MIDI Automation Systems (Part 1)
How Good Are They?
Over the next two issues Graham Hinton tests five mixer automation systems which all provide control of audio signal level by means of Voltage Controlled Amplifiers (VCAs) and are accessible to some degree by MIDI. This month it's the turn of the Akai MPX820 and Simmons SPM8:2.
Several products have appeared recently under the broad category of 'MIDI Automation Systems', and although they approach the application in different ways they are still sufficiently similar to survey them together and examine their relative merits and absolute fitness for the task. The products we shall be scrutinising over the next few months are not necessarily competing with each other, they fulfil different functions and vary greatly in price, but all provide control of audio signal level by means of Voltage Controlled Amplifiers (VCAs) and are accessible to some degree by MIDI. Graham Hinton puts them to the test.
Although some of these products are claimed to be automation systems, I would prefer to term them automation components, maintaining my stand against marketing opinion that an audio automation system must locate at least one tape machine to be worthy of the name. In this light, I shall be assessing the Akai MPX820 and Simmons SPM8:2 as MIDI-controlled submixers this month, and the J L Cooper MIDImation, IMD Megamix and Twister PAC as MIDI VCA retrofits for non-automated mixing desks, next month. How well do they perform their audio function and how well can they be integrated into a home or semi-pro MIDI studio system?
When I approached the various distributors for samples of their wares I was greeted with scepticism: "But it's the best VCA on the market," they cried, "why do you want to take one away and measure it? What do you think you'll find?" Well, in the past I've been burned and so have others, so now I'm cautious. Complex systems cannot be judged on the basis of a half-hour salesman's demonstration and you only really appreciate what a piece of equipment can or cannot do after attempting to use it yourself in earnest. So, before looking at the products themselves, it would be useful to explain some of the background and give some insight to the problems associated with VCA control.
The development of VCAs themselves goes back to the early seventies. There are two basic principles and two uses have polarised. There is a transconductance amplifier which can be traced back in transistorised form to analogue computers and the early Moog and ARP synthesizers, which was first integrated as the ubiquitous (and notorious) CA3080 and later refined by Ron Dow of Solid State Microtechnology (SSM) and Doug Curtis of Curtis Electro-Music Specialities (CEM) as a variety of general purpose analogue synthesizer building blocks without which the E-mu, Sequential Circuits and Oberheim products would not have been possible. The other type is the logarithmic multiplier exemplified by dbx VCAs (but since David Blackmer of dbx owns the patent, this is not an available option for anybody else) and without which professional automated mixing consoles would not have been possible.
Broadly speaking, synthesizer VCAs have been developed to optimise DC integrity, low control voltage feedthrough, low power supplies and consumption (for battery operation) and minimum external components and adjustments. They are more useful if they invert the signal when used with an op-amp and have a maximum gain of unity. The noise level nowadays tends to be in the fairly respectable -80dB range, and who the hell cares about a mere 1% distortion when you're filtering a sawtooth wave? On the other hand, professional audio VCAs are optimised for AC integrity, low phase shift, low distortion, high power supplies (for headroom) and a precise gain characteristic. They are more useful if they do not invert when used with an op-amp and have gain if required. The noise level tends to be in the highly respectable -90dB range, which is unlikely to be bettered in the foreseeable future - not with silicon anyway.
There are only two pro-audio VCAs available in IC form: the dbx 2150 series, of which the 2151A is the best performance part, and the Aphex (B + B) 1537A. Both were originally developed for outboard dynamic processing effects equipment. There are higher performance devices, but they are hand-built from selected transistors into modules rather than ICs, need external trimming components, and are consequentially very expensive. There is a larger choice of synthesizer VCAs but only two manufacturers, SSM and CEM, offering a variety of input, output and control structures and a slightly reduced, but still respectable, audio specification. Japanese synthesizer companies sometimes make their own proprietary devices but mainly use the American ICs. In fact, it is utterly amazing that so many manufacturers bother to make so many high quality VCAs at all for such a relatively small industry and market, when the same silicon could be used more lucratively making memories or gate arrays. Analog Devices tried to get in on the act a few years back but gave up because it was too difficult (!) and they would get no real return for their investment. Personally speaking, I never want to hand-match a pair of transistors again so please keep making them, guys.
