For its Autumn 2001 release, the BBC Radio Collection tasked series producer Mark Ayres and scriptwriter Sue Cowley with their biggest task yet... The restoration and release of the twelve-part Hartnell-era epic 'The Daleks' Master Plan', plus its single episode teaser 'Mission to the Unknown' as a CD box-set. In this in-depth technical article, Mark describes the process of taking this huge project through from raw off-air recordings to fully finished CDs. The article is presented in the form of an audio diary, which Mark posted as a series of updates on the Technical Forum as the project progressed...
Diary Entry 1
Four different sources of off-air recording are available to me, plus two optical film tracks from the surving episodes 5 and 10. My first task is to assess the quality of all these recordings...
New transfers direct from original tape newly-loaned by David.
Microphone recordings - mic placed in front of TV speaker. Tape speed 3.75ips - resulting in better quality to tape. Tape stock was "International Electronics" 600' acetate-backed on a 5" spool. This was a cheap tape stock commonly available from (amongst others) high-street hi-fi chain Laskeys. Despite its cheapness, good quality results could be achieved on this stock - indeed I used it myself in my 'teens! Despite some mains hum (easy to remove), and the sound being a bit "roomy", quality is generally good.
"Mission to the Unknown"
Runs "fast" - this can be corrected. Despite some mains hum (easy to remove), and the sound being a bit "roomy", quality is generally good. Recording features trail for "The Munsters" before episode, and complete opening and closing titles. Tape also features part of a radio review (of unknown origin) of "Doctor Who and the Daleks" at the end (also transferred to my archive).
Episode Eight - Volcano
Runs slow - can be corrected. Some mains hum, and recording is also extremely muffled compared to "Mission to the Unknown". Complete with opening and closing titles.
Allen Wilson Recordings:
Allen taped a complete run of episodes from "Galaxy Four" to "Destruction of Time" (i.e. all of the adventures "Galaxy Four", "Mission to the Unknown", "The Myth Makers" and "The Daleks' Master Plan"), after which he ceased recording. I have already used short clips from his recordings of "Galaxy Four" and "The Myth Makers" in the preparation of those episodes for audio release.
All are microphone recordings, very roomy, with frequent movement, whispering, and fits of coughing in the viewing room. Intelligibility is generally quite good, however. Tape speed 3.75ips, tapestock is Grundig brand.
One interesting thing is that although the recordings were made on a standard 1/4-track tape machine (this presents the user with two tracks which can be used either for two separate mono recordings or for a single stereo recording), Allen never seems to have realised that he could have turned the tapes over, got himself another two tracks, and doubled his tape stock!
All recordings are rather hummy, and feature prominent scan whine from the TV tube. Note some additional points below.
Most episodes (except as noted below) have opening and closing titles intact (barring the occasional clipping of a couple of note at the start of the opening titles or end of the closing titles). The advantage of this over some of the other available recordings is of course that cliffhangers and reprises are complete.
Episode Two - Day of Armageddon
This episode is missing the opening titles, and the last couple of minutes after Chen's line "I, Mavic Chen, give you the core of the time destructor" (this due to tape running out, after Allen had cut off the opening titles to try to fit 3 episodes on one reel).
Episode Four - Traitors
This episode is missing the opening titles (the cliffhanger reprise is intact, however - the only recording we have of this reprise). The end titles are also missing. Despite these omissions, this recording is actually Allen's best overall.
Episode Five - Counterplot
Quality is back to usual after the exceptional "Traitors". Opening titles and cliffhanger reprise are missing, as is the last few minutes (again due to trying to fit three episodes on a tape - you would have thought he would have learnt his lesson!).
David Holman Recordings:
Microphone recordings. Intelligibility generally good, with some exceptions.
"Mission to the Unknown"
Opening titles intact, no closing titles. Generally good.
Episode One - The Nightmare Begins
No titles at either start or end. Episode boundaries are all-but obscured.
Episode Two - Day of Armageddon
No titles at start (episode boundary obscured); closing titles intact.
Episode Three - Devil's Planet
Opening titles intact; no closing titles. There are a number of dropouts due to tape damage, and considerable interference at times.
Epsiode Four - Traitors
Opening titles and cliff-hanger reprise missing. Closing titles missing.
Episode Five - Counterplot
Opening titles and reprise missing. Very distorted start due to to recording level too high. Closing titles missing.
Episode Six - Coronas of the Sun
Opening titles and reprise missing. Cliffhanger and closing titles intact. Considerable peak distortion at times.
Episode Seven - Feast of Steven
Opening titles intact. Closing titles missing. Some peak distortion.
Episode Eleven - The Abandoned Planet
Opening titles intact, closing titles missing. Some considerable peak distortion. There are a number of dropouts due to damaged tape, and some varying levels (the recording level appears to have been altered a couple of times during the recording).
Graham Strong Recordings:
Intelligibility generally very good - as I understand it these are the first "line" recordings (i.e. recordings made with the recorder linked directly electronically to the TV rather than via a microphone) Graham made of the programme. These recordings tend to be the highest quality overall, but suffer from a fluttery dropout throughout, probably due to the quality of the recording media.
Episode Eight - Volcano
Start of opening titles missing, but start of episode is intact. Cliffhanger intact, most of the closing titles are intact.
Episode Nine - Golden Death
Very start of opening titles missing, but start of episode and cliffhanger reprise is intact. Some interference at times.
Episode Ten - Escape Switch
Opening titles and cliffhanger reprise intact. Cliffhanger intact. Most of closing titles intact.
Episode Twelve - Destruction of Time
Opening titles nearly complete; cliffhanger reprise intact. Ending and closing titles intact.
Episode Five - Counterplot
Quality good, except for serious wow and flutter over first part of episode.
Episode Ten - Escape Switch
Quality good, except for serious wow over opening titles.
Diary Entry 2
In the first of these diaries, I documented the various sources I have for The Daleks' Master Plan (with the exception of the Richard Landen recordings - more on these in a later entry!). After a few days collating these sources, it's time to start remastering.
Normally, when doing these releases for the BBC Radio Collection, I do the remaster and the voiceover mix all as one Pro Tools file. But on this occasion, with 13 episodes to do, I would find my hard drives bursting if I did not find another way to do it. So I'm remastering the episodes first and writing them to new files, then I'll overdub the voiceover to those.
I decided to start with the easy ones (!): episodes 5 (Counter Plot) & 10 (Escape Switch). In both cases, it made sense to use the film recordings as my main masters. Despite the fact that they are covered in optical film noise which would not have been present on original transmission, they are - by definition - professional standard line recordings from the original tapes. The Allen Wilson recordings of these episodes are not in contention, while the David Holman recordings - although good - are slightly "crushed", and have quite a lot of peak distortion at times.
Another reason for using the film recordings as the masters was the ease of making them resyncable to the video masters in future.
Steve Roberts provided me with a new DAT taken from the optical tracks. As previously noted, these are not at all bad (certainly better than the soundtracks of the episodes on "Daleks - The Early Years" would lead you to believe). Counterplot is quite noisy at times, with many optical pops. Escape Switch is pretty clean all the way through - very few pops, just the standard "sheen" of optical noise.
First job was the one I do as a matter of course - replace the opening and closing titles music with a direct digital copy from the original masters. This is not as straight forward as it seems. For a start, the music has to be faded into the programme as it was originally. This can be done by cross-fading into the original (trying to avoid a phasing effect on the dissolve) or, as here, by fading out and then up on the programme. To avoid an abrupt change in quality, I lay some background noise from the programme across the join. In both of these episodes, the effects at the end of the programme tail out over the closing titles, so this has to be rebuilt by taking some jungle effects from elsewhere in the programme (the end of part 5) or from the beginning of the next episode (the end of part 10 - I stole some effects from the David Holman recording of the start of part 11).
The main problem with both episodes is the "wow" (slow-ish variations in playback pitch) over the starts of the episodes. For Escape Switch, this was only unbearable on the opening titles - no worry, as I replace these from the masters as a matter of course. But for Counterplot, the effect continues for nearly three minutes, almost to the point where the Doctor, Steven and Sara get caught up in the Molecular Disseminator. It is difficult to say how the effect got onto the films. The film recorder couldn't have been out of alignment, as the sound and picture are sprocket-locked and if the film recorder changed speed, it would lose lock with the video player. Likewise, if the video playback machine was unstable, the film recorder would have lost lock. So, either the video playback machine was picture-stable but wowing over the audio heads, or the audio must have been laid off to a non-sync medium and back at some time. Baffling.
