Chapter 7
Working with timecode

As this edition went to press, the EBU/SMPTE Task Force for Harmonized Standards for the Exchange of Programme Material as Bitstreams issued its final report. It identified the following requirements regarding timing and synchronization:

1. The reference signal for both digital and analogue signals will be analogue colour black. SMPTE RP154 has been enhanced and is being re-issued as a standard. It will include the option of including VITC, and (for 59.94 and 60 Hz related systems) a sequence of ten field identification pulses to assist with the locking of 24 Hz related signals and the five-frame sequence implicit in the relationship with 48k Hz.
2. Absolute Time Reference for studios should be, where possible, the Global Positioning System (GPS) or equivalent (GLOSNAR), with redundancy receivers. Timecode standard should be the revised SMPTE12M (still under revision at this time).
3. For situations where GPS reception is not possible (some locations or moving vehicles) because of poor reception, an alternative time reference will be required.
4. For fixed locations GPS should be implemented as the time reference.

LTC characteristics

The European Broadcasting Union recommends that timecode generators/regenerators that are built into a recorder or a system should have an output level capable of being varied between 0.5–4.5 V peak-to-peak, and that they should have an output impedance of less than 30 Q. This is to permit compatibility with the usual operational practices within audio installations. The details of timecode tracks and levels for the various magnetic tape formats have been covered in Chapter 4.

Companding systems (e.g. noise reduction) should not be used when recording LTC on an audio track. This is to reduce the possibility of cumulative distortion of the waveform. It is also for this reason that the EBU recommends that the response of an audio system to pulses as well as to steady-state tones or pink noise should be specified, together with the traditional parameters. The pulse edges will be maintained as long as the electronics handling timecodes have a bandwidth wide enough to read at least the 3rd harmonic of the fundamental frequency of the pulse train (around 2 kHz), and are phase-corrected to correct the distortion brought about by the differentiation effect on replay.

LTC crosstalk

The LTC signal is composed of pulses having finite rise and fall times, with a 'sin²' shape to the transition. This shaping ensures that the harmonic content of the wave is kept low. Consequently, crosstalk should be a problem only if poor operational practices are employed. However, the fundamental frequency of the pulse train is one to which the human ear is very sensitive.

A blank track should always be left between timecode and the nearest audio track on multi-track machines. On ¼in twin-track tape machines employing centre-track timecode, the audio tracks should be the NAB option, as discussed in Chapter 4. This allows a suitable guard band between timecode and audio to minimize the possibility of crosstalk.

Many operators have reported hearing apparent LTC breakthrough when off-tape monitoring during recording, particularly on location analogue recorders. Later replay has shown no evidence of timecode breakthrough, so what has happened? The apparent breakthrough has been due partly to electromagnetic coupling within the wiring looms connecting the record and play amplifiers to their respective tapeheads, and partly to the lack of a Dolby decoder in the off-tape monitor path of a location VCR.

Despite wave-shaping, the LTC signal does contain a significant proportion of energy in its harmonics. To drive sufficient current through the inductance that forms the record head, a proportionally greater voltage than for a non-inductive load is required. The comparatively low inductance of the wiring loom, and the high amplification factor of the replay amplifier, give a level of timecode that can be heard when off-tape monitoring during recording. Figure 7.1 illustrates the process.

Much has been said about breakthrough of timecode onto audio tracks. The reverse can apply, with low-frequency audio signals affecting the timecode track on multi-track machines. To prevent this, if a guard track cannot be left, do not record audio containing a significant proportion of low-frequency energy next to the LTC track.

Regeneration of timecode

Whenever a tape is copied together with its longitudinal timecode, it is imperative that the timecode be regenerated. There are a number of reasons for this. Firstly, there is the differentiation effect on LTC replay. This distortion will be aggravated with each successive generation. Fig. 7.2 illustrates the point. The IEC specifications for the waveform shape of both LTC and VITC are given in Appendix 2.

