MIDI In Control
MIDI is regularly used to perform tasks as varied as switching notes on and off, storing patch libraries and controlling mixes, but can it control sound itself? Vic Lennard looks at the uses of MIDI noise gates.
From its humble origins as a means of allowing synthesisers and sequencers to "talk" to each other, MIDI has become a studio control system that will also control audio signals.
IMAGINE YOUR MULTITRACK recorder, be it a humble Portastudio or a Fostex E16, as an integral part of your MIDI system. Fade-ins, fade-outs and general audio levels are under the control of your MIDI sequencer. Only the vocals and acoustic instruments are actually recorded onto tape. An attractive scenario? Well, the technology is with us now as more and more software writers are taking advantage of the real-time control facilities the equipment manufacturers are offering them. We're not only talking synthesisers here, but MIDI signal processors too.
AT THE LOWEST level, a MIDI sequencer will record the timing, pitch and length of notes by breaking them down into Note On commands, which occur when a key is depressed, followed by a note off command, on its release.
Each of these constitute a MIDI event which takes less than I millisecond to send. In technical terms the information sent by the keyboard takes the form "9n NN VV" for a note on and "9n NN 00" for a note off, where "9" shows that the event is a Note On, "n" represents the MIDI channel, "NN" is the key number of the note (its pitch) and "VV" represents the velocity. These three bytes are sent in hexadecimal (base 16). Fortunately you rarely need to get involved in the mathematical side of MIDI (especially if your sequencer happens to be computerbased and has a visual editor).
Apart from the black and white bits, most MIDI keyboards have various performance enhancers which are actually numbered MIDI controllers - such as the modulation wheel (MIDI controller 1), volume slider (controller 7), sustain pedal (controller 64) and pitchbend wheel (which is in a category of its own). Other MIDI controllers include channel pressure (or aftertouch), which certain keyboards can send, and less common examples like balance (between two sounds) and pan (across a stereo spectrum). Depending on the sequencer, most of these can be recorded along with the note information or by themselves, which means that controller information can be generated after the notes have been recorded and set to the same MIDI channel as the notes - why worry about getting that awkward pitchbend right while playing when it can be dealt with separately later? Although it appears that this method of working will take up a lot of sequencer tracks, they can be mixed together once the recording has been satisfactorily concluded.
Many computer-based sequencers allow you to insert controller values in an event editor - a list comprising all of the data that has been received at the MIDI In port. This can be extremely tedious (especially in the case of pitchbend, whose values range from -8191 to +8191 for the full wheel movement - imagine inserting those one at a time) but if an approximate performance could be recorded and then visually edited to taste, the result would be both acceptable and quick. Hybrid Arts' MIDITrack ST series of sequencers allow you to draw controller curves and dynamic envelopes by using the mouse, for straight lines, or freehand - although this has to be carried out while the sequencer is static. C-Lab's Creator/Notator package for the Atari ST takes a totally different point of view by having a real-time MIDI Generator page with 16 sliders, one for each MIDI channel, which can actively alter any of the controllers as the sequencer is playing, and records these movements to boot, giving instant results which can then be repeated. Steinberg's up-and-coming Cubit program appears to have a similar approach to Hybrid Arts with graphic editing of controllers, and even Iconix, long since defunct, had a basic facility of this sort.
RELIABLE SYNCHRONISATION devices for locking a tape machine to a sequencer by using a timecode on one track are now quite commonplace. Two different varieties exist. The first is tape sync, which uses a start signal and clocks from that point to drive the sequencer in time with that tape, but has the disadvantage that the tape has to be rewound to the beginning of the song each time to pick up the start command. The second is SMPTE timecode, which incorporates timing accurate to either 1/25th or 1/30th (depending on the system) of a second within the code recorded onto tape. This is then either injected straight into the computer program via one of the ports (subject to the manufacturer's hardware), or is converted into MIDI Song Position Pointers and clocks to keep an account of how many 96th notes have passed since the beginning of the song - hence permitting synchronisation from wherever the tape is started.
In this way a sequencer can be run in time with a multitrack on mixdown as an alternative recording approach to laying everything down onto tape. The advantages are obvious - no need to commit yourself to finalised sounds and treatments on multitrack tape, or even to a song arrangement. But there are drawbacks too - you'll need a sizeable mixing desk to cater for a couple of multitimbral expanders, a drum machine or two with individual audio outs and a couple of signal processors - and each instrument can only be doing one job at any time in the mix; if you need two DX7 sounds, you'll need two DX7s.
"The advantages are obvious - no need to commit yourself to finalised sounds and treatments on multitrack tape, or even to a song arrangement."
But what of a signal recorded on tape?
APART FROM PROCESSING MIDI data by time-shifting (quantising) or pitch transposing, sequencers can control the level of output of a synth by altering the velocity or volume. The signal processing equivalent of this is a noise gate which works by shutting off any signal below a threshold level (derived from either the internal signal or an external trigger), but opens once this is exceeded, using a predetermined attack time to reach its fully open state. This is followed by "Hold" and then "Release" (sometimes called Decay) phases. And in keeping with old analogue terminology this profile is called the "envelope" of the sound. Alternatively a gate can work in reverse, open unless the threshold is exceeded. This process is termed "ducking" and is regularly used in radio broadcasting to allow a DJ to talk through a piece of music. But what if these two ideas could be incorporated into a single unit with the triggering provided by MIDI - a MIDI-controlled noise gate?
There are such units in existence, the most comprehensive of which is the Drawmer M500. This unit is actually far more than a noise gate, but let's concentrate on its functions in this context. A specific MIDI note, or range of notes, can be used to trigger each of the two channels, which will then follow the envelope you have set. Keeping the envelope in the Hold phase and using MIDI timing data, the gate can be opened from a start command and then triggered periodically after a certain number of MIDI clocks. Then life starts to get rather interesting...
