'Sea Devils' Scratch Repair

When BBC2 repeated the classic Jon Pertwee story 'The Sea Devils' in 1992, episodes one, two, three and five were sourced from relatively poor quality NTSC standards-converted master tapes. A PAL copy of episode five actually existed, but the tape was too badly scratched to be transmitted. Steve Roberts tells the story of how this episode has now been restored to its former glory….

'The Sea Devils' was made in 1972, on the then-current two-inch Quad videotape format. Although capable of very high quality recordings, the format has one major weakness - it is a segmented recording. Unlike modern analogue formats, where a single video field is recorded onto a single helical track on the tape, Quad uses a high speed head moving almost at right-angles to the tape travel, with each field consisting of approximately twenty tracks on the tape. This is the cause of the characteristic 'banding' effect often seen on archive material that has not been aligned correctly for replay. The segmented recording technique is also prone to show up longitudinal tape scratches of the sort easily caused by dirt trapped against one of the stationary parts of the videorecorder's mechanism. In modern recorders, such a scratch shows up as a single long horizontal line, but on Quad it shows up as over thirty short lines spread all over the picture. This is extremely annoying and very difficult to detect and compensate for in the videorecorder.

The scratching on episode five of 'The Sea Devils' begins about eleven minutes in, as a very deep scratch which becomes less noticeable towards the end of the episode, but never quite disappears. For this reason the PAL master was considered untransmittable and so a copy converted from a recovered Canadian NTSC tape was broadcast in its place.

A couple of years later I decided to have a look at the possibility of repairing the damage to the PAL master. The BBC Film and Videotape Library released a copy to us on D3 digital videotape - effectively the same generation as the quad, as D3 suffers very little duplication loss. Close examination of the damaged portion showed that there were thirty two separate dropout lines visible on the screen. However, they always stayed on their respective television lines, with a deviation along the line of approximately 6uS (a visible TV line is 52uS long). Using a genlocked Commodore Amiga microcomputer, it was possible to generate a series of white lines on a black background, which when superimposed on the damaged pictures completely covered the dropouts. This signal could then be used as a 'key' to insert small sections of repaired 'fill' video into the damaged areas.

To generate the 'fill', I borrowed the technique of dropout compensation used by videorecorders to try to mask tape dropouts. The technique relies on the fact that picture information does not vary too much from line to line - any area in a TV line will look fairly similar to the line above it. Thus by replacing the damaged section by a copy of the information in the line directly above, a fairly good repair can be made. This was done by passing the damaged video through a Digital Video Effects generator. This allowed me to move the picture down by precisely one line, the result of which was recorded onto another D3 tape. This was run in synchronism with the damaged tape, and the key signal used to drop the fill into the damaged areas. Although this technique worked fine most of the time, I was not happy about the way it showed up on diagonal lines in the picture. I decided to refine the process slightly, by filling the damaged section with a fill signal generated by taking the average of the lines immediately above and below the damage.

The DVE process was repeated to generate another tape, this time moved up by precisely one line. These two tapes were run together in synchronism though a vision mixer to generate a third tape, which appeared to be blurred vertically. The important point is that the part of the line corresponding to the damaged area was now free from dropout and consisted of an average of the lines immediately above and below. By using this tape as the fill signal, a very much better result was obtained than from the single line method - in fact the result is so good that you have to be very close to the screen and know exactly what to look for before you can spot the repair.

The repaired version, on D3 videotape, was technically reviewed at the Film & Videotape Library and is classified as the transmission master tape for this episode. For reasons that escape me, it appears that this tape was not used for the BBC Video release of this story and an inferior version originating from an NTSC copy was used instead.

Copyright Steve Roberts, 1997