Tape machines suffer from an effect known as 'scrape flutter' or 'timing jitter'. This takes the form of small but rapid variations in tape speed superimposed on the nominal play speed. It is caused by the friction of the tape-to-head contact combining with the slight elasticity of the tape to bring about a stick-slip effect similar in many respects to that of a violin bow scraping over a string. This effect is more pronounced if the tape-to-head pressure is increased as a result of increased tape tension or misalignments in the tape path, and as the longitudinal tape speed drops. This jitter may be small, but becomes worse with succeeding generations.

When replaying material off VTC or VCR machines, variations in tape speed which may be of little or no consequence for audio result in highly unstable video. These variations are corrected by a 'timebase corrector' (TBC). The TBC will also maintain the special timing relationship between the colour sub-carrier and both line and field synchronizing pulses (touched on in Chapter 1 and to be dealt with in detail later). The timing jitter on LTC recorded on the third audio track of a C-format VTR or one of the longitudinal tracks of a Hi-Band U-Matic will not be corrected unless dedicated electronics are provided. Component analogue machines provide regeneration as a matter of course.

Signal levels coming off analogue tape machines can vary, particularly if the tape concerned has undergone a high number of passes of the head block. (Try recording a steady-state tone and replaying it after a few passes of the head; there will be observable variations in level on a meter.) In addition there is always the possibility of dropout on the tape causing momentary but severe level reductions on code which may cause it to be misread. Tape hiss also becomes worse with successive generations, and this can eventually cause code to be misread.

For the above reasons, raw longitudinal timecode coming off tape should be fed to a regenerator before being passed on to another machine during dubbing. A regenerator may provide the following facilities:

1. Reshaping: restoring the LTC signal to its proper form. This is usually done by evaluating the zero crossing points of the replayed waveform, and using these points to generate a steep-sided waveform. This waveform is then passed through a filter to restore the sin² shape of the transitions.
2. Reclocking: extracting the original clock frequency from the LTC waveform and, using the incoming zero/one pulse train as data, forming a completely new pulse train. Timing jitter is eliminated, but longer-term variations in the incoming data rate (perhaps due to variations in play speed of the tape) can be followed, usually over set limits - say ± 6%. Figure 7.3 illustrates.
3. Retiming: When the lock range of the timecode unit is exceeded, retiming takes place. This involves full decoding of the incoming serial LTC data, followed by re-encoding.

Because the regenerator has to examine each complete timecode word for validity prior to re-encoding, a process that will also require a finite time, there will be a one-frame delay (1 LTC word) between timecode into the regenerator and timecode out. This delay must be compensated for within the regenerator for the time information bits in the timecode word if the correct colour framing relationship is to be maintained, as it will have to be if frame-accurate editing is to be performed from these bits. If time information is also being carried in the user bits within the word, it is important that the 'plus one frame' process is applied to these bits as well. A good regenerator will have this option. If it is switched on, and non-time data are being carried in the user bits, the regenerator may well increase the last digit by 1. The 'add one frame' process on the user bits may be enabled/disabled deep in the menus or on internal DIP switches.

Adjusting for the decoding delay

When a timecode reader/regenerator performs a complete retiming operation, the sequence (illustrated in Fig. 7.4) is as follows.

The incoming LTC signal is first decoded into blocks of serial data representing the 0s and 1s of the timecode word. The synchronizing word is also examined to determine whether the code is running up or down in time. Depending on the result of this examination, the data will be modified to compensate for the decoding delay. Under the control of a clock (perhaps driven from external sync pulses) this data train is then fed either to another register or to a microprocessor, from where it is read out to another biphase mark encoder. The LTC reader/regenerator performing a full retiming operation will continue to clock out timecode at standard rate, within small limits, no matter what the speed of the tape supplying the timecode data, as long as the code can be read (one manufacturer quotes a reader / regenerator as reading over a range of one-fiftieth of standard speed to eighty-times standard speed), though this will depend on die ability of the machine to replay the signal at these speeds. Obviously, if the data rate deviates too far from standard, then words will either be lost (play speed too high) or repeated (speed too low).

