Linux Audio Development

JulesVD writes "There is an article from Linux Journal about the latest plans for Linux audio functionality from the first developer's conference in Germany. Developers from more than a dozen countries attended this successful conference, representing organizations such as SuSE, Linux Audio Systems, Stanford University, IRCAM and Centro Tempo Reale. Topic discussions included in-depth presentations of the rapidly evolving Linux sound system, a look at the details of programming for professional audio standards and a survey of recent applications and audio-centric Linux distributions." Mmm...interesting reading (blantant plug for cool program), but I think the most important question is will it make Scrubby happy?

Comments

Sound Support

[rf0]

ALSA is actually quite nice. It works on more cards than OSS, allows duplex on some cards and is still backwards compatible with the OSS API. However I still feel that audio support on Linux could be improved. For example in my latop Intel soundcard which seems to only play at 44 Khz if using xmms. Now I know this is a software problem but it would be nice if it worked out the box.

Not a major issue but annoying

Rus

Re:Sound Support

[0x0d0a]

ALSA is actually quite nice. It works on more cards than OSS, allows duplex on some cards and is still backwards compatible with the OSS API.

Agreed, ALSA *is* nice.

ALSA's biggest drawback is the project policy that software mixing should be done in userspace (presumably by a separate project). They feel that a user should use hardware mixing *or* software mixing provided by a sound server like arts, esd, JACK, etc. This produces less kernelspace code, which is good, but means that Linux cannot handle using hardware mixing until all channels are exhausted, then fall back to software mixing for additional channels.

Esd and arts do not provide solid enough latency and sync to be the sound server in this scenario. JACK may be enough -- haven't used it -- but it's not in common use outside of the audio content creation software area.

OSS/Free has a similar approach. Hardware or software sound mixing, no hybrid approach.

A mixer capable of supplying such mixing would need to intercept all ALSA calls and use realtime scheduling. Kinda nasty. It always seemed that adding mixing and resampling code to ALSA would be easier.

It *is* a bit of a thorny problem, though.

Re:Sound Support

[Anonymous Coward]

0x0d0a (568518) wrote:
> ALSA's biggest drawback is the project policy that software mixing should be done in userspace (presumably by a separate project).

Since 0.9.1 ALSA can do mixing itself, with the so called dmix-plugin. You can throw away soundservers now, except of course JACK which is a lot more than a simple mixing sound server.
-- fbar

Re:Sound Support

[paulbd]

LSA's biggest drawback is the project policy that software mixing should be done in userspace (presumably by a separate project)

This just is not true. ALSA now has the dmix plugin that handles software mixing in user space all by itself. Its very, very efficient and has no impact on latency (though it can't offer JACK-style sample-synchronous execution). dmix makes regular software mixing "servers" irrelevant, and JACK fills the remaining needs.

A pro audio platform would be cool...

[japhar81]

Something like Logic or Cubase on Linux would absolutely kick ass. My powerbook beats the hell out of windows for music apps, but an open-sourced suite would be worth it to switch. You have no idea how often I wind up thinking, geez, if only logic had an extra slider here to control this or that or the other...

Re:A pro audio platform would be cool...

[fruey]

It's called Ardour, look for it at Sourceforge.

-Simon

Re:A pro audio platform would be cool...

[mickwd]

A guy called Austin Acton has put together the Mandrake Audio Workstation HowTo for Mandrake 9.1.

It uses packages contributed to Mandrake 9.1 to build an audio workstation (including a low-latency "multimedia" kernel) - using URPMI to simplify package dependency issues.

Quote from the HowTo: "You can setup a professional quality audio workstation in an afternoon or less, with Mandrake Linux. No compiling. No text editing. No dependencies. It's this easy.".

I don't know enough about computer audio to comment further, but you might be interested in checking it out.

Re:A pro audio platform would be cool...

[blinder]

Ardour is nice... on paper. Compiling it is a whole other story.

I'm affraid its just a bit too immature right now to be considered a serious contender to ProTools because only the geek-elite can really get this thing compiled (which I sometimes consider myself at different times of the day) but I have yet to get this thing compiled.

If they would just release a stable build or just a stable distribution that could be built (ala configure/make/make install) then it it would be a serious contender.

I see from Ardour's mailing list that folks do have it up and working... ugh... guess I'll try again tonight to get it built :)

Re:A pro audio platform would be cool...

[Doc Hopper]

I ran into this same problem often with Ardour and others. I gave up, and installed Gentoo. Although it takes a bit of "geek eliteness" to get Gentoo installed (knowing how to compile a kernel, mount filesystems, work with Grub, etc.), it's worth the effort.

Once Gentoo is installed, installing Ardour is a one-liner:

# ACCEPT_KEYWORDS="~x86" emerge ardour-cvs

If your system is slower than 1GHz, check on it every few hours. You needn't do much else while it builds Ardour and all the dependencies to get there. I recommend starting this right before you go to bed, by morning it might be done. Gentoo is a bit like baking a turkey. Yeah, it takes a very long time to cook, but if you took the time to prepare it right, the result is beautiful and tasty. Unless, of course, you're a vegetarian...

That's really what it broiled down to with me. I'm far more patient with allowing stuff to build automatically over hours/days and not using my main PC (lucky me, I have three) for hours/days than with sitting in front of the console for hours working out dependencies by hand.

Aaah Sweep.

[gowen]

Sweep development:

Two messages in five forums

When the forums outnumber the posts, you know you've ound a real classical sourceforge project.

the drivers need to work. period.

[Anonymous Coward]

Anything less than complete parity with Windows drivers FEATURE FOR FEATURE is unacceptable. Linux is STILL not there.

I installed Windows XP after 9 years of running Linux (various distros) because I was tired of only being able to use half the features of my hardware.

Re:the drivers need to work. period.

[QuackQuack]

I get around this problem by buying only well supported hardware.

But what's really annoying is when features aren't supported on Linux for non-technical reasons. For example, Matrox has written Linux drivers for their graphics cards, but the Linux drivers don't support TV-out. Why?

It's apparently because Linux doesn't have Macrovision support, and I guess Matrox doesn't want to be sued by the MPAA for releasing drivers that "enable piracy". So Linux users can't have a feature that has many legitimate uses, just because someone might tape a DVD onto a VHS tape with it.

Re:the drivers need to work. period.

[stratjakt]

Matrox has written Linux drivers for their graphics cards, but the Linux drivers don't support TV-out. Why?

It's apparently because Linux doesn't have Macrovision support, and I guess Matrox doesn't want to be sued by the MPAA for releasing drivers that "enable piracy". So Linux users can't have a feature that has many legitimate uses, just because someone might tape a DVD onto a VHS tape with it.

That's a good conspiracy theory. Another theory is that linux cant dynamically add/remove display devices in userland at run time the way windows can.

You'd either have TV out enabled all the time, or CRT all the time, or both, but you wouldnt be able to switch at will.

It could also be because of the MODELINE nonsense in your xf89config. TV out has some pretty specific frequency requirements, and perhaps the ability to tweak those while enabled would damage the card, and then void the warranty - but if you voided the warranty with a matrox driver, you dont void the warranty, so they have to continually replace cards for all the 'tweakers' out there.

Or it could be that you're looking a gift horse in the mouth. Want all the features of the vid card to work? Fire up emacs and get a-crackin'.

BTW, IIRC macrovision is a hardware feature, not a software one. There is no 'linux support' to be written, save flipping it on via a hardware register. They're only required that it be present, it can be disabled as easily under windows (a registry tweak for most cards) as it could be under linux.

Re:the drivers need to work. period.

[Anonymous Coward]

I understand where it's coming from, but that's a dangerous attitude to take. The only reason Linux has problems with hardware support (along with any other would be OS), is that hardware vendors will not release development info. They want to keep it under NDA with Microsoft.

I have worked very hard in the past to get this info from vendors in the past, both legitamately and by using my company and inventing "fake" projects just to get the data. The fact is, if you are not microsoft, they do not want to talk to you unless you're prepared to layout serious money. We're not talking $1,000 or even $100,000 either.

The only correct answer, as one poster explained, is to only buy well supported hardware. Reward the companies that play fair.

Re:the drivers need to work. period.

[Gordonjcp]

I can do TV out on a plain vanilla Geforce 2MX 400 - no TV out at all. The RGB cable is simple - just connect the VGA RGB to the SCART RGB then mix the HSync and VSync signals and feed them into the SCART composite in.

There is a special modeline you need, which puts it into a 704x256 mode (really!) scanning at the required 25Hz. It would be pretty simple to adapt this for 30Hz for USian TVs.

As for switching modes on-the-fly, this has been possible for a very, very long time. Try pressing to switch modes...

Drivers?

[Drathus]

Isn't the largest issue with any support on Linux still the fact that very few companies are willing to put in the time to create Linux drivers let alone have have the decency to release the information on the chipsets to the public to allow 3rd party drivers to be created?

Maybe what should be looked into is the creation of a vested interest just to deal with sound card companies and try and negotiate accessability to chipset information to create drivers for their cards for other OSes than they're willing to develop for in-house.

Re:Drivers?

[rf0]

The only company that I know is producing commerical drivers for linux is 4Front Technologies. No idea on the quality though
Rus

Re:Drivers?

[Drathus]

Well, they don't need to be commercial drivers per-se (where the drivers are sold as-is) but rather a commercial body (who knows, maybe sponsored and funded by the for-sale Linux distros) which would negotiate and create the drivers to distribute back to the community which funded them.

They'd just be the central group who'd be dealing with the companies to get the information.

Re:Drivers?

[Adnans]

Contrast with the monolithic kernel that linux has. Creative would need to participate and coordinate with linus et al from start to finish.

Erh no! Most, if not all drivers in ALSA were written without any interaction from the kernel folks. ALSA is now integrated in the 2.5.x kernel, but that doesn't mean driver developers will have to deal with Linus et al. They just deal with Jaroslav, the ALSA maintainer. All mainstream cards are already supported by ALSA. If a company doesn't want to provide docs it can always choose to write and distribute their driver on their own.

In short, the monolithic kernel is an albatross around linux' neck when it comes to wanting hardware support from the manufacturers.

Nonsense.

-adnans

Re:Drivers?

[Nosher]

Creative are actually pretty good about supporting their soundcards under Linux (http://opensource.creative.com/) with their EMU10K1 project. I have an Audigy running in Mandrake 9.0 and it sounds pretty good (not much difference to an Audigy Platinum EX I've got running in a Windoze PC - both run through a decent amp and good speakers). All it takes is a bit of compilation (of the driver) and a change to /etc/modules.conf. I even use an Audigy in a custom "suitcase PC" running Mandrake to do the occasional disco (line up a few tracks and head for the bar :-). It has been frustrating though that it's taken *sooo* long just to get to this point though...

Re:Drivers?

[Adnans]

And ALSA is the way every card will work from now on? Or will it change to something else with kernel 2.6?

The ALSA API in kernel 2.5.x will be the same one in kernel 2.6.

Vendors like standards and specifications. They dont like researchers and academics and expiriments.

Vendors look at the bottom line. If there's enough incentive they will write against any API (*cough* Windows *cough* :-).

-adnans

Re:Drivers?

[Dysan2k]

#1) Releasing specs on chips like the EMU101k (for example) hurts their market position. That stuff is trade secrets, and patents and whatnot. nVidia doesnt just give out schematics for their new FX chips, after all. Nor does creative (or whomever) want to hand out their fancy-dancy sound chips.

I'm a little curious just because I've not had to do any driver hacking except fixing/tweaking, but is it really possible to duplicate someone's chipsets with the driver info released? I mean, as far as I knew, you just need the timing, addresses, and buffer sizes to fire up most pieces of hardware. Is there something that I'm just missing? (Honest question here folks.. not trying to troll or anything)

After dealing with Linux sound for 3 years.....

[mao che minh]

"Topic discussions included in-depth presentations of the rapidly evolving Linux sound system..."

I hope so. After working with Linux for three years I have come to expect little in terms of audio. Hell, I was taken completely by suprise when the Redhat 8.0 install actually had a "play sample sound" button. It was like first time I witnessed scaling effects on the SNES. Inspiring.

Yet a long way to reach DirectX....

[WARM3CH]

Yet there is a long way to have an API with DirectSound/DirectSound3D/DirectMusic/DLS features in Linux. (I don't know about EAX support in Linux, anyone can help?) Sound is one of the major obstacles for the games to come to Linux....

Re:Yet a long way to reach DirectX....

[Sloppy]

I don't know about EAX support in Linux, anyone can help?

You get EAX if you run Linux on a 386sx or better. The PPC, Alpha, and SPARC ports don't have it yet, though.

(PS: I apologize.)

Mixing...

[ElGuapoGolf]

I may be wrong, but whatever. It seems to me that Linux *really* needs a decent stream mixer.

I hate not being able to play more than 2 sounds at once (and that's only because that's supported in hardware by my card, my old card could only play 1). Neither esd (does anyone use this anymore) or artsd cut it. They're too laggy to be usable for games, and in order to increase their response time, you have to increase their priority, thus slowing everything else down.

Say what you want about Windows, but at least it gets this right.

So, anyone know a soundcard that will let me play mutiple streams WITHOUT having to use esd/artsd, and is decently well supported under Linux? Anyone? BTW, can we keep it under $100 (USD) if possible?

Re:Mixing...

[Lface]

Trident 4DWave NX/DX based cards are excellent! Not sure if they're still available though. For about $20 you get a card with S/PDIF out and at least 16 stereo pcm devices with ALSA. I bought a couple of them at Hoontech Taiwan (just checked. no longer available from their site).

No, they are no longer available, and haven't been for a while either.

The SoundBlaster Live is an example of a cheap soundcard that does hardware mixing.

Cool

[IamTheRealMike]

The most amazing thing about the Linux audio scene is how well organised it's been since day one. All the key players work together, and are putting together a seriously sweet pro audio architecture. The same people who write ALSA collaborate on Jack, LADSPA (which looks like an open alternative to VST plugins), write sample editors - the works.

It's too bad that the desktop audio scene is such a shambles by comparison - in 2003 there is still no standard way of mixing and resampling (sound servers). What I expect will happen at this point is simply that once 2.6 is rolled out, distros will simply start shipping with the ALSA dmix plugin which mixes at the hardware level by writing into a sound cards DMA buffer (iirc), making most desktop uses of sound servers obsolete (though they can still be useful for network transparent audio and jack style synchro).

Oh, and GStreamer kicks ass :) It's messy right now, but I think Linux multimedia could become really great.

VST and ASIO

[blueworm]

Linux needs VST, and most importantly, ASIO support before it can even begin to be remotely considered for professional use.

Re:VST and ASIO

[IamTheRealMike]

ASIO is a Windows specific hack around the lack of low latency audio on that OS I think, so it doesn't apply to Linux/ALSA which has some of the lowest latencies around with the right kernel patches.

I have to say...

[SatanicPuppy]

I've had a better audio experience with Linux than I ever did with windows.

Case in point: I recently was bequeathed a SB Audigy card (Platinum...Oooooooo.) with no driver disk.

So whay you say! I can download the driver no problem, you say!

NAY! I say, they have restricted the downloads to driver "upgrades" only. If you don't have the original, then you get NOTHING! I had to go rip off a damn copy of the original driver CD to use a physical piece of hardware. Severely annoying.

This is in windows. In LINUX, I found the driver and it worked perfectly. Took like 3 minutes. It was GREAT! No pops or crackles, just pure wonderful SOUND!

My Name is SatanicPuppy, and I'm a switcher.

=P

no VST on Linux

[onash]

I can't switch to Linux on my desktop machine because I would have to stop making music. I haven't found any sequencers that support VST instruments or even a VST host for Linux yet, so this talk that musicians can easily switch to Linux doesn't really make sense to me.. unless all those who make music on Linux have tons of hardware synths and don't need soft synths like poor me :P

If anyone knows about any open source Sequencer with planned VST support, let me know, I would love to help. I searched Sourceforge for Linux VST.. and found nothing.

Re:no VST on Linux

[Adnans]

If anyone knows about any open source Sequencer with planned VST support, let me know, I would love to help. I searched Sourceforge for Linux VST.. and found nothing.

http://boudicca.tux.org/hypermail/jackit-devel/thi s-month/0020.html
Google around for details...

Re:I have a single question

[Consul]

A popular program, under active development, is Rosegarden4.

I haven't used it much, but I was able to get it to run stably, and talk to my soundcard just fine. Many people on the Linux Audio Users mailing list like it a lot.

Creative Open Sources their drivers

[brunes69]

Not only do they create drivers for their chips (SBLive! and Audigy series, OpenAL), they release the code as Open Source. The driver sin the Linux kernel came from Creative, not some 3rd party. Another reason to support Creative (as if having the best stuff wasn't enough of a reason)

http://opensource.creative.com

Re:My head hurt

[Doc Hopper]

Most people who are familiar with GNU/Linux audio development refer to the "Open Sound System/Free Version" as "OSS/Free". Unfortunately, this is often confused with "Open Source Software/Free Software", which is also sometimes abbreviated "OSS/Free".

Even better, I often talk about running something in "UML", or "User-Mode Linux", at work. Until I explained, many people were thinking I was talking about "UML", (I think it's an abbreviation for) "Unified Markup Language". They thought it was amazing I could do all this cool stuff with kernels and isolation using a prototyping & flowcharting language.

When in doubt, use the full name first and the abbreviation later...