The Three Doctors

Until the end of October 2002, the fortieth anniversary year was supposed to kick off  with the tenth anniversary special, bringing three incarnations of the Doctor together  to fight a renegade Time Lord. However, BBC Worldwide's Marketing department have other plans for this release, so it has moved back to a November release instead.

Jonathan Wood, senior colourist for BBC Resources, explains the processes the story had to go through to make it ready for its DVD debut...


The tenth anniversary story of Doctor Who, 'The Three Doctors' exists in its original form on the four transmission Quad tapes from 1973. They had been transferred to D3 as part of the archive project some years ago and these composite digital recordings were dubbed to component Digital Betacam via the new Transform PAL Decoder developed by the BBC Research Department. This produced amazing clean pictures free from PAL artefacts, providing a great start for the restoration work to produce the DVD master. Having worked recently on the 1980's story "Resurrection of the Daleks" it was obvious what a step down the later 1-inch format was in terms of image crispness, particularly after several tape generations.

As usual the pictures were passed through the grading system to improve the colour balance and levels, while at the same time applying a small amount of video noise reduction with the DVNR. As expected there wasn't that much to iron out on the studio material once an overall setting was found, whereas the film inserts required a complete re-grade. This would also include "fighting" against the original, unsubtle grading tweaks mid-shot on the original transfer. Once a complete episode was programmed then the edit through the grading system would begin onto what should become the master tape. At this stage that the odd flash frame or poorly aligned invisible edit would either be removed or repaired. Again the film inserts (mostly in the first two episodes) were particularly poor with 2-3 frame "bounces" on the edits, completely ruining the illusion of monsters appearing and disappearing. However, it became apparent that there were more complex problems with the film inserts than was first thought.

A Polygon telecine machine had been used to replay some of the film into the studio (episodes 1&2), and unfortunately this type of machine had the effect of smearing the film frames together giving the impression of a standards conversion. The start of the story begins with an exterior sequence shot on film and after examining the frames closely it seemed that one field of the video frame would be a blurred mix between adjacent film frames, whereas the other field was unblurred. It was initially hoped that it would be a simple case of putting these film sequences through a Snell and Wilcox ARC-100 Aspect Ratio Converter (which includes a facility to 'filmise' video), to effectively throw a field away (the bad one) from each video frame of the Polygon sequence, the result being clear movement with proper frame by frame motion. Unfortunately, experimentation proved that this process was unsatisfactory with this particular material.

Further examination revealed that the Polygon telecine was free running, meaning that the blurred field would at some point change from being field 1 to field 2 of the video frame, with the several frames over the transition having both fields blurred. This meant that about 50% of the time the result via the ARC looked like an out-of-picture-phase telerecording of a film insert (like "Ambassadors of Death") giving a double image to all movement and looking worse than the original. The idea of making two copies of the sequences, one from field 1, one from field 2 and then editing the two together appropriately was tried but was far too laborious, and choosing the wrong frame to edit caused either a stutter or jump in the motion. The importance of getting back most of the original discrete film frames was not purely to provide less smeared motion, but a necessity in terms of clean-up. As usual the 16mm film inserts looked like they'd been dragged all around the floor before being telecined and were absolutely covered in dirt. With the frames merged together, much of the dirt and sparkle would consequently appear on two or more adjacent frames meaning that an automatic clean-up device like the DVNR wouldn't touch them even on a very high setting. Manual clean-up on the scratchbox would also be impossible without discrete frames in a correct phased sequence. Some sort of automation was required that would go through the film sequences field by field and output a new sequence using the best source for each film frame.

At this point, the problem was handed over to Ian Simpson in the 3D and Video Effects department to see if it would be possible to automate the process of selecting the 'clean' fields and outputting them as frames. Using a software package called Illusion running on a Silicon Graphics platform, Ian built a system which examined each pair of fields and selected the best one. This process relied on the fact that the unwanted blurred field contained less high frequency detail than the unblurred field. In order to weigh each field's detail against the other, Ian's system first adds a massive amount of white edge enhancement to each field, then subtracts the original field from the enhanced field. This gives an outline view of the image, showing just the edge correction - which is obviously more pronounced in the unblurred field as there are more defined edges. Each field is then blurred massively so that it becomes a flat grey field, the average picture level of which is directly proportional to the amount of detail in the original field. By using the higher level grey to switch through the appropriate field, the good fields are selected automatically and the blurred field discarded. Where both fields are blurred, the system selects a mix of both fields to its output. Each selected field is then interpolated up to a full frame.

The result was a sequence which now had correct frame based motion, with the many of the frames unblurred which allowed for DVNR to do some processing followed by Scratchbox work. About 850 blemishes were removed on the film inserts for episode one, with the first two shots of the story alone requiring 100 fixes each! A further 600 in episode two, and about 1100 between episodes three and four. There wasn't such a problem with the preparation for clean-up on the last two episodes as these had the film inserts replayed from a MkII Twin-lens telecine which didn't mix frames together. However, they did suffer from luminance and geometry differences between video fields causing flicker, so these inserts were passed through the ARC which this time worked a treat. Although most of one field is thrown away there's no discernible difference in resolution because of the nature of the source material, the result being a much more stable looking picture. In the "new" Polygon sequences there were some points where the movement stuttered due to similar frames being output (as the telecine was running slower than 25fps) so a further edit was required to remove these frames where possible. The results aren't perfect and indeed there were points on some motion where areas of the picture had to be mixed back with the original so that there wasn't apparent double imaging, even though on close inspection this would of course re-introduce the blurred frames. However, the resulting pictures are much improved over the originals.