Machine-to-machine operation

If timecode is to be transferred to another video recorder, or a digital audio recorder, both machines must be locked to common syncs, though these could come from the replay machine, via a TBC. If a VCR is providing guide video for sound post-production, the synchronizer(s) driving the audio machine(s) must be fed with the same syncs as the VCR. Any 'stand-alone' timecode regenerator should also be fed with these syncs, because the reprocessed timecode must be generated at the correct rate. For example, a VCR can be locked to a time address supplied from a sound station. As soon as the VCR has locked up, both it and the sound station will lock to reference syncs via the master synchronizer.

Regeneration of damaged code

A regenerator can be used to repair damaged timecode if it is run in a jam-sync mode. As long as it receives some usable code, and is run at exactly the same rate as the off-tape data, it will regenerate the missing addresses. Even if real-time code is being regenerated, as long as good code is present at the start of each take, and sufficient pre-roll time is allowed, the regenerator will recognize good code and regenerate it. Do remember (and this cannot be too highly stressed) that whenever you are regenerating timecode, the regenerator and player must be fed with common syncs. Do not assume that the player is locked to station syncs – it may not be, especially within an edit suite.

Momentary and continuous jam-sync

When regenerating damaged code the regenerator should be able to operate in both momentary and continuous (auto) jam-sync modes. In momentary jam-sync the regenerator will examine the incoming timecode for a few frames, then regenerate identical time data, locked to syncs. Should the incoming code fail, the regenerator will continue generating contiguous time addresses which will increment up at frame rate. Should the incoming code have any subsequent discontinuity in the time addresses, this will be ignored in momentary jam-sync mode, the regenerator carrying on from the last contiguous time address.

In continuous (auto) jam-sync mode, a regenerator will again examine the incoming code and regenerate identical time addresses. It will also continue contiguous time addresses if incoming code fails. If, however, there is a discontinuity in the time addresses the regenerator will re-jam to the incoming code. Figure 7.5 illustrates the differences between the two modes.

Using VITC

Most UK companies incorporate VITC on lines 19 and 21. It is as well not to put it too early in the field interval as when 1 in VTR machines are playing at other than standard speed, perhaps using a Dynamic tracking (DT) or Auto sense tracking (AST) head for track following during postproduction, the new relationship between video head and recorded tracks will result in more than the usual 12 lines (10 lines in NTSC) being lost. The timebase corrector will replace these lost lines, but it cannot replace timecode that has not been scanned by the video head. Of course if the C-format sync head option is employed the 18.75 lines (15.75 in NTSC) recorded by this head will supply the missing data.

At the time of writing no particular lines are specified for VITC; it is thus perfectly possible to end up with two different sets of VITC on different line pairs, and also to over-record one line (or both) of a VITC pair with fresh code. Should this happen, the reader will not be able to cope with the ambiguity. It will need to be reprogrammed to read either one or both lines of an unambiguous code. It is perfectly satisfactory to set a VITC regenerator to read one VITC line and one non-VITC line, but in this case it should be set to continuous (auto) jam-sync, and fed with corresponding external syncs to protect the data in the event of dropout. It is particularly important that internally regenerated code should be recorded on the correct lines during any editing operation. The VITC lines setting of the generator should also be checked, particularly if the intention is to put a second set of time addresses on tape. Some VITC generators have the options of generating a single line of VITC or a block of lines.

Timecode corruption

The causes of timecode corruption have been mentioned briefly earlier (pp. 23-7). They are different for LTC and VITC, so will be examined separately, with possible remedies suggested.

LTC data corruption can be due to a variety of defects on replay; dropout has already been mentioned. VCR machines run at extremely low longitudinal tape speeds. If timecode is recorded onto an unmodified audio track, with perhaps a relatively poor frequency response, even a small amount of dirt or oxide on the replay head concerned can cause problems. Regular cleaning of the longitudinal timecode heads will prevent many such problems.

