On Friday 12 December 2008 20:34:21 Sergei Steshenko wrote:
> On Fri, 12 Dec 2008 20:24:25 +0000
>
> Paulo Moura Guedes <mo...@kdewebdev.org> wrote:
> >I don't know about ALSA.
>
> So, ALSA, as well as many Linux applications, have a choice of qualities
> for resampling, and I bet you won't hear the difference between middle and
> high quality.
>
> There is no "OS resampling", resampling is a well researched field, and
> resampling quality does not depend on OS, it depends on the chosen
> algorithm.
>
> Again, no serious sound recording company records at CD sample rate, they
> record, I think, at 96kHz - the ADCs at 192kHz do not perform that well.
>
> So, every CD you listen to is resampled.
That's correct, CD quality is 16bit 44.1kHz, which should be much lower as the
recording rate as you said.
Nowadays, it's not so dificult to find (pay) for files with 24bit 96kHz, but
that
is the result of resampling as well.
There is nothing one can do about that though... only make it worse, which is
what I'm trying to avoid ;)
BTW, Benchmark DAC1 resamples internally to 110kHz:
"The process of upsampling does not inherently improve sound quality during
D/A conversion. However, Benchmark converters re-sample for a very specific
reason: jitter immunity. Benchmark converters use a proprietary clocking
system (we refer to it as UltraLock). It works like this... The incoming
digital signal is immediately re-sampled by an ASRC (asyncronous sample rate
converter). The ASRC, as the name implies, is not syncronized to the clock of
the incoming digital signal. Therefore, its performance is independant of the
quality of that clock. In other words, it doesn't matter if the signal came
from a cheap transport with cheap cables, or from a $10,000 signal chain. The
large amount of jitter caused by the cheap transport and cheap cable will be
moot with respect to the ASRC process. The output of the ASRC is then clocked
to an on-board clock with extremely low jitter and strategic sheilding and
board traces. The output of the ASRC can be configured to any sample rate that
we choose, including the original sample rate. However, we dictated the re-
sample rate as 110 kHz because it is the highest sample-rate at which the
digital interpolation filter of the D/A chip will operate optimally. The ill-
effects of the digital interpolation filter at higher-then-110 kHz include pass-
band ripple (non-linearities in frequency response) and inferior attenuation
of stop-band frequencies (which results in aliasing). Therefore, the D/A
performance is optimized by maintaing 110 kHz. Many converter designers have
since employed similar topologies, but use lower re-sampling frequencies, such
as 96 kHz. By resampling to 110 kHz, the low-pass filter of the ASRC and D/A
are moved as far up as possible as to not infringe on the analog bandwidth of
the audio"
"On the question of: Why does the DAC1 re-sample to 110 kHz? Here is why: it
is the highest frequency to maintain the full oversampling of the D-A chip.
EVERY D-to-A chip on the market cuts the oversampling rate in half to
accommodate 192 kHz. This will also implement a different type of digital low-
pass filtering which is inferior to the filter used at and below 110kHz. This
is
also why most recording engineers don't use 192 kHz. The higher bandwidth
seems appealing, but the stat-of-the-technology is such that 192 kHz
conversion is actually inferior to 96 kHz. Also, the DAC1's oversampling ASRC
and resulting 110 kHz sample rate reproduces 96 kHz signals much more
faithfully then a D-A converting the original 96 kHz signal. This is because
the Nyquist frequency is on the slope of the filter (attenuated, but not
completely). This is undesirable for two reasons. The first reason is the
Nyquist frequency is not faithfully converted to analog (ie, the analog
bandwidth of 96 kHz conversion is actually less then 48 kHz). With the DAC1,
the full bandwidth of a 96 kHz signal can be faithfully reproduced. The second
problem with 96 kHz conversion is the frequencies at and above Nyquist (48 kHz
and up) are not completely attenuated, so some aliasing and imaging will
occur. With the 110 kHz upsampling and conversion in the DAC1, the frequencies
below 55 kHz are not in danger of being aliased."
Thanks,
Paulo
------------------------------------------------------------------------------
SF.Net email is Sponsored by MIX09, March 18-20, 2009 in Las Vegas, Nevada.
The future of the web can't happen without you. Join us at MIX09 to help
pave the way to the Next Web now. Learn more and register at
http://ad.doubleclick.net/clk;208669438;13503038;i?http://2009.visitmix.com/
_______________________________________________
Alsa-user mailing list
Alsa-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/alsa-user