A VCA is intended to be equivalent to an audio pot, a signal goes in one terminal and a fraction of it comes out another determined by a control voltage. Obviously all that silicon is going to introduce some more noise and distortion than just a pot by itself, but only the type of people that can detect the effect of the Earth's magnetic field and whether there is an 'R' in the month on their precious metal speaker cables are likely to quibble. The rest of us are going to be pushed to tell the difference between a normal audio pot attenuating a signal and a modern VCA being controlled by a voltage from a pot.
Ah, if only life were that simple. All of the automation systems being examined are not controlled by straightforward control voltages derived from individual pots, they are controlled by control voltages from a DAC (digital-to-analogue convertor) under microprocessor control - and here lies the root of all evil!
It takes about three VCAs in series before the static distortion effects become noticeable, and a large mix before noise and modulation begin to become perceptible. Next time you feel like wasting large sums of money in studio fees, try doing the same mix on the channel (VCA) and monitor (not) faders of a 56-channel SSL mixing desk, this being about the only known way of making a direct comparison. For those unable to indulge in such an extravagant test, take it from me that it does sound different, however carefully you set up all the levels, but it's hard to say exactly how or where.
This, however, is only a static comparison and all the nasties happen dynamically. When a computer system is inserted between our hypothetical pot and VCA, the control voltage is sampled and reconstructed both in level and time. Many people are now familiar with such problems when applied to both sampling synthesizers and MIDI delays. Well, it's the same old stuff but in a slightly different form. A pot is both infinitely variable and has to go between all intermediate values when changing (unless the wiper jumps off the track). A sampled pot has a number of steps determined by the ADC resolution and a sampling rate determined by system design or simply how much work the computer has to do. The effects of this manifest themselves as sluggish response if the pot is sampled too frequently (too much data overwhelms the system) and the dreaded zipper noise when not frequently enough or with insufficient resolution, or both. Imagine the sound of a zip fastener and you will imagine the unpleasant effect of audio level changing in discrete jumps. Having seen both (American) reviews and people moving faders and then claiming there to be no zipper noise, I must point out that zipper noise manifests itself as a modulation effect on the sound, rather than a hiss or buzz, and is more obvious on long, sustained pure tones, such as a piano. You will not notice zipper noise during silence or unpitched percussive sounds. One manufacturer recently demonstrated a programmable equaliser using only organ music so that the 'clicks' were less noticeable, so beware!
To prevent this side-effect the VCA control voltage has to be varied by either counting the DAC between all intermediate values, if there is enough resolution, or filtering the output with a simple low-pass filter. Both of these slewing mechanisms take time and so a compromise has to be reached between step resolution and the fastest fade time. If the VCA must perform audio mutes as well, the slewing must be defeated on 'Cut' and 'Return from Cut'.
In the type of 'MIDI automation system' under consideration here, there is often only one ADC (analogue-to-digital convertor) sampling all the inputs through a multiplexer and one DAC refreshing all the control voltages through multiple sample and holds. So we have an input scan rate and an output refresh rate, both of which have to be optimised to create an illusion of continuity. In an expandable system, we get an additional complication of reduced performance as the number of channels grows and the load on the computer increases. If the number of sample and holds is very large, say 64, the output may droop between refreshes. It may only be a small amount, but it constitutes a subtle form of modulation and it is present on all channels feeding the mix. Also, the processor may get so busy keeping the sample and holds alive that response to other parts of the system suffers.
It is not really necessary to have more than 7-bit resolution when reading a fader or knob (ie. more than 128 positions) unless the fader is the long-throw type (> 100mm) or non-linear, which it will be if it's the normal one already in your mixing desk. To all those critics who knock MIDI for being only a 7-bit system, if you actually bother to read the MIDI specification you'll find that it is perfectly capable of sending 14-bit data, as indeed it already does for pitch bend. Given that we have 7- or 8-bit linear resolution of control, the steps have to be distributed sensibly over an 'audio law'; don't forget that the attenuation calibration of a fader is non-linear anyway.
It is also necessary to maintain a good resolution around the normal operating level of the VCA. There is a marked difference to both fades and mix balance between having 0.3dB and 0.1 dB steps within the 0dB to -20dB range. With a linear dB/volt law (which VCAs tend to have), the first is almost achievable with an 8-bit DAC, where the latter requires more than 10-bits. If not slewed, steps of 0.5dB in gain correspond to zipper noise clicks with a maximum level of —24dB (dependent on where the step actually hits the audio waveform), whereas steps of 0.1 dB give a maximum click level of -40dB. The next time someone tells you that you can't hear half a dB, they either have never mixed, can't do the maths of decibels, or are trying to sell you an inadequate automation device - possibly all three! Just try metering a level with a real VU or PPM meter instead of a crude LED bargraph when attempting either a smooth fade, an exact balance, or (even worse) stereo positioning, and then tell me that 0.5dB steps are OK!