LTC can be corrupted on a location shoot by a hum-loop, especially if an unbalanced line is being used to send the code from camera to recorder. Make sure the generator and any associated equipment are fed from the same mains outlet. To minimize the risk of data corruption from mainsborne spikes, interference suppressors should be fitted. They are expensive, but the expense will quickly be repaid in the reduction of down time, and the minimizing of the risk of a system crash which could result in total loss of data in disc-based post-production equipment. To minimize the risk of induced mains hum appearing on LTC data, and possible RF interferences, the signal should be distributed via screened, balanced lines, ideally with screened transformers or opto-isolators at either end of the circuit. Electronically-balanced inputs may not provide sufficient electrical isolation. Radio links are sometimes used for sending timecode between camcorder and audio recorder, and these are susceptible to fading. If the recorder has momentary (one-time), or, better, auto jam-sync facility, use it. If using momentary jam-sync, make sure that timecode is running (and incoming) before jamming to it. This might seem obvious, but it is traditional in film to start the sound recorder before running up the film camera (film stock is more expensive than audiotape). If code is not running, either it will not be recorded at all, or the wrong time addresses will be recorded. With film, there is the risk of the camera running at the wrong speed if incoming timecode fails.

To minimize the possibility of loss of code on recording, the following procedures should be adopted:

1. Use a balanced link for cable feeds of timecode.
2. If mains-powered equipment is being used, try to connect camera and recorder to the same mains point, or use an isolating transformer to feed 'floating' mains, with no earth reference, to the equipment.
3. Check all cables regularly. This will minimize the risk of connector failure.
4. If using a radio link, consider using a 'diversity' system to minimize the risk of fading.
5. If feeding timecode from a master to a number of slave units, do not loop through. If one connector fails, all downstream devices will lose code. Use a distribution amplifier (Figure 7.6).

• 6 If the recorder has no visual display of incoming code, try to check the incoming level or use a separate display unit.
• 7 When using a field audio recorder, if possible use continuous or auto jam-sync mode. The use of jam-sync on location will be discussed in Chapter 8.

Dealing with LTC corruption

Timecode readers/regenerators can cope with corrupted LTC in a number of ways, and there are a number of associated techniques for replacing corrupted code with fresh code. Selection of a particular technique will depend on the exact circumstances but all techniques are based on one of the following:

Momentary jam-sync restriping

If the recorded code on a videotape is badly corrupted it may be possible to use a good section of code to start up a generator in momentary jam-sync mode. The video format used must allow timecode to be laid down independently of video and control track (perhaps a longitudinal audio track can be used). The videotape recorder should be set to record just LTC and put into play at the start of bars at the head of the tape. The timecode generator, which must be locked to the off-tape, timebase corrected video, will jam to the off-tape timecode as long as just a couple of frames can be read. As soon as this happens put the VTR into record mode only on the LTC track (Figure 7.7). The regenerated code laid down on tape will consist of a contiguous series of time addresses. If there are discontinuous sections of timecode on tape the process will have to be repeated at the start of each timecode block. This can only be successfully done if each recorded section has sufficient run-in to accommodate the operation, which will involve the jamming of the regenerator, and time for the operator to confirm the regeneration and throw the timecode channel into record. As long as the corruption was not due to faulty tape (unlikely if the video is still usable) this will solve the problem. As an alternative to locking the regenerator to the off-tape video, the VTR and the timecode regenerator may both be locked to a common source of video sync pulses, perhaps from a common SPG. A good quality timecode generator, even free-running, will have a stability of better than 1 frame in 11 hours, but without reference syncs the code cannot be guaranteed to be colour-locked. If using an internal time generator the VCR should still be locked to reference syncs.

Auto jam-sync restriping

If 'time-of-day' code was used at the time of recording, the regenerator should be placed in the auto or continuous lock mode. As soon as it has received valid code, it will generate a contiguous sequence of time addresses. It will continue to generate contiguous time addresses in the event of the incoming code being corrupted, and will update when a discontinuous time address arrives.