Anyway, how to fix it?
My initial unfeasable
thought was to use a modulated delay to track the effect in inverse, so removing it. As I
say, unfeasable - syncing the modulation would be a nightmare. So I was left with the only
other option - to rebuild the first couple of minutes of the programme from other sources.
Having replaced the opening titles, I took the first few seconds of Sara Kingdom's dialogue from the 16mm. (An aside - this was actually a 35mm playin of the previous episode's cliffhanger but is not actually the same take that ends part 4. It was restaged at the end of part 4's recording for playin at the start of part 5's, with a hold for the new episode's opening captions to be superimposed). The wow is not particularly noticable on the dialogue, so I got away with it! But when the first music cue comes in over the episode titles caption, it's terrible. So, as I happen to have a copy of the music masters from this story in the archive (I'm using them on another project ;-)), I relaid that music cue from the master, dirtying it up a bit to match the original, and rebuilding the effects behind it.
Out of the music cue, I decided to take the next couple of scenes from the David Holman off-air recording (I couldn't take the music cue from here, as it's very distorted due to David still setting the record level). This avoids the unwanted pitch modulation on the beeping of the disseminator (it should be a constant pulse, but on the 16mm it wavers up and down by +/- a semitone), and on the next music cue, which starts with a flute mimicking the beeping. The recording was carefully repitched to match the original so that it will maintain picture sync should we ever copy it back to the video master. I also used a rather complicated equalisation curve to remove the scan coil whine from the TV tube (this is a mic recording), reduce the mains hum, and match the characteristic to the 16mm track. As Sara enters the disseminator and points her gun at the Doctor and Steven, I cross back to the 16mm for the rest of the episode, the fault being cleared by now. The final thing to be seen to here was the fact that the transitions into and out of the David Holman recording were very obvious, it being an off-air copy of the original magnetic soundtrack. I solved this by adding a loop of optical film noise behind the entire clip - a bit brute-force, but it works!
The rest of the cleanup was pretty easy, but laborious. I spent around 6 hours drawing optical pops out of "Counterplot", and 2-3 hours working on "Escape Switch", it being a lot cleaner to start with. I also corrected a couple of miscued sound effects which would otherwise confuse the audio-only audience. The complete episodes were then treated to a moderate broadband noise reduction process to take the background noise down a bit.
All in all, an average of about a long day per episode. Not bad.
Now - Mission to the Unknown awaits!
Diary Entry 3
Mission to the Unknown is, as we all know, a one-off Doctor-and-companions-less Doctor Who episode bridging the gap between Galaxy 4 and The Myth Makers. It acts as a scene-setter for the epic 12-part The Daleks' Master Plan which follows The Myth Makers, and will form a single "bonus" CD as part of the forthcoming boxed set of The Daleks Master Plan on audio.
I have three recordings to choose from here. Allen Wilson's is really out of contention, as it is far too roomy. But David Butler's and David Holman's recordings each have their own advantages and disadvantages. The Butler recording is much "closer", warmer, but somewhat crushed, while the Holman tape is more open (more of the listening room is apparent) but contains a subjectively better frequency response - but also a lot of dropout and quite a bit of cross-channel interference. In the end, the decision as to which one to use will come down to which one cleans up better...
First job, though, is to correct the recordings to duration, due to the fact that no two tape recorders run at exactly the same speed. Without a picture reference, of course, it is impossible to get better then +/- 2-3 seconds for the length. We have the BBC official timings but, believe me, these can themselves be out by a second or so either way.
The David Butler recording has complete opening and closing titles present so can be used as a reference. It runs to around 23.50. We know from the BBC archive (and with thanks to Andrew Pixley and David Brunt whose reference books prove, as ever, a godsend) that the transmitted duration was 24.42. So a bit of digital varyspeed was employed to correct the duration of each recording. This is a fine adjustement - a pitch difference of just 1 cent (one one-hundredth of a semitone) can mean a difference in playing time of nearly a second over the length of a Doctor Who episode. All sorts of complicated mechanical and electronic devices have been created to keep separate media elements in sync, but the offair recordings contain none of the relevent codes and sprockets, so it's always going to be a guestimate.
There is another problem, and that is that not only do no two tape recorders run at exactly the same speed, but nor did most domestic tape machines of the 1960's run at a constant speed within themselves. As one tape spool fills up and the other empties, it puts differing strains on the mechanics and causes the machine to speed up or slow down (professional tape machines have very complicated tension-balancing mechanisms to avoid this). Also, a slight variation in mains frequency or voltage (very common, even now) can cause the machine to change speed (most of today's machines use internal crystal or feedback-loop locking). It transpired that the Allen Wilson recording of Mission exhibits this problem particularly - starting at a considerably lower pitch than it finishes - and the Holman/Butler recordings are not immune. If the length-corrected episodes were to be offered up against any future rediscovered prints, the sync would drift in and out by a few frames during the course of the episodes.
Having adjusted the two recordings to the correct length, I placed them on adjacent tracks in ProTools, and started to throw plugins at them to see which one would form my basic master. Both started to clean up very well, but it soon became clear that in all respects except one, the David Butler recording was ahead: it was warmer, closer, and pretty much clear of the dropouts that plague David Holman's tape. It also suffers none of the interference from other sources that Holman's tape contains. The downside is that it is rather muffled - sibilant sounds ("s" and "t") are unclear, and no amount of equalisation can resurrect sounds which are simply not recorded on the tape. Nevertheless, the decision was made.
With the master recording chosen, I replaced the opening and closing titles music (rebuilding some effects over the transition out of the opening). Then I manually repaired any dropouts in the recording (by patching in sound from elswhere to fill any gaps), drew out some clicks and pops, and tidied up a couple of poor edits (these would have been present on the original transmission, but are distracting when listening audio-only). Around 15 separate bands of equalisation were applied, both for noise removal (hum and tube whine) and to closer match the original transmission characteristic. Then some light broad-band noise reduction. Nearly there.
The lack of high frequency
response in the recording was still bothering me, so I decided to try to take advantage of
the fact that the Holman recording is brighter. Using the basic restoration parameters I
had already applied (noise removal and so on) I further heavily equalised the Holman
recording to remove all but the sibilant sounds, and synchronised it with the Butler tape.
Bingo! In fact, due to the sync-drifting effect described above, it was necessary to
resync the two recordings line-by-line - about a day's work. Some frequency-conscious
gating was the final touch which helped blend the two sources together.
The lack of high frequency response in the recording was still bothering me, so I decided to try to take advantage of the fact that the Holman recording is brighter. Using the basic restoration parameters I had already applied (noise removal and so on) I further heavily equalised the Holman recording to remove all but the sibilant sounds, and synchronised it with the Butler tape. Bingo! In fact, due to the sync-drifting effect described above, it was necessary to resync the two recordings line-by-line - about a day's work. Some frequency-conscious gating was the final touch which helped blend the two sources together.
The final stage was to bounce the restored recording down to a single audio file for later use.
PS. I was asked why I didn't use the Spectral Extraction function in SoundHack to isolate the transients in the David Holman recording. There are a number of reasons for this. Firstly, SoundHack (while very useful) is an offline processor. This means that you have to set the parameters and let it crunch the numbers. Spectral extraction on a single episode could take up to 6 hours (OK - faster than this if I did it on the G3 processor, but still time-consuming!). Secondly, it's very much trial-and-error: you have to set the parameters, run a bit, listen to it, adjust the parameters, run a bit, and so on. And because the transients on these recordings are so smeared anyway, it would take hours to get an acceptable result. As I said, the EQ method worked, and was quicker (either way, I'd still have had the sync problems).
PPS - additional notes from Technical Forum discussions on detecting duration:
I have found with 625 line off-air audios that you can often detect recorded line scan frequency with a frequency analyser - I used the built on one within Cool Edit. It's normally pretty attenuated but you can usually see a spike.
