But even before you got established in New York you found yourself back in the British Isles.
Yes, Daniel Lanois was to be producing an album for Peter Gabriel. They were going to record in what was actually a cow shed on Peter's property near Bath, England, and originally Daniel thought he'd bring an engineer in for the mix. He knew I'd up and come to America, so he thought I had the courage to tackle the task. Once they got going on the project they called me to come over and help record as well.
A cow shed.
Yes, the worst thing is that it really was very cold.
When I arrived there were two analog 24-track machines, a Studer A80, and a Studer A80 shell that had been modified by a local electronics wizard, with its own audio cards and transport controls. A month before I arrived, as Dan explained it to me, and the day before the band arrived for basic tracking, Peter had installed a synchronizer to link the two machines, and somehow the two machines weren't talking to each other properly. The stock Studer had an FM card and the modified Studer had a DC card, and the synchronizer thought they were both FM cards, so there were incorrect pulses being sent to the second machine. The way Peter was working, he had a demo of each song with piano, maybe a Prophet pad and a Linn drum machine, that he would put up on the B machine, which was the modified machine, and he would play that to the musicians in the studio. They would then play along with that in their headphones, and record all their parts onto the A machine. They'd also copy some parts across from the demo to the A machine. They'd do a couple of takes, say on "Sledgehammer," and instead of leaving the demo reel up on the B machine they would take the first set of performances and put that bit on the B machine, so the musicians could then hear that in their headphones along with the demo rhythms-all that information was getting transferred across to the A machine along with the second set of performances. And they kept doing that, so they could constantly reference quickly back to a part they'd just played.
They were recording new takes while hearing the previous ones, and running two machines all the time.
Yes, and what happened was that the two machines were slipping ever so slightly. With each pass the problem got compounded, so that when you took tape one, and locked it up to tape six, they drifted. Add to that the fact that Studer A80s are notoriously unreliable at the front and back of the reel, and it made for a real mess. There were parts of each performance that you simply couldn't sync together.
But they had all these great performances, and it became my job to corral them together. Dan and Peter sat down and built up lists of what they wanted to get, say from tape six to tape one on the same song, and we planned to somehow edit these pieces together.
I arrived around the end of May, and the first part was getting familiar with the songs. There were musicians coming in during the day, and Peter was working on vocal ideas, so we had to try to isolate a performance that would be the master track of the song, and somehow keep tracks open so that we could edit in the pieces that we wanted from other takes. Late at night, after Peter had left, we'd work on retrieving the tracks we needed.
The musicians were overdubbing to parts that weren't final?
Yes, but with Peter-and this is probably true to this day-you're never really sure of what the final part is going to be, because he's constantly writing the lyrics, and seeing what will work. Sometimes he'd come up with a great lyric, and there'd be some part that didn't work with the vocal-so we were constantly changing.
Finally, I decided the best way to proceed would be to edit the multitracks into a more structured form and put it on one tape machine that we knew was reliable. In New York I'd used the Mitsubishi 32-track, and I proposed that we should get one. Then I set about editing whatever needed to be done-if there were four reels per song, I would have to do the same edits on each reel and hope they all were good. The process of getting stuff back from slave reel number five to master reel number one involved sometimes lining up the 2-inches and flying between them manually, bit by bit. In July we got an AMS sampler with 14 seconds of sampling time, so we could actually sample four or five measures of music and fly it in that way, which helped. By September we had everything over to the 32-track machine.
But when you mixed you ended up locking three machines anyway.
Yes, Peter felt that some things sounded better coming off the analog, particularly certain percussion elements, and he was right. But then, we never really sat down and mixed, you see. Most records get broken down into periods-pre-production, tracking, overdubs, the mix. With Peter, we never really entered into those phases. We were always mixing. So by October, we'd put up a mix, with whatever overdub had been achieved that day, and I would always store the mix on the SSL and take down documentation. Then, at some point we'd decide, "That's a really good reference for that track," and from then on we'd always reference back to that particular mix. So, if all of a sudden we were having background singers come in, we'd recall the mix and do all the patches, crosspatches, etc., so that they would hear the best possible interpretation of the song. By the time the album was "mixed," the mixes had already been going on for three or four months. And, of course, the patchbay was a mess. You couldn't see anything.
Do you remember how you recorded Peter's voice?
We used a tube 47. Peter always maintained that he'd used a handheld 57 in the control room on his previous records. Dan and I were skeptical, so we did a test. We set up a bunch of mics, and we blindfolded Peter and had him sing into each and tell us which one he liked the best. It came down to one, and he was sure it was the 57. When we took the blindfold off, of course it was the 47.
But there was a funny sound to that 47. It had much more air than you would normally expect-very pleasing sounding, but not as much bottom as you'd expect. We had our maintenance guy, Neil Perry, take a look and it turned out that one of the cables had a little nip in it. It wasn't getting full contact on the shield, and when he reconnected it back up it sounded much fuller-but the presence that we loved was gone. So we rigged up a system where we plugged the output of the mic into the patchbay, and we took the shield off a regular patch cord and used a mult of the two things; we brought the mic up on two faders to duplicate that sound-the regular 47 and the dropped shield version of it-and we'd balance between them.
What about the rest of the chain?
Peter had an old Decca compressor, kind of like an LA2A, very smooth, very slow. I set it up for a minimum amount of compression. I chose not to use a lot of EQ, mostly because I knew we'd be dropping in lines further down the road. We'd have EQ on the monitor side.
What outboard equipment did you have?
We had an AMS DMX1580 and an RMX16, an EMT 140 plate, and we used the studio as a chamber. I recall a Sony DRE1000 reverb unit, four Decca compressors, a couple of LA2As, and tape slap. Not a lot of outboard-we printed effects as we went along. The console was very small; I think just 56 in. We only had six echo sends, because I was using so many of the small faders for tape returns, and we had to constantly figure out different ways to get into devices. We would often chain effects together-send a signal to the AMS for delay, then out of that into another delay, like the Deltalab DL2 with the extra delay module that gave almost a second of delay time. We'd do a lot of things that would shimmer back and forth, then come out of that and patch into a reverb: By sending one thing you could get three effects.
Did you know you were working on a great record?
We knew it was good, but I don't know that we knew it was going to be one of those hugely successful records, because it took ten months and by the time we were finished we were just relieved to be done.
I sent a copy of the record to my friend Randy Ezratty in New York, and after he listened, he said, "You're going to get so much work from this, you have no idea." And he was right. The So record was very pivotal for me. I'm sure without it I would have had some success, but that tended to put me in a different place. Since then I've gone through periods that got a little dry, but for the most part people say, "You've done that record and you've done this record, therefore there's a level of consistency between all your work." Which is really what you want people to look at-you want them to judge you from your whole body of work, as opposed to just the last record that you did.