Using Timecodes (Part 3)
Tonmeister Course lecturer, Francis Rumsey, explains how timecodes are used to synchronise audio to video in Part 3 of our series.
Having looked at the various ways in which timecode relates to video, we can proceed to investigate the role of timecode in the locking of audio machines to video. There are a number of factors unique to this subject, and these generally relate to the imposition of 'video' structuring on a free-running medium like analogue audio.
We have already discussed the need for a control track on VTRs (video tape recorders), as a means by which the tape's speed can be locked to an external reference, and as an internal reference for the machine itself. With VTRs, the timecode track has secondary importance as far as speed control is concerned, and is more important as a positional reference on the tape, for location of relevant sections.
Free-running analogue audio recorders do not traditionally use any form of reference track; their speed control relying entirely upon the accuracy and stability of, say, a crystal reference oscillator which governs the capstan. It is for this reason that timecode becomes so important when locking ATRs (audiotape recorders) to VTRs because the timecode is the only means by which a synchroniser may know the physical passage of tape. (Roller guides may issue pulses when moving, but these will only serve as a rough guide during fast wind modes.) Thus the timecode becomes a virtual control track for the ATR, imposing a 'framed' structure onto a previously non-segmented recording.
This is quite a difficult concept for video engineers to grasp, many of whom find it incredibly difficult to imagine a tape recorder without a control track or speed reference. I have met a few who never realised that there wasn't anything of this sort in the conventional audio studio!
It is an area where misunderstandings arise because video machines do not need timecode to remain locked together, whereas there is no other way to lock ATRs than to use a timecode of some sort. For this reason, the timecode recorded onto the ATRs in a system must be almost flawless, having no jumps in it, and no gaps; whereas that used in a video editing system, for instance, will not always be as critical. To clarify further, once VTRs have locked up initially, they switch over to the video reference and CTL track for speed control, whereas speed control for the audio machine must be constantly applied by the synchroniser with reference to the timecode track. Any 'glitches' or number jumps in the ATR timecode will confuse the synchroniser.
You may remember that, in the first section of this series, I pointed out the need for very tight lock when locking two audio recorders together. It was because any short term variations would result in a phase-shift (which might be audible) between the audio tracks on the two machines. Locking of audio tracks to a picture does not demand the same short-term lock criteria, because the brain is more tolerant of short-term changes of sound in relation to vision than it is of phase variations of sound. It is quite possible that the sound could move by a considerable fraction of a frame without being objectionable, as long as the longterm drift was not increasing.
This slackness of short-term lock tolerance is helpful to synchroniser designers in preventing the transfer of wow and flutter from picture to sound. It is generally true that the video or film transport will have more wow and flutter than a high quality ATR, so it is useful for the ATR not to have to follow every slight variation; rather to average out the flutter and follow any gradual speed changes.
As far as long-term changes in offset are concerned, it is a generally held opinion that differences of up to a frame can be tolerated between sound and picture, although to me this is too great. You will have noticed the effect on film or TV, when the speech does not match the speaker's lips. The industry criteria for film work is half a frame in twenty minutes for free-running, crystal-locked cameras and recorders, and this is adequate because rolls of tape and film have to be changed regularly, allowing for re-synchronisation.
Certain modern synchronisers allow for alterations in relative offset of audio machines which amount to fractions of a frame. This can be termed bit-offset, because the change is usually in multiples of one timecode bit, which is one-eightieth of a frame (one timecode frame is composed of 80 bits of digital data). You can probably see that this would be of no use to VTRs, because the minimum useful division of time in a TV picture is one frame, making the shift of one against another by less than that amount absolutely pointless.
Bit-offset can usefully be employed to fine-trim the positions of sound effects and the lip-sync of sound with mouth movements. It might also be useful for adjusting fine phase between two locked ATRs with similar material recorded on them, or for editing of audio at tricky points.
During the initial pre-roll period, in which all the machines are locking up (synchronising), it will be the normal requirement that lock is achieved as fast as possible, and this usually results in audible wow of the sound while the synchroniser is adjusting the speed of the capstan. After this, any modifications to the speed of the slave machines should ideally be inaudible, and thus performed over a longer period of time. An example of such would be the transmission of a simultaneous broadcast on TV and radio, where it would be imperative that no audible shifts of pitch in the sound were caused due to problems with timecode.