Very true. I've also used this method before (and by detecting the frequency of the mains hum!).
By comparing the frequency of the peak of spike with the expected line scan rate you can work out the correction to be applied to bring the recording back to the correct speed.
Correct running duration = playback duration * (detected line frequency / expected line frequency)
With recordings from 405 line TX, the line scan frequency is a lot lower (10.125kHz) so it should be easier to detect.
It is indeed - and I use a deep notch at line frequency to remove the scan whine from the recordings anyway. But there are a number of factors which make this process as hit-and-miss as any other. Firstly, there's the drift-in-tapespeed problem. Whereabouts in the tape do I sample the line frequency to get the most accurate result? (I frequently have to dynamically move the eq notch - automated in ProTools - to track the change in tape speed and keep locked onto the line frequency during the course of an episode!). In addition, the flutter in the recordings means that the line frequency meanders anyway. As I pointed out in the "diary", a difference of a single cent in pitch terms (1 cent = 1/100th of the semitone) can mean a difference in duration of 1 second over the length of a Dr Who episode; 1 cent is the equivalent of around 1/4Hz - so it's all very fine!
PPPS - Further to above
I suppose the best way is to divide the episode up into sections and calculate a compensation to be applied to each section. By getting the line frequency correct throughout the episode you would be assured of the correct duration and also correct content pitch. This would also then avoid the need to dynamically move the eq notch.
Well, yes and no. The speed
is varying all the time - it's not sudden changes, but a steady variation - so where do
you cut the sections? What would be needed is a constant dynamic tracking of the pitch and
constant correction. Interestingly, the technology is becoming available to do this...but
you'd still need a stable reference.
Well, yes and no. The speed is varying all the time - it's not sudden changes, but a steady variation - so where do you cut the sections? What would be needed is a constant dynamic tracking of the pitch and constant correction. Interestingly, the technology is becoming available to do this...but you'd still need a stable reference.
Agreed that flutter is a problem. Perhaps the answer would be some sort of automated dynamic compensator which constantly tracked line frequency, but this might be tricky to integrate.
Yes, as mentioned above.
Yes, as mentioned above.
Maybe it's possible to create some sort of macro to automate the process I explained above and take it to it's limit - with the episode divided into very short sections. All this would take a lot of time and effort though!
Indeed. You can *always* get
it better. But there comes a point where life is too short (and budget is too small!).
Indeed. You can *always* get it better. But there comes a point where life is too short (and budget is too small!).
At least with line frequency cross-talk it's constantly present and provides a crystal locked reference that you could use if you wanted to go to such ends. Mains hum could also be useful as it is relatively stable over periods, but the absolute change in pitch is smaller than at line frequency so differences are more difficult to detect. A 1 second change in overall episode length corresponds to a 0.0347 Hz shift at mains freq but a 7.03125 Hz shift at 405line freq.
Indeed - nice to see that
someone else understands the math!
Indeed - nice to see that someone else understands the math!
Diary Entry 4
Just a brief entry this one. My
previous post on the sources available for restoration of "The Daleks' Master
Plan" omitted the Richard Landen recordings - for the simple reason that I had not,
at that stage, obtained copies.
I now have copies of Richard's recordings of episodes 3 (Devils Planet) and 11 (The Abandoned Planet). These will perhaps be of use in patching the problems evident in the David Holman recordings of these episodes.
One of the main points of interest in Richard's recordings is that they are, already, copies. When Richard taped the episodes off air, he did as other recordists did, and manually edited the intervening titles and cliffhanger reprises from the recordings. He later regretted this, and copied all of his episodes down, reinserting the titles and rebuilding, as best he could, the cliffhangers. This effort was not particularly successful - the copies are not very good in themselves, and the edits are, again, pause-control efforts. This contributes to Richard's recordings being, sadly, the poor relations here - very noisy and lacking in dynamic range and frequency response. Nevertheless, I have often used his tapes for patching (The Highlanders and The Celestial Toymaker particularly), so they are a welcome addition to the store.
Diary Entry 5
With The Daleks' Master Plan episodes 5 & 10 (taken mainly from the film recordings) and "Mission to the Unknown" out of the way, it is now time to tackle the bulk of the story. Episodes 1-4, 6, 7, & 11 will be taken mainly from the David Holman recordings, so it makes sense to do these in one large batch. Here, the David Holman recordings are backed up by a complete set of Allen Wilson tapes, and Richard Landen recordings of episodes 3 & 11.
First job, again, is to get the
episodes to length. As has been discussed before, there are a number of ways to tackle
this. For the film recordings, there is not a problem as they are taken directly from a
locked source. For Mission to the Unknown, I had a complete offair recording including
opening and closing titles, plus the BBC offical timing, and a comparison of these yealded
a reference for varyspeeding the offair recording back to the correct length.
For the episodes under consideration here, there are a number of problems. Some of the Allen Wilson recordings have complete opening and closing titles, but some don't. The integral speed of them can vary quite a lot, and a very quick spectrum analysis showed that they are all running at different average speeds; similarly with the Holman recordings, though not as bad. So I decided to resort to the horizontal line-frequency analysis method I had rejected for Mission. Due to the flutter (small pitch/speed variations) and numerous small dropouts present in these recordings, it is actually quite difficult to get a decent reading of the scan frequency, which is why I tend to use the length-comparison method more often - simply it averages better over duration.
Some of you might be wondering quite what I mean by the "horizontal line-frequency analysis method". Simply, the scan coils in the TV have to scan each individual line that makes up a TV picture by deflecting an electron beam across the phosphors on the TV tube, making them glow. As the coils deflect the beam, they emit waste energy which is in the audible range - a high-pitched whine. The scan frequency for 625-line, 25-frame PAL will be 625*25 = 15,625Hz. For 405-line, 25-frame black-and-white TV (as used in the UK until the late-1960's) it is 405*25 = 10,125Hz. Our ears tend to tune it out after a while, but the offair audio recordings often pick it up "beautifully". By analysing the frequency of the whine on the offair recordings, we can compare it to what it should be, then varyspeed the recording to correct it. (Eventually, I always filter it out if the recordings when doing the remastering).
So for these episodes, I chose quiet sections of the audio and heavily filtered them so as to remove everything below about 9kHz. To get a decent lock it was also necessary to heavily boost everything *above* around 9kHz. The filtered and boosted scan whine was then sent to a frequency counter. Despite the fact that the frequency would only register to at best +/- around 15-20Hz, it was possible to get an approximate reading. I took a number of readings across each episode. If they only varied by a few Hz, I took an average, and used this to calculate the required varyspeed in cents (100ths of a semitone). This ranged from -35 to +23 cents. The files were then processed.
My ears told me, however, that the Holman recordings of episodes 1 & 6, and the Allen Wilson recording of episode 11, had more severe problems, being that the episodes change in pitch over time. For the Wilson recording, the episode rose in pitch by nearly a semitone over its full length. I sampled the line frequency at a number of points in each episode and graphed the results to make sure that the pitch change was broadly linear. In each case (bearing in mind the tolerances of the readings!) it was. Here, a simple pitch change would not work - the pitch correction had to vary in a linear fashion over the length of the episode.
This is actually quite easy
to do. I opened the episodes in a little utility called SoundHack, entered the starting
pitch offset, the ending pitch offset, and then got SoundHack to draw a straight line
between the two, showing the pitch variation needed at each point in the file. Then I hit
"process" and went off to have a couple of cups of coffee. Processing each file
took about an hour. (I must have got it right, as my final restoration of "The
Nightmare Begins" - more in the next diary - was exactly on the nose to the reported
This is actually quite easy to do. I opened the episodes in a little utility called SoundHack, entered the starting pitch offset, the ending pitch offset, and then got SoundHack to draw a straight line between the two, showing the pitch variation needed at each point in the file. Then I hit "process" and went off to have a couple of cups of coffee. Processing each file took about an hour. (I must have got it right, as my final restoration of "The Nightmare Begins" - more in the next diary - was exactly on the nose to the reported transmitted duration.)