On the studio material video dropouts, flashes etc were fixed utilising the vision mixer, as were impulse interference spikes on the camera when a sparking device is used by Sergeant Benton in episode two. As usual these sorts of problems are field based, appearing on only one field of the video frame. However, as no video tape machine yet allows for single field editing, many required that a digital clone of the material be made before fixing. This allows the good field from the same frame to be dropped back in afterwards (after changing the edit timing) in case it gets "corrupted" due to the fix, as sometimes the pasted material that's been aligned to cover the fault is in totally the wrong position for the other field in the frame. Of course this is only necessary if there's some movement (or a panning shot) creating a difference between each video field. There were quite a few instances of changing line length between cameras, producing excessive blanking down the right hand side on occasions. However, there's not much you can do other than centre up the image, and with it changing quite often (and sometimes dissolving through to other shots) no time was  spent on this aspect.

The explosive effect used in the story (when something is transported to the black hole) was simply a film insert with a white on black graphic luminance-keyed over the studio shot at the appropriate time. However, as usual this was a dirty piece of film which meant that dirt and sparkle was added to the studio output prior to the "explosion" (as soon as it was keyed on - in this case, live in the studio gallery without the benefits of frame-accurate VT editing). Also, once the film graphic had become full frame the vision mixer would dissolve through to white (varying shades of) but only after several more dirty frames of film had gone through. The film insert, as with the others, was out of picture phase and by it's nature impossible to clean up (apart from an obvious tape splice), but the keyed sparkle has been removed and the transition to white remade frame-by-frame. Some CSO keying was used in the story on shots like a disappearing doorway in Omega's complex, and the Brigadier opening the door of the transported UNIT HQ. Both had some noisy edges and these were simply "blasted" with DVNR to reduce the fizzing effect and painted back with various mattes on the vision mixer so as not turn the rest of the picture into a smeary mess. The results are never going to be great but it's now less obtrusive on the eye. A sequence near the beginning of episode two, where the key had been left switched on into the next scene, causing black dots to appear on the faces of Jo Grant and Doctor Tyler, was manually touched up by Peter Finklestone using Commotion. This was still a little 'bitty', so it was then DVNRed to smooth out the correction.

Episode three had some noticeable Quad banding due to a head clog, which gradually got worse, lasting about five minutes, then slowly disappeared. Unfortunately it's very difficult to get the Quad tape back for a re-transfer but luckily  a quad protection dub of the episode had been made in 1981 and that in itself had been dubbed to D3. This copy was viewed and did not have any head clogging, and only exhibited a very slight increase in noise and ringing being another tape generation down. It was certainly preferable to the banding so this material was inserted via the grading system and DVNR to match what was done previously.

The opening titles were remade using the master Pertwee titles transfer and re-keying of the captions using the originals as a key source. The closing titles of episodes aren't remade but often have some clean-up done on the film background, but in this case each episodes end sequence suffered from the same problems as the other film inserts. The problem here was that any filmising via the ARC to allow for clean-up meant that the electronically produced captions keyed over the film became edgy, losing half of their vertical resolution. So once the backgrounds had had a quick clean-up, a key was derived from the captions themselves (pre ARC) in order to partially re-key them back onto the filmised sequence. This meant that the background was now properly frame based (without flicker etc) but the field integrity of the video captions was retained. This process did throw up some instances when either the captions cut on the wrong field or changed within active picture area which required some further fixing.

Once the new masters had been locked to length, DAT copies for audio clean-up work by Mark Ayres were produced. Mark explains the processes he had to put the audio through...

"The first thing I do, as a matter of course, is to replace the opening and closing titles music from my digital copy of the original master. This can be quite tricky were the titles fade out into the opening action of the episode, or dissolve through from the last scene. Careful positioning and crossfading to avoid phasing is employed here. For the openings of episode 2-4, where the music fades into the action, it is often possible to copy up part of the preceding episode's cliff-hanger (as on the original edit) to give a clean "in" -  as here. This also saves a tape generation by redoing the original reprise digitally.

The soundtracks are then put through a gentle decrackle filter to remove little electrical clicks (of which there are many in a couple of these episodes). The crackle filter is also quite good at catching some of the peak distortion present at times. Careful listening is then done to make sure that all clicks have been caught - some larger ones, and some additional distortion, is manually drawn out of the waveform where necessary. Some peak distortion on a couple of loud shouts is impossible to remove, however. It is also important to check that the filter does not remove sound that is required and, indeed, here it softened some of the radiophonic effects such as the Doctor's Geiger counter and the antimatter monster sparking in episode two. The crackle filter is gently backed off at these points (by programming a curve on the "reduction" parameter in the workstation) and the sound is checked again.

The next stage is a noise filter. These episodes have quite a loud tape hiss all the way through, so a gentle broadband noise reduction is employed. Further listening to the now relatively noise-free audio reveals some low level low frequency bumps on some edits (usually caused by a slight mistiming in the switching of the erase and record heads on analogue editing machines). A gentle low frequency shelf filter is ramped in at these points - again, this is all programmed on the workstation so that the filter is only employed where necessary.

Once this is done, the audio is copied back to freshly-formatted DAT cassettes (maintaining the correct episode timecode) and sent back to the BBC, where it will finally be copied back to the video masters ready for DVD mastering.
"

All this has now taken place and the assets for this disc have been delivered to BBC Video.

Extras for this release include:-

 

Copyright Jonathan Wood / Mark Ayres / Steve Roberts,  22 April 2003