Inferno

Hot on the heels of  'Genesis of the Daleks' comes the epic seven-part story 'Inferno', felt by many to be the jewel of Jon Pertwee's first season, if not his entire tenure as the Doctor. Held back until now so that the BBC Research and Development department's Reverse Standards Conversion process could be perfected,  'Inferno' boasts a double-disc release with a full package of special features.


Although this story  exists as 16mm black & white film recording negatives in the BBC library, the preferred colour videotape copies were 525-line NTSC 2" Quads recovered from Canada. These NTSC tapes were standards conversions made in the 1970s for sales before the 625-line PAL masters were wiped. We knew that, before any restoration work could practically be attempted on stories existing only as conversions, some sort of reverse process would be necessary to give us the original field/frame structure. A standards conversion back to 625 causes all sorts of motion problems (as seen on the VHS video release) with smeared frames making clean-up work extremely difficult. Thankfully Jim Easterbrook and James Insell at the BBC R&D department managed to turn Jim's preliminary investigations made some years ago into a viable 'Reverse Standards Conversion' process for the BBC-designed converter used back in the 1970s. Helped also by improved technology and the increase in available processing power we have today, the fine tuning of this process has resulted in the best possible recovery of this material in a form suitable for restoration work. (For further background information on the RSC process please see the article on 'The Claws of Axos'.)

After receiving the 625-line RSC-converted tapes from BBC R&D, the RT's Jonathan Wood was given the job of getting the pictures into the best possible shape before the final manual cleanup. The inherent problem in the new RSC'd 625-line masters is noise due to the various electronic stages and analogue tape generations that the signal has gone through. Far from being consistent this appears as a "noise pumping" with the original 50-fields being calculated from the sequence of data spread across the 60-field source. Also, it appeared that the chrominance was worst affected, with hue and level flicker being evident on picture. After making various test recordings via the grading system and DVNR , Jonathan concluded that ideally the luminance (black & white signal) and the chrominance (colour) should be processed with noise reduction separately. This would enable the use of a far higher level noise reduction only on the colour signal, which would have otherwise caused undesirable effects on detail and movement contained in the luminance. Through experimentation it was found that the usual noise-reduction side of the DVNR (which does have separate adjustments for chrominance and luminance) wasn't fully effective on the rather granular noise and interference patterns on this material, with high levels of NR causing the usual associated movement smear and flatness to the image. Instead, the ASC (advanced scratch concealment) part of the device (which usually takes care of minor dirt & sparkle in an image) was found to "bite" into this grain at certain contrast settings and seemed to work much better in stabilising levels and concealing the noise patterns. However, with no way to split the different levels of processing desired between the luminance/chrominance signals in one pass, the only method available was to produce two sets of Digital Betacam tapes - one for each component. Full DVNR "blast" for the chrominance set of tapes and then a more appropriate setting (found by experiment viewing a monochrome picture only) for the luminance was found, which hopefully maintained most of the source detail. Then, in a three-machine edit these tapes were synced and the processed luminance/chrominance signals re-combined onto a third set of tapes. The DVNR was kept in the video mode except for the luminance during a film insert in which case it was obviously set to "film" with adjustments for dirt size depending on the action - more static shots having a larger "dirt" size applied. Whilst generally very pleased with the results after dealing with the noise this way, an amount of "fizzing" on edges does return on movement as the noise reduction algorithms give up due to the increased motion between video fields.

The resulting combined noise-reduced master then formed the source for the grading stage, with its stabilised colour signal allowing for a more accurate re-adjustment of the skewed hue from the 525 dubs as well as the usual shot-by-shot re-grade of the film inserts. As usual it seems with Doctor Who 525-line material, many of the episodes had taken on a generally yellow bias to the picture. This required correcting, not just  to give a more uniform colour throughout the seven-part story but also to aid the illusion created by RSC of working with the original PAL material. As there isn't a frame of this story existing in its original form there is no original colour reference, therefore the result is Jonathan's personal take on the correct colour balance. Having said that, there were many variations within studio scenes between shots so there is no doubt that differences between the 1970 colour cameras caused by drifting and so on have been "ironed out" in the process. The hues should now appear more natural with other colours more vivid, such as Liz Shaw's dress which is now more of a Royal blue instead of the turquoise hue on the NTSC copies.