(Incidentally, processing only took about an hour because I bumped the files across to my G3 PowerBook for processing - my aging desktop Mac would have taken 6-7 hours! For anyone who is interested, SoundHack is a great swiss-army knife for sound designers. For you PC-types, tough, it only runs on real computers (Macs!), but for us Mac-users there is an additional bonus - SoundHack is free (www.soundhack.com)).
The last files to be corrected were the Richard Landen recordings. These being of lesser fidelity, the scan whine had not actually been captured in the tapes, so they were adjusted by comparing a representative section from each with the same section from one of the corrected Holman tapes.
So, now to the restoration proper!
PS - on the frequency
Wow! Now that is clever -
using one of the things you want to get rid of to aid the very process of getting rid of
it! A solution worthy of the Doctor himself!
A solution worthy of the Doctor himself!
One hopes so.
One hopes so.
Did you invent this technique yourself?
Well, I suppose so -
although it's obvious, as someone else on this forum also suggested it before I got around
to describing it.
Well, I suppose so - although it's obvious, as someone else on this forum also suggested it before I got around to describing it.
Come to think of it, how many of your tecniques are original/heavily refined versions of pre-existing ones? Are you steadily inventing/discovering?
I've been doing this kind of
thing for a few years now, so one picks techniques up and refines them.
I've been doing this kind of thing for a few years now, so one picks techniques up and refines them.
I used a similar technique on the final episode of "The Massacre". Here, the volume level of the episode rises and falls on the tape, and this needed counteracting. So I had to find some kind of reference. What I did was to heavily filter the audio so that I was left with just the mains hum inherent in the recording. I then fed this to the side chain of a compressor set to track in inverse - everytime the mains hum decreased in volume, the compressor increased the gain to bring it back up again. The main audio was then fed through the signal path of the compressor and post-filtered to remove the mains hum. It wasn't perfect, but went a long way to making the episode half-listenable!
Boy, am I giving away some secrets, here!
I'm glad to see you're giving the line scan method a try with this one. Does the extra effort involved pay off?
Yes, and no. As I've said
before, I only use it if I have to - i.e. if I don't have an absolute length reference I
Yes, and no. As I've said before, I only use it if I have to - i.e. if I don't have an absolute length reference I can trust.
I'm sure the duration will now be within a second or two of the original.
Actually, my episode 1
turned out on the nose (and this was the one I had to do the sliding pitch correction
on!). My episodes 2 & 3 (using the line scan frequency method) are about 4 seconds
under, and 1 second over respectively. Now, I could adjust for this, and rerun the
varyspeed, but frankly, that's close enough, and life's too short - especially seeing as
one can't be 100% sure that the BBC reported timing is itself correct!
Actually, my episode 1 turned out on the nose (and this was the one I had to do the sliding pitch correction on!). My episodes 2 & 3 (using the line scan frequency method) are about 4 seconds under, and 1 second over respectively. Now, I could adjust for this, and rerun the varyspeed, but frankly, that's close enough, and life's too short - especially seeing as one can't be 100% sure that the BBC reported timing is itself correct!
I guess it is an obvious method - it just needs someone to have inspiration to think of it!
As I say, I've used the
method before, but I do not consider it to be more accurate than any other, simply because
of the difficulty of getting a stable frequency reading from these old tapes - it's always
going to be +/- anything up to 50Hz or so averaged across the total duration.
As I say, I've used the method before, but I do not consider it to be more accurate than any other, simply because of the difficulty of getting a stable frequency reading from these old tapes - it's always going to be +/- anything up to 50Hz or so averaged across the total duration.
Have you still had to move the position of the eq notch around within the episode? What's the typical sort of variation that you've found in line freq within an episode?
It can hover up and down by
anything up to (as I say above) 50Hz or so. On these, I've managed to even out the playing
speed pretty well from the start (so far!) - and I've only occasionally had to move the
notch a few Hz to catch a slight tape snag.
It can hover up and down by anything up to (as I say above) 50Hz or so. On these, I've managed to even out the playing speed pretty well from the start (so far!) - and I've only occasionally had to move the notch a few Hz to catch a slight tape snag.
Diary Entry 6
Restoration of Episodes 1-4, 6, 7 & 11 continues...those of you wondering why there has been such a gap between the last diary entry and this one should now have a greater idea of the amount of work involved here!!!
As described previously, all episodes were examined with the audio equivalent of a fine-toothed comb, clicks removed, and dropouts repaired where possible. Notch filters were employed to remove scan whine and mains hum where necessary, some roll off at each end of the frequency spectrum helped with hiss and rumble, and a general eq curve closer matched the sound to the original recording. Opening and closing titles music was replaced from the masters. The following will concentrate on particular problem areas with each episode.
Episode One - The Nightmare Begins.
The very start of this episode presented the first challenge. The commonly- heard Holman recording is slightly clipped at the start, and missing the opening titles, a couple of seconds of effects, and the first note of the opening music cue. The Allen Wilson recording is complete, but of much poorer quality. I rebuilt the opening using a combination of these two recordings, plus a copy of the opening music cue taken from Tristram Cary's masters.
One line from the character Lizan ("I'm going to do some work now, even if you aren't"), had some bad interference over it on my copy of the Holman recording. The Wilson recording was unsuitable at this point for use as a patch (it is quite muffled, noisy, and the change of ambience would be too distracting). So I manually cleaned up the line as best I could - it wasn't perfect, and the last two words were still slightly lost. A day or so after after I'd done this patch, I decided to have a listen to the (ahem) telesnap reconstruction of this episode (thanks Robert Franks) to see how they had got around the problem. I was surprised to find that the distortion was not present at all - either they've done a *very* good job fixing it, or they had a clean copy to start with. Assuming the latter, it seems that my dub of the Holman tapes may have some additional problems. Cheekily, I nicked that line from the TS copy, and heavily treated it to match my own restoration (I counteracted some of their over-brightening of the sound, and ran some hum removal on the clip) - I hope they don't mind! I also hope this problem will not reoccur.
This episode (and parts of others following) also suffers from some distant cross-channel interference. I edited around this as much as possible. There is also a sudden rise in level by around 5dB towards the end, after which there is a fair amount of distortion present. The gain rise was counteracted, but there is little that can be done about the distortion, sadly.
I rebuilt the cliffhanger to its original form using the Wilson recording as reference. Note there was no music "sting" going into the closing titles - this only features on the reprise in episode 2. Neither is there the whispered line "Steven...Katarina" from the Doctor, as he sees the Daleks surrounding the TARDIS - it is in the camera script, but not said. (Note to Andrew Pixley - the closing titles are actually 52" long).
Episode Two - Day of Armageddon
We have no complete off-air reference for the cliffhanger reprise at the start of this episode, but it is obvious from the script that very little is missing. I have rebuilt it as I think it would have been - and all that is audible is the Doctor moving through the forest toward the TARDIS. The closing cliffhanger (Brett about to launch the Spar, leaving the Doctor behind) is intact on the Holman recording, so I used this as a pattern and rebuilt it so as to lead in to the replaced closing titles. (The closing titles are actually around 35" long).
Episode Three - Devil's Planet
The opening is intact on the Holman recording, so replacing the opening titles was an easy matter. The closing cliffhanger (Kirksen grabbing Katarina) is slightly cut on both the Holman and Landen recordings. I rebuilt it to its original form using the intact Wilson recording. (The closing titles are around 35" long).
Scene 3, set in the Dalek control room, is a conversation among Daleks. The Holman recording suffers from bad interference here, and some severe tape damage causing the loss of part of the dialogue. I decided to replace this entire section from the Richard Landen recording, as replacing just the problem words would have been distracting. To avoid too much of a tonal shift, I removed the relevant section from the Holman recording but looped the control room background to fill the gap. The Dalek voices were then filtered to remove the background, and dropped in over the Holman loop.
The following scenes, and many throughout this episode, are smothered in similar dropouts and tape damage. Dropouts were repaired as and when possible by filling the gaps with bits of audio taken from elsewhere (this was great fun - not - where scenes were set in the Dalek control room, as the rhythm has to be kept consistent). Areas of tape damage were replaced with the corresponding sections from the Landen or Wilson recording, depending on which was the better at each point (I found generally that the Dalek voices were better from the Landen, human dialogue from the Wilson). This was done again by cutting out the offending section, rebuilding the background behind, then dropping the replacement in over the top, eq'ing it so as to match as closely as possible. Where possible, I crossfade between the original audio and the patch so as to disguise the join. (I used similar techniques in part two of The Celestial Toyroom, which also suffers from tape damage causing sections to be missing).
