Hello,

Just saw the changelog, any idea when the extra sox recipes will be integrated?

ver 0.19.3 (2014/11/11)
* fix distorted audio with soxr resampler


--
Best regards,

Frederic Vanden Poel

Klinkt Beter
www.klinktbeter.be
[email protected]

+32 473 79 36 91
Blauwe Paal 45
9230 Wetteren



On 3/11/2014 10:54, Frederic Vanden Poel wrote:
On 3/11/2014 8:37, Max Kellermann wrote:
add new soxr recipes: libsamplerate equivalent SOXR_LSR0Q, SOXR_LSR1Q
>and SOXR_LSR2Q

Tried to summarize this info, but it's still long:

add new soxr recipes: libsamplerate equivalent SOXR_LSR0Q, SOXR_LSR1Q and SOXR_LSR2Q

Since 0.19 mpd implements soxr, which is much faster than libsamplerate (effiency x5 measured on a celeron). Most power efficient cpu's such as atoms and celerons cannot handle libsamplerate samplerate_converter "0" (best sinc), and are limited to samplerate_converter "1" (medium sinc).

While the new samplerate_converter "soxr very high" provides a faster alternative, it sounds a little more harsh. soxr provides recipes that emulate libsamplerate settings which are SOXR_LSR0Q , SOXR_LSR1Q and SOXR_LSR2Q, which remove this harshness.

This patch includes these better sounding libsamplerate equivalents:

samplerate_converter "soxr best sinc" uses the SOXR_LSR0Q recipe, and is an alternative to samplerate_converter "0" samplerate_converter "soxr medium sinc" uses the SOXR_LSR1Q recipe, and is an alternative to samplerate_converter "1" samplerate_converter "soxr fast sinc" uses the SOXR_LSR2Q recipe, and is an alternative to samplerate_converter "2"

"soxr best sinc" sounds much better than samplerate_converter "1", and uses a fraction of the cpu load: example for a 24/96 .wav file upsampling to 705.6 kHz on a celeron 847, which my ifi dac supports:

samplerate_converter "soxr best sinc" consumes 20% cpu on one core
samplerate_converter "1 " consumes 95% cpu on one core




--
Vriendelijke groeten,

Frederic Vanden Poel

Klinkt Beter
www.klinktbeter.be
[email protected]

+32 473 79 36 91
Blauwe Paal 45
9230 Wetteren

_______________________________________________
mpd-devel mailing list
[email protected]
http://mailman.blarg.de/listinfo/mpd-devel

Reply via email to