Sudden and short errors may occur in the recorded timecode on either audio or video, and these will become the points at which minor adjustments in sync could become necessary. These errors can be due to drop-outs on the tape, inadvertent gaps in the code or other human bodges.
Synchronisers vary in the facilities provided, but most allow the user to select fast or slow re-lock depending on the use. Others may allow manual or automatic switching after initial timecode lock to the counting of CTL or tachometer pulses, or to the counting of timecode sync words, so that number jumps would not matter.
Now that we have the facility for locking more than one audiotape recorder (ATR) to a video recorder (VTR) using a synchroniser, we can extend the concept of electronic editing to the audio domain.
Video tape editing involves the copying of sections from various source tapes to a master in the right order, rather than physically splicing the tape. Now that time code is present on our audiotapes it will be inadvisable to cut them, because this would destroy the numerical sequence of the code, therefore a similar procedure must be applied to that used in video.
A lot of the audio editing performed on multitrack tape for video productions will involve the dubbing of effects, music, voice-overs etc, at the relevant points in the picture. Here, the synchroniser can be used to set up the relevant offsets between the effects tape and the multitrack, lock them together before the effect occurs, and drop the multitrack into record at the desired point. This will allow for rehearsals of the dub, and alterations to the timing by adjusting the offsets between machines. In a similar way, two-machine editing could be performed between audio recorders to assemble effects in the correct order for dubbing, using the synchroniser to line up the material and control the record drop-ins.
A few points must be borne in mind when attempting electronic edits with ATRs, because of the way in which analogue audio is recorded. Although most professional machines have 'gapless drop-in', this does not provide a good enough join for cutting inaudibly in the middle of music. This is mainly due to the need for the tape machine's bias signal to be ramped up at the drop-in point, so that a click is avoided, and, although this ramp will be optimised for the fastest drop-in without a click, it will still result in an imperfect join. Consequently, this technique can be used for editing during short gaps between sounds, such as that between two sentences of speech, but not for accurate musical editing, and this is usually adequate for post-production work.
The method is workable with pictures because VTRs perform the drop-in when the rotary head which records the video is in the vertical interval (the time in which the spot scanning the TV screen is travelling back to the top of the picture). It also works with most digital audio systems, because these are segmented recording formats with which the drop-in can be timed correctly.
Quarter-inch 2-track tape recorders have recently been enhanced by the addition of a timecode track which runs down the centre of the tape, in the middle of the guard-band between the two audio tracks. This allows for them to be synchronised while retaining stereo capability, and also compatibility with non-timecoded tapes, apart from the following exceptions:
All centre timecode ATRs have heads physically conforming to the NAB standard, which has a wide guard-band. DIN standard heads have a narrow guard-band which does not allow enough room for a timecode track without creating adversely high crosstalk (see Figure 1).
It follows that both NAB and DIN recorded tapes without timecode may be replayed on a NAB centre timecode machine, but that NAB tapes with timecode will not be playable on DIN machines due to the timecode being picked up by the audio gaps. NAB timecoded tapes are perfectly playable without crosstalk on NAB machines without a timecode track. Remember all that if you can!
Such tape recorders always ensure that the timecode is recorded physically coincident with the audio at any point in time. Therefore, it doesn't matter where the head is which records the time code, as long as the advance or delay within the machine is adjusted accordingly: this eliminates the differences in head positioning between machines, so that tapes can be replayed on any make of equipment.
Certain manufacturers use two heads to record and replay the code, whereas others use just one. This depends on whether the machine is capable of advancing the numerical value of the timecode, as well as delaying it. With a two head machine, only capable of delaying, the time code replay head is placed prior to the audio replay head, so the code is picked up first and then delayed until that portion of tape passes the audio replay head (Figure 2).
The time code record head is placed after the audio record head and the input timecode is then delayed until the time-coincident audiosignal passes the timecode record head. Single head machines will still delay on replay, but they use a software regeneration process to advance the code in record mode, so that the value of timecode recorded is before the actual input value.
Of course, the simplest solution is to have the timecode record and replay gaps in the same tape head as the audio gaps, but so far the resultant crosstalk has proved too bad...
Feature by Francis Rumsey
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!