The other major problem with this episode is that it is beset by some severe crackling for lengthy periods. Clicks are fairly easy to remove, generally being quite well-defined and very short. Crackle is somewhat different, being rather more drastic distortion of the waveform over greater periods of time. I threw everything I could at this, to no avail, so I have just cut around it as much as possible, using patches from the Wilson and Landen recordings to cover where appropriate, and reducing the affected areas to less than half. I could probably have removed it all by hand eventually - but I have other things to do over the next few months! I notice that Michael Palmer (who did the telesnap version of this episode) removed the crackles by a combination of drawing them out and heavy equalisation. But I feel that the side effects of this (throwing out most of the wanted signal as well, the fluctuations in background noise level and so on) are worse than the problem they are trying to cure - Palmer's version is, to my ears, unpleasantly peaky with a very narrow frequency range, and very tiring to listen to. I suspect that the only devices that would help here would be the Cedar modules - these (Dehiss, Debuzz, Dethump, Declick, Decrackle) would make my life considerably easier generally but, sadly, their use on these projects is beyond our budget. (I did discuss this with a colleague who also works in remastering. He agreed that reducing the crackles even marginally, and even using Cedar, would be a time-consuming, expensive, and not totally successful operation - there are some problems which just won't fix). The only other alternative would have been to have used the Allen Wilson or Richard Landen recordings as the basic masters, or to use far more of these recordings as patches but, believe me, I think it's better generally to put up with the crackles!
Episode Four - The Traitors
Once again, the very opening of the episode presented the first challenge. The cliffhanger reprise was re-performed in studio for the continuing scene with slightly different dialogue and effects, and an added music cue for the episode title. (The episode opens on the rising roar of the Spar's engines, a music cue starting after a few seconds over the titles, then the dialogue starting a few seconds later still). The David Holman recording lacks the reprise, instead offering a hard cut into the new material at the end of the previous episode. The Allen Wilson recording is complete, but of quite poor quality, especially over the opening effects and music as Allen settles down in front of the TV! At this stage, I do not have Sue Cowley's linking script in front of me (she's late delivering!), and I felt that I might have fairly sparse narration over this opening sequence. (This is confirmed when, later the same evening - Monday 14th May - I checked my email and found Sue's script waiting for me). I was also aware that this will be the first episode on the third CD of the set, so the quality is a worry. Obviously, I had to use the Allen Wilson recording up until the point where the Holman recording catches up, but I decided to rebuild the very opening using the original recording of the music cue from Tristram Cary's master, and the original sound effect of the Spar taking off from the effects master. Tricky - but successful.
A few minutes into the episode, after the Doctor's eulogy over Katarina, there is a 33-second telecine sequence of Katarina and Kirksen's bodies floating away into space, accompanied solely by a piece of electronic wonder from Tristram. Sue has suggested leaving this narration-less, and I agree. But, although this sequence *should* be accompanied solely by Tristram's music, a fair amount of dropout and transmission noise has hitched a ride. So I re-laid this from the music master. The fact that it is suddenly so clean makes it more poignant, somehow!
I also replaced a later music cue at the start of the scene where Chen is examining the map in his office, rebuilding the studio background noise behind it. Interestingly, a big dropout in the music at this point is also on the music master - so the tape damage here happened very early on and was reproduced faithfully later; as it's electronic music, I offset it against itself so that it filled in its own dropouts!
Much of this episode it beset by tiny electrical clicks - a tedious 10 hours was spent removing them.
I removed an off-camera cough (it's present on both Holman and Wilson recordings, presumably from a member of the crew, or cast on another set) from the scene with Chen, Lizon and Karlton.
The ending of this episode is clipped on both extant recordings, so one has to make an educated guess, using available clues, as to precisely how the cliffhanger worked. The (ahem) telesnaps guys seem to think that the music cue that starts part 5 was also present at the end of part 4; I disagree. Other episodes in this serial end on fairly downbeat cliffhangers, with no music cue - music is added during the recap at the start of the next episode for the titles. Also, Holman and Wilson tended to cue the tape to the very end of the episode, just before the theme music comes in, ready to record the next episode - and there are no music cues before the edit on either recording. One last clue - the music cue that starts episode five is on episode five's music master, not at the end of episode four's - this would suggest that it was written for the later episode. In the script, the next episode caption is superimposed over the shot of Sara, immediately after her "...and aim for the head" line; there would have been neither need nor time for a music cue here, so I am assuming that the theme music started immediately, and that's how I've rebuilt it.
(An aside - the cliffhanger ending of this episode was played in to the start of the next episode's recording from 35mm film - but is not actually the same take that ends part four. The cliffhanger was immediately restaged at the end of part four's recording for playin at the start of part five's, with a hold for the new episode's opening captions to be superimposed in place of the mix to the next episode caption and closing roller. This technique was also used on other episodes.)
Episode Six - Coronas of the Sun
The opening cliffhanger reprise again presents the first problem. Once again it consists of a film playin of a retake of the end of the previous episode. The Holman recording is missing the reprise; the Wilson recording is complete, including the opening titles, but of poor quality. The actual lines said are identical in both takes, but the pauses between lines are slightly longer on the reprise, and the intonation is slightly different. I felt it would be ridiculous to use the lesser quality Wilson dialogue, so used the Holman recording of the end of part 5, but recut it so as to match the timings of the reprise. I finished by mixing in some of the effects from the Wilson recording (so that you can hear the rumble of the Dalek casings at the start of the episode). Again, I replaced the opening incidental music cue and rebuilt the jungle atmosphere behind (this due to dropout and clicks on the Holman recording - there was little point in repairing them when I have the original materials to hand).
One other music section was also repaired using the masters, the Doctor approaching the sentry at the Dalek ship.
The Holman recording of this episode suffers from severe distortion at about 11.20 for about 1.12. This appears to be due to the recording level on the tape machine being suddenly increased for reasons unknown. There is very little that can be done about this. I have patched the moment where the distortion first comes in with the relevant line (the Doctor's "Come on my dear, let's get on with our mineral while we've still got the time") from the Wilson recording, and the perspective change works as the Doctor is moving away at that moment anyway. But to hold the Wilson recording for the entire period of the distortion would be too much - it is very distant, and rather muffled and indistinct. So I have level-balanced and cleaned up the bulk of the distorted section and left it as was. As with episode three's crackles, some things you just have to live with.
The rest of this episode was relatively clean, bar some occasional rustling distortion and some low frequency tape-head bumps. These were cut around as much as possible, the thumps being largely taken care of by the low frequency rumble filter.
The end of the episode is intact on both the Holman and Wilson recordings, which gave me an easy reference for replacing the titles correctly.
My total duration turned out to be just 2" more than the reported original duration. (The closing titles are actually 31" - not 25" as previously reported). This episode was one of the quickest restorations so far - 1 day!
P.S. - at this point, I rang Sue Cowley to ask if she has worked out quite why this episode is called "Coronas of the Sun"? There are no Coronas, and no sun...she doesn't know. She asks the assembled throng of the technical forum. They don't know either - there are some inventive replies, but the consensus seems to be that the title is just a load of dingoes kidneys.
Episode Seven - The Feast of Steven
The opening of the episode is intact on the Holman recording, so no problem there. Generally, there are were many little clicks to remove, and some unpleasant peak distortion to try to control.
This illustrates one of the main problems with these recordings: in normal conversation, the occasional peak is not too troublesome, as any distortion produced is of small duration and relatively easily controlled. Yet there are scenes here which consist entirely of hysterical shouting - the pictures would undoubtedly aid comprehension, but it can be very difficult to make out what is being said in the melee at times. This problem is quite common in battle scenes in early Dr Who, and the dance floor sequence in "The Celestial Toyroom" was a similar nightmare. It is very difficult to remove this distortion, but it can occasionally be softened by equalisation or by redrawing the most troublesome areas. A quick listen to the Wilson recording (generally unusable due to its over-openness) suggests that at least some of the distortion at these points was introduced in the transmission chain, and merely exacerbated by the recording - and the waveform looks remarkably clean considering the audible peakiness, again suggesting that it is ingrained, rather than a unique feature of this copy. I suspect, though, that we are going to have to add extra narration to this episode to explain what is being said (and to cover some of the worst distortion, which is really painful to listen to) - and this episode, being largely visual anyway, will already use a lot of narration.