The M500 can record a piece of music up to 20 seconds long and extract its dynamic envelope by using an algorithm for "reading" the input gain (scanned 256 times during the recording in 0.5 dB steps), and shaping an envelope from the resulting data and it can keep 16 of these in memory at any one time. This means that a brass stab, for instance, can be precisely reproduced from a block chord without having to worry about getting the feel right. And there's more. The output volume of the unit can be controlled using MIDI controller No. 7, allowing a sequencer to change the output level of the M500 at precisely the right moment; a preset fade of up to 99 minutes can be initiated by setting any controller between 65 and 95 to "on" - just enough time to hide the tape hiss and the vocalist saying "Is that it then?".
Two other aspects of the M500 are worthy of note: Firstly, a MIDI velocity value can be calculated from the input gain of a signal - a snare drum for example - and placed with a note value to create a MIDI Note On event which can trigger a MIDI sound source - in this case it would be either a drum machine or a sampler. Secondly, any buttons pressed on the front panel (yep, no knobs) can be recorded by the sequencer and reproduced at mixdown, and as the M500 is a multiple processor with gating, compression, panning and so on, effects can be gauged before mixdown without the pressure that the final mix invariably causes.
Another MIDI-controlled noise gate worth a mention is Studiomaster's IDP1. This lacks many of the more sophisticated routines of the Drawmer unit but does have one very interesting facility - given the correct MIDI channel, Note On or aftertouch data will initiate the attack phase while the corresponding note off will cause the hold/release cycle to come into operation. A similar operation can also be initiated by using the pitchbend wheel with the attack being imposed when the wheel is in any position other than the centre, and the hold/release phases taking effect when the wheel is recentred. By judicious setting of the envelope, fade-outs can be programmed. In this case, the release cycle may take a maximum time of 50 seconds. The IDP1 is also a multi-processing unit incorporating MIDI control in its compressor.
"How about recording harmony vocals over six tracks and then bouncing them onto a stereo pair while synchronising fade-ins and fade-outs, panning and so?"
For our final example of MIDI-controlled gates we'll take the Brooke Siren Systems (BSS) DPR 502. This shares the technique of note on/off for triggering the attack and hold/release phases with the IDP1, but has another MIDI function; a Note On event can be generated after the attack time with the MIDI channel and note numbers being adjustable, and the velocity set at either 64 or via a conversion from the input gain. The length of the note is dictated by the hold time and the Note Off is then generated.
So much for the theory.
THE DRAWMER GATE has the one function that the other examples lack namely that of controlling the gain of the unit by MIDI volume. Here's how that function could be used in a practical manner. Using the two audio channels separately and connected to two of the group outputs on the mixing desk, the sequencer can record controller No. 7 from a keyboard normally assigned to the volume slider - while a musician is playing. Take the example of a guitarist playing with a different feel from one section to another and whose general level is quite variable. The usual method of dealing with this situation is to patch in a compressor, which won't just change the level, but also alter the sound by reducing the dynamics. Changing the gain of the noise gate, however, will maintain the relationship between loud and soft notes while ensuring that any extraneous noise (which guitars specialise in creating) is kept to an absolute minimum. The guitarist can then go for The Big Take confident that all he has to worry about is playing the correct notes.
This technique would not be suitable for a vocalist because the dynamic range of the voice is so wide, and each performance will be different according to the feel imparted. Instead, try recording to tape as usual, but when the track is complete, run the vocals into the noise gate (using group outputs if necessary), and again record the changes to any of the necessary settings on the sequencer. The advantage of this approach is that the procedure can be repeated over and over until a satisfactory result is obtained.
The ability to create a Note On with velocity from the level of an incoming audio signal also opens up interesting possibilities. This would have the effect of retaining the original velocity sensitivity which would be very effective if you have a sampler with crossfade velocity (like the Akai S900) which controls the mix between loud and soft sample, subject to the velocity, so changing the texture of the instrument from beat to beat in as subtle a manner as the sound source will allow. The recently released Roland R8 drum machine works on a similar principle.
Creating rich backing vocals could also take a rather neat turn. How about recording lush harmonies over, perhaps, half a dozen tracks and then bouncing them down onto a stereo pair while running them through a MIDI gate? Synchronised fade-ins and fade-outs, panning for vocal effects and so on are all possible with a little imagination. Testing the result before committing it to tape will give better results, especially as the threshold setting is bypassed when in MIDI mode, so preventing any accidental triggering.
THERE ARE ALREADY a variety of MIDI-controlled mixers on the market - the Studiomaster Series II has MIDI muting of input channels and auxiliary returns; the Simmons SPM 8:2 can save 64 settings, each containing individual three-band EQ, level, pan and auxiliary send levels. However, neither of these desks approach the method of using a MIDI noise gate as detailed above.
The requirement is for a multi-channel unit dependent on the desk size, which can be programmed using specific MIDI controllers with a two-way interface (MIDI In and Out) for real-time applications through a sequencer. This would keep costs down by dispensing with a hardware remote control unit and special software for the computer. Using such a device, thought could be put into the manner of mixdown from the very start of a song with progressive ideas being stored in the sequencer. The most expensive parts of such a system are likely to be the VCAs; here there are two possible options: accept that the unit is going to be built specifically for semi-pro use and keep the cost down by conceding that noise will be audible but barely so, or wait for a manufacturer to design cheaper, high-quality VCAs. Personally I'd like to see a serious semi-pro unit that makes sensible compromises on performance and noise but opens the doors on one of the most exciting, as yet untapped applications of MIDI.