Linux: It's Not Just For Computer Geeks Anymore

[Cheez]

gutturalpiss, so it appears that both Ardour and Hydrogen still has VST issues. Well, I'm glad to hear that the Linux community is going strong and I certainly hope they will eventually sort out the DAW issues. With 2/5 and 3/10 success with VST is less than even the bare minimal unacceptable level. But I'll keep a watch on it's development. It appears at this stage, it is not near where it should be in terms of DAW apps. Since you already have Linux running, do you mind trying out Linuxsampler and tell me how that runs on your system? There are free gig files you can download to try out the sampler.

Actually, I seem to recall sometime in the past (during their development phase), Opensynth (Open labs) considered using Linux as it's OS. Then they have decided to drop it and do Windows instead.

I would really like to try out Linux. But problems: 1. No time, 2. Require another boot partition to my already dual booted system - and that will require Partition Magic which will unfortunately takes up background resources. In my current setup, because I'm using XP and Pro in 2 partitions, I don't need third-party partition software. My DAW partition is stripped to the bare minimal with XPLite Pro (no IE, any hint of explorer shell within Windows had been totally removed - extremely fast and stable as a result; it's like XP engine on a 98 shell).

 

[bongman]

Since musicians don't usually have time to spend on twiddling with IT espeically if they have a project right up the cornder (and their DAW crashes etc), there may be a market (although small) for an IT person to help them solve this problem. Think you can be this person?

oh boy, I think another person that is doing this is Mr. JOEL68, I've met him and he said he helped quite a number of people with their MIDI and computer setup. actually I'm as good as any other user... can do only lah...

BTW, I didn't know so many geeks here, good to know that !!!

 

[gutturalpiss]

thor666: heh. I guess they gotta attend to those things. The guys should atleast provide a GUI installation (including GRUB) like the official Ubuntu, and also auto-mounting features especially for USB drives and NTFS. Still new..grabbing the official distribution and then downloading the Ubuntu Studio packages should be better, although u miss out the good theme and configuration.

bongman: closet geek here :lol:

cheez: Not Hydrogen. I was talking about Rosegarden. Hydrogen is the most stable and feature-packed, 0 problems. Alright. Installed QSampler on my working system as currently I have no testbed. Pre-packaged audio distributions come with all available stable applications, such as this. Take note LinuxSampler by itself is command-line only, since its a back-end, so JSampler and QSampler are the front-ends using different GUI toolkits (JAVA and QT). Read that it also is a good server application for networked audio workstations, and was made to suit such situations best.

I had some gigasamples lying around, most short ones for drum kits and keyboards/pianos..the interface is straight-forward, looks-wise JSampler is better since its using Java. There is a page to edit instruments/gig files. I ran with the sample loaded up to RAM, around 600MB+-. This one is a full electronic guitar complete with its own effects such as reverb, chorus, delay, distortion etc. I am able to edit the parameters, 2 controls each..Playback is good. Untweaked Linux platforms have bad audio playback quality, not a good thing. Future kernels will revise that deep inside kernel-space so then it will not be a problem anymore. My system is lightly patched, so it's all good.

I didn't know whether LinuxSampler was fully optimized for multi-core (smp). So i took out 1 out of 2 total GiBs of system memory to see if there were any bottlenecks. This sample is well within the mid to high-end range, natural enough for non-musicians although there are synthesized hints, so it should be tasking. Sure enough, i faced lags. I killed some external processes, and there still were lags, my HDD was ticking. So in the end, I figured out SMP support wasn't there. Read around, it was right infront of my face. They have SMP and clustering planned for future releases.

Overall, its role as a sampler is quite alright. To add to that, its role as a gigastudio alternative is possible, since the author has outlined that most of the work in fact goes into that. It would probably take some more time to benchmark and truely test its advantages and usability, so for now I can't say much except that I had no issues with it. Apparently the sources are platform-independent, but whether someone has made a windows binary is the main question. Mac there is. If not u can compile it. Version used: 0.4.0 source, compiled with architecture- and system-dependent features

http://www.linuxsampler.org/features.html = good outline of features

OT: http://www.youtube.com/watch?v=xOvP6JMW5ew (uninteresting video of rosegarden)
http://www.youtube.com/watch?v=ypCTmtIa_DQ (new pc/mac/linux commercial)
http://www.youtube.com/watch?v=72HcTx2yAds (humour)

 

[Cheez]

What was the latency like when using Linuxsampler? Gigastudio's low latency was due to various factoirs: 1. using their own GSIF drivers (not ASIO), 2. running at kernel level. I'm not sure if Linuxsampler is able to get around these issues, and I'm not familiar enough with Linux - whether it is possible to run at the kernel level or even if doing that is necessary at all.

Running the sampler as a standalone playing sounds is one thing. Using the sampler within a sequencer is another issue. There may be added latency. Have you tried Linuxsampler within, for example, Hydrogen or Rosegarden? How many notes polyphony were you able to achieve before you get into clicks and pops? Have you tried maxing out the RAM to see how many samples you can load?

 

[seekz]

You just made me download ubuntu.

Will give my verdict soon.

 

[gutturalpiss]

Ok i tried with both my onboard sound and m-audio entry level card, was surprised that the intel-hda ich7 chip could maintain 3ms for each session and 5ms for final (stable, no fluctuation), but heavy tasks give me clicks and pops although there were few "Xruns".

JACK is the host, so one would configure and start jack first before other apps. I had Hydrogen, Rosegarden, QSampler and Ardour all start once jack was up. I programmed the drum tracks first with Hydrogen (1.3ms), moved them to Rosegarden to do some fine tuning (1.4ms). Loaded up a few samples in QSampler, small ones for keyboard and effects, took about 1.2GiB of memory (1.4ms). I don't have much samples and used a basic synth controller..so can't really report on that. Imported to Rosegarden, got ready to do final render with Ardour. Polyphony iirc was 64 at 24bits throughout for Hydrogen/Sampler. Again I don't have a good ear to notice any decrease of polyphony as I'm not a professional heheh. I finally brought them in to Ardour and multi-tracked with my guitar, latency was stable at 1.5ms, but low enough for serious work. Notice that there weren't any significant drop-outs for latency, will get to that later.

The configuration of JACK is what counts. Basically u specify some parameters and that will determine what kind of performance u will get. So there's abit of trial and error if hardware is not professional standard (if it is, u can just apply the best options). The reason why latency is stable is because of kernel-level operation, yes. What really counts are real-time kernel preemption, schedulers which one can configure when compiling a kernel (that is if u don't run a pre-configured DAW distribution in which case its already done), and few tweaks to whatever audio system that resides (sometimes ALSA+JACK is prefered, so one also has to configure ALSA for some tasks and so it will not interfere). All these will make sure that u have a good latency without a single drop-out, although that greatly depends on hardware and software configuration.

Testbed is more of a common mid-range desktop rather than a workstation:

Intel Core 2 Duo @ 3.4GHz CPU
2x1GB A-Data @ CL4 RAMs
320GB Seagate 7200.10 HDD
Gigabyte DS4 mainboard
M-Audio Delta 44 sound card+i/o device on 32-bit PCI bus
No monitors, desktop Microlab speakers output from M-Audio device :lol:
Gentoo Linux operating system

So it looks like there needs to be some form of tweaking to set up a Linux DAW, and possibly needs more attention than MacOS or Windows initially. But projects like JAD (http://www.jacklab.org) will very soon change that.

 

[Cheez]

That's very helpful information. Thanks for taking the time to test them. The latency is surprising low, even lower than some Windows setups. With that latency, it is almost not noticeable.

I didn't quite catch whether you tested latency for live-playing of the samples. This is the crucial bit for samplers, not so much of the playback (since playback is automated anyway). The programming part (actual playing of samples) need to have low latecncy or we will have problems sequencing.

 

[gutturalpiss]

No problem. This is the first time I'm trying out a sampler in Linux, and also one of the few times I'm dealing with a sampler so I don't have much experience. Last time i tried out Gigastudio, so-called pirated version of it to check out what it was :lol: I think the gig samples i have are from then but are still v3. I kinda like this, but QSampler has now crashed 3 times and required a full system reboot once otherwise it fails to start again.

Yes sorry i wasn't clear, i meant the actual programming part of QSampler with a midi controller. Of course the main point of low latency is to make sure real-time syncing doesn't go wrong, and even if its somewhere between 2 to 6ms i think it is acceptable? Once u work at kernel-level, as is also the case for Gigastudio and maybe some Windows DAW apps, latency will fluctuate less since its nearly close to hardware levels.

Do note that my system is the result of many trials and errors, so I definitely can't say my test is proof of Linux DAW capability out-of-the-box. I have not tried out the pre-packaged distributions like Ubuntu Studio, JAD (i was only alpha testing for a while), or even those mentioned in this article except Dynebolic, so maybe one can expect the same results from those since they are of professional configuration. Also, most of them look like they come from the Win 98 days, so I don't really favour that kind of material. JAD and Ubuntu Studio stand out because of their modern look and feel.

Another notable project is Planet CCRMA by Stanford University, from their Center of Computer Reasearch in Music and Acoustics: http://en.wikipedia.org/wiki/Planet_CCRMA

This article is also a good read about what can be expected from Linux as a DAW solution and who may benefit and who may not, although its old, also details the project called AGNULA which is like Planet CCRMA: http://www.soundonsound.com/sos/Feb03/articles/linuxaudio.asp

 

[dee]

CCRMA in Singapore in August to run sonic arts workshops

hi.
FYI, the guys behind planet ccrma distro will be in Singapore in August to run a series of sonic arts workshops at Republic Polytechnic:
http://rp.sg/ccrma/