I removed a bit of pre-echo "print-through" of Blossom Lefavre's scream so that it is the surprise it is meant to be - previously you heard it faintly before it happened. This was done by covering the faint scream with a bit of TARDIS noise pasted in from elsewhere.
The ending was intact (bar a couple of seconds of TARDIS background) on the Holman recording; I rebuilt it to length using the Wilson recording as reference.
Episode Eleven - The Abandoned Planet
The opening is intact on both the Holman and Wilson recordings (though not on the Landen), so I am curious as to why the (ahem) Telesnaps folk used the ending of part ten instead...(the reprise was a telecine playin of a retake of the end of the previous episode, so it is not - Hartnell's performance especially - the same).
Nevertheless, the opening is messy, with a nasty edit from the titles into the film recording. Although this is as originally transmitted, I have tidied it up by adding some studio noise and TARDIS background to the transition. One small part of the reprise (as the Escape Switch is pressed) was patched from the Wilson recording (due to varying recording level at that point on the Holman). I then replaced the music cue over the episode titles (a repeat of the Katarina's death music from "The Traitors"), leading into the landing of the Dalek timeship.
As with Episode Three (Devil's Planet), there were many areas of severe dropout caused by irreparable tape damage. In addition, there are areas where there are varying levels and severe over-modulation (as with Episode Six - Coronas of the Sun). These were patched using similar techniques as employed on the earlier episodes.
Two particular problem moments were the death of Gearon, and part of the sequence of Steven and Sara trying to locate the TARDIS in the jungle. In both cases, distortion made the Holman recording unusable, yet neither the Landen nor the Wilson recordings were that inspiring either. The former was patched with sections taken from each, making the best of a bad job, the latter using a heavily-eq'd patch from the Landen, overlaid with some additional jungle effects from the effects master to ease transitions. I'll also cover both sections with some additional narration to distract!
At times, the recording level actually drops considerably as well; counteracting this inevitably leads to a higher noise floor, but I felt that the Holman recording was still the best in any event. Even with the occasional rises in noise floor, I have still managed to keep it to an average of about -40db, around what it would have been originally, so I'm not too disheartened.
For reasons of clarity, I corrected one miscued sound effect towards the end (the roar of the delegates' spacecraft engines comes in early).
The ending of the episode is truncated on the Holman recording, slightly truncated on the Landen, but complete on the Wilson. I rebuilt it to match the original using a combination of all three recordings and some original Varga jungle effects.
With the basic restoration
process completed for all these episode, fine-tuning of the noise reduction process and eq
was performed for each episode one-by-one, checking levels (especially the relative levels
over patched sections), and bouncing them down to final master files for further use. This
took a further day-and-a-half. All makeup materials are also archived in case of
With the basic restoration process completed for all these episode, fine-tuning of the noise reduction process and eq was performed for each episode one-by-one, checking levels (especially the relative levels over patched sections), and bouncing them down to final master files for further use. This took a further day-and-a-half. All makeup materials are also archived in case of disasters.
Now I have to start work on the final batch of three episodes: 8, 9 and 12, from the Graham Strong tapes.
Diary Entry 7
The final set of episodes to be
restored are those taken from Graham Strong's recordings: episodes 8, 9 and 12.
First off: getting the episodes to the correct pitch and length. Neither the David Butler recording of episode 8, nor any of the Graham Strong recordings, have audible line-scan frequency, so using the line-scan frequency detection method on these is impossible.
The Allen Wilson recording of episode 8 is problematic, as the speed is very unstable - the detected line scan frequency wanders up and down between 10020Hz and 10150Hz. Getting a stable lock is impossible, and trying to track this would be very time-consuming - and a waste of time considering that I do not intent to use this episode as a master. Episode 9 is similarly troublesome - detected line frequency here ranges from 9035Hz to 10080Hz; episode 12 hovers between 10090Hz and 10185Hz approximately. The best idea seems to be to forget using these as any kind of reference.
The David Butler recording of episode 8 (Volcano) is complete, including opening and closing titles. My transfer of it runs to 25.39. The published duration (with thanks to Brunt and Pixley) is 24.42 - a digital varispeed of +66 cents sorted this out. I then used this as a reference to bring the Graham Strong recording to the correct pitch: +11 cents.
Graham Strong's Episode 9 has the
start of the opening titles and the end of the closing titles clipped. I could do a
correct-by-length by taking the published durations of the opening and closing titles,
subtracting that from the published total episode length, then adjusting the duration of
the remainder to fit. Yet the music durations on cue sheets can be extremely inaccurate
(believe me, as a composer who relies on duration-calculated royalties for a proportion of
my income, I know!), and already I have found discrepancies of up to 10 seconds in
reported durations of themes on Master Plan episodes. The best way to make a guestimate
was to add the clipped bit to the front of the opening titles, extrapolate the closing
titles to an obvious fade point (they are just starting to fade as Graham's recording
ends) and use that as a reference. As a result, I adjusted this episode by +5 cents. Using
the same technique I adjusted episode 12 by + 14 cents.
My first disappointment was on listening again to the Graham Strong recordings which, in any event, are going to have to form my main masters for these three episodes. As I wrote before, the intelligibility of these recordings is generally very good - as I understand it these are the first "line" recordings (i.e. recordings made with the recorder linked directly electronically to the TV rather than via a microphone) Graham made of the programme. However, as I also noted previously, the recordings suffer from a fluttery dropout throughout, probably due to the quality of the recording media. The noise floor is very high, and was obviously going to take some work - I suspect that the recording level on the original tapes (remember that I am working from DATs made at the BBC from those originals) was very low. This fluttery dropout is present on many recordings from Graham (including "The Abominable Snowmen", where it also presented problems), but this was possibly the worst I've heard. A quick comparison with the tapes of "The Tomb of the Cybermen" (which as I write this have just arrived to help with the DVD remastering of that tale) showed them to be chalk and cheese - the "Tomb" recordings are very good. I'd be interested to know what tape stocks Graham used - I suspect that the difference in quality may be in part attributable to that.
Nevertheless, work must continue! Rebuilding the beginnings and ends of episodes was an easy task, as all are intact. I replaced all the titles music, and also replaced the TARDIS takeoff sound at the end of part 12 - we might as well end on a note that in some way approaches hifi!
The following details particular problem areas with each episode.
Episode Eight - Volcano
Problem One - most of this episode is beset by a regular click, once every second. As to how this got there, your guess is as good as mine. My heart sank when I realised this: the duration of the episode (minus titles) is around 23.26 - that's 1406 seconds, i.e. around 1406 clicks, plus other irregular clicks...
Interesting fact: at the beginning of this episode, set in the Dalek Control Room, the studio grams operator plays the TARDIS background in error - it fades out and into the Control Room background hoping no-one will notice. I did not try to correct this error.
Episode Nine - Golden Death
The cliffhanger reprise used the same Dalek voice takes as the episode 8 cliffhanger, so this section was copied up. I replaced the first music cue and the TARDIS materialisation from masters, also a later music cue building up to the arrival of the Monk's TARDIS, also replaced. Another music cue replaced was the comic cue marking the Monk's exit from his TARDIS. I again rebuilt the studio noise behind these patches.
Lots of clicks, lots of dropouts. I mixed in a tiny section of the Allen Wilson recording at one point to help fill in some deep dropout. The main problem was some rustling interference at times. Again, impossible to remove entirely - I just had to cut around it as much as possible, although careful setting of the noise reduction also helped enormously. This episode is also particularly noisy generally.
Episode 12 - The Destruction of Time
Last episode! And the quickest to restore so far! Some deep dropouts filled in where possible (including two by mixing in some of the Allen Wilson recording), and a fair number of clicks.
Again, with main restoration performed, each episode was double-checked, levels set, noise reduction parameters tweaked, and bounced down to a new file.