Neither of the above remedies will work effectively for long stretches of corrupted code from an analogue audiotape, as without good code to control the speed of the tape via a synchronizer there is bound to be a degree of slippage due to the friction nature of the drive, even if tacho pulses are being used to control the speed of the transport.

Recording code onto an audio track of a VCR

If it is not possible to lay timecode track down independently of video and control tracks, it may be possible to record regenerated code onto a spare longitudinal audio track. The timecode output should be fed to the spare audio record input. The VCR should be set to record just on this spare audio track, and put into play at the start of bars at the head of the tape. As soon as the regenerator has jammed, the recorder can be put into record and regenerated code will be recorded onto the spare track. If it is not possible to regenerate the code, a separate timecode generator, locked to the off-tape, timebase-corrected video, can either be momentarily jammed to a good section of code, or if all else fails, be set to approximately the same time as indicated in the tape's log sheet. In either case, code should be locked to video syncs. If a slate was put at the head of each take, the difference between the re-recorded code and the original code (the offset) can be calculated for the purpose of post-production.

If it is proposed to do colour-framed edits during post-production, the timecode generator must be locked to video from the VTR in order that the 8-field (4-field in NTSC) sequence is followed by the fresh code. This also means that the generator must be reset at the start of each new recorded sequence.

Dubbing with fresh code

If regeneration or restriping is not possible (perhaps because the LTC is totally corrupted or missing, there is no VITC present, the format does not permit independent recording of LTC, or both audio tracks have programme material on them which cannot be lifted off), there are a couple of options possible. The first of these is to dub the programme material across to another tape, relaying timecode at the same time. The video material should be copied via a timebase corrector fed from station syncs or a sync pulse generator. Fresh timecode can be added from a generator locked to the same source of syncs. If a proper log was kept at the original recording stage it should be possible, if somewhat tedious, to stripe the copy tape with close approximations of the original code. Of course, this method will result in an additional generation of tape, though perhaps the opportunity could be taken to 'lay-off' the audio together with this fresh code, if it is intended to post-produce the audio anyway.

Control track editing

If the worst comes to the worst, and if nothing subtle is required in the way of editing (no 'invisibles' etc.) then it may be possible to edit using the control track pulses. There are distinct problems with this form of editing. There will be slippage caused by the misreading of the pulses as the machines stop and start, which will usually prevent frame-accurate editing. This will be particularly so if the edit is previewed a number of times before the edit is done. There will be no colour framing so there is the possibility of picture disturbance over the edit point.

All the above assumes there is no usable VITC code on the tape. Matters become much easier if VITC is present since any regenerator can be fed with this form of the code, in which case the regenerated LTC will have time addresses that correspond exactly to the VITC recorded on the video tracks. This will permit the dubbing or restriping of identical longitudinal timecode. It will also allow the audio tracks from a machine where the separate recording of timecode is not possible to be dubbed onto a separate audiotape recorder, together with LTC regenerated from the original VITC, for post-production.

VITC corruption

VITC is far less likely to be corrupted than LTC, being recorded as a part of the video signal, where the FM method of recording reduces the effects of partial dropout. It is also recorded twice per field, four times per frame, so if one VITC word is corrupted the chances are the other three words will be valid. If tape dropout is so bad that all VITC words in a frame are affected then the programme video is likely to be unusable too.

Note that since the VITC word is recorded as a part of the video it cannot be restriped without dubbing the video at the same time.

Record-run and time-of-day codes

There are two modes in which timecode can be recorded. In the 'record-run' mode the timecode generated represents the cumulative recorded duration of the tape. The use of record-run mode has the advantage that the time addresses recorded on tape are contiguous, that is they increment up in sequence, with no gaps in the time addresses. In the record-run mode many VCRs will backwind slightly when put in the pause mode from record, with the tape being held against the head drum. When they go back into 'record', the code (and the video) will pick up from where they left off, maintaining continuous (i.e. no gaps in the signals) and contiguous (i.e. no gaps in the addresses) timecode. This option may well be of benefit in situations such as newsgathering, where events may preclude pre-roll time, since the editor can pre-roll from the preceding scene. This contiguous sequence of numbers might also be of use during post-production, as the editor knows that the difference in time between two timecode addresses on tape also represents the difference in tape time. The disadvantage from the operator's point of view is that holding a recorder (including a camcorder) in this mode is costly in terms of battery life, as the head drum continues to rotate.