The separation of the luma and chroma information had a second benefit in allowing an adjustment to the chroma timing. In the world of analogue video signal paths and tape formats, sometimes the timing of the two picture components becomes altered causing colour bleeding effects where an area of a particular hue no longer exactly matches the corresponding luma component. This can be a subtle fault but it does have the perceived effect of slight image softening and poorer quality. It is very noticeable on faces, which tend to have an unnatural deepening in colour saturation at one side and a corresponding monochrome area on the other. In extreme case, such as  coloured control panel lights, the colour element can appear to be quite separate from the underlying image of the light. In episode seven, not only the more common horizontal timing problem was visible, but also a vertical one. Therefore the chroma path was routed through the Pogle grading system and by changing the timing of the Pogle the colour was shifted back into position.

Although the 625-line structure has been accurately interpolated from the 525-line source (knowing which lines were dropped in the original conversion) there is a certain amount of line jitter in the result that is seen mainly on edges and highlights. This was found to be worse in episode five and can be distracting to the eye on certain shots when it occurs on background detail. Rather than apply a general softening and anti-aliasing across the entire image (which would affect all detail whether it's jittering or not) some time was spent using a pattern key on a digital vision mixer to isolate certain areas or highlights. Using pre-read on the Digital Betacam VTR these areas were then replaced with a vertical softened version, with adjustments to positioning and size of the matte per scene. This doesn't combat the problem completely but lessens the effect so that the eye isn't drawn to these highlights quite so much. On certain locked-off shots an alternative method was available, taking a frame-based freeze from the scene and partially mixing this back over an affected area of zero motion - this maintains the 'live' feel of the picture instead of it being an obvious freeze. More manual work on this jitter problem would be undertaken later on in the manual repair stage at SVS. One last RSC side-effect (due to the noise inherent in the original recordings) is that about 70% of the scene cut transitions suffer with a field or two of break-up. This can take the form of a hue displacement or a noisy combination of the preceding and following shot images. Therefore it was decided that during the manual de-blob stage the material should be treated as we do a film recording and  a frame removed on every scene transition in order to produce clean-cuts throughout. Something that not only benefits the general viewing of the material but also an improvement on the MPEG encoding process for the DVD, allowing the correct placement of I-frames.

During the restoration of "The Ambassadors of Death" for the VHS release, some film cans were called up from the library containing the specially printed stings for that story with zooming captions. At the same time a one-light Spirit transfer was made of a reel which turned out to be a 16mm version of the Pertwee opening titles with a printed dissolve through to footage of lava flows. This was of course the original film insert for the start of each episode of "Inferno" with the keyed and de-focused story captions added electronically via the studio. The original transfer of the faded film was resurrected and the lava footage graded to have the same feel as versions existing on the RSC video masters. Rather than living with the poorer quality 16mm reduction of the main Pertwee titles, the 35mm digital master version was used as normal, with an electronic dissolve across to the 16mm "Inferno" version just prior to the original optical dissolve into the lave footage. However, in order to make this transition invisible a slow dynamic distortion of the 35mm version was applied via an aspect ratio correcter (after Pertwee's face has disappeared) so the geometry matched the 16mm version at the dissolve point. This provided a high quality textless version of the "Inferno" opening to be used as background by SVS for a faithful re-construction of the electronic captions later on.

One thing noticeable about "Inferno" compared with stories existing on their original PAL tapes, apart from the previously-described noise due to standards conversions, is how badly the episodes are affected by dropout. Normally on a PAL tape of 1970 vintage, you might expect there to be around one dropout every ten seconds or so. Presumably due to the multiple tape generations and origin of the source, "Inferno" had one or more dropouts every three frames on average. The PAL-NTSC-PAL conversion also makes such dropouts more visible and damaging to the picture than on an original PAL tape. Repair of these dropouts did not require any more special skill or techniques than usual (essentially painting out the damage with a section from an adjacent area or picture frame) but proved extremely laborious.

