MIDI recording, or 'sequencing', offers many operational advantages and convenience features when compared to conventional audio recording. After working with sequencers for a while, you start to take for granted the ability to jump instantly to a new location, or seamlessly cycle round the same few bars while you work on a part. Sophisticated edit facilities allow so much freedom in recording — you can go for 'the ultimate take' knowing that any little errors can easily be corrected. Going back to working with tape again can seem like a backwards step, especially when you're sitting there with an idea in your head, frustratedly waiting for a rewind to finish so you can get it recorded. With a MIDI recording, you can easily edit an instrumental part, or change its sound, right up to the final mix, but once something is on tape, that's what you've got, unless you're prepared to re-record it.

ESSENTIAL

Whatever its limitations, tape remains an essential part of most recording setups, for there are still many instruments that cannot be simulated at all successfully via MIDI. Guitars are an obvious example; brilliant though Jan Hammer's impersonations are, they too are instantly recognisable as a simulation by even the untrained listener. If you are using vocals, of course, then tape is obligatory; even if you intend to eventually use them as samples, you will almost certainly find it necessary to record them on tape in the first place, simply for convenience — recording vocals directly into a sampler is extremely limiting. The obvious step is to tie the two systems together, using MIDI for what MIDI does well, and an ATR (audio tape recorder) for normal audio recording.

With the sequencer 'locked' to tape via a form of code recorded on one of the audio tracks, its MIDI recording tracks can be treated almost as if they were an extension of the multitrack. The output from the MIDI voice modules can either be recorded on tape, if you have enough tracks, or remain as separate sources right through to mixdown, thereby retaining all the flexibility that MIDI offers. This method of operation also allows a much smaller multitrack format to be used, as there is generally less need for many conventional audio signals.

SYNCHRONISE

MIDI-to-tape synchronisers have been around almost as long as MIDI. In a combined MIDI/audio system the tape is always used as the master transport, as it is considerably easier to externally control the speed of a clock-based system like a sequencer than to accurately govern a mechanical system such as a tape machine's capstan. With the ATR as the master, the obvious place to put synchronising data is on one track of the tape itself, for it is the only thing in the system with an absolutely fixed relationship to the other tracks.

All MIDI recording devices (drum machines, hardware and software sequencers) send out and receive timing clock information in order to be able to run in sync with one another — the 'MIDI clock' marks out time with a regular series of pulses related to tempo, running at 24 clocks per quarter note (96 clocks per 4/4 bar).

The frequency of the MIDI data stream, however, far exceeds the bandwidth of a tape recorder, and cannot therefore be recorded directly. The earliest MIDI-to-tape sync devices simply read the MIDI clock pulses and converted them into a signal at audio frequencies. A clock pulse is simply a regular transition between two levels, but as the absolute level at which the code will be recorded is not predictable, it is necessary to find another means of representing this waveform. Frequency Shift Keying (FSK) is used; this is a robust signal, modulating between two different frequencies to represent the pulse edges, and thus independent of level.

BASIC FSK SYSTEM TECHNIQUE

The method of MIDI-to-tape sync with a simple FSK system is to program your sequencer or drum machine with a guide rhythm or backing part which incorporates any tempo changes that might be needed. The entire song is then run in real time on the sequencer, with its MIDI output feeding the MIDI-to-tape sync FSK unit. The MIDI clocks are converted to FSK, which is recorded on a tape track; conventionally, the highest-numbered track is used. On replay, the sequencer (set to External Clock) must first be 'primed' with a Start command; it will then sit there in a 'waiting for clocks' mode. When the tape is run, the FSK is re-converted to MIDI clocks, which can then drive the MIDI sequencer in sync. Any tempo variations in the programming will be as faithfully reproduced as the MIDI clock resolution allows, as the FSK pulse rate has been generated from the original clock data.

Yamaha's simple (but effective) YMC10 is one of the most commonly encountered examples of this generation of sync unit, with very cheap second-hand units often available from users who have upgraded to a timecode-based system.

You now have a sequencer or drum machine that will start in time with the tape, and remain in sync with it for as long as there is a sync signal on tape for it to follow. The limitation is that the sequencer has no idea where it is — it is merely advancing one time division for every clock it receives. Operationally, this means that you must always run the tape from the beginning, which, as any multitrack user knows, is not the method by which multi-overdub recordings are best made. The best way to work around this problem is to record a 'guide track' sub-mix of your MIDI parts onto another track of the recorder. The MIDI sync hook-up can then be ditched while you work on the tape parts, giving you the freedom to start wherever you like, and drop in and out at will.

