Timecode

Timecode in audio operations refers to the SMPTE/EBU time and control code which is used for synchronising audio with video and for locating sections of programmes. Linear Time Code (LTC) is most commonly used, although Vertical Interval Timecode (VITC) may be encountered in television operations. Both versions use an 80-bit binary code to identify video frames by hours:minutes:seconds:frames (hh:mm:ss:ff), and may be conformed to any of the four standard frame rates – 24 fps (film), 25 fps (PAL & SECAM TV), 29.97 fps (SMPTE drop-frame, used for NTSC TV) and 30 fps (used only with monochrome US-standard TV and some older video recorders e.g. Sony PCM-1630).

The IPS Guide to Timecode.

What you need to know is this:

Regardless of the timecode framerate you record on location, this does not affect the playback speed of the audio itself when working in PAL (625).

1 second at 24fps, 25fps, 30fps (and even at 47.5 fps if that was possible) is still one second. Period.

The Golden Rules.

· If your project is for UK TV Transmission (camera running at 25fps) then you should record timecode with a framerate of 25fps.

· If the (film) project is for a 525 territory (USA, Japan etc.) then timecode should be 30 NON DROP. Camera running at either 24fps or 30fps.

· If shooting with a video camera (either SD (525 or 625) or HD) then the timecode rate of the audio should always match that of the camera.

· If the camera is using Aaton or Arri timecode, then the audio framerate should match that of the camera (sounds obvious, but not always the case!).

Always jam the recorder to the camera. An ‘Origin C+’ or equivalent should be used.

Date in UserBit

DD-MM-YY-YY. For example 1st Feb 2015 would be 01-02-15-xx

There is no point putting the sound roll number as the last digit of the UserBit. Just write it on the paperwork etc.

Early software versions of the DEVA ignored the date in the incoming userbit when jamming to an external source. Always make sure the system date in your DEVA is correct. Beware when changing timezones.

Always resync the camera and recorder every 4 hours or so. Ideally at every magazine or battery change as well.

And, in an attempt to give the guy doing the sound sync half a chance, a mic tap (dare I suggest you do it in front of the camera) now and again helps enormously if there is a camera-to-audio timecode offset.

·If recording on DAT or ¼” then select FREE RUN on the timecode generator.

·The sample frequency to video sync relationship is IMPERATIVE. Digital VTRs record digital video and audio. In a SDI (Serial Digital) bitstream the audio and video are embedded within the same signal so the video and audio MUST be locked to each other.

Camera 24, TK25

Audio timecode MUST be at 25fps (otherwise the dubbing theatre will get very upset or, if the editor is doing audio sync in Avid, then if the audio timecode is 24fps the Avid will assume the project is NTSC and similarly get very upset). It is possible to gearbox the timecode, but this will introduce lip sync errors and should be avoided. Alternatively the location DAT could be cloned to a 25fps tape and that timecode and audio used for the remainder of the post process, but that introduces additional costs for the production.

NTSC

With NTSC, the actual picture framerate is 29.97 fps. Therefore, in order to maintain sync with the film, the audio is slowed down by 0.1% to match the telecine framerate of 23.976 fps. A system known as 2:3 pulldown (usually incorrectly referred to as 3:2) is applied to the pictures.

Pulldown explained

Clearly, film (or HD Video) shot at 24fps cannot just be transferred to tape at 30fps as the speed, pitch etc. would be totally unacceptable. Yes, NTSC is actually 29.97 but round numbers work much better to the brain.

So, in order to get the 24 frames of film to fit into 30 frames of video, the telecine scans the film as follows:

The first frame is scanned twice to produce the two video fields needed for frame 1 of the TV signal.

The second frame is scanned three times and the three fields make up frame 2 and the first field of frame 3.

The third frame is scanned twice and the two fields go to make up field 2 of frame 3 and the first frame of TV frame 4

Finally, film frame 4 is scanned three times and the three fields make up the second field of video frame 4 and the two fields of frame 5.

Therefore, by a clever method of scanning the film, four film frames make up five frames of video.

Repeat the process until 24 film frames have been scanned and the result is 30 frames of video.

In reality, the NTSC TV system runs at 29.97 fps (a 0.1% speed reduction). Therefore, to make the film fit, the TK machine runs at 23.976 fps (24 – 0.1%)

So when audio is involved, the location audio is slowed down by 0.1%

That is why on some recorders there are alternative sample rates of 48048, 44144.1, 96096 and 88288.2 Hz. Never EVER record at these sample frequencies as it will cause CHAOS in Post Production. I know, because it has happened to me in the past. If for any reason you are asked to record at these frequencies, ask why, and get the request IN WRITING. Arse covering is always a good thing.

If the camera originally ran at 30fps then 2:3 is not needed but the film in TK is still slowed down to 29.97 at the transfer to tape stage.

HD

HD is a different kettle of fish. With HD VTRs it is possible to play back the tape at other framerates than the one the original was acquired at.