Restoration of all episodes is now complete, but this still leaves four stages of the production process to go: preparation of script for voiceover session, the voiceover session itself, editing and mixing of final voiceover with restored episodes, and final CD mastering. These stages will be described in forthcoming articles.
But before I get to the
first of those final stages, there is one other most important thing to do. All restored
episodes are backed up to CD-ROM, as are all restoration session files. The episodes are
also compressed as mp3 format filesand cut to CD-ROM as a double-backup.
But before I get to the first of those final stages, there is one other most important thing to do. All restored episodes are backed up to CD-ROM, as are all restoration session files. The episodes are also compressed as mp3 format filesand cut to CD-ROM as a double-backup.
Diary Entry 8
(Note that this article necessarily contains spoilers!).
With restoration of all 13
episodes complete, it's now time to prepare for the voiceover recording session.
With restoration of all 13 episodes complete, it's now time to prepare for the voiceover recording session.
Firstly, all episodes are
transferred to two 2Gb Jaz disks - Mission to the Unknown and Master Plan 1-6 on one,
Master Plan 7-12 on the other. The Jaz drive is disconnected from my desktop Macintosh and
reconnected to my PowerBook (Macintosh laptop). Having been using a full-on ProTools
system for which I paid a small fortune up until now, the next stage will be performed
using the *free* version of ProTools! This gives me all the editing and main mixing
features of the big system, but is limited to 8 tracks (no problem here!) and does not
have all the DSP (Digital Signal Processing) bells and whistles (again, they are done with
I make up three "sessions", based on the copied episode groups as above. These feature the episodes placed in a row with a few seconds between each one, and will form my final mix sessions after the voice recordings.
I then go through Sue
Cowley's linking script (which has been delivered as a Word file attached to an email),
placing "markers" in the ProTools files to mark the in and out points of each
section of voiceover. (ProTools only allows up to 200 markers in a file, so I have to work
in a number of 2-episode sub-sessions). If I think I will have to create a bit of space
for a voiceover passage in the final version, I mark it with an "*". Until now,
my concern has been technical - I haven't really cared what the Dalek Supreme is saying,
as long as there was not a dropout or click in the middle of his speech. But now I follow
the plot, reading along with the linking script, and my job slips sideways from being
mastering and restoration engineer, to being producer and script editor. Sue has done a
great job with the links, giving form to the story. I have to try to be that innocent set
of ears, hearing the story for the first time, finally dotting i's and crossing t's:
sometimes I may think that a piece of narration is unnecessary; sometimes I'll find myself
lost and feel the need to add a bit, or there'll be a long silence in the track that I
would like to leave a little less empty; sometimes I'll think "what does he look
like?", and I'll add a bit of extra description. Being able to look at a graphical
representation of the episode on the computer screen is a great help. Sometimes, of
course, a minor change in one place has knock on effects for other voiceovers. I also know
what words might trip Peter Purves up, and how fast or slow he can speak, so sometimes I
split passages up and move them around a bit to avoid having to add space and destroy the
rhythm of a scene. If I feel a major structural change is necessary (here, it's not) I'll
call Sue. Even after both of us have examined the story in this way, we can still miss
things (we did not describe the Rill in "Galaxy Four", we both missed pointing
out precisely where, when and how Steven is injured in "The Myth Makers", and
Michael Stevens and I enlisted Sergeant Rugg into the wrong service - mistakes happen).
I then go through Sue Cowley's linking script (which has been delivered as a Word file attached to an email), placing "markers" in the ProTools files to mark the in and out points of each section of voiceover. (ProTools only allows up to 200 markers in a file, so I have to work in a number of 2-episode sub-sessions). If I think I will have to create a bit of space for a voiceover passage in the final version, I mark it with an "*". Until now, my concern has been technical - I haven't really cared what the Dalek Supreme is saying, as long as there was not a dropout or click in the middle of his speech. But now I follow the plot, reading along with the linking script, and my job slips sideways from being mastering and restoration engineer, to being producer and script editor. Sue has done a great job with the links, giving form to the story. I have to try to be that innocent set of ears, hearing the story for the first time, finally dotting i's and crossing t's: sometimes I may think that a piece of narration is unnecessary; sometimes I'll find myself lost and feel the need to add a bit, or there'll be a long silence in the track that I would like to leave a little less empty; sometimes I'll think "what does he look like?", and I'll add a bit of extra description. Being able to look at a graphical representation of the episode on the computer screen is a great help. Sometimes, of course, a minor change in one place has knock on effects for other voiceovers. I also know what words might trip Peter Purves up, and how fast or slow he can speak, so sometimes I split passages up and move them around a bit to avoid having to add space and destroy the rhythm of a scene. If I feel a major structural change is necessary (here, it's not) I'll call Sue. Even after both of us have examined the story in this way, we can still miss things (we did not describe the Rill in "Galaxy Four", we both missed pointing out precisely where, when and how Steven is injured in "The Myth Makers", and Michael Stevens and I enlisted Sergeant Rugg into the wrong service - mistakes happen).
I pay particular attention to the cliffhangers. The transition between episodes three and four, as Kirksen stows away and attacks Katarina, worries me particularly here. Sue has, for the very good reason of wanting the cliffhanger to be a total shock, left any explanation until episode four. But I feel that by this time our audience will be thoroughly confused; they will not know how Kirksen got aboard the ship without going back and listening to the later half of episode three another couple of times. It's largely the fault of the soundtrack, which is not entirely clear, so some important lines get lost. Further, in order to put the explanation where Sue has placed it, I'll need to make a space in the track to avoid treading on some important lines - but making a gap here would destroy the hectic dynamic of this very dramatic scene. Part of the problem is that on audio we do not have the little visual hints that were on TV, so I've added a couple of narrative prods - hopefully it's still a surprise, but at least when it happens, you immediately get it, without it having to be explained afterwards.
When I get to episode three, there is a patch I made (around where the crew realise that they are being drawn to the planet Desperus), that sticks out like a sore thumb, suddenly. This concerns me, and then - with one of those sudden flashes of inspiration - I remember that I already restored this passage from film (it is one of the few clips from this story still existing) for the Ice Warriors "Missing Episodes" release. I dig it out of the archive, and paste a few seconds into my new master making, in effect, a 72 edit of that episode's restoration! Sometimes, you get so close to a project that the obvious solution does not immediately occur to you! This will have to be saved to a new backup in uncompressed and mp3 formats.
Now, to the voiceover
Now, to the voiceover recording...
Diary Entry 9
It's time for the voiceover recording. The most we've previously done in one day was eight episodes, when Peter Purves and I recorded "Galaxy Four" before lunch, and "The Myth Makers" (rather more slowly - the beef and beer pie was *very* nice!) afterwards. This time, with 13 episodes to do, we're going to spread the session over what we think will be two long days.
First, I pack up the gear:
1 Apple Macintosh Powerbook Computer
1 2GB Jaz Drive
1 JoeMeek JM-47 Meekrophone
1 JoeMeek VC3v2 ProChannel
1 Sony TCD-D3 Portable DAT Recorder
1 Boss 8-Channel Mixer
1 Microphone Stand
1 Popper Stopper
2 pairs of Headphones
...plus numerous cables to
connect it together and power it all, various mic clips, an AKG C-451 microphone as a
spare, a pile of DAT cassettes, the Jaz cartridges with the episodes on (mustn't forget
those), two copies of the linking script, and my copies of the camera scripts (in case we
suddenly lose the plot!).
...plus numerous cables to connect it together and power it all, various mic clips, an AKG C-451 microphone as a spare, a pile of DAT cassettes, the Jaz cartridges with the episodes on (mustn't forget those), two copies of the linking script, and my copies of the camera scripts (in case we suddenly lose the plot!).
Next, I kiss the wife and kids, load the gear into the car, and set out on the drive to Peter's.
On arrival I set up the equipment in Peter's upstairs office. This is quite a large room and away from the road (which is, thankfully, quiet anyway). This is the room in which we recorded the previously-mentioned stories. "The Massacre" was recorded at Peter's previous house, while "The Celestial Toyroom" was recorded in the drawing room downstairs. I prefer this upstairs room; the drawing room had a troublesome echo...
