Getting into Video (Part 2)
PART 2: Timecode - much talked about but little understood - is of increasing importance to musicians and recording engineers. Whether you are synchronising audio to video, or MIDI to multitrack, at least a working knowledge of SMPTE/EBU is required. In the second part of this series, David Mellor explains what timecode is and how it is used.
Timecode - much talked about but little understood - is of increasing importance to musicians and recording engineers. In the second part of this series, David Mellor explains what timecode is and how it is used.
Few audio recordings these days are made without the use of timecode. Even if the instruments are played entirely by hand without the assistance of a sequencer, the automation on the studio mixing console may need to be synchronised to the music. If video is involved, then the audio tape needs to be synchronised to a video machine. Moving up a further level of complexity, in an installation with multiple video machines, cameras and video effects units, all the machines must be synchronised to a house reference sync generator, but fortunately that is an aspect of the story that the sound engineer can ignore with a reasonable degree of comfort - so far.
Machine synchronisation is, in the modern technological world, a big deal. We just cannot live without it. Fortunately, using synchronisers is fairly straightforward. Designers have incorporated clever software in their machines to handle the complex technical side of sync, leaving the user free to concentrate on the artistic and creative aspects of production. But an understanding of the underlying principles is, as always, a great help in getting the best out of any system.
At the heart of synchronisation is timecode. This is the stuff that resides on each tape and each video cassette - and provides a reference for sequencers and other computer-based equipment - that allows a synchroniser to locate and run each machine in a precise lock. So what is timecode, why does it exist and where does it come from? Read on...
In the beginning of timecode history there was video tape, and by 'tape' I mean two-inch wide tape on reels - video cassettes are a comparatively recent development. As I briefly outlined in the first article of this series, video tape was originally seen not as a means of editing, but as a time shifting device which enabled a TV programme to be recorded and transmitted at a later date, or possibly even more than once.
But sound editing was already a common procedure, and had been since the days when radio programmes were recorded on 78rpm discs. Dub edits, from disc to disc, were made to assemble a programme. It soon became obvious that if it were possible to edit video tape, TV programmes could be assembled in the same way, with the advantages that would bring.
Unfortunately, video tape is not as easy to edit as its audio equivalent: a video recording is discontinuous, that is to say it consists of a series of individual frames, so it is not possible to cut and splice the tape in the same way as you can with audio tape. Unless the join is made at exactly the right point in the frame, the picture will break up.
One method that was employed to find suitable splice points was to make the recorded video waveform - though not the picture itself - visible by painting a suspension of fine magnetic particles onto the tape. Subsequently, electronically controlled dub editing, from machine to machine, was made possible by recording a series of control pulses onto a separate track of the video tape. However, this technique was still prone to problems due to tape dropout, and the fact that the pulses did not identify exact frame numbers. Edits were therefore approximate at best.
The basis of a more satisfactory means of video editing, in fact that which is still standard today, was provided in 1967 when the American Society of Motion Picture and Television Engineers (SMPTE) devised a system which would label each frame on a video tape with a unique code-timecode. Timecode is a sequence of pulses, digitally encoded with time-of-day information, which can be recorded onto an audio track of the videotape. By this means an electronic video editor can determine the exact position on a piece of video tape, and make frame-accurate joins.
Timecode starts life as electrical pulses created by a timecode generator. The generator converts these pulses into a waveform which looks, to an audio tape recorder (or the audio track of a video recorder), just like a typical audio signal, and can therefore be recorded in the same manner. There are probably an infinite number of ways in which this audio signal could be created, but in fact the method employed is called 'bi-phase modulation', and it has a number of important advantages over other possible techniques.
Each timecode frame (which on a video recording corresponds to a single video frame) contains a binary 'word' consisting of 80 bits. At the start of each of these bits, the voltage of the signal changes from high to low, or from low to high. If a bit is to represent a binary '1', then there is a second transition halfway through the bit period. If the bit is to be a binary '0', then the voltage remains constant until the start of the next bit period. Figure 1 shows a sequence of timecode bits, and their binary values.