PLAYING BY THE RULES

It's hard to believe now, but there was a time when everybody working with both MIDI and tape had to work in the way I've just described, because that's all there was. There are a couple of 'rules' worth observing; one is to be sure to record the guide track with the sequencer driven from the code on tape, and not to record it at the same time as the code. With all basic FSK sync units, there is a slight delay in the conversion process, and a guide track laid 'live' will cause all the material you subsequently record on tape to be ahead of the synced parts when the two are eventually re-united.

Even after you have recorded the guide track from the code, check it again, with the MIDI gear running in sync, monitoring the guide track at equal level, at the same time. You should hear a 'flanging' effect, with a random cycle, as the two identical signals wander in time slightly on either side of each other — anything else, and you've got unreliable code. This is worth spending time on; if you lose sync on a simple FSK system, there is practically nothing you can do without more sophisticated equipment — all the effort that you put into the the tracks on tape will be wasted.

The second rule is not to record the guide track next to the code. Ideally, you would leave an entirely blank 'guard track' next to it, but in systems with a small number of tracks, that is obviously impractical. What you are actually trying to avoid is the possibility of a corrupting pulse from 'spiky' material with a high transient content being read as crosstalk by the code-track's head-gap; as guide tracks generally include fairly prominent drums, sticking this next to the code is asking for trouble. Although one of the principles of FSK coding is to be independent of absolute level, real systems do tend to have fairly precise requirements; some experimentation with recording and replay levels is usually necessary to find the reliable operating 'window'.

THE SONG POSITION POINTER

The easiest way to improve on a basic FSK sync system is to incorporate, within the data stream, a means of telling connected devices not just when to start and how fast to run (which is all that the MIDI clock does), but precisely how far into the piece of music they are at any time. The MIDI specification incorporates this, in the form of Song Position Pointers. SPPs are able to uniquely identify any point in a song by incrementally registering groups of six clocks — effectively, it is counting in 16th notes (24 clocks per quarter-note divided by six gives four Song Position Pointer divisions per beat). If the positional information is encoded with the clock data in the FSK signal on tape, the combined system can be started from anywhere and still run in sync

The PPS1, and subsequently the PPS100, from JL Cooper are perhaps the best-known units to operate in this way, and they offer a major operational advantage over basic FSK by allowing the sequencer to become an extension of the recorder, and avoiding the need for a guide track. The procedure for recording the sync track remains the same as for basic FSK, with the sequencer sending out its MIDI clock stream, conveying all tempo information in real-time. Unfortunately, some early MIDI devices couldn't actually read Song Position Pointers, which made them distinctly second-class items as soon as 'intelligent' SPP synchronisers became available. Yamaha's onetime professional flagship sequencer, the very expensive QX1, was actually among these — I know, I had one. No drum machine or sequencer could be marketed without this facility today, but if you are offered a second-hand bargain, be sure to check that the model is not one of the small number with no SPP capability.

MIDI MERGE

One potential difficulty that arises with this setup is that the MIDI In socket of the sequencer must now always be occupied by the output of the synchroniser. This leaves you with nowhere to connect your keyboard when you want to record some more MIDI parts running in sync with the tape, which was, after all, why you were interested in the system in the first place. You have two choices; either buy a stand-alone merger (available fairly economically these days, especially the Phillip Rees, and Anatek Pocket series units), or make sure that you get a sync unit which incorporates a MIDI merge facility. It doesn't even have to be a full-spec merger; it is only required to merge one data input with timing messages.

"Because timecode merely 'divides up' the tape into uniquely identified frames, the SMPTE-to-MIDI convertor must do all the work of calculating the tempo information for the song, before sending a Start command, followed by MIDI clocks at the required rate."

SMPTE/EBU TIMECODE

SMPTE timecode's origin in the film and television industries has been well covered in these pages many times; fortunately, the timecode-to-MIDI sync user needs to know little of the data structure, you only need a grip on the principles to use timecode. SMPTE is often referred to as a 'time-of-day' code, because it is expressed in hours, minutes and seconds, sub-dividing further into frames and bits. For the casual user, the situation is slightly complicated by the four different frame rates available. Most SMPTE-to-MIDI units support all of them, and which one you use for SMPTE-to-MIDI work doesn't actually matter, so long as your synchroniser is set up to read the same type as you have put on tape. The reason for the four different types is that in film or video applications, the timecode format must match the picture frame rate, to provide a unique identification for every picture frame.

Film uses 24fps (frames per second), whilst UK TV uses 25fps (half the 50Hz mains frequency). To be strictly correct, 25fps timecode really should be referred to as EBU, rather than SMPTE, but the more general term is often used for convenience. American TV uses 30fps (half their 60Hz mains), or 29.97fps for their NTSC colour system. We needn't go into the reason for this horrible number except to say that it is a bit of-a lash-up to cope with an unforeseen problem. It is usually synthesised by using the fourth type, 30fps 'drop-frame'. This accepts the short-term error of using the basic 30fps rate, but 'drops' (discards) two frames at the beginning of every minute, except the tenth minute, to prevent the error becoming cumulative. Believe it or not, this maintains longterm sync between 29.97, and 30fps frame-rates.

Unlike FSK systems, SMPTE/EBU timecode contains no tempo information; indeed it is normally 'striped' onto the tape before anything else, perhaps even before the tempo of the song is decided. An FSK multitrack will be striped with sections of code corresponding to (and unique to) each song; a SMPTE tape will normally be continuously striped for its entire length. The code can be generated starting from any value, but it is fairly conventional to start from just before 01:00:00.00, or 1 hour — the colon dividers are often used to help differentiate the hours and minutes columns, and there is sometimes a two-digit fifth column to the right, for 'bits'. Once again, you don't need to know the precise method of encoding, except that it is independent of absolute level (in fact, it registers data by the presence or absence merely of a frequency transition, in either direction, within a given period, and can thus survive phase reversal; it can also be read, or generated, backwards).

SMPTE-TO-MIDI CONVERTORS

Because timecode merely 'divides up' the tape into uniquely identified frames, the SMPTE-to-MIDI convertor must do all the work of calculating the tempo information for the song, before sending a Start command, followed by MIDI clocks at the required rate. If you start the tape after the beginning of the song, the convertor must calculate the distance into the piece, and then send a Continue message, followed by MIDI clocks. To do this, it needs to be given a certain amount of information about the song. It needs to know the 'Start Time'; comparing this time in memory with incoming code will tell it when to send a MIDI Song Start command, and output the clock stream. It also needs to know the tempo; not just the starting tempo, but any changes in tempo, and their precise locations. These may be already programmed into the sequencer, but unlike the FSK-based system, they will not have been used to generate the code, and thus will have no influence on proceedings, for in a SMPTE-to-MIDI system, it is the synchroniser which is in charge.

Most basic SMPTE sync units, of the type I am describing, will allow you to enter tempo change information to a resolution of a bar — which means that they also need to know the time signature of a piece of music, in order to be able to calculate bars from beats. Setting up a piece of music with frequent tempo changes can be a lengthy and tedious process, and if you enter one of the changes in the wrong place, the result will be loss of sync, for the SMPTE-to-MIDI unit only calculates on the basis of the information you give it. Even the most basic synchroniser must therefore have a display, not just to show incoming code, but for entering the setup data, and also data entry keys.

An obvious advance, rather than having to enter this information manually every time you use the synchroniser, is to devise a means of storing it. Two of the first devices to make SMPTE-to-MIDI sync really affordable were the XRI Systems XRI3, and the Nomad SMC1, which, respectively, offered saving and loading of the setup info as MIDI Sys Ex and an FSK data dump (to tape). The Nomad, however, had no means of manual entry of the information as a back-up, so if your data failed to reload at any time, you were sunk. For a time, the Roland SBX80 was practically ubiquitous in studios, as one of the most reliable and easy to program, SMPTE-to-MIDI convertors of its day.

SMPTE/EBU can thus be seen to have its advantages, not least that it is a 'standard', and its disadvantages — there can be no doubt that, for simply locking a sequencer to tape for a song preprogrammed with numerous tempo changes, smart FSK is a great deal easier to use.

The next generation of SMPTE-to-MIDI synchronisers, however, adopted the principle of 'learning' the tempo information from the sequencer, and applied it to timecode operation — sophisticated professional units like the Friend Chip SRC/AT were 'universal', whilst others, like the C-Lab and Steinberg systems, were dedicated to their respective MIDI recording programs. Dedicated software-controlled systems have a number of advantages; all the information that the synchroniser requires already resides within the host computer. The only thing you need to tell it is the start time. Secondly, all the sync information is saved as part of the song file; the two can not be accidentally separated.

More significant, however, is the possibility of bypassing the process of converting to MIDI clocks and SPPs altogether. Although the timecode can be read as soon as the tape is moving past the head, a basic SMPTE-to-MIDI sync unit will often take a whole bar to get the sequencer going. A major improvement in pick-up time can be achieved by interfacing directly with the computer's processor; the MIDI sources are then started practically instantaneously. This level of performance really does encourage the use of the sequencer as 'virtual tracks', driving MIDI sources that are not printed to tape during the mix (although in professional situations they are often printed subsequently, to a slave reel if necessary, in order to make the recording truly 'portable', so that you don't have to assemble exactly the same pile of MIDI gear to mix it at another location.