Power supply back-up

'Time-of-day' (sometimes erroneously called 'real-time') code will run from the moment it is set, and will continue to run no matter what mode the recorder is in, as long as power is available. Most recorders contain an internal back-up power supply, either in the form of a large internal capacitor (2 farad or so), or a NiCd battery on permanent trickle charge from the d.c. supply; which will continue to run the timecode generator for a short time while batteries are changed. Most recorders will continue to run timecode even when they are switched off, as long as the batteries are in circuit, until they are exhausted. One audio field recorder will run timecode for up to three months when left switched off with a fresh set of batteries. However, a word of warning would not be amiss regarding the ability of the internal backup battery to keep the timecode generators running while main batteries are changed. The internal batteries will of course go through the charge/discharge cycle each time the recorder is turned off or has its main batteries removed, or if the batteries go flat, and then is turned on again with fresh batteries. There is a limit to the number of cycles a NiCd battery can undergo before its efficiency drops, and with an elderly internal battery there is the possibility that the recorder may not keep time of day code running with the power off for main battery changing. In field recorders employing large capacitors to maintain the timecode generator (the Nagra IV-S TC is one such), time is required to charge up on first use, or after a prolonged power failure. This recharging time may take a few tens of seconds. During this period the timecode generator will not be available for setting. It is a good operational practice to have fresh batteries ready before removing the old ones, as this will minimize the drain on the backup system. If timecode is lost while changing batteries in a recorder with an internal battery back-up, even if the time taken for the change is short, it is likely that the back-up batteries are life-expired. They are designed to have a long life - 5 years or so - but they will eventually fail. Get them replaced in good time.

Setting the timecode

On some VCRs, including those in camcorders, it is possible to record two sets of timecode. Record-run code can be set into the time-address bits of the timecode word (both LTC and VITC), and time-of-day code can be set into either the time-address bits of both LTC and VITC, or into the user bits of the VITC code. It is thus possible to have both codes present. Figure 7.8 illustrates the facility as provided on the Sony DVW-700P. The REAL TIME switch provides ON / OFF facility to record time of day code in the user bits; the SET position allows the setting of the code in this mode. The F-RUN/SET/R-RUN switch provides free (real-time) run or record-run facility in the time-address bits of both LTC and VITC; the SET position allows setting of both time address and user bits.

Multiple machine continuous jam-sync

If a number of recorders are to run with time-of-day code there are two options open: 'slave lock' (continuous jam-sync) or 'one-time lock' (momentary jam-sync). If slave lock is being employed then one machine or timecode generator should be designated 'master', and LTC feeds taken off this, via distribution amplifiers (DAs), to all other recorders. If composite video is being recorded, all items of equipment, including any master timecode generator, must be genlocked to the same pulse generator. This will either be a camera feeding the master recorder, or a stand-alone SPG. This requirement is necessary for two reasons. Firstly, if the timecode generator and the recorders are not locked to a common pulse train the timing of each may well drift with respect to the others, or their individual timecode/video frame relationship will drift, causing major problems if timecode is to be used in post-production. Secondly, the 8-field (4-field in NTSC) relationship between timecode and colour subcarrier can be established only if the code generator is fed with reference video. In this respect, note that a feed of sync pulses alone (e.g. mixed syncs) is not sufficient, as these contain no colour information. A feed of 'Colour Black' or 'Black and Burst' will be required at least, though a feed of video from the master camera will also suffice. On no account should a system be slaved to timecode supplied from an analogue audiotape machine, as this will have no established relationship with the colour framing sequence. Shooting to audio playback is possible, and is covered in Chapter 8.