For example, a 24fps tape can be played back at 23.976 or 25fps or even 29.97 (the VTR in this case plays back at 23.976 and adds 2:3 pulldown).

Similarly a 25fps tape can be played back at 24 or 23.976 or (again with 2:3 pulldown) at 29.97. Some of these permutations do not support audio playback!

Therefore the speed at which the material was shot may not be the speed at which it will be transmitted!

Conclusion

My suggestion for all of the above is to try to find out as much as possible as to the final delivery of the programme before anything is shot. Pretty impossible I know, but at least you will have done some homework.

Q&A following on from my IPSNET posting:

Q. Why do I have to use an external timecode ‘box’ to lock an audio recorder to a VTR?

A. Broadcast VTRs have never had battery-backed timecode generators. Early machines did not have timecode at all. When timecode was introduced, it was distributed from a central source. Imagine if all VTRs had a battery backed timecode generator and the resulting tapes from a multi camera (or whatever) football match had timecode generators that differed by anything from 1 frame to 30 seconds? Which VTR would you use for logging? Suppose the VTRs were separated from each other and you tried to synchronise all the timecode generators? Imagine how many VTRs there will be at (say) the Olympics.

Timecode incidentally has to be locked to video as well, so some camcorders may refuse to work if the external timecode feed to them is out of sync with the genlock.

Q. Computer based editing systems like continuous timecode with no breaks. Has this been addressed?

A. Basically I think the misconception of having continuous timecode on tapes is just the editors being lazy (meaning they can start digitising and go to the pub). Certainly I have come across problems where a digi beta camcorder has introduced timecode jumps by a frame or so on in-camera edits and, as the Avid sometimes does not detect the timecode discontinuity and therefore assigns an incorrect timecode to the pictures, when the conform happens, the edit is wrong because the timecode in the EDL is different from the timecode on the master tapes.

Q. With the new flavours of file-based recorders, do we really need different timecode frame rates?

A. Hmmm… I am bound by a NDA on this. Suffice to say, work is being done to ‘make life easier’. Can’t say much more than that at present.

Q. Flex files and EDLs

A. Film Log Exchange files are usually generated in the telecine suite. They look like this. There will be an event each time there is a slate.

000 Manufacturer Aaton No. 021 Equip Indaw Version 4.65 Flex 1005 010 Title IBS FLEX FILE EXAMPLE

012 Shoot Date 16-07-02 Transfer Date 18-07-02

100 Edit 0001 Field A1 PAL 110 Scene 12 Take 1 Cam Roll A6 Sound D.02 11:58:08:17.0 200 16 25.00 000003 000054&03 Key EASTM KU506263 004358+03 300 Assemble 00000003 At 03:00:02:13.0 For 00:00:43:08.0

What it all means:

Lines 000, 010 and 012 are the file headers

Line 100 is the event number. Don’t worry about ‘Field A1 PAL’

110 contains the Scene/Take, film camera roll details, sound roll details and sound timecode

200 Film size (16/35). Telecine speed (in this case 25fps, but could be 23.98, 24, 29.97). The 23.976 gets rounded to 23.98 to save a character. Next is the film roll number (3 in the example) followed by the Keycode. Keycode is put on the film at the manufacturing stage. 54&03 is the duration of the event in feet and frames. The next section gives the full Keycode value (EASTM KU506263 004358+03).

Keycode (in case someone does the sums) is actually printed on 16mm film every six inches so the duration above is actually 54 (half-feet) and 3 frames (27 feet and 3 frames).

300 is the VTR details. Assemble means the machine was in Assemble edit mode. VT roll 3. The clap is therefore at 03:00:02:13 and the event lasts for 43 seconds and 8 frames. The audio timecode in line 110 is the ‘clap’ frame and the Keycode is the Keycode on the film at the ‘clap’ point.

The flex file is loaded into the Avid (or Lightworks or later versions of Final Cut Pro) and the videotape is digitised using this information. During the editing, the numbers are tracked so an audio EDL can be generated and the audio conformed from the location DATs.

If a copy DAT was made at the sync stage (the timecode on this DAT exactly matches the timecode on the videotape), then a flex file is not needed as the video EDL can be used to conform the audio. A flex file is only usually needed if the project was shot at 24fps.

Edit Decision Lists

001 011 V C 16:38:30:03 16:38:55:05 10:00:06:02 10:00:31:04

The EDL was invented for videotape. Film was still being edited on Steenbecks. An EDL assumes the audio is on the same tape as the video. In the example above, it is event 1. The source (rushes) tape number is 011. It is a video only edit (the V could be replaced by A1, A2 etc. for an audio edit) and it is a video cut.

The ‘in’ and ‘out’ timecodes for the playback tape follow, then the ‘in’ and ‘out’ timecodes for the edited tape follow. Therefore this shot is six seconds and two frames into the programme (assuming the start is at 10:00:00:00) and is 25 seconds and 2 frames in duration.

This article by Jeff Booth of Soho Images Film Laboratory & Digital Post was first published in the June/July 2004 edition of Line Up.