The prime candidate for restoration on each episode was the opening title sequence, as it provides an opportunity to impress from the off! The versions on the full episodes suffer terribly from lack of clarity and muddy colours with banding, exacerbated by the heavy reds and oranges of the volcanic eruptions. It was clear that going back to the source film would be of great value, but this presented a further problem. Each episode had a story title, writer credit and episode number electronically keyed over the film, with each slide beginning blurred and slowly coming into focus. Reproducing this effect required some thought. One concern was that electronic defocusing of images is often unconvincing, and consideration was given to reproducing the original technique of a caption slide and camera optics. However, this was clearly a non-starter as there would be far too many variables making it almost impossible to replicate the originals reproducing the lettering would be easy, but what focal length was the camera lens? How far away from the caption card was it? How large was the caption card? In the end, the effect was created electronically after all, in After Effects using a combination of blur filters, keyframing the effect to match the original version. Remaking the titles also allowed us to have a proper start to episode two, which has always been damaged and unplayable on the tape.

Episode one opens with a film sequence, which typifies similar problems throughout the story. There is flicker caused by a the old twin-lens telecine equipment and the usual solution is to deinterlace the picture, rebuilding the image from only one field. The resolution of such film sequences is so low that no true detail is lost. However, with the cyclical noise of the RSC process, it is not possible to keep the same field as the source all the time as every few frames you will throw away a "clean" field and base your frame on a "dirty" one. Analysis showed that there was a repeating pattern of clean, dirty and "hybrid" fields, a "hybrid" being a frame where each field looks very noisy in isolation but the noise tends to cancel out when both are viewed together. To create as clean and flicker-free an image as possible, each sequence was deinterlaced twice, using field 1 and then field 2. These were then loaded into a paintbox system and each "dirty" frame was replaced with the "clean" one. Every five frames there would be a hybrid, where the two deinterlaced frames were blended together equally, cancelling out much of the noise. The end result was a sequence without flicker, and based only on the clean fields on the tape.

Some model and effects shots in the story were slowed down during production, usually to 2/3 or half speed, probably with a polygon telecine although it is difficult to be certain as the standards conversion process mimics the characteristic artefacts of blended fields. These were approached in the same way, but duplicate frames were removed, again keeping only the best, and the sequence subsequently retimed using Furnace to match the original speed.

Although it had been planned to remove dirty cuts at scene changes, in actual fact in most cases it was possible to repair distorted fields, so most episodes lost only three or four frames. Minimising the number of cuts would also make Mark Ayres" job slightly easier as he would have fewer audio repairs to make when conforming his work.

Episode five was edited to match the original broadcast version in the UK (see deleted scene listed in Extras section below). A viewing copy of the b/w film recording was used to make sure the edit was reproduced exactly as Barry Letts made it in 1970.

Throughout the story several small offlocks were fixed. Episode five had a larger one lasting three fields which could just about be covered using in-betweening software and a bit of retouching, but episode 6 had an offlock which provided a much bigger challenge about two seconds of major distortion, with no fewer than 33 frames either completely missing or so badly corrupted as to be unusable. The scene, as Greg Sutton talks to Liz in the hangar, could certainly not be fixed solely using computer animation. A VHS copy taken from a screening in Ohio in 1983 showed severe damage to the tape but with only five frames completely missing. This was probably due to the NTSC Quad recorder and timebase corrector used for that broadcast being able to lock to the tape a little better than the one used for copying to D1 in the 1990s. Looking at a copy of the episode shown in Australia and New Zealand in the mid-1980s shows that BBC Enterprises solution was simply to edit the damaged section out of the episode and oddly enough the narrative doesn"t suffer at all. However, such an option was always our absolute last resort. In the end a more radical and adventurous route was taken to repair the damage. The b/w film recording negative of the episode was called up and the relevant section transferred on the Spirit by Jonathan Wood, with attention paid to matching the picture size and grading as closely as possible to the tape. This was then imported into After Effects and a new composite image was made of five layers (centre image, left, right, top and bottom borders) with each layer requiring different degrees of distortion. This provided a match close enough to allow a relatively innocuous dissolve from the VT to film recording. One interesting side-note is that, usually, film recordings have less image than VT as they are zoomed-in slightly. However, the "Inferno" film recording had more information at the edge of the picture than the VT the explanation lying in the way the original PAL to NTSC conversion was done in 1970, with different line lengths being handled not by digital resampling (as we would do today) but by simply chopping off the beginning and end of each scan line! The geometrically-matched composite was then VidFIRE processed and colourised using surrounding frames as a reference. In fact, given the quality of the film recording compared with the VT, the colourising had to be "roughened up" slightly in order to match the surrounding material. The end result of the repair is acceptable and, perhaps, best regarded as a "Sutekh"s Hand". Hopefully most people won"t even notice it but, once you know it"s there, it"s easy enough to spot.

