Minneapolis, Minnesota is about 400 miles north-west of Chicago. The young man who would become Bob Dylan came from around there, but probably the best known musical resident these days is Prince. Not far from the Great Lakes and the Canadian border, it gets pretty cold in winter. It's also home to a commercial recording studio which may not be as famous as Prince's Paisley Park complex, but is probably the first anywhere to depend on Linux-based software at every stage of the music production cycle.
Ron Parker started Mirror Image Recording Studios around 20 years ago with brothers Dana and Bill Bailey, after they moved to the city with the bands they were in at the time. Dana Bailey also worked with Prince and The New Power Generation as MIDI and keyboard technician on two world tours. Later, NPG vocalist Rosie Gaines partnered Dana as a songwriter, creating tunes including the speed garage anthem 'Closer Than Close'. As is all too often the case, selling hundreds of thousands of records around the world did not produce proportional financial returns for the artists involved — and years later, Dana and Rosie are still waiting to be paid the royalties on that particular record.
Mirror Image is based in a large late-Victorian house, where the studio is in the basement and the top three floors provide living accommodation for visting musicians. Apart from a pre-production area, there are four acoustically designed and tuned rooms in the studio — two live rooms, one control room for tracking and another for mastering. As an independent operation, the partnership maintains a schedule which is by necessity very busy, and has recorded around 400 clients in the last decade alone. Ron Parker is largely responsible for the introduction of Linux to the studio's computer systems, and fits research and testing on the new software in the gaps available between sessions.
Like many medium-sized studios, Mirror Image made a gradual transition from analogue to digital, eventually ditching tape formats for the flexibility of computer-based hard disk recording. Parker explains: "We began producing with a half-inch eight-track tape machine in 1984, upgraded to one-inch 16-track and then 32 tracks of ADAT before moving into hard disks. We got started on those around 1999 with a couple of Yamaha AW4416s and Macintosh computers running Digital Performer interfaced with Mark Of The Unicorn 2408s. We're currently using a Tascam DM24 as the primary digital console, which interfaces to our computers with TDIF and ADAT." A Soundtracs Solo desk is also kept handy because of the quality of its preamps and EQs, but the AW4416s are now due to be retired from recording duties.
The current Mac platform is based around a G4 machine running OS 9.2. As well as MOTU's Digital Performer, the studio runs Logic from Emagic and uses BIAS Peak as a WAV file editor. Now, a Linux system will replace the Mac at the core of the studio, offering a number of new technologies which Parker hopes will transform the way the partnership works. The main Linux machine is a dual AMD Athlon 2600+ with 1GB of RAM, plus an RME Hammerfall 9652 card with 24-channel ADAT and stereo S/PDIF I/O. The ADAT interfaces are routed via a patchbay to the DM24 console and the MOTU 2408 on the Mac system, so audio data can pass easily between the old and new systems.
At any one time, 30 or 40 active jobs are stored on an ICP Vortex four-channel SCSI RAID controller, which currently hosts two RAID5 channels with a total of 10 Seagate 18GB hard drives. While Apple has only recently started to offer this technology with the Xserve RAID, these multi-disk storage systems are common in the Linux world, where they are used for instant access to all kinds of 'mission-critical' data. Parker adds: "It's common to have clients drop in unannounced during someone else's session and request a CD of their songs. I've written scripts that enable me to burn CDs for any client with three keystrokes. If I get an autoloader for the CD burner, I won't even have to get out of the chair any more."
RAID systems can be configured in various ways for increased performance over a single drive or data security, but the RAID5 design is considered a good trade off between speed and hardware redundancy. If one drive fails for any reason, the work in progress should be safe. There is also an Rsync mirror (automatic backup) to a 120GB IDE hard disk in another Power PC machine to make absolutely sure that work is not lost. Once projects are finished, Parker uses a SCSI DAT drive to archive the complete set of data from the job.
A fast local area network runs throughout the entire building, which means the seven musicians currently living upstairs could connect their own computer-based DAWs to the main studio RAID array and get on with their project, even when they aren't in the studio itself. This alone is potentially a massive boost to efficiency, since visiting artists will need to book the live rooms only when strictly necessary. Tasks such as reviewing, editing or adding MIDI parts to their work can now be carried out in the comfort of their own rooms, in their own time.
The Linux environment at Mirror Image is based around the JACK low-latency audio server, the Ardour DAW and the Rosegarden MIDI + Audio sequencer (see my article on Linux and music in SOS February 2003, available on-line at www.soundonsound.com/sos/feb03/articles/linuxaudio.asp). As an early user of Ardour, Parker thinks he probably produced the very first full album to be made using the program — when it wasn't even a beta version. "I knew it would be an almost intolerable technical challenge, but felt the proof of concept would be good for me and the Linux professional audio community. A feature that I really appreciate that other DAWs don't have is the Sound File Database (SFDB). It's a database interface to the file system — a useful tool for anyone who intends to use large sample libraries."
Ardour has now matured to the point where Parker can consider it for more projects, although there are still a few glitches to work out as of the second beta release, which will lead to version 1.0. But the potential for efficient, stable and highly flexible technology offered by the Linux system keeps Parker motivated ("I'm no good at benchmarking, but I have put this system under some real-world stress tests and the performance is incredible"). Despite claiming to be a non-technical user when it comes to computers, Parker has been able to engage directly with Linux audio developers. Parker's many hours spent behind a mixing desk have provided useful experience that the developers need to get the software right, so it's a two-way collaboration. This partnership effort has helped the development of new applications that will complete a pure Linux setup at Mirror Image, from first take to finished CD.
The Mixed Approach
Of course, not many studios would ditch their existing computer setup entirely and start from scratch with Linux. In the field of networking, Free Software systems proved their worth at specific tasks such as web serving and firewalls before becoming more widely used. There's no reason to suggest that the recording industry would be any different, and so we are likely to see Linux used at first in mixed computer platform environments.
"It's mostly about flexibility, and low latency — I hope the stability is proven eventually. In my setup the HDSP 9652 is really just a low-latency audio router with 26 physical inputs and 26 physical outputs. The mixing and clock control I get from Thomas Charbonnel's hdspmixer and hdspconf programs make it almost magical for me. I think this box completely replaces and improves upon some of the commercially available ADAT signal routers out there at a fraction of the cost.
"I believe I will eventually be able to add a second HDSP 9652 into the box and clock them both from the same external word clock source. I'll then get a 52-physical-input, 52-output audio router with 104 software inputs, should I ever need them. I'd be able to mix 104 audio tracks in the card without using CPU overhead — just sending audio data — and hooking in low-latency Linux soft synths all on the same machine. This is way beyond where I think I could ever get with Pro Tools."
Knecht seems happy with the mixed approach, and will continue to choose the best solution for his needs, regardless of platform. "I think JACK and its ability to connect together different applications and computers, and to maintain low latency while doing it, is the best thing Linux has going in the pro audio space today."
The first thing that Parker noticed to be missing from Linux was a professional-quality mastering application. While there are audio editors and effects plug-ins available for Linux that can be used for mastering work, at the time there wasn't a program dedicated to the task. Parker explains why he wants to be able to control every aspect of the mastering process: "For many years we sent our mixes out of house to be mastered. I don't know how many albums we've produced over the years — maybe around 200 — and the number of demos and single songs is much more than that. In the early days, before we understood mastering, clients would complain about the results. I didn't know what the problem was. It was very stressful — I thought our mixes sucked.
"A couple of years ago I got off my lazy butt and started designing and building a mastering studio. The first time I mastered a mix in the new room, I was blown away. Within several minutes I stopped working and started listening and hitting the bypass switch. Something very important happens when mastering in different rooms and with different monitors than those used for recording and mixing. The acoustical properties of the mastering room and its monitors cause entirely new stories to be told. How many inexperienced engineers feel 'fed to the wolves' when the client asks 'Why does it sound so good here but sounds like crap in my car?' I gotta raise my hand — been there, done that, don't ever want to feel that way again. Mastering has been shrouded in a mystique that's born from ignorance. I can make that claim because ignorance is my personal experience! I maintain that it's possible for any aspiring mastering engineer to learn how to achieve the sonic potential for a mix."
Now the Linux audio community is working on a mastering tool which will work with Ardour, Rosegarden and the other JACK-enabled applications. Steve Harris, who has written many free effects for the LADSPA plug-in standard, is the lead developer on the JAMin project, the name of which stands for JACK Audio Mastering. JAMin is a software module, rather than a direct replacement for complete proprietary systems such as Sonic Solutions or SADiE. At this early stage in its development, it features 30-band and 1023-band equalisers, a spectrum analyser, a three-band peak compressor and a look-ahead brick-wall limiter. Planned additions include multi-band stereo processing, parametric EQ, a loudness maximiser, and preset and scene capabilities. While JAMin could be run on a mixed-down stereo track, it is designed for the user to be able to reach back into the multitrack mix and fix the root cause of problems that conventional mastering techniques have to work around — in real time.
Parker comments: "With a multitrack source, if the kick drum is 5dB hot, we can adjust the just the kick -5dB. With stereo files as a source we might have to adjust 250Hz by -5dB, but that affects all the instruments. In our studio we export JAMin and Ardour with SSH (secure shell) and control both applications from the mastering room. The Ardour mix is routed to the control room mixer, while the stereo buss is routed via S/PDIF from the control room console to the mastering room console."
Mirror Image are reusing one of their AW4416s to receive the stereo mix in the mastering room, which outputs to Mackie HR824 monitors. "If the multitrack mix needs to be adjusted, I can make changes from the control room mixing console or from within Ardour in the mastering room. So the mix is running in both rooms at the same time."
As far as the user is concerned, JAMin looks a bit like a collection of VST mastering plug-ins gathered together into a single interface, but the dedicated nature of the tool should mean that CPU use is more efficient. Another benefit of basing JAMin on the JACK audio server is a flexible choice of audio sources — the inputs to JAMin could be live analogue or digital audio, or the outputs of any JACK-compatible software.
Harris explains: "The design of JACK means that we can use the recording and sequencing capabilities of other tools like Ardour, Muse and Ecasound and the synchronisation facilities of JACK to allow JAMin to do the mastering without needing a complete sequencing and editing environment. You connect the outputs of your sequences or inputs of your soundcard to JAMin's inputs and you send the output of JAMin to any JACK-capable recorder. I like this because it enables me to use the familiar interface of Ardour, while doing processing that goes beyond anything that's practical with plug-ins.
"There is some discussion about whether we will support direct exporting of data ready for glass mastering. Currently the easiest way is to export the JAMin-processed data from Ardour as a disc-at-once Red Book master. Allowing this to be done inside JAMin would make the final production more integrated, but would complicate the user interface and detract from the simplicity slightly."
Harris says the collaborative development process used to create JAMin is typical of a small Free Software project. "There's a mailing list which the users and developers all subscribe to and discuss features, ideas, bugs and so on. The development work is divided up pretty naturally according to our experiences — Jan Depner and Patrick Shirkey mostly worked on the user interface, Jack O'Quin on the concurrency code and me on the DSP code. Everyone chips in and argues about what features we should or shouldn't have, and where such and such a menu should go. The input from experienced engineers was very important; feedback on features and user interface layout let us develop more quickly and made sure we were always going in the right direction. Knowing that there's lots of people waiting to use the software is always motivating too."
Of course, some mastering engineers might not want to take responsibility for the state of the multitrack mix, and could resist the introduction of this sort of technology into their work. Parker isn't bothered by the prospect: "My attitude is, fine and dandy — let 'resistant' mastering engineers tell an artist who just maxed out their credit card for a $3,000 production budget that they need to return to the mixing studio to cut the kick drum by 5dB when the mixing studio costs $50 an hour and the mastering studio costs $100 an hour. I love music and working with artists and the more the artist and I accomplish, the happier I am. When I encounter 'resistance' I bristle with intolerance.
"The primary drawback to the multitrack backend solution is that mastering isn't about getting one song to sound great; it's about getting 12 songs to sound great together. Ideally this would be accomplished by having all the songs on one timeline — putting 12 multitrack song sessions on one timeline isn't reasonable, but I do think there will be a solution for this."
Free Software: Software that is licensed in a particular way to allow the freedom of developer collaboration and user modification, Free Software is usually offered for download from the Internet as both binary and source code at no cost, but doesn't have to be. Also known as Open Source, although just having access to source code doesn't necessarily allow you to do anything with it.
GNU GPL: The GNU Project's General Public Licence, the terms and conditions under which the Linux kernel and many other Free Software projects are released. Prevents Free Software from being passed off as proprietary, using author copyright as a legal safeguard — you can't just add a few lines and pretend you own the program.
JACK: Low-latency audio server software which can connect a number of different applications to an audio device, as well as allowing them to share audio streams. Client applications can run separately from the JACK server, or as plug-ins to it. JACK does a similar job to Core Audio on Mac OS X, but also benefits from deep-level system tweaks only available to users of Free Software.
Linux: Usually taken to mean GNU with a Linux kernel, a UNIX-like Free Software operating system. Applications for Linux systems are often Free Software, but there are plenty of proprietary programs available for it too. Most popular on servers and in high-end computing, but becoming more widely used on desktop hardware.
RAID: Redundant Array of Independent Disks, used to improve the data security and/or performance of a hard disk storage system. RAID is a workaround for the unfortunate fact that hard disks are mechanical and will inevitably break down from time to time. Tape storage can be an alternative for backup applications, but does not offer random access — it cannot jump instantly to the particular part of the data you want.
SSH: Secure Shell is a remote control system for networked computers, which allows log-ins from one machine to appear on another. It's a bit like Telnet, only encrypted for security — otherwise, anyone on the network could potentially take over your computer.
It's all very well being at the cutting edge, but if Parker's setup won't work with what anyone else is using, then Mirror Image could find itself marginalised. The key problem of compatibility constrains all studio software — just like software in other industries — and tends to mean that professional users settle on the same small collection of applications, regardless of their limitations.
There's little point in a DAW system which allows the flexibility to get the sound right until the final stages of production if incompatibility means that engineers and producers have to work with a two-track mix in a legacy format. But the second of Parker's 'missing' applications attempts to address just this problem — and a contemporary challenge to the business model of independent studios too. Ardour developer Taybin Rutkin is working on a session exchange application originally suggested by Parker, which is being sponsored by Mirror Image. It will perform peer-to-peer exchanges of Ardour projects across the Internet, merging just the differences from either side — which should save a lot of IP bandwidth. Parker comments: "It was titled Ardour Exchange but we're thinking that because it's separate from Ardour and performs Session Exchange tasks it could be titled SEX." That should certainly get attention, although prospective users may have difficulty finding the software on search engines...
"Every musician in the world has a pre-production studio in their bedroom. The small-to-medium studios are getting killed because a lot of our work is being done on computers at home. But none of these musicians has our acoustically treated rooms and recording gear at home. With SEX, musicians can do pre-production at home and then contract studios for drum overdubs, or whatever. Of course SEX will let us all be creative together too, and it will be Free Software licensed under the GNU GPL."
The principal cost of Mirror Image's migration to Linux has been in Parker's time. As something of a trailblazer, his investigation and deployment of free software has probably taken much longer than the same process will for the studio owners who come after him, thanks in no small way to the effort he has put in to its development. The Internet — in particular, the mailing lists for specific Linux applications and the more general 'linux-audio-user' list — has enabled an unprecedented level of collaboration between users and developers of studio software.
When asked if trying to work with Free Software has created lots of problems for the studio, Parker responds: "No problems. Challenges are another story. Linux-based solutions are highly configurable, but that capability introduces a learning curve. You probably don't have to learn a lot until you want things done your way."
Nevertheless, Parker has fitted the migration into a busy recording schedule, and has been able to reuse some computer hardware that was previously put to very different tasks. "The Digital Server case I got for free. The guy had about 40 of them and they all had HP Alpha servers with installations of the Oracle database. The loading dock had a pile of SCSI drives, SCSI controllers, network cards and other stuff that were all exposed to the rain. Referring to the stuff in the warehouse, he said 'You can have all 40 of them. I need to get rid of this crap!'
"What I find encouraging about Linux based audio is the freedom —not the 'free' in the financial cost sense. In a Free Software development model, when developers don't place a priority on requested features, cash donations can be offered. These cash donations are excellent investments because they inherit the attached freedom and become available to everyone. As the user community grows, more features should be requested and paid for."
However, Parker is careful not to oversell the benefits of Free Software: "I don't think Linux audio solutions are all things to all people, yet! However, I am hopeful that developments during the next couple of years will cause us to become more productive and creative than the people using proprietary software. We've used several other digital recording systems over the last 20 years. As an engineer and studio owner, the only one I am inspired to invest time and money into is Linux. I never had a software developer give me any meaningful time until I stumbled upon the Linux audio community.
"With the release of Ardour beta 2, my tests convinced me that it was time to select a couple of new jobs for start-to-finish production with Linux. This also meant I would have to teach my partners how to use Ardour, JAMin and Rosegarden." Parker looks forward to Linux studio software maturing to the point where its reliability matches that of established Free Software projects. "The beta 2 version of Ardour that I am using has some routing problems, but I can probably get around them. The point is that this solution enables the simultaneous execution of numerous non-trivial audio tasks that many studios probably aren't performing. Perhaps what's most interesting is that these tasks are affordable, stable and with a little savvy technically achievable. The truth is, if I can do it, anyone can."
JACK sound server
JAMin mastering tool
Linux Audio User mailing list
Mirror Image Studios
Sound & MIDI software for Linux