Multiple machine momentary jam-sync

It is, of course, perfectly possible to have the video and audio recorders run time-of-day code independently of each other. The timecode generators on the video recorders should be set from a master generator, remembering first to connect the video genlock feed from the master to ensure correct colour framing. An audio recorder can be fed momentarily with timecode from one of the video recorders to lock its generator. If it is an R-DAT machine it should also be genlocked to the video signal. Once this is done, all recorders will then run timecode-locked to the video. An analogue audio machine, though not synced, will generate timecode with a high degree of accuracy, but should be re-jammed from time to time to minimize any drift.

Control tracks and tacho pulses

When neither LTC or VITC can be read, perhaps because the shuttle speed is too high, audiotape machines will read 'tacho' pulses generated from a transducer coupled to a rotating idler wheel within the tape path. These pulses will be converted into an estimate of elapsed time. There is the possibility of slippage due to the friction nature of the drive, particularly during start/stop operations; and during the shuttle process the microprocessor controlling the operation will probably slow the machine down from time to time so that the timecode track can be examined. If the timecode recorded was time-of-day code there will probably be gaps in the number of sequence time addresses. The tacho pulses cannot take account of these gaps, so when the machine slows down to examine the code the microprocessor may see that it has overshot the time called for, so will put the machine into reverse shuttle. Again it will slow down to examine the code, and again will see that it has overshot. The process repeats itself all over again until the operator calls a halt to it. The answer to the problem is good logging of the recorded material so that the operator can instruct the machine to shuttle to just before the gap in the incremental sequence of addresses, and take it over the time-address discontinuity by hand.

R-DAT recorders, as we saw in Chapter 3, can read timecode at the highest shuttle speed. However, they shuttle so fast that some edit controllers, unless programmed with the ballistics of the R-DAT machine, cannot easily cue them up on a particular time address. This makes locating a pre-roll time tedious. This can be avoided by the use of an 'emulator'. This device will make the machine appear to have the characteristics of a VCR. Video recorders, as we have seen, can be set to read control track. They can be set to read this if neither LTC or VITC can be read. The same arguments apply regarding gaps in the timecode address sequence as with audio machines. The control track will contain colour framing information.

Telecine machines will generate 'biphase' pulses as a matter of course. These are not timecode, but are the equivalent of control track pulses. As this pulse train is directly linked to the rotation of the sprocket reels, there is no slippage.

Digital audio synchronization

Digital audio signals are composed of discrete samples. The sampling process is controlled by a highly accurate clock. Operations that require digital audio signals to be mixed or connected between different machines must be in synchronism. This is done by ensuring that the machines' internal clocks are locked together to bit accuracy. Locking is often achieved in a large installation by means of a master clock generator. The AES recognizes two standards for these clocks: Grade 1 clocks are accurate to within 1 part per million (ppm), Grade 2 clocks are accurate to within 10 ppm. Grade 1 clocks are for use in studio centres or when a high degree of accuracy is required. Grade 2 clocks are suitable for single studios. A digital audio recorder will usually have a number of sync input options, as well as a highly stable internal clock. If it has to interface with a video machine, it should be locked to video black and burst, either directly if it will accept such an input, or via an external sync clock genlocked to reference video. The video frame rate and audio sampling rate must be correct, and in particular the sampling rate on replay must match that on recording (avoid having the internal clock running at 44.056 kHz on replay if the recording was made at 44 kHz). Digital audio machines rely on their internal clocks to control their transports. If the clock is running at an incorrect rate, the audio machine will drift out of sync with video in replay. Figure 7.9 illustrates some possible methods of achieving lock between digital audio clocks and video syncs.

If a recording has been made with an unlocked digital recorder, the situation may be recovered by replaying unlocked, and resyncing sound with pictures as necessary. This, however, could be time-consuming (and expensive), and might not be possible with a long continuous recording (e.g. an opera) which cannot be resynced.