Dave Stewart: Creating A New Album From Archive Material

Dave Stewart and Barbara Gaskin (below) working at Spaceward Studios in the early '80s.

Dave Stewart's career has spanned several generations of music technology, and for his latest project he faced the challenge of bringing his old multitracks and MIDI sequences into the computer age.

About a year ago, I finished making an album. No big deal: practically every musician one runs into on the Internet seems to be in the same position. However, putting aside its musical merits (or otherwise), my project has the dubious distinction of spanning several eras of audio, which posed some 'interesting' (in the Chinese sense) challenges when it came to finishing it. While mixing the tracks, it occurred to me how things have changed since I started recording in the late '60s, and I thought it would be instructive to chart some of that evolution.

A Whiff Of Oxide

In the beginning, there was analogue. Throughout the '80s, I and my vocalist partner Barbara Gaskin spent many happy months in the studio recording on 24‑track, two‑inch tape — ah! I still remember its smell (the tape, I mean, not the studio). That all changed one day in autumn 1988  when the owner of the East Anglian recording facility we were working in strode into the premises and announced out of the blue to our young engineer that the studio was about to close and he was henceforth out of a job. The engineer, Owen Morris, went on to produce hugely successful albums for Oasis and the Verve, so no harm done there. But for us, the closure posed a dilemma: we'd been regularly hiring commercial studios since 1981, but now many were struggling to survive. I didn't fancy looking around for a new work environment, only to see it shut down six months later, so we simply bought more kit and carried on recording at home, occasionally hiring outside facilities when necessary.

Our domestic setup then was based around a 16‑track tape recorder, supplemented by much off‑tape sequenced MIDI material.Following the closure of Spaceward, a Fostex E16 became the centrepiece of the Stewart/Gaskin studio. I enjoyed the physical business of handling tape: fiddling around with splicing blocks, razor blades, leader tape, pinch rollers, take‑up spools and isopropyl alcohol (used to clean tape recorder heads, not for recreational purposes, you understand) made me feel like an old‑school studio pro. But my love affair with the brown ribbony stuff turned sour when I was mixing a song called 'Your Lucky Star', a tribute to the maverick producer Joe Meek. The tape machine decided to make its own zany contribution by playing at slightly the wrong speed, creating a tuning discrepancy between the recorded material and the off‑tape sequenced keyboards which Mr. Meek might have enjoyed, but I didn't. I became paranoid, and from then on always recorded an A=440 tone on tape at the top of every track so I could check playback speed.The Dave Stewart tape archive. Part of it, anyway.

Since the late '80s, I'd been mixing onto DAT tape (RIP), which sounded great, generated no tape hiss, required no calibration, failed to grow sonically dull after repeated plays and invariably ran at the right speed. By the end of 1993, the time had come to ditch analogue tape and go digital. At that stage it would have been prudent to check out Notator Logic (as it was then called) or one of the other emergent PC‑based recording systems, but I was wary of computers; having started out playing in a band, operating a mouse and keypad on a daily basis felt, to me, uncomfortably close to working in an office. I opted instead for three Tascam digital eight‑track recorders (a pair of DA88s and a DA38), which could be synchronised and used as a 24‑track recording system.

Eight Ace

These digital eight‑tracks had a great sound and a sweet top end, but they brought their own problems.Recording 24‑track required three separate tapes, each of which lasted for nearly two hours and had to be formatted in real time prior to use. Originally designed as camcorder tapes, these Hi‑8 cassettes used 8mm tape that was thin, flimsy and prone to creasing. More than once, I ejected a tape and found a loop of tape trailing outside its plastic case and snarled up inside the player. Result: severe audio dropouts and brain‑ache.

Welcome to the digital age: three synchronised Tascam multitracks allowed simultaneous and noise‑free 24‑track recording, but brought with them their own frustrations.

Premature head wear proved to be a major issue, and when one DA88 needed an expensive replacement head after only 240 hours' use I put my foot down and refused to pay for it. Why, users asked, were the recorders' heads wearing out so quickly? Tascam responded by making veiled accusations about the "abrasive” nature of the brand of tape I'd been using, despite it being recommended by reputable pro audio dealers. Somewhat rattled, I cloned all of our DA88 tapes onto an alternative make which was a) in short supply and b) expensive.

I got into the habit of backing up overdubs after every session. On the occasions when I had to retrieve audio from a back‑up tape due to dropouts on the master, I'd make a back‑up of the back‑up. As a result, my DA88 tape collection grew at the speed of a rabbit population. This proliferation was accelerated by my habit of creating new slave tapes for different overdub sessions. Keeping track of their contents required a great deal of housekeeping. I'd avoided turning into an office worker, but at times I felt like a librarian.