I hope by now that you're beginning to see that the application of VCAs is a veritable can of worms - and I haven't even got warmed up yet! Rather than fill a book with all the various details of exactly what can go wrong with controlled audio, I will simply refer the more inquisitive and dedicated masochists amongst you to a booklet actually written on the subject by Solid State Logic, entitled The Future Of Audio Console Design. And just when you thought it was safe... it also covers digital audio.
There are a great many myths about certain makes of VCA sounding different, and for the sake of completeness the VCA type is listed in the comparison chart for the products. However, I stress that it is far less relevant than the type of control system employed, for the reasons already mentioned. When people talk about the "so-and-so sound", they are often subconsciously thinking of the 'feel' of the control action rather than the actual tone quality of a sound. It is only because there have been a limited number of products around to demonstrate the various different combinations of VCA type and control system technique that such red herrings have appeared to confuse the real issues.
So now we're beginning to see analogue synthesizer-type technology and ergonomics applied to audio level processing tasks more than ten years after the Sequential Circuits Prophet V polysynth did it for musical instruments. Why has it taken so long? Anyone remember the Roland CPE-800 Compu-Editor that appeared in 1981? It died. So have all other retrofittable automation systems for mixers. Why?
The answers are complex but can be condensed by saying that, although superficially appearing to employ the same technology, musical instruments and mixing consoles are actually worlds apart, having fundamentally different operational requirements, and that nobody relishes buying a piece of equipment from one manufacturer and then tearing it apart to fit parts from another, because when you get the inevitable problems neither manufacturer wants to know. Yes, I know that sounds a bit obvious but it is true. Also, such products were ahead of their time. When MIDI was first introduced, common personal computers were the likes of the Acorn BBC B, which were clearly not up to the type of things that an Atari ST can do for the same money nowadays. I would regard an Atari ST, or anything similar, as the minimum requirement for a controlling computer for the types of automation devices under review. A threshold had to be passed and now has been. With such personal computers and so many MIDI synthesizers and effects units now available, there is a need for a component to compliment them and enough manufacturers appear to think that it's now time to try it again.
Consider some of the aspects of programmable and assignable control surfaces that have evolved on synthesizers. Some time ago, various manufacturers realised that all those knobs just had to go if the products were to stay (barely) affordable and so a Prophet V evolved into the D50, but you can still see the lineage. The only reason the ergonomic evolution could take place was because most of the synthesizer controls were redundant a lot of the time and, as the cynics tell us, 90% of the users didn't know what they were for anyway and never edited the preset sounds. Nobody has dared to reduce a pitch bend wheel to a pair of up/down buttons - they simply and obviously would not get away with it. The same is true of a mixing console and you only have to imagine (or remember) dealing with a feedback howl blasting your monitors or half your audience into oblivion to see why a DX7-style editing interface is simply not on.
Comparisons are also made between modern effects rack technology and mixers. Effects are by definition just that - effects - peripheral unity gain devices, whereas the mixing console is probably the single most important piece of equipment in any studio, large or small, and has to be evaluated to a different set of standards. It has gain and everything goes through it, maybe several times, so the compromises made on other equipment cannot apply. The audio path must be as transparent as possible and distortion, often a virtue on everything from a Marshall amp to a Fairlight II, cannot be tolerated.
So now we all know about VCAs, what happens when you take a handful, add a dash of microprocessor control, throw in some knobs, jacks and LEDs, stir, bake at gas mark 4, pour into a case and season with MIDI? Well, we might get an Akai MPX820 or a Simmons SPM8:2. Both devices perform an 8 into 2 submixer function, but with a completely different approach. Both devices have been reviewed in this magazine before, so regular readers will no doubt know what they look like and how they are operated.
Looking back at Paul Lehrman's review of the Akai MPX820 in the March 1987 issue, after having one to play with for a couple of weeks, I find it difficult to believe that he could possibly be so charitable. If a manual 8 into 2 mixer resembling this, with naff EQ, one effects send and no mutes cost £399, it would be regarded by the audio world to be a spud. The fact that it costs a thousand pounds more for MIDI control is astounding, especially since the one thing it is supposed to do doesn't work properly: if the MPX820 receives two different MIDI program changes in short succession, and by that I mean a quaver rather than 320 microseconds, it simply ignores the second one. Luckily, I received an English manual rather than the pidgin Swahili version Paul Lehrman had to put up with, and learned that the unit transmits and receives MIDI System Exclusive blocks, but for what, Akai aren't letting on. The only MIDI controller recognised is Volume, which controls the master fader level. However, as this is a memorised parameter once the MIDI code is received, it constitutes an edit and the exact volume level has to be hunted for with the front panel fader before manual control can be regained by the operator. Try that with sound going through it? No way!