I've often been asked why we do these recordings "on location" rather than at a studio. It came about out of necessity. While we were setting up "The Massacre", Peter had an unfortunate accident which left him confined to a wheelchair for many weeks, making it very uncomfortable for him to make it to a studio. I was already thinking about a mobile recording solution anyway: even if we were to do the voiceover recording in a hired studio I would still be doing pre- and post-production in my own facility (sadly I do not have a voiceover booth) so would need to transport the session. The Powerbook-based studio was the obvious solution, and worked so well that we've stuck with it.
We place a comfortable high-backed office chair in the centre of the room for Peter, with a low table a few feet away for the equipment, in front of a comfortable sofa for the engineer! The Meekrophone is supported on a boom-stand in a shockmount, and protected by a mesh Popper Stopper (to stop explosive p's and b's "popping" the mic).
I chose the Meekrophone after a long search for a microphone which was both cost-effective (good microphones can cost thousands of pounds), and which also had a warm vintage sound to it which blended well with the original soundtracks. Its output is fed into the VC3 ProChannel, which adds some subtle compression and enhancement to the voice. The output of this is split two ways, to the audio input on the PowerBook and the input of the DAT machine. The DAT machine records the voiceovers "wild", purely as a back up - I haven't had to resort to it yet, but it's best to be safe. Yes, I use the inbuilt analogue-to-digital converters on the PowerBook. Some engineers would wince at this idea but, although I would not wish to record a symphony orchestra through these converters, they are fine for voice use.
The output of the PowerBook is fed, along with the output of the DAT machine, to the Boss mixer. This is used purely for monitoring, its "left" and "right" outputs feeding two pairs of headphones via custom cables. This allows me to monitor the original episode recordings alongside the recording of Peter's voice through the PowerBook, while Peter hears the original episodes from the PowerBook mixed with his voice looped through the DAT. There are two reasons for this - if Peter stops hearing his voice, he can warn me that the DAT tape has run out, and it avoids the slight monitoring delay that occurs when monitoring through the computer - fine for me, but off-putting for the artist.
Now, the session can start. We go through the episodes passage by passage, recording along to the soundtrack. If multiple takes are necessary, that is fine: I have two "keeper" tracks onto which I can drop previous takes. This way I keep all takes made. Occasionally we might make further changes to the script. Peter is very quick: at least 50% of the voiceover is first take. It takes some courage to accept this, and not to redo things for the sake of it but, with a break for a salad lunch, we do "Mission to the Unknown" and episodes 1-8 of "The Daleks'Master Plan" by about 6pm. At that point we call it a day and go out for dinner! Reconvening the following morning, we have the final four episodes completed by lunchtime. So much for expecting two long days...
With the recordings in the can (or
should that be "on the hard disc") I backup all the files to a Jaz cartridge,
pack everything up, and make my way home.
With the recordings in the can (or should that be "on the hard disc") I backup all the files to a Jaz cartridge, pack everything up, and make my way home.
Now, the edit!
Diary Entry 10
After the voiceover recording, all files are transferred back to the main Macintosh (and the fully-fledged ProTools system) from the PowerBook. Now, it's time for the final two stages of the production process.
I go through the episodes,
voiceover by voiceover, editing Peter Purves's takes to produce the final master. Most of
the time, what we've recorded fits exactly as intended. But, sometimes, I may feel that he
said a particular word, or groups of words, better in an alternative take, so I can cut
around and mix and match as I go through. I can also move voiceover takes around if I feel
they actually work better in a slightly different position. On a couple of occasions, I
even time-compress a passage to help it fit in the gap allotted to it. On others, I'll
make a bit of space for a voiceover by looping a bit of background atmosphere. There are
rules I impose about this: I can make space where there is just background atmosphere, but
I cannot lengthen or loop any music (this would have rights implications, for one thing),
nor will I interrupt the flow of a dramatic scene if making space would destroy the rhythm
of the dialogue - it is always important to maintain the integrity of the original
episode. As I work through, I also balance the voiceover against the soundtrack. I can
edit 3-4 episodes a day like this.
I go through the episodes, voiceover by voiceover, editing Peter Purves's takes to produce the final master. Most of the time, what we've recorded fits exactly as intended. But, sometimes, I may feel that he said a particular word, or groups of words, better in an alternative take, so I can cut around and mix and match as I go through. I can also move voiceover takes around if I feel they actually work better in a slightly different position. On a couple of occasions, I even time-compress a passage to help it fit in the gap allotted to it. On others, I'll make a bit of space for a voiceover by looping a bit of background atmosphere. There are rules I impose about this: I can make space where there is just background atmosphere, but I cannot lengthen or loop any music (this would have rights implications, for one thing), nor will I interrupt the flow of a dramatic scene if making space would destroy the rhythm of the dialogue - it is always important to maintain the integrity of the original episode. As I work through, I also balance the voiceover against the soundtrack. I can edit 3-4 episodes a day like this.
At the end of this process, I go out for a walk to clear my ears, then put the mixes up on the studio speakers at moderate volume and work through the episodes again, fine-tuning the mix, making sure that each voiceover can be heard - that each passage is neither too loud nor too quiet against the soundtrack. Then, once again, each episode is bounced down to a new final master file.
It is at this point that I really do feel, absolutely, that I am about to go mad. The masters are finished, but it is necessary to listen through to each one, checking for faults, and making final decisions about where to place the CD track IDs - these are dropped into the file as markers. With thirteen episodes to do, this takes the best part of another two days - and a lot of coffee.
Now, finally, I can make up the five Compact Disc masters. I use a program called WaveBurner for this (made by Emagic - and I even co-wrote the manual!). This allows me to see the track markers and indexes against the audio and make any final alterations. We work on the basis of each scene being marked by a track change - although I will often replace track IDs with CD "indexes" in the case of very short or interlinked scenes where separate tracks would confuse more than they help. The final tracklists are exported as text files from which I can derive PQ charts (timing information charts for the CD pressing plant), and tracklists to send to Michael Stevens for preparation of the booklet. PDF files are also prepared of the voiceover scripts and full tracklists. For technical reasons which I won't go into here (I have to retain some surprises), I had to export the CD master of disc one as an image file and cut that disc in an alternative program - Toast CD-DA.
Four copies of each disc are made: a master and safety for BBC Worldwide, a safety for my own archive, and a copy for Sue Cowley (a total of 20 discs - and each one must be checked). All makeup files are also archived to CD-ROM (this project has generated some 23 archive CD-ROMs!).
I now go and collapse in a gibbering heap...
...but there is no rest for the wicked. "The Tomb of the Cybermen" and "The Hitch Hikers Guide to the Galaxy" are vying for my attention, the next audio releases will soon appear on the horizon and, one day, I must prepare some more music releases and, even, get around to actually writing some music myself which - after all - is supposed to be my day job!
At the end of the process I of course spot things that I would like to change; you can never get things exactly right. But I think that "The Daleks' Master Plan" works very well - even episode seven, despite the irredeemable distortion.
Numerous thankyous are due, not all of which will find their way onto the CD packaging, so (alphabetically)...
For supply of the original
off-air recordings without which none of this would be possible: David Butler, David
Holman, Richard Landen, Graham Strong and Allen Wilson.
For supply of the original off-air recordings without which none of this would be possible: David Butler, David Holman, Richard Landen, Graham Strong and Allen Wilson.
For additional help, advice,
hard work or tolerance: David Brunt, Sue Cowley, Kathryn Evans, Robert D Franks, Piera
Johnson, Jason May, Michael Palmer, Andrew Pixley, Peter Purves, Steve Roberts, Michael
Stevens and Paul Vanezis. And of course, my wife Nicki and sons Adam and Samuel, and the
saner members of the RT forum for interesting discussions (at times!).
For additional help, advice, hard work or tolerance: David Brunt, Sue Cowley, Kathryn Evans, Robert D Franks, Piera Johnson, Jason May, Michael Palmer, Andrew Pixley, Peter Purves, Steve Roberts, Michael Stevens and Paul Vanezis. And of course, my wife Nicki and sons Adam and Samuel, and the saner members of the RT forum for interesting discussions (at times!).
Copyright Mark Ayres, October 2001