Weighed against these drawbacks were some distinct advantages: the recorders were highly portable, and with two in the car boot we could drive to North Wales and record a 30‑piece choir, take the slow, stomach‑churning Swansea‑Cork ferry across the Irish Sea to record our guitarist in his Bantry studio, or pop into a London studio to record acoustic piano. We also paid repeated visits to our drummer friend Gavin Harrison (who now plays with Porcupine Tree and King Crimson), to transfer overdubs between our DA88s and his Logic/Mac system. And thereby hangs a tale...

Sync Or Swim

It goes without saying that when transferring audio between a digital tape recorder and a hard disk recording system the two have to be perfectly synchronised, otherwise your overdubs will end up out of time with each other. Technology makes this easy, right? Wrong.

Drummer Gavin Harrison and myself hired and/or borrowed a number of different makes of synchroniser and MIDI interface in our attempts to accurately sync my Tascam DA88s and his Logic 5/Mac system, but Logic consistently refused to respond to the audio timecode (aka EBU) they generated. In the end, we stumbled upon a just‑about‑workable system by trial and error: it involved running MIDI timecode from the DA88 into a MOTU MIDI Express interface connected to Gavin's Mac G3 via Firewire. The secondary question of clocking the digital audio was solved by sending word clock from the Mac's MOTU 2408 MkI soundcard into the DA88. If we neglected to do that, loud audio glitches rended the air.

This somewhat counter‑intuitive lash‑up produced solid, no‑drift lock between my digital tape recorders and Logic. The problem was that when monitoring the pilot click-track recorded on both, we'd hear a positional discrepancy which, infuriatingly, was never the same twice. This 'timing gap' was sometimes in the region of a 16th note, at other times close enough to produce a flam. Once in a while, unpredictably, the two clicks would sync so tightly that they would phase‑cancel, enabling us to complete the transfer. That joyous moment rarely occurred until after a dozen or so false starts and much profanity. This was not the sample‑accurate sync we'd hoped for.

I asked many colleagues about the problem and none of their suggested solutions worked any better. The problem appeared to lie principally with Logic, though I can't substantiate that. Fortunately, such transfers are not something one needs to do on a regular basis, but if you still have multiple ADAT or Tascam eight‑track project tapes in the attic that you're planning to archive to hard disk one day, be aware of this sync issue.

Biting The Bullet

It was Gavin Harrison who pointed out to me that hard disk recording had made it possible for drummers to edit their performances, something keyboard players had been able to do for years using MIDI. Watching him perform three highly accurate back‑to‑back drum takes and then comp his favourite bits into a master performance in no time at all made me realise how powerful this hard disk recording business was. Doing the same thing on two‑inch tape would have taken all day.

It was time for a change. Smash-hit drum & bass tracks were now being recorded by 15-year-olds in their bedrooms, on cheap home computers, and I had to face the fact that my once cutting‑edge gear was antiquated. Things reached a head in 2001 when we were comping a vocal on a DA38: I set up an automated trial drop‑in and auditioned it three or four or five times — perfect each time. When I hit 'record' to execute it, there was a loud click at the drop‑in point on playback.

That was it. I bit the bullet, bought a Mac G5, a soundcard and Logic 6, gritted my teeth, learned the jargon and rode the learning curve. It was a tedious business at times, but there were immediate rewards. MIDI sequencing in Logic was far quicker and easier than it had been on my old hardware sequencer and I now had the keyboard player's Holy Grail of perfectly integrated MIDI and audio. There was, however, one fly in the ointment: all our current songs were recorded on multiple DA88 slave tapes. How would I assemble their contents in Logic?

In the end, it was the prospect of moving house that finally spurred me into action. The thin walls of our two‑up, two‑down Orwellian terraced property meant that we could hear every utterance of our next‑door neighbours, and from their side of the drainpipe my running commentaries down the talkback mic to Barbara ("Sounded good, best yet — wanna try one more? OK, rolling...”) must have sounded like a madman's monologue. Not an ideal environment for a recording studio, really. We decided to head for the hills and "get our heads together in the country”, as musicians used to say annoyingly in the '60s.

Mind Over MIDI

Before fleeing the metropolis, I transferred all our DA88 tapes into Logic three at a time, using my MOTU 2408 MkII interface's three TDIF digital connections. Some songs were spread over as many as 20 slave tapes and required multiple passes. Mindful of the dodgy sync we'd experienced at Gavin's, I took care to include a digital clone of each song's pre‑recorded pilot click track with every pass, so that I could keep a handle on the timing. I set up an automated drop‑in in Logic four bars before the song start, which meant that all the audio files started at exactly the same place regardless of where the overdubs occurred in the arrangement. It was tedious to have to go right back to the top to transfer percussion parts that didn't come in until the playout, but since I hadn't yet created tempo maps and bar charts for the songs, this was the only way to keep the audio lined up.The old order: in the days when backing tracks involved a lot of hardware MIDI sound sources, a Roland MC500 and SBX80 were used to sequence and synchronise them to tape. The later Roland MC80 was a far‑sighted purchase allowing MC500 data to be converted back to MIDI files.

As mentioned earlier, I was in the habit of using a lot of off‑tape sequenced material, so I needed to recreate all this MIDI data within the Logic songs. Up to a point, this was comparatively easy. In a rare moment of foresight, realising that no computer program would ever be able to read data from my Roland MC500 hardware sequencer, I'd bought a Roland MC80 sequencer, which could read MC500 floppy disks and convert said data to a MIDI file. This handy 'black box recorder' (which had sat on the shelf gathering dust for years) saved the day. Before long, I had MIDI files of all my old sequencer data for each song running happily in Logic. The problem was synchronising this data with the audio recordings.

My old system relied on a Roland SBX80 sync box: the song tempi, time signatures and timecode offset were programmed in the SBX80, which read audio timecode (aka EBU) from the DA88's sync card and sent MIDI clock and song position pointers to the MC500. I often change tempo mid‑song to make things feel organic, so the bpm figure was rarely static. Replicating the SBX80 program data in Logic should have been a simple question of copying the time signatures and bpm figures, but unfortunately it wasn't that straightforward. An obscure technical fact, which occasionally causes sync problems, is that not all devices compute tempi in the same way (I'm reliably told that this is due to differences in rounding off numerical decimal values, known as 'representation errors', a subject explained in mind‑boggling detail by Wikipedia).

By comparing the timecode readouts of Logic and my old Roland SBX80, I could see that they had a different way of measuring tempo. For example, in Logic, bar 17 of a 98bpm song starts at timecode position 00:39:04:47 (minutes, seconds, frames, bits), whereas the SBX80 places it 31 bits earlier. Since a 'bit' is only half a millisecond long, a few bits variance either way doesn't matter much, but over the length of a song, such timing discrepancies soon become obvious. I got round the problem by making micro‑adjustments to the Logic tempi, so that bars occurred in precisely the same place as before. Consequently, passages originally programmed at 128bpm now ran at the exciting new speed of 127.796 beats per minute. That's Numberwang!

Blood Sample

Once the MIDI, song tempi and bar lines were sorted, I could finally turn my attention to the errant audio files. I lined up the pilot clicks from each DA88 three‑tape pass and noted which was the earliest, then used Logic's sample editor to figure out how many samples later the others occurred. Having established this figure, it remained to trim all the other files from the same pass by the same amount. Once I'd done that, my recordings were now all playing in time with each other again. Hallelujah!

Unfortunately, that wasn't the end of it. Though in perfect sync with each other, the audio files sounded early to Logic's click, and by looking at its waveform, I could see that the pre‑recorded DA88 pilot click was now well in advance of the beat: comparing it to a live, Logic‑driven MIDI version of the same thing, I could hear a pronounced flam. As this click was the only verifiable timing link between the old tape playback and the new Logic songs, I had to abandon scientific theory and slide all the on‑screen audio regions to the right until they matched the live click. The size of the timing shift varied from song to song, but averaged about 80 milliseconds.

Having overcome all this mathematical madness and successfully entered the modern world of audio, I was left with the relatively simple task of making musical sense of the hundreds of overdubs I'd transferred into Logic. A lot of them were now redundant: guide backing track and drum-kit submixes created for overdub sessions, early guide vocals with first‑draft lyrics, pre‑bounce multiple backing vocal tracks, guitar and piano out‑takes, alternative keyboard solos, and so on. Sorting through this lot made me realise how I'd unwittingly made a rod for my own back in this project: there were just too many scattered options, but now I could see them all assembled in front of me, it wasn't hard to identify the viable ones and junk the rest.

Box Of Tricks

The question arose of whether to mix 'in the box' or break the audio out into a desk and mix on faders. I had sufficient desk channels for the latter, but opted to mix in the box for two reasons: it was simple and it was repeatable. It may be that the audio would sound subtly better if run through a desk's analogue circuitry, but I knew from experience that if we wanted to quickly remix a song with (say) the vocals a couple of dB louder, recreating the analogue mixer settings would require an enormous amount of note‑taking on my part, with no guarantee of perfect recall. By comparison, loading the Logic song, adjusting a single on‑screen fader and doing an off-line mix bounce would take five minutes, and the backing track would sound verifiably identical.

The fruits of all this hard work: Dave Stewart and Barbara Gaskin tour Japan with guitarist Andy Reynolds (left) to promote the album that took two decades to make.