The most immediately obvious aspects of the sound control are that the response is sluggish to the point where a fader can be manually whipped up and down quickly and then you hear the sound do the same thing. There is also a clearly audible and annoying zipper noise. Both of these undesirable features point to the front panel knobs and faders being scanned too slowly by the microprocessor. This was confirmed by the zipper noise being less noticeable with external MIDI Volume changes from a Yamaha MCS2.
On measuring the MPX820's control knob scan rate I found it to be 55 milliseconds, and I would regard the 60Hz SMPTE frame rate at 33ms to be the slowest acceptable. To put this in perspective, most modern synthesizers scan the pitch bend wheel every 5 milliseconds and the other controls at around 10ms. To give it its due, the Akai MPX820 is quiet, even with all faders set at 0dB, but you do have to try hard to design a mixer of this simplicity with a bad noise figure. Its main faults seem to be the editing system and the frightening way that such a simple, familiar-looking object gets completely out of control. After all the big fuss about it being the world's first MIDI-controlled mixing desk, I did expect something that wasn't quite so... well, boring. If a lesson is to be learned, it's that a device such as this shouldn't look exactly like the conventional mixer that people have learned operational skills with.
The Simmons SPM8:2 has most of the faults of the Akai model and worse, but somehow I can't dislike it so much and the fact that it has been squeezed into 1U and is below half the price for about the same audio quality means I can forgive it a lot. The one thing that I can't forgive it, especially as it calls itself a MIDI programmable mixer rather than just plain programmable, is that it only receives MIDI program changes and does not accept running status! This means that simply running your finger along the program/patch buttons of most synthesizers will confuse it - and you.
To be honest, I was actually looking forward to trying this unit out - with its brave, assignable control panel - and it was only after four weeks that I was much more conscious of its limitations. Most of the applications of this unit will find it tucked away in a rack, mixing a bunch of expanders with the memories probably never used, and since nobody else makes such a useful device it will sell, maybe.
As soon as you touch this mixer you realise that there is something seriously wrong with the knobs, apart from them not having a white 'centre' line and being too fiddly. Changing level feels like winding up six foot of elastic with a lead weight on the end - nothing happens for a while and then suddenly it hits you in the back of the neck. The cause of this effect is the knob scanning rate, which is a horrendous 140 milliseconds!
More difficulties come when you try using a channel as an effects return and regenerating the effect with its send after a little EQ. This is not a stupid, bloody-minded thing to attempt (everyone I know does it all the time) and yes, I did expect to get a feedback howl before I got the levels just right - the only problem was how. For a start, the SPM8:2's overload LEDs come on after clipping rather than just before, and because they double-up as the channel selector, if that channel overloads the LED goes off instead! It is very easy to lose track of what is happening and adjust the wrong channel. The first eight knobs double as channel level controls when none of the select/overload LEDs are lit, which is achieved by using the left/right arrow buttons to move off either end. If you forget which end you moved off, pushing the wrong button won't get you back, creating a further delay before you realise the mistake. Vital seconds while the voice coils of your speakers are melting.
What this really brings home is that in any programmable/multi-function control system certain important functions, in this case level and overload indication, must be dedicated.
The preset up/down button is one piece of plastic positioned over two switches and if pressed straight on, ie. both switches, it flattens the EQ. This is a lot more difficult than it sounds and is a total disaster area! To move from one preset to another it is necessary to select intermediate presets, which may have completely different balances. If the EQ is flattened by mistake, it can only be restored by moving to an adjacent preset and back again thus losing any other editing. In trying to measure the noise performance of this device I attempted to set all eight channels with flat EQ, I sometimes got as far as channel five without changing the patch (can anybody beat my score?) and after a solid hour gave up!
Both of these products - the Akai MPX820 and Simmons SPM8:2 - illustrate perfectly violations of certain ergonomic principles.
Both have preset memories, but both work differently from nearly every other piece of equipment with memory stores on the market, which is annoying when they would be used in the same rack and an operator would have to change editing styles constantly. What's wrong with the four button Inc, Dec, Store and Recall/Compare system?
