You are here

Mixing Atmos: David Kosten

Mike Oldfield's Tubular Bells - Immersive Mixes By Sam Inglis
Published July 2023

Mixing Atmos: David KostenPhoto: Lexie Morgan

David Kosten’s first commercial Atmos project was an immersive mix of one of the greatest albums of all time: Tubular Bells. No pressure...

With much to learn, and much to unlearn, most mix engineers would want to ease themselves into Atmos by taking on a few low‑key projects. That didn’t quite happen for David Kosten. Instead of an obscure indie EP or television soundtrack, he found himself charged with Atmosifying Mike Oldfield’s 1973 masterpiece, Tubular Bells.

Thanks to his work with artists like Keane and Bat For Lashes, as well as his own Faultline project, Kosten is a very well established producer, but he cheerfully admits to having no track record with Atmos. So how did he land such a high‑profile, high‑pressure debut gig? “I co‑produced and mixed Steven Wilson’s last record [The Future Bites], and I sat in on the Atmos mixing side of it. And that’s really where my Atmos journey began. I was sitting at Dolby HQ, watching this record go from two channels to a whole bunch of channels and loving it. And thinking, ‘Oh wow. Pretty much everything I’ve ever worked on would’ve benefited from working in this three‑dimensional format.’

“I think every artist would know from working with me that I’m obsessed with talking about three dimensions, and front‑to‑back perspectives. So I always talked in those terms, but Steven Wilson was actually physically doing it. For a long time, he had been saying, ‘David, you really should be making, at the very least, your own music in this format because it’s so well suited to it.’”

Filling The Gap

As part of the team behind The Future Bites, Kosten was then nominated for a Grammy Award for immersive sound. Nevertheless, he understood that he had a lot of work to do to get up to speed as an Atmos mixer. “The next step was putting the speakers in, which I did about a year ago, then getting my head down to learn it. And there’s a learning curve.

To equip his West London studio for Atmos mixing, David Kosten supplemented his main ATC speakers with a Focal surround rig.To equip his West London studio for Atmos mixing, David Kosten supplemented his main ATC speakers with a Focal surround rig.

“It’s not as complex, technically, as I actually thought it might be. I understood that pretty quickly, and I’d already had enough background from talking to Steven. But what I needed to do was work intensely on a lot of tracks, so I truly understood how to present music spatially, coherently and enjoyably, without it coming across on speakers as a collection of disparate sounds. And I worked on things that are already produced, went back to old mixes and converted them to Atmos, did some dummy runs with my own stuff. And Caroline Hilton was unbelievably helpful and incredible with her time and knowledge. She really is the person who got me to this point where she would give me very honest feedback, and tell me when she didn’t think something was working, and really helped a lot.

“It’s pretty good being able to give your music to Steven and her and a few other people for them to critique and say, ‘You know what you’re doing now.’ And so by that point, I felt like I was ready probably for a low‑key first proper gig.

“So, then I got my low‑key first proper gig!”

Back In Time

Although Mike Oldfield himself remixed Tubular Bells in 5.1 back in the early 2000s, everyone agreed that the starting point for the Atmos mix should be the original stereo album. Kosten’s first and arguably biggest job, therefore, was to recreate the original stereo mixes as closely as possible, in order to generate stems that could be used for an Atmos presentation.

“My intention was that I wanted people to be hard pushed to tell that this wasn’t done in 1973. It would just happen to have a time machine and be able to put it into spatial. I wanted it to sound exactly like the 1973 version. No excuses. There is no decision in terms of panning or EQ or effect or anything that I made. Every single decision that I heard on those stereos had to be duplicated as exactly as I could first. So I was essentially trying to mirror what the guys did in 1973 before I did anything to do with Atmos.”

This, it turned out, was harder than it sounds, one reason being that important parts were simply missing from the surviving multitracks — among them the album’s titular tubular bells. Although there were several bell recordings on the multitrack, none of them was the part used on the original album mix. Where did the bells on the album itself actually come from? Had they been overdubbed during mixdown? Or did Mike Oldfield erase them from the multitrack later? We’ll probably never know, but it meant a lot of work for Kosten, who had to finesse that section of the stereo master into his multitrack.

“Weird things happen when you don’t realise that your record is going to be an iconic classic!” says David. “Maybe you’re a little bit more laissez‑faire with masters, and there’s no undo button on a two‑inch 16‑track. And he was literally filling every second of that tape.

“One of the things I was faced with early on, as well, was that there was no multitrack available for the end of Part Two, the Sailor’s Hornpipe. So I had to use the 5.1 from 2009 for the end section because there was no multitrack from it. But then, obviously, that needed to be edited in perfectly to match the originals.”

This Pro Tools session was the starting point for recreating the stereo mix of Part One. The original stereo master is at the top, followed by the 16 tracks from the original recording, and lastly some additional elements that were unearthed during the 2009 remix.This Pro Tools session was the starting point for recreating the stereo mix of Part One. The original stereo master is at the top, followed by the 16 tracks from the original recording, and lastly some additional elements that were unearthed during the 2009 remix.

Exploding Tracks

Tubular Bells was recorded to 16‑track, two‑inch tape, and considerable ingenuity had been required to capture all of Mike Oldfield’s musical ideas. Each track on the multitrack typically contained six or seven different parts, often featuring completely unrelated instruments. “Every second of every track has got something on it to get these ideas that he had out. So he’s looking for space everywhere.”

Since he was working in Pro Tools and thus not subject to the same limitations, Kosten split each musical element out to a separate track. The original 16‑track recording thus ballooned to well over 100 tracks, most of which would eventually become stems to be placed within the Atmos sound field. But before that could happen, there were plenty more hurdles to be overcome.

Famously, each side of Tubular Bells is a continuous piece of music that lasts over 20 minutes, and is not broken up into separate songs. However, it turned out that the multitracks for each side actually differed in length from the album as released, either due to speed variations on tape machines or edits made to the master tapes. Much painstaking listening and editing was needed to recreate the exact timing of the master.

“I had to do all those edits to then match the timing of the multitrack. So I used the structure of the original stereo and then I edited the multitrack to that, because I wanted to really be able to A/B between them. The timings of these things just never quite marry, they’ll be tape machines running at a very slightly different speed and it’s almost impossible to match.”

Once all the parts from the multitrack had been split out to individual Pro Tools tracks, the session ballooned to over 100 tracks in size.Once all the parts from the multitrack had been split out to individual Pro Tools tracks, the session ballooned to over 100 tracks in size.

Download this ZIP file to access larger, more detailed views of the Pro Tools sessions.

Package icon

Match & Mix

Once the basic arrangement had been recreated, and unused parts and studio chatter had been trimmed out, the next challenge was to rebuild the original mix. For an album recorded in 1973, you’d expect this to be fairly straightforward; engineers were in the habit of committing everything to tape during recording, so pushing the faders up on the console would usually deliver something recognisably similar to the finished album. Not so with Tubular Bells.

“I realised that, across getting on for 50 minutes of music, what I was actually listening to was the most astonishingly brave and uncompromising time spent in a recording studio that you could imagine. And the fact that it turned into this monstrously successful thing is mind‑boggling really when you realise the extreme processing that was going on. What I had to do to duplicate those mixes was not gentle. They must have been just following their noses and thinking, ‘I want it to sound like this. How do we do that?’ By turning that knob — a lot!

“It’s so uncompromising. And the first thing I realised as soon as I listened to those raw sounds was how in so many places the mix of the original record was extraordinary.”

So radical was the EQ and other processing in places that it was hard to identify which sound on the multitrack was actually responsible for the end result. In one instance, for example, a full‑range Hammond organ track had been high‑pass filtered at the mix to the point where only a thin buzzing sound remained.

David Kosten: The first thing I realised as soon as I listened to those raw sounds was how in so many places the mix of the original record was extraordinary.

Never Standing Still

“What I think was also extraordinary,” adds Kosten, “was how dynamic the mixing is. There’s no point in the entire album where a sound is just left alone. The famous pianos more or less just sit there and do their thing, but then there’s other layers of piano that come in and those have different processing: they’ll have a different reverb, they’ll be panned in a different spot. But it’s dynamic in the sense that guitars will pan all over the place left to right and back again. Or they’ll sit over on the right‑hand side, but they’ll get extreme wet/dry changes all the way through. And you hear it so clearly on the record.

“For instance, at the end of Part Two there are three or four different guitars, and you hear them totally wet, it’s almost down a hole, down a ravine — and then it comes back and it’s right next to you. All of that stuff had to be automated, listening on headphones, doing it a second or two at a time.

“There are also big organ stabs — I think he wanted them to sound like brass stabs, but he had a Hammond. And every single one had a different EQ and a different level. It was unbelievable how they evolve. A dozen or so stabs and they were all different.

“There is nothing conventional about any of this. I think at some point there’s five or six bass guitars playing at the same time. You don’t get that very often. So what does one do with that? You are just using your experience as a mixer to try and maintain separation in the same way that they did when they were making their stereos.

“There was also a decision to make as to what is good hiss and what is bad hiss. Good hiss, dirt, grime, mistakes: all of that is good if it’s on the original multitrack and got used in the final mixes. But I didn’t try to emulate the hiss of the original stereo quarter‑inch tape. I thought that that would be a hiss too far. I did actually have to recreate hiss here and there and do funny things to try and seamlessly make sections work. There was one bit where they’d obviously used a processor that was very hissy, and it didn’t sound exactly as the original did if I didn’t add some hiss. I’ve got an amazing old Binson echo out there, which sounds fantastic but also has the right kind of hiss. So it’s my go‑to hiss creator. If you want something to transport you into a particular era, it’s pretty amazing how well that works.”

Source To Stem

Although the sources on the original multitrack were nearly all recorded dry and in mono, effects such as chorus, delay and plate reverb had been used copiously at mixdown. In David Kosten’s recreated mix, therefore, the sources all yielded stereo stems with effects printed to them. Sometimes, in fact, he found it necessary to combine two or three different effects to recreate what was done on the original. These were often printed to separate stems in order to open up further options for Atmos panning. The aforementioned Hammond stabs were a case in point.

“There’s three different reverb effects, plus all the EQs and things. I printed those separately and then moved the Hammond so you get your stab and then the decay of it floods from front to back. So if you listen to it side‑on, you’ll hear the reverbs shoot past you. So that was a bit of fun, but the actual sound of those Hammond organs needed all of those different effects to get as close as I could to the original.”

The 50th Anniversary reissue of Tubular Bells includes not only David Kosten’s Atmos mix, but also his recreation of the stereo mix, alongside the original quad mix and Mike Oldfield’s own 5.1 remix.The 50th Anniversary reissue of Tubular Bells includes not only David Kosten’s Atmos mix, but also his recreation of the stereo mix, alongside the original quad mix and Mike Oldfield’s own 5.1 remix.

Creating stereo rather than mono stems also meant that stereo motion panning — an important element of the original mix — could be baked into the stems. “By doing it that way, you’re then maintaining the stereo imagery that Mike had, and then you can choose where to position it front and back. I like the simplicity of that system, rather than trying to mix it from the original multitrack and go straight into Atmos. It doesn’t matter how you get there if you’re creating something brand new. But this was so much about maintaining the integrity of those original sounds that this was the way.

“All the panning that I was doing was embedded in that stereo, so essentially, the Atmos part of it was more to do with... I feel like storytelling is the best way of describing it, even though it sounds horribly pretentious. That part of it was where I spent the time in the Atmos mixing. It wasn’t in the panning because that had already been done. I made the odd little nudge of EQ here and there, but pretty much was happy with how the sound was printed. There’s a little bit of automation that I did; for example, there’s a piano that rushes towards you. If you listen to that on a surround system, you’ll hear the piano start behind you and rush towards the front. So that was just drawing in some automation for that. But mostly all the panning choices were baked into those stems.

“I felt it was more important to just be able to know that the sounds were right, and then just allow it to breathe enough so it almost hovers around you. I didn’t get into extremes, there’s no fun with spinning sounds or anything like that. I’ve basically copied every single panning choice from Mike Oldfield.

“It was the placement of those stems that I spent a lot of time thinking about. If it’s a strongly melodic element, or if it’s a storytelling element that you want to focus on, that usually suggested that it was more in front of you than as a special effect behind you.”

Objective Approach

Dolby Atmos is a hybrid format that combines one or more channel‑based surround ‘beds’ with mono or stereo ‘objects’. Objects are packaged with their own automation data, which is decoded on playback to determine their position and movement. Although he rarely used Atmos motion panning as an effect, Kosten chose to treat his stereo stems as objects, rather than routing them into the bed. “The control over the precision, the pinpoint ability to put it in a specific space, is much more what I wanted. And you just don’t have that same control using a bed. Also, of course, for the binaural headphone mix, you can specify a different Atmos distance setting for every single stem as well. And by having them all separate, it meant that you could be much more selective about that.

“I’ve heard awful binaural mixes where the lead vocals have the wrong setting, or they put too much reverb on their rebuilt stereo mix, or whatever. And I’m full of complaints about lots of Atmos mixes that I don’t like. But by giving it a stereo mix that I was happy with, it felt fine. But obviously, I did the binaural mix from 80‑odd stems, and it’s a very unusual thing to have that control. Normally, we’re presented with a tiny number of stems, and that’s what most Atmos mixers are faced with. So they’re having to do all sorts of things to try and make the Atmos mix sound reasonable. But here I was able to avoid that.”


Many engineers talk up the differences between stereo mixing and Atmos mixing, but David Kosten is emphatic that a correct stereo mix is the key to a good Atmos mix, especially when it comes to remixing a classic album like Tubular Bells. “I hope people can hear that it was done with the attention to detail that a fan would want. I’m pretty sure there’s been rebuilds of tracks where that’s not been the case, or fans have heard things and gone, ‘Well, it doesn’t sound much like the original. Why does it sound so different?’ It’s not to do with the Atmos part of it, it’s to do with the agony. That’s the amount of work that it takes to recreate what was going on in a studio 30, 40, 50 years ago. That’s quite hard. So I have respect for anyone who tries it.

“This is all about honouring the choices that were made in 1973, but giving it access to a new, more immersive space. That was the whole point of it. So I didn’t even want to get into the idea of creating something new for this piece. It was much more about how you maintain the integrity of that original stereo master, but gently bring it into 2023 without appalling anyone. And I’m sure it’s going to have lots of people appalled, but my hope would be that anyone listening to the Atmos version of this would be able to hear the love and detail that I threw into it, because it was a hell of a labour of love to get to something that was even close to the original stereo.”

Track & Trace

Although Tubular Bells was recorded to 16-track, the track sheets that David Kosten received were for a 24-track machine. The riddle was eventually solved thanks to the contribution of Manor Studios engineer Philip Newell, who takes up the story...

"I believe I recognise my own handwriting there, as that was the track sheet that I prepared before the 1975 quadraphonic mix. And no, there was never any 24-track transfer.

"When Alan Perkins and I mixed the quadraphonic version, we had the advantage of being able to use The Manor's newly installed Allison Research 65K / API VCA-fader automation system, but it required two tracks to always be available to bounce the automation data between on each update. So, for the quadraphonic mix, we played the 16-track masters on an Ampex MM1200 with its 24-track head block fitted, and what you see on the track sheet is the way that the 16 and 24 tracks correspond.

"In the left-hand column of the sheet are labelled tracks 1 to 24, but at the left-hand edge of the next column are hand-written the numbers 1 to 16, which show where the 16 tracks best aligned with the 24-track head block. The eight continuous horizontal lines drawn on the track sheet are the tracks of the 24-track head-block which are positioned mainly in the guard-bands of the 16-track heads. So, these tracks were free to use for bouncing the automation data between, although we probably chose to use the two tracks (perhaps different for different sections of the mix) which were least sensitive in terms of the crosstalk — so that the data would not bleed through into the music tracks.

"The 24-track machine reproduce amplifiers had probably been lined up to the tones on the 16-track master tape (and not to a normal full-track test tape), so that the reproduce levels of the non-perfectly-aligned tracks would play back at the correct Dolby A level, for the correct noise-reduction tracking.

"Incidentally, the quadraphonic mix first appeared in a discrete four-channel form (ie. not in the earlier SQ matrix form of the various vinyl disc releases) on the SACD, released in 2000."