Note that it is not the absolute voltage levels that determine the value of each bit; it is how often the level changes. This means that if the timecode signal is accidentally phase reversed, probably through the incorrect wiring of a connector, it still has the same meaning. It also means that the information can be read both forwards and backwards. I shall explain when and why you might want to read timecode backwards in a later article in this series. Table 1 shows the meaning of each of the 80 bits in the SMPTE (as opposed to EBU) timecode word. Note that although 'SMPTE' is often used as a generic term in relation to timecode, there are in fact several types of timecode which I am about to explain, and that the different types will use these 80 bits differently.
As you can see, the content is simply time information, plus user bits (which may be used for station or reel identification etc), plus a couple of flag bits which I shall explain shortly. The sync word is simply a sequence of bits which is used to maintain synchronisation and to determine whether the tape is travelling backwards or forwards.
Although we Brits were fortunate that the founding fathers of the US of A decided to choose our language - it was actually a close run contest between English and German - we were not lucky enough to have our 50Hz mains frequency adopted as well. For kettles and hair dryers this doesn't make much difference (although the voltage difference does!), but for early TV systems the frequency at which the mains voltage alternates provided an all-important frequency reference.
Thus in Britain our TV system runs at a frame rate of 25Hz (half the mains frequency), whereas in America and a few other countries, the frame rate is 30Hz - half the 60Hz mains frequency.
Since timecode provides an identification for each frame of the picture, as well as a sync reference, it follows that American code must count up to 30 frames, but ours in the UK must count up to 25. Therefore, there have to be at least two incompatible types of timecode. In practice, there are four!
Whilst the original US standard of 30 frames per second (fps) was adequate for black and white broadcasting, when the NTSC colour system was developed in the States it was found necessary to reduce the frame rate slightly to 29.97 frames per second to maintain good picture quality. To provide each frame with an accurate time-of-day reference would therefore require messy decimal places, and extra bits in the timecode word to encode the information. To avoid this, it was decided to count up to 30 frames before incrementing the 'seconds' count, as with regular SMPTE code - but since this would mean that the timecode clock would now run slow, certain frame numbers were omitted so that the clock would catch up from time to time.
What happens is that at the beginning of every minute, two frame numbers are dropped - hence the description 'drop frame' timecode. This doesn't provide a total fix, so every tenth minute, the frames are not dropped. There is still a slight error over the course of a long time period, but not a significant one as far as an individual programme is concerned.
The result is that American timecode - SMPTE code - has two varieties: drop frame and non-drop frame, running at 29.97fps and 30fps respectively.
The two European colour television systems (PAL and SECAM) were both developed several years after NTSC, and managed to avoid the drop frame problem. Timecode for these two systems runs at a straightforward 25fps, and is known as EBU (European Broadcasting Union) timecode.
The fourth type of timecode is geared towards the requirements of the film industry, rather than TV. Film runs at 24 frames per second and therefore needs 24fps code. All of these four types of code are, unfortunately, incompatible, and if they are to be operated together, a special convertor must be used. Audio Kinetics make such a unit, known as the 'Gearbox', for fairly obvious reasons.
Going back to Table 1 for a moment, you will see that Bit 10 describes whether SMPTE code is drop frame or non-drop. Bit 11, labelled 'colour frame', is used purely in video editing, and is irrelevant for audio purposes. For the sake of curiosity, I can say that in the NTSC and SECAM systems the colour information is encoded in a sequence which covers two frames. The correct relationship between the frames must be maintained in video editing. The colour frame bit places a marker at an appropriate position in the sequence. In PAL, the sequence covers four frames.
The usual source of timecode is a timecode generator. I say usual because a company called Prosonus have recently released a compact disc containing an hour's worth of code, but this is useful in only the simplest of applications.
In television production, the timecode generator will probably be set to the actual time of day. It must also be synchronised to the video sync pulses produced by a reference sync generator which will supply sync to all cameras, recorders and effects units in an installation. Precise synchronisation of the timecode to the video frames recorded on tape (plus or minus one line of the picture) is essential or there could be a situation where a single timecode word is evenly split over parts of two frames, creating an obvious source of confusion.