There is no readout or indication of any of the stored parameters, which makes it difficult to form a mental image of the mixer, something done almost subconsciously when glancing at a conventional desk. As the knobs are nearly always in the wrong position, by definition, the fact that they look like normal non-programmable settings only presents more confusing information that is difficult to mentally screen. I've never had any trouble editing a DX7 or an Oberheim Xpander with all their complex parameters, but I found that I couldn't remember more than three or four set-ups of an 8 into 2 mixer! As they could not be given a title, by the next day I had forgotten what I was using them for as well. I attribute this to not having any patterns to use as mental images, even if that pattern is an LCD readout.
Both of these submixers operate on a 'snapshot' and 'fade time' principle (fading from one static mix to another) without access to any individual controls. I doubt the validity of storing a fade time fixed with a snapshot, as it tends to restrict the use of the snapshot or only works in one order. It would be far more useful to be able to instruct the mixer to crossfade to a certain snapshot at a certain (independent) rate. You also need a considerably larger number than 64 snapshots if you want to make a movie - try a few zeros on the end. As these devices do not transmit any Controller movements over MIDI - they simply can not be recorded - and as they do not internally record the information, they are simply not automation components.
I'm not sure what I did expect from the world's first MIDI-controlled mixers, but it was certainly a lot more than these offerings. I would expect it to be reasonably MIDI-controlled something (like the Yamaha DMP7) but with headroom and knobs on. Well, okay, half the MIDI implementation of the DMP7... oh, alright, a tenth then as long as the levels and mutes can be remote-controlled. Is that such a lot to ask?
Part of the trouble is that we have all been spoilt by cheap digital synthesizers and multi-effects units and when returning to analogue devices where all of the circuitry has to be there all of the time (instead of time multiplexed) something has got to give. On the other hand, we've never had it so good for analogue circuitry. Ten years ago I would have killed for the analogue components that are commonly available now. Luxury! Ten years ago the Prophet V came out and revolutionised analogue synthesizers within months of the Z80 microprocessor being commercially available, with 75 sample and holds to refresh the tuning accuracy and with only two 1K EPROMs and 1,152 bytes of RAM. Try telling that to young people today and they'll laugh at you!
Seriously, the reason that I keep harking back to the Prophet V synth is that it is essentially the same technology as that used in these submixers, but ergonomics and components have advanced considerably since then. To sum up, a sentence from the Future Of Audio book, written in 1985, seems very apt:
'The freedom provided by programmability creates a number of exciting opportunities to improve the man/machine interface - and almost unlimited possibilities for getting this wrong!'
Well, here are two of those possibilities. Less said, sooner forgotten.
As we are all spoilt by cheap digital synthesizers and effects and (groan) we've just had more for Christmas, and now they're giving digital reverbs away with petrol, how does a MIDI-controlled mixer really fit into the scheme of things?
Well, we keep getting more and more outputs on new gear and a lot of them are stereo, so a submixer with assignable stereo channels would have to be a good thing. The trend in synthesizers seems to be incorporating EQ and effects in the output stage, so really the crude analogue EQ that best belongs in a car radio should be left there and can we have more effects sends instead?
I can see two distinct scenarios: (1) using a MIDI submixer to tie up all the loose outputs in a largely synthesizer-based studio and sending a stereo feed to a conventional mixer used in a conventional multitrack way; (2) using a MIDI submixer to control the non-MIDI sound sources and a conventional mixer to tie up all the loose MIDI-controlled outputs. Both are possible ways of working and in both cases all the MIDI gear would be controlled by a multitrack sequencer - so what does the submixer do? Firstly, MIDI-controlled mutes so that recorded material can be integrated with MIDI-controlled rhythms and sequences, which requires response times accurate and repeatable to, say, 1/96 crotchet. You might want more, but you wouldn't be happy with less. Secondly, subgroups and master levels should be assignable to external standard MIDI controllers for recording manual edits and fades in with the sequence.
It's very early days for MIDI mixers, a bit reminiscent of both early MIDI and early MIDI effects. Anyone want a non-dynamic, uni-timbral, analogue MIDI synthesizer? Or do you know how it felt to buy a Yamaha D1500 a week before the SPX90 came out? It's that sort of time.
Next month, I shall be examining the three retrofittable VCA systems: the Twister PAC, IMD Megamix and J L Cooper MIDImation.
Part 1 | Part 2
Gear in this article:
Feature by Graham Hinton
Previous article in this issue:
Next article in this issue: