the MIDI usage and three months later haven't heard anything. I only hope they serve their customers better.
J.L. Cooper products have not always enjoyed a reputation in this country for professionalism and reliability, but after looking at the other products it was a relief to find something that at least did what it said it could. The MidiMation system really needs the SAM (SMPTE Automation Manager) and the SAM Disk unit to get the best out of it, as this takes on the responsibility of getting the fader levels right at drop-in points and interfacing to SMPTE timecode, so the full system ends up quite pricey and lacks the quality that would normally go with that price.
The MAGI unit is a 3U rack-mount case containing a microprocessor, DAC and sample & holds that connect to separate MAGI VCA units via ribbon cables. Each VCA unit contains up to 16 VCAs in a 1U rack-mount case. The largest system of 56 VCAs would thus take up 7U of rack space plus a further 2U for the SAM and SAM Disk units. Surprisingly, the MAGI contains mainly empty space, where the VCA units are crammed and yet have an individual mains power supply.
The rear panels of this suite of units are extremely busy with audio, mains, control voltage, MIDI and RS232 connections and it's no surprise to learn of mains hum problems. Using MIDI as the control link between a computer and a piece of audio processing equipment is normally a good idea, as the opto-isolation comes for free. As soon as other digital ground paths are introduced, as in both the Twister and MidiMation, all the advantages are jeopardised.
There are two types of remote fader units for use with MAGI, an eight channel bank switched unit and a larger one which may have 16, 24 or 32 faders plus four subgroup faders. The faders are positioned very closely together with no room for a scribble strip and I experienced parallax problems when using it placed over the mixing desk to which the VCAs were inserted. Cheap carbon faders are used on the grounds that no audio is going through them, but in automated mix systems the faders get used a lot more than in a conventional one, especially if they can be reassigned, and as it is the main human interface, 'feel' is an important consideration.
However, in terms of sound control the MAGI + VCAs was the only piece of equipment I looked at which did what I expected of it. There were no clicks or zipper noise on fades or mutes and yet the fade response time was not heavily sluggish either. This is also the only system that had SMPTE control as an integral part of it, rather than relying on Song Position Pointers via an external SMPTE-to-MIDI convertor.
All of these automation systems have only 8-bit resolution, which can give no more than 0.5dB accuracy and with the MIDI implementations only have 128 steps (64 in the case of the Megamix), whereas professional systems are 10- or even 12-bit accurate with at least 250 steps.
Professional systems are normally SMPTE frame accurate, but no better, despite constant demands for at least quarter-frame accuracy on mutes. Although a MIDI-based system could potentially offer the resolution expected of any sequencer, none of these systems have been designed with this as a goal. The Twister could provide this, but clicks. The MidiMation MAGI does respond to quarter-frame accuracy on a small system, but degrades to frame accuracy in a large system and the faders and buttons aren't read faster than frame rate anyway.
Both the MidiMation and the Twister have auxiliary programs that display the fader positions on a computer screen (like the Megamix). This is not actually a lot of use. In fact, in all three systems a personal computer is not really employed to any great effectiveness, considering the current sophistication of software sequencers, for example, and that a perfectly good system from which to crib ideas has been sitting around for over ten years. All of the useful computer assistance features are missing: no track lists or cue lists; no ability to define, edit or use cue points; and no programmed drop-ins and drop-outs.
What I would far rather see a computer screen used for is to actually design complex crossfades between cue points by drawing a multi-channel graph of level against time, rather than keep trying to perform the fades and edits until you get them right. In other words, let's see the computer assist in performing operations that would otherwise be difficult to perform manually, rather than just sit there waiting for you to get it right by repeated performance.
If you have a collection of MIDI synthesizers under sequencer control and one goes down on a session, it's not the end of the world. If the sequencer or the mixer goes down it is. However, two of the UK distributors of these products had no circuit diagrams, service manuals or spares for them, leaving one wondering what happens when one goes wrong?
If you've got the impression that I'm not impressed with this batch of products then you're absolutely right. Eight VCAs in a box no more constitute an automation system than eight VCOs in a box make a polyphonic synthesizer. All the little bits that add up to the whole are lacking. Despite this current situation I do believe that a MIDI Automation System should be possible that would actually exceed the performance of the current professional systems.
Several things are necessary before that can happen, the most important being a standard (and cheaper) way of locating tape machines under the control of a software package running on a personal computer. The controlling software and the audio hardware would probably be produced by different companies, each with the correct expertise, just as sequencer packages and synthesizers are. Such a system would strongly exploit the inherent features of the MIDI Time Code specification defined by the MIDI Manufacturers Association, instead of just adapting MIDI in a private badge way.
The whole idea of MIDI Time Code devices is that a distributed manufacturer-independent system of devices can be pre-loaded in a defined way over normal MIDI cables, which are then used to pipe the timecode around. When each device recognises its own cue points, at the right time it takes appropriate action.
Products based on MIDI Time Code must now be past the gestation period. We've already seen Digidesign's Q-Sheet program for the Macintosh appear, which is a step in the right direction. A well designed, analogue submixer implemented as an intelligent MIDI Time Code peripheral is what I'm waiting for.
Part 1 | Part 2
Gear in this article:
Feature by Graham Hinton
Previous article in this issue:
Next article in this issue: