Linux Music

Choice of music computer

So you really want that Mac, for its cool interface, great applications, and the nice low latency of music programs. But your slim budget only allows a clunky commodity PC with just enough RAM to keep Vista from dying, and then you find yourself restoring Windows from scratch for the seventh time?

You are ready to try the third option: Install Linux, and never look back!

First you have to realize that running Linux is not a choice of either Windows or Linux. You can have both Windows and Linux installed on the same machine at the same time, and decide at boot time which OS you want to start. So you can still run all your expensive proprietary software on the Windows installation, but boot into your Linux-based experiment when you feel like it. This is called "dual-booting". All popular Linux installers will assist you in installing Linux as a dual-boot alongside Windows.

As an added bonus, Linux is a superior crash-recovery tool next time your Windows installation decides to break down again. From your Linux installation you have access to the Windows harddisc, even when Windows itself refuses to boot, and you can make all the necessary backups of precious data, before restoring Windows. Then after restoring Windows, boot back into Linux again and restore the backed up data into the original Windows folders. I personally built a backup and a restore script for my 13-year old daughter, since her Vista kept crashing all the time. With these two scripts, she was able to backup complete iTunes settings, reinstall Vista, and reinstall the iTunes data several times, entirely without my help!

Linux these days is a mature desktop oriented system, with an intuitive user experience, that appears more similar to Mac than to Windows. And then Linux is both very reliable, open ended, and has a strong and enthusiastic community, that provides on-line answers to most problems that you may encounter. Linux is no longer a nerds-only OS.

Last, and not least, Linux is blessed with an enormous amount of quality software, free for all to install, use - and even to modify, if that pleases you. This also holds true for music production. This page is dedicated to the use of Linux as a platform for producing music. My reference Linux distribution (or "distro" in Linux lingo) is Ubuntu Studio AMD64, as this is the distro that I run. But most of the software, OS kernels and configurations apply for most of the common distros.

A typical Linux desktop running Ardour DAW system.

Latency and realtime kernels

There are countless Linux distributions out there, and enough choices to confuse any new-comer. Therefore a recommendation for a distro is in place.

What you will be looking for in a music production environment is an OS that has a pleasant user experience, access to all the necessary music software that you need, easy maintenance and properties that makes it specifically suited to run music software. Now, all operating systems are based on a kernel, that schedules CPU time to the programs running on top of it. In Linux this is known as the Linux kernel. Most distros come with a general purpose kernel, suited for general purpose tasks. But while the mainstream kernel can be used for music production, true musicians will be concerned about latency. For music production we want something that gives us low latency so we can multitrack, and still stay in sync.

There is a special branch of Linux kernels that can provide this - the so-called realtime kernels. These kernels are compiled with special patches that allow music program users, to get instant access to CPU time, whenever the music programs find it nessecary. These programs can then keep up audio processing with minimal delay, and therefore provide the much desired low latency.

Realtime kernels form the basis for a couple of distros specifically aimed at multimedia production. Studio64 is one such distro, while I personally have used Ubuntu Studio for several years. But even with standard distributions, such as OpenSuSE, Fedora, RedHat, Debian and most other, it is usually possible to supplement the normal kernel with a realtime one. Just as with dual-booting, you can decide on boot time if you want to run with an ordinary kernel or a realtime one. The programs running on your installation will work equally well, no matter which one you use, except the realtime one gives you better latency. This page for instance was written on an Ubuntu Netbook Remix distro, with the original kernel replaced with a 2.6.31-9-rt kernel. Look for a kernel that ends with -rt.

When you choose to install Studio64 or Ubuntu Studio, the realtime kernel is used by default, and you don't have to think further about it, which is one good reason to choose one of these distros.

Music software on Linux

There is a lot of prejudice surrounding open source software. It is a common belief that open source and free software cannot be as good as "professional"/sommercial software, or that it is not as reliable. This is not true at all. Consider that there is Linux hidden inside many pieces of qear that you use all the time. If you have an Android device, you are in fact running Linux already. And your home file server is very likely to run it too. And in any case, major webservers, like those used for Amazon, Google and similar big operations run Linux. Even in the world of music, many devices, such as the Korg OASYS workstation, are based on Linux too.

Also, most of the open source software out there is written by software professionals anyway, and often without commercial constraints or unreasonable deadlines. This means that open software can be a private room where the software developer can put his best skills at play, and do it in a quality that lives up to the developers often high personal standards. Second, the software being open will immediately be exposed to a community that can provide instant feedback, so bugs get ironed out fast. Third, there often is a feasible business model behind open source software. The major software, such as OpenOffice and even Linux itself, is backed by large companies such as Novell, Oracle, IBM, Intel, AMD and recently even Microsoft.
A DAW system like Ardour is financed through a voluntary donation system. While less predictable, is still manages to support the devolper. In any case, the quality is as good as any commercial software.

For music production you need software that provides some of the same functionality as Logic, ProTools, Cubase, VST, ASIO etc. and has drivers for all sorts of accessories and formats. You also need to have access to effects plugins, virtual instruments drivers for sound cards, MIDI and a system to route sound between programs.

All this comes as standard with an installation of a dedicated multimedia distro like Ubuntu Studio. Here is a description of the most commonly used open source music production software, all pre-installed in Ubuntu Studio, and a description of what it does:

ALSA
- This is a piece of software, that is installed per default on most Linux distros. It is a driver layer, that connects the physical cards installed on the PC to the program layer. The amount of sound devices supported by ALSA is enormous, and most common sound and MIDI cards are actually supported, even though it does not say so on the product package. Even so, before investing money in a sound card, it is always advisable to do a quick Google search, to see if the card of your fancy already has ALSA support. A pretty safe bet is to get M-Audio cards, as they have really good support. In-depth information is available from the ALSA project homepage. which also has a soundcard compatibility matrix.

JACK

- The self-referential name means "JACK Audio Connection Kit". Why do we need another audio layer now that we have ALSA, you may ask? Well, ALSA connects to the cards allright, but you also need something like ASIO or the old Mac OS9 OMS system, to be able to route sound signals between concurrent audio applications. You also need to be able to route MIDI between applications, and finally you need a master clock signal common to all programs, to sync everything up.
JACK does all this, and it does so with virtual no-latency! In other words, JACK is your internal software patchbay and sync master. There are several front-ends for JACK, to build and save patch setups. The one shown here is the default one from the Jack control interface itself. Info about JACK is available from http://jackaudio.org/

LADSPA
- Short for Linux Audio Developers Simple Plugin API. This API is a standard that allows audio plugins in LADSPA host applications. The major audio processing programs all have direct support for LADSPA, but if you want to run effects external to your audio applications you can also host them using the program JackRack that can load a set of plugins and make them available to routing from other applications through JACK. There is a very large set of effects plugins available for this API, that cover the essential areas, such as reverbs, compressors, equalisers, simulators, phasers, choruses... you name it. Also specialised plugins such as phase inverters, comb filters, spatialisers, gates and much, much more is available. Information is available http://www.ladspa.org/. Try a google search for "ladspa plugins" and you will see the amount of plugins available for free.

Rosegarden

- Still the best MIDI editing program for Linux, though this may change soon with Ardour 3.0. Rosegarden is also capable of sound recording and editing, but in this respect it is much simpler than Ardour. The MIDI capabilities are nice however. The program syncs to JACK, so that it may be slaved to the digital audio workstation program, which is likely to be Ardour. This is necessary, as Ardour still does not have MIDI capabilities. Rosegardens MIDI editing lacks a few quantization features I remember fondly from Cubase, but is otherwise a great program. If you are intimidated by the complexity of Ardour, you will be able to make great music with Rosegarden alone. The project home page is http://www.rosegardenmusic.com/

Ardour2
- This is the king of open source digital audio workstations and your most important reason for selecting Linux as your music platform of choice! There is nothing much missing from this impressive package. Recording can be made in any selectable resolution up to 32-bit floating point precision and 192kHz sampling frequency, if you like and your A/D-converters are good enough. You can mix to stereo, 5- or 7-channel surround or any other number of channels. It has busses, sends and effect plugins. It produces exports with CD TOCs. Faders and effect parameters are fully programmable. And so on. There are many other really pro features. Ardour also forms the core of several commercial products, testifying to the quality of the package.
The full integration with JACK means, that you can send signals out of Ardour, into other programs or external effects and back into Ardour again for further processing.
Ardour 3.0 will for the first time introduce MIDI editing capabilities, and given the high standards set by Ardour 2.8, the MIDI support is likely to be of the same high standard. I cannot wait. Until then you may sync Ardour 2.8 to Rosegarden to record MIDI programmed sound. Info about Ardour can be found at http://ardour.org/.

NetJACK

- A very cool spin-off from the JACK project. NetJACK is a service that will enable audio routing between a number of PCs across an ordinary ethernet network. This means that you can build a very large software production setup, where you dedicate different machines for specialised tasks. Some pieces of software, reverbs for example, are particularly CPU demanding, and works best if they monopolise a CPU. Other programs are very memory intensive, such as sampler programs. Finally, there are programs only developed for Windows, that have to be executed either in a Windows environment, or under the Wine API (see below).
With NetJACK, you can dedicate machines for these purposes as you like, and still control everything from one master Linux machine.

NetJACK routes both MIDI and audio, so your old patch chords are now replaced with cheap computer network cables, while your audio patchbay is a network switch! Pretty cool, I think!

Netjack information can be found here and here.

LinuxSampler
- This is a really ambitious open source project, building a piece of sampling software that is compatible with the Gigasampler format. This means that the software gets access to huge libraries of samples in this format. You can build and create your own instruments too, and save them in the Gigasampler format.

The software is native client/server based, which means that it is built specifically to run the sample playback engine as a server on a backend machine, through a computer network, while control of the sampler is done through one of a selection of possible client GUIs from yo ur desktop. LinuxSampler has a Java client GUI called JSampler for instance, and the LinuxSampler package itself has another called QSampler. This means that you only have the controller GUI taking processing cycles on your main desktop, while all heavy CPU and memory intensive action happens somewhere else.

Control from GUI to server happens across a standard TCP ethernet network, and is configured directly in the client and server itself.
DSSI
- This is a newer API for audio plugins on Linux. DSSI works in a similar manner to VSTi for Windows and Mac, but has its own set of Linux based plugins. Still, we would all like to have access to all the VSTi instruments available for Windows. This is also possible with the dssi-vst package, that provides a host for plugins developed for Windows such that they may be executed under Linux. Dssi-vst builds on the more general Wine API that enables many more Windows applications to execute under windows. With dssi-vst you can run most af your favourite synth plugins with full MIDI and audio integration with JACK - which means with other Linux music applications.
Wine
- A very important project that provides a nice "compatibility layer", that enables a large number of Windows programs to be executed on a Linux machine as well. This comes in handy for instance when you have a program like your ancient Win95 editor for the Korg X5 synthesizer. With Wine the editor runs and integrates through the underlying Linux layer. Running Windows programs under Wine is a bit hit-and-miss. This is not really Linux' fault, but has more to do with the fact that many Windows programs exploit undocumented or badly documented features of the Windows API. This makes it very hard to make Wine fully behave like Windows, but Wine is rapidly improving all the time, and more and more Windows programs actually work under Linux. Wine is a standard package on most popular Linux distributions.

More advanced hackers notes

First notice that any tweaks you make under your Linux installation, will not affect the Windows installation you might have on the same machine in any way. Your Windows installation will be totally unaffected and unaware that you use the same hardware for something as heretic as running a different operating system! Of course, changes to the BIOS will affect Windows, but this has nothing to do with tweaking Linux anyway.

Much as gamers tune their machines to optimal performance, Linux music users can also benefit from tuning their machines for optimal music use. There is a nice command-line tool, or "script", that you can run on your machine, that will measure all sorts of settings on your machine, and report parameters that need tweaking for optimal performance. I have put the script here, although scrictly speaking it is not really mine: realTimeConfigQuickScan.pl. (To run it, you may have to install a script language called Perl first). Mind you, you don't need to tune your machine to start running music software, but once you get familiar with music on Linux, you might start considering it.

The script is run from a Linux command line, and will come up with a number of suggestions for system parameters you ought to tweak. These include io settings for hard disc access, settings for memory locking, real-time kernel settings and a number of other things that can make a machine run better and faster for music use.

There are other settings to play with. Personally, I have installed two M-Audio cards on the same machine. One is a Delta 1010, and the other an Audiophile 24/96. Both share the same chipset, and so they may be daisy-chained in ALSA to appear as one large sound card with 12 channels in and 12 cahnnelse out. If the second soundcard was replaced with another Delta 1010, I would have 16 analog channels in, 16 analog out plus 4 digital channels in and out. This requires editing of a settings file called ".asoundrc", which can be found in the users "home"-folder.

All user settings for a Linux user can generally be found in the "home" folder "/home/username", and, as in the instance with the ".asoundrc" file, the settings file often has a "." as its first character.

Background courtesy wallpaperstock.net

Choice of music computer

Latency and realtime kernels

Music software on Linux

ALSA

JACK

LADSPA

Rosegarden

Ardour2

NetJACK

LinuxSampler

DSSI

Wine

More advanced hackers notes