You are not logged in.
Pages: 1
As far as I know ALSA supports 4 sound formats, S16_LE, S24_LE and S32_LE and FLOAT_LE
My soundcard plays S16_LE and S32_LE fine, and for FLOAT_LE and S24_LE fail to play at all.
I found it a bit surprising that I can play 32-bit format but not 24-bit format, isn't 32-bit format superior to 24-bit to begin with? (Either way I have an X-Fi Titanium which supports 24-bit format, just not on Linux, fuck Creative!) Then there is Floating Point/double precision which is even better than these others.
Since I can force ALSA to use a specific format, I was wondering if I should force it to use S32_LE or S16_LE or just leave it at whichever format the source wants to use? And also if there would be any point in using S24_LE at all if we already have S32_LE which works fine.
Last edited by rabcor (2014-07-10 23:44:18)
Offline
Where did you get this information? Take a look here, they are probably limited by the hardware side.
Dmix will be used along the signal path with a high propability, which leaves you with 24 bit either way. What is your output stage (in a global sense)? Or what is your whole setup?
Ask Creatives engineers why 24 bit was left out. If I had to implement it, I would do it at the cost of three times the latency compared to 32 bit, although I look at it from a "more components, not just one X-Fi chip, with no other limitations" perspective.
Last edited by emeres (2014-07-11 00:26:33)
Offline
Well I didn't get this information anywhere per say, I just tried forcing these settings in my own config to see the results.
So you're saying 24 bit has 3x the latency of 32 bit? But 32 bit really is the best right?
I ended up forcing S32_LE since S16_LE could lead to bugs for some suorces (Pepper flash is one), S24_LE gave no sound and FLOAT_LE gave no sound either.
Last edited by rabcor (2014-07-12 15:35:06)
Offline
So you're saying 24 bit has 3x the latency of 32 bit? But 32 bit really is the best right?
No, I meant that if I had to implement 24 bit resolution support for the hardware codec/controller/dsp that already supports 32 bit resolution, I would go for the trade off of 3 times the latency, compared to 32 bit for the sake of simple implementation (I assume some hardware restrictions here, simply filling 24 bit data to 32 bit with zeros would be easier). That however is a comparison of latency (strict delay in this case) between the data being send and processed, so the actual latency difference in audio might not exist at all. This remark has nothing to do with alsa whatsoever.
"The best" is not empirical. What is your setup? Do you know what mass inertia is? Quality is a difficult topic and one has to consider the whole signal path/chain: beginning from the source, through all connections and amplifiers, ending with speakers or headphones. Is 32 bit better then 24 bit? That depends on the circumstances (like emi, elements along the signal path that alter the signal etc.). If you only take resolution in consideration, then yes, since it is more precise representation of the original signal.
Last edited by emeres (2014-07-12 16:01:20)
Offline
Pages: 1