On the audio front, things were a little more straightforward for Mark Ayres, although the multi-generational aspect of the masters meant that the quality was not as good as many of the stories he has worked on. The majority of his time was spent on de-noising and de-clicking, equalising and correcting levels. However, he was able to rebuild the soundtrack to the special opening title volcano graphics using the original sound effects elements from the BBC Radiophonic Workshop library. Mark also mixed the commentaries, which was complicated by a technical fault that had gone unnoticed during the recording session, affecting one of the mics in the stereo pair. For this reason, the commentary is presented in mono.

Mark's greatest and perhaps most enjoyable challenge for this release had, in fact, nothing to do with the restoration of the episodes' audio, but instead allowed him to wear his musician's hat. Steve Broster wanted some appropriate music for his "UNIT Family" documentary, and thought that Dudley Simpson's cue for the start of the major battle scene in episode one of "The Ambassadors of Death" might be appropriate. Sadly, the original tapes of this music are long-lost, so Steve asked Mark if it would be possible to re-record it, a thrown-down gauntlet that Mark was only too happy to pick up. With the original written session scores for this story also missing, Mark recreated his own by carefully notating the music while listening to the original scene. That done, he re-recorded it using samples of the appropriate original "real" instruments (our budget, sadly, not stretching to the booking six musicians for a session).

 

The commentary for this story features script editor Terrance Dicks, producer (and director on some of this story) Barry Letts, Nicholas Courtney (The Brigadier) and John Levene (Sgt. Benton). John Levene was recorded solo-mic on episodes two and five some months earlier than the commentary for the other five episodes, to take advantage of a rare visit to the UK from his home in Hollywood.

Can You Hear the Earth Scream? is a 35 minute documentary produced by Steve Broster, looking at the making of Inferno. It features interviews with Barry Letts, Terrance Dicks, actors Caroline John, Nicholas Courtney, John Levene, Ian Fairbairn and actor/stunt arranger Derek Ware.

The UNIT Family - part one is a 36 minute documentary  from Steve Broster and Richard Molesworth, looking at the first half of the 'UNIT family' that characterised the Third Doctor's Earth-bound stories. Featuring interviews with Barry Letts, Terrance Dicks,  UNIT creator Derrick Sherwin, and actors Nicholas Courtney, John Levene, Caroline John and actor/stunt arranger Derek Ware, the documentary is an affectionate look at Pertwee's first season through the eyes of those people most intimately involved in creating the 'family'.

Visual Effects Promo Film is an excerpt from an early attempt to 'sell' the experience and facilities of the BBC Visual Effects Department to new clients. This short film features model effects filming from the Doctor Who stories 'The Ambassadors of Death' and 'Inferno', a watertank sketch from 'Marty Amok' featuring Marty Feldman, model shots from 'The Caves of Steel' and the filming of the forced-perspective model pier for the explosive finale of the missing 'Doomwatch' episode, 'Survival Code'.

Deleted Scene - a short deleted scene which was never seen in the original BBC transmission of the story, featuring a rather too obvious Jon Pertwee as the voice of a radio presenter. Although this scene features in the overseas versions of the story and in previous VHS releases, it was decided to remove it from the episode and present it separately in order to be truer to the UK transmission.

The Pertwee Years Intro is a short introduction to the story's final episode from Jon Pertwee, originally presented as part of BBC Video's 'The Pertwee Years' VHS release.

The 1971 Doctor Who Annual and clippings from the Radio Times are included as PDF files for PC and Mac users.

As always, a Photo Gallery and Subtitle Production Notes round off this release.

Copyright Jonathan Wood, Steve Roberts, 24 January 2006. No reproduction allowed without written permission.