Actually, I have done that test already in the course of my original FLAC player shootout. See this old thread on FLAC audio quality. But I'd like to repeat it, with the new version of VLC.

To summarise the results of the original shootout was, that there was no difference between FLAC and WAV for most players, only VLC was good with WAV, and also with Ogg Vorbis and other formats, but it was the worst player of them all with FLAC, and much worse than with WAV.

So it seems, not only usage of the same components, but quality of implementation, internal architecture, and, as was mentioned here already, efficiency in usage of system resources, may influence the result.

gargamel

Hmm, if the library for decoding FLAC is the same and this should make all players sound the same, then old and new VLC versions should provide identical audio output. For FLAC they definitely don't, as far as I can tell, but I suggest to those of you, who claim there cannot be any differences, because there must not be any differences in audio output quality, a little experiment: Install old VLC (1.7 or so) and new VLC (2.0.5) on the same hardware, and compare FLAC audio quality.

gargamel

Does VLC do anything to the audio between decoding it with libflac and actually playing it? From looking at Tools->Preferences->Audio, it would appear that the answer is... it can. Replay gain might be turned on, for example, and the volume might be set higher than the system default. There are also several dozen audio filters that I would assume you haven't turned on.

If you want to check if resampling is being done, you can check by opening the messages window (Tools->Messages) and setting the verbosity to DEBUG.

So, I grabbed my Audio Technica ad700's, a couple of beers and ran through a bunch of my favorite flac songs on different players.. I couldn't really tell the difference between any of the players. Thanks for introducing me to 'play' though, and giving me an excuse to rock out.

Audio quality on Linux is mostly affected by:

• weird sound servers (arts, esound, pulseaudio), which try software-mixing and are not good at it, so make sure, your player outputs directly into ALSA.
• weird AC97 codecs, which resample everything to 48 kHz (most older sound cards from the PCI era are affected, including the well-known SoundBlaster Live)
• and of course sub-par hardware drivers

I was about to post pretty much this.
Any FLAC decoder will output the same or it is not a FLAC decoder.
However, what then happens to the sound will be very important. Personally I can cope with Pulse Audio, just, vut I'd ditch it in a second if I could isten to the oytput of more than one audio source through ALSA.

You actually can, although your audio hardware has to support that.

I'll admit I've not spent a lot of time trying but it certainly hasn't worked out of the box, or with some messing around, on the few cards I've had.
It's always frustrated me that under Windows audio "just works" but Linux means faffing around with ALSA and possibly Pulse, oh, and ESD or whatever it is you have to use under WINE because for some reason your Pulse and ALSA conflict, etc.

Sorry that was off-topic but I had to vent.

I never used Pulseaudio. Back in the old days I used FreeBSD's newpcm, which natively supports in-kernel-mixing like OSSv4. Since 2006 HD audio codecs have hardware mixing available, so the ALSA issue is gone.

Sound servers like Pulseaudio are a thing of the past.

So, without wanting to drag this OT too much, which Distro's allow you to ditch Pulse and still have a mixer? I do mean they allow you to use any source be it Flash, mpd, Kaffeine, VLC or anything else at the same time and not just a subset of specially tweaked apps.

That should be working out of the box (playing more than one audio source through ALSA). If any application is outputting through OSS, however, then ALSA will accomodate it through OSS emulation, which can only play one audio source at a time. The problem, therefore, is that any application outputting through OSS will block the audio device. I usually deal with this by commenting out the load_oss_modules lines in rc.alsa, which disables OSS emulation. Then applications that want to output audio will look for an OSS device, not find one, and fallback to ALSA.

Also, jtsn, my impression is that ALSA does software mixing, which works regardless of your audio hardware.

I think it's the ALSA fallback I'm seeing problems with. I admit I haven't played with things recently but I'd always find one of the applications I wanted didn't play ball -- then it's back to Pulse.
Thanks for the input though, I'll take another look at getting rid of Pulse.

When stuck on incapable hardware, you could try OSSv4, it has a kernel-mode software mixer and allows multiple applications to output through the /dev/dsp standard OSS interface.

By default ALSA offers no software mixing. This was the reason for the sound server plague in the first place.

Sure it offers software mixing. From http://alsa.opensrc.org/Dmix :

Eric

I dunno...I got the vlc 2.0.5 just now and opened it, xine, and audacious and played the first 40 seconds of a song (the same song) in each, 4 times (each) and there *is* a slight difference.

To *me*, xine had the crispest sound, everything seemed 'equal' and I could hear the snare cleaner than the others. VLC was second but compared to xine it got a little 'heavy' in the midrange. Audacious does okay, but compared to the other two it seemed a tiny bit...muffled, or muddy...I guess is the best I can describe.

This was on my slackware 14, with alsa, a SB Live (damn thing's got to be 12+ years old, lol), cheapo koss earplug-type headphones plugged into a logitech z-680 with all the levels set at the center and the effect at 'music'. The mixer is kmix and I have the treble set at 88, the bass at 65, master at 100 and pcm at 93.

The 'differences' I heard between these apps was just barely discernable but still *is* discernable, IMO. I forgot until this waa all typed out to try 'play' <mutter>.

According to ksysguard, vlc used the least amount of RAM just sitting there not playing anything - 4.6 (19.1 shared, whatever that means). Audacious was next with - 10.8 (10 shared). Xine last with - 21.5 (11.7 shared).

None of this really means much as all my stuff is cheapo (except the logitech setup), but figured I'd put in my two cents worth to muddy the pool.