For purely audio purposes, the timecode can be free-running (ie. does not have to be slaved to any more fundamental clock source), and is of course used as the basic timing reference in a studio. Normal practice is to 'stripe' (record) one track of the tape from beginning to end with code before any programme recording takes place. In audio, the real time of day is likely to be irrelevant and generation usually takes place from 00:00:00:00.00 or 01:00:00:00.00 (hours:minutes:seconds:frames.subframes - subframes being 1/100ths of a frame). Note that 00:00:00:00.00 is an important time - midnight. If possible, timecode should not cross midnight on a reel of tape (ie. start before and end after 00:00:00:00.00) as it might confuse the synchroniser reading the code and send the machines searching in the wrong direction.
Since timecode is recorded on tape in the same way as any other audio signal, it follows that setting the correct levels is important. Achieving a good signal-to-noise ratio, in this case, is not usually as important as maintaining a good quality of code. Timecode is a pulse waveform, and is therefore full of sharp corners which need to be accurately recorded to ensure correct decoding. Sharp corners on a waveform always indicate strong high frequency content, and although tape recorders may have a frequency response extending up to 20kHz and above, they do not maintain their performance at all recording levels. The higher the level (and the slower the tape speed), the harder it is to record high frequencies. Therefore, timecode should be recorded at a few decibels below 0VU. In practice, the only way to find the correct level is to experiment.
Recording timecode onto multitrack tape throws up another problem - crosstalk. The nasty high-pitched shrieking sound that timecode makes could not be less like music, nor more objectionable to the ear. Even a small amount of leakage onto the adjacent track or the track beyond can make a recording unusable. Therefore, a recording level must be found which enables correct decoding, yet keeps crosstalk to a minimum. A further means of reducing crosstalk problems is the convention of always recording timecode on the highest numbered track of a multitrack recorder - eg. track 24 on a 24-track machine. This ensures that there is only one adjacent track onto which crosstalk can leak.
Still on the subject of crosstalk, it is also important that timecode never gets anywhere near the mixing desk - or at least not while you are recording or replaying music. Many mixing desks, in the interests of low cost, do not have particularly good crosstalk performance. Music crosstalk is not such an inconvenience, since the leaking signal is generally related to the music being routed through the desk in the proper way. Crosstalk is of more concern to broadcasters as they may be processing several totally different signals at the same time - music, commentary, communications etc. Timecode crosstalk, in this situation, can be disastrous. Wiring for timecode should run direct from generator to tape recorder, and direct from the tape recorder to the timecode reader.
Reading timecode should be a straightforward matter, provided it has been recorded correctly. Unfortunately, even with modern tape formulations, dropouts still occur, and a dropout on the timecode track means a momentary loss of synchronisation.
Timecode readers - which may be separate units or incorporated into the synchroniser - come in all quality levels, from 'give me good code or I won't play with you any more' types up to sophisticated units which can cope with badly mangled code. Basic timecode readers need to receive code with good sharp edges, free from dropouts. If the code disappears, even momentarily, sync will be lost. Better quality readers will cope with code with rounded edges, such as you are bound to get from the audio tracks of VHS recorders. Also, there may be a 'flywheel' feature where the reader will fill in for any missing or damaged code until good code is found again.
However, there is no substitute for having good code in the first place, and following two simple rules will avoid most problems. Rule 1 is to record at the correct level, as found by experiment. Rule 2 is never to copy code without regenerating it. As we shall see in later installments, it is often necessary to copy timecode, but if this is done by hooking a cable between audio out and audio in from one machine to another, the nice sharp edges of the code will be lost.
Regenerating timecode involves passing code from the playback machine through a device (probably incorporated in the timecode reader or generator) which will decode the timecode numbers, make sense of them and compensate for any discrepancies, and then re-issue perfect code to the recorder. A moderately satisfactory alternative is to 'reshape' the timecode. This means that the sharp edges are restored, but dropouts and errors are not compensated for.
So far, I have outlined what timecode is, how it is recorded and how it is read. What you can do with it must wait until the next installment of 'Getting Into Video', along with a description of some of the more sophisticated problems that may be encountered. But basically, successful timecode operation depends on good equipment and proper procedures. As we shall see, the synchronisation of machinery such as video and audio recorders is considerably more complex than syncing a MIDI sequencer to tape.
Highly recommended further reading is the 'Timecode Handbook' published by Cipher Digital - a manufacturer of timecode-based equipment.
Future Film Developments, (Contact Details).
Feature by David Mellor
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!