HI-RES SYNC

The current generation of timecode-to-MIDI synchronisers now tend to offer much higher tempo resolution. Previously, transferring from one type of synchroniser to another, mid-project, was asking for trouble. With synchronisers allowing tempo to be set only in whole-number bpm values, some quite wide variations could occur, with no means of resolving them. A unit such as the SRC/AT or C-Lab's Notator/Creator-integrated Unitor synchroniser, however, resolves tempo to four decimal places of a bpm. At this level of resolution, and with timecode values accessible down to individual bit level (1/80th of a frame), it is possible to accurately re-sync to any constant tempo source, simply by ear, if necessary. A timecode-to-MIDI sync unit of this sophistication makes the MIDI synchronisation process virtually transparent in operation, offering all the advantages with none of the disadvantages. As it is a standard format, you can also use the same timecode track for a mixing automation system, and for a transport synchroniser.

MIDI TIMECODE

MIDI Timecode (MTC) is an extension of the MIDI protocol, encoding SMPTE time-of-day messages into the MIDI data stream itself. The SMPTE-MTC unit dispenses with MIDI clocks and SPPs altogether, converting incoming timecode directly into MTC, allowing higher resolution, much as do dedicated software systems. MTC offers much simpler operation too, with all tempo derivation being handled by the sequencer itself, and it gives near-instantaneous pick-up — effectively all the advantages of the integrated systems (at present, all dedicated to particular pieces of software), but in a potentially 'universal' format.

It has a number of other highly advantageous features, such as the ability to encode cue points and setup data, but above all, it should eventually be able to offer real compatibility in the MIDI synchronisation area.

Reservations about the possible effect of MTC on the already-taxed MIDI bandwidth are somewhat defused by the proliferation of parallel MIDI distribution add-ons to many systems. Despite the many advantages, however, MTC support is picking up relatively slowly after three years or so, and not enough MIDI gear presently supports MTC to make people sufficiently aware of the possibilities, particularly in the UK (it is growing much faster in the USA).

"To the novice, MIDI-to-tape sync may seem fraught with difficulty and potential pitfalls. It is not; it merely requires some simple choices to be made."

INTEGRATION

The future of MIDI-to-tape sync undoubtedly lies in further integration of the two systems. Tascam's MIDIizer offered MIDI-to-tape sync, transport remote control, and tape transport synchronisation, all in one unit, but the recent agreement on the MIDI Machine Control protocol undoubtedly points the way to the form which future integration is likely to take. Although the first steps were somewhat fragmented, with Fostex incorporating SysEx control of their R8 recorder from within Cubase, followed by a third-party interface for Tascam recorders, for both Cubase and Notator/Creator, the MMC agreement should ensure compatibility in future developments. The basic concept allows a tape transport in a MIDI sync setup to receive its transport instructions via standardised MIDI data, generated from the controls of a sequencer. By incorporating a 'local-off' mode for these controls, the transport commands can be issued over MIDI, without directly instructing the sequencer. When the tape transport responds to a Play command, it starts outputting timecode, which will then drive the sequencer as normal.

So what have you gained from this convoluted instruction route? You have gained the use of all the sequencer's sophisticated transport control features: cycle-play, location to specific bar-beat references, programmable drop-in... these can all be imposed on the tape-machine. Yes, even automatic drop-in, because full integration should allow control within the software of anything you can switch from the front panel of the recorder.

There are further possibilities for MIDI Machine Control. It doesn't actually have to be a sequencer issuing the commands — Allen and Heath's new GS3 desk offers function keys which can be programmed to issue any MIDI message. Effectively this section of the desk can function as a 'universal remote controller', which can be used with any MIDI-compatible tape recorder. We are sure to see more of this sort of implementation.

CHOICES

To the novice, MIDI-to-tape sync may seem fraught with difficulty and potential pitfalls. It is not; it merely requires some simple choices to be made. A basic, non-SPP, FSK setup is still very usable, provided you are prepared to accept the operational constraints detailed above, and if you're on a tight budget it is still a lot more fun than no MIDI-to-tape sync at all. If you can make the step up to smart FSK, try to make sure you get a merger in the system, or much of the advantage will be lost—with SPPs and a merger, you have all you need for full 'virtual track' operation.

Real SMPTE/EBU timecode is essential if you aspire to any form of professional operation, or even just want to keep abreast of the latest developments. The latest generation of timecode-MIDI synchronisers are a pleasure to use, unlike their early counterparts, providing previously undreamed of flexibility and sophistication.

In an integrated system, it is the addition of some real instruments that always seems to bring the MIDI sources to life. With good programming, and a proper context, there is no reason for MIDI parts to sound 'sequenced' at all. With the present artistic climate putting the emphasis firmly on 'real' instruments again, efficient MIDI-tape sync is an essential ingredient in the musically tasteful use of MIDI.