13 Jun
2005
13 Jun
'05
10 a.m.
Erik de Castro Lopo:
I was primarily testing speed, but both libraries where using the sinc
routine, so...
I recently did a lot of benchmarking between libsamplerate and mus_src
in clm/sndlib. My result was quite interesting, the fastest mus_src sinc
resampler where a lot faster than the fastest libsamplerate resampler.
I think most people would agree that speed is not the most important
aspect when measuring the quality of a sample rate converter.
With libsamplerate, I can state that the three sinc
based converters
have the following characteristics:
SNR Bandwidth
SRC_SINC_FASTEST 102.42 dB 80.23 %
SRC_SINC_MEDIUM_QUALITY 98.99 dB 90.68 %
SRC_SINC_BEST_QUALITY 97.43 dB 96.96%
where SNR is signal to noise ratio and Bandwidth its a percentage of
the theoretical best bandwidth (ie half of the minimum of the source
and desination sample rate).
He did, and Bil Schottstaedt used this algorithm as well for mus_src.
When I compaired mus_src with width=5, and libsamplerate with
SRC_SINC_FASTEST, the first one is much faster, I don't know what kind of
SNR/Bandwidth you get with your way of measuring when using mus_src with
width=5, but the source is easy to read so I guess you can figure it out.
--
A third program is the original sinc-resampler
from Julius Smith:
http://ccrma-www.stanford.edu/~jos/resample/
I don't know how this one performs compaired to the other two though.
libsamplerate the same algorithm as this one. I think Julius O. Smith
developed this algorithm.
13 Jun
13 Jun
10:49 a.m.
New subject: [linux-audio-dev] Re: Looking for fast integer resampling code
Kjetil Svalastog Matheussen wrote:
I was primarily testing speed, but both libraries
where using the sinc
routine, so...
Yes, but you can design a sinc filter with 10 coefficients and one
with 10000. The one with 10000 coefficients will have a steeper
transition band and a better signal to noise ratio.
When I compaired mus_src with width=5, and
libsamplerate with
SRC_SINC_FASTEST, the first one is much faster, I don't know what kind of
SNR/Bandwidth you get with your way of measuring when using mus_src with
width=5, but the source is easy to read so I guess you can figure it out.
The SNR and bandwith cannot be determined by reading code. Measurement
is the only option.
Erik
--
+-----------------------------------------------------------+
Erik de Castro Lopo nospam(a)mega-nerd.com (Yes it's valid)
+-----------------------------------------------------------+
"To me C++ seems to be a language that has sacrificed orthogonality
and elegance for random expediency." -- Meilir Page-Jones
11:18 a.m.
New subject: [linux-audio-dev] Re: Looking for fast integer resampling code
On Mon, Jun 13, 2005 at 10:49:38PM +1000, Erik de Castro Lopo wrote:
The SNR and bandwith cannot be determined by reading
code.
Measurement is the only option.
It's perfectly possible to calculate this, and it isn't even
very difficult. In the case of a sinc filter, everything is
determined by the window on the complete sinc function that
defines your filter, and the ratio of the two sample rates.
In fact the 'measurement' you describe in the presentation
you referred to is not really a measurement. It is 100%
numerical and there's nothing physical involved. So it's in
fact a long winded way to do the calculation.
A Vulcan could probably do it while reading the source :-)
--
FA
6:46 p.m.
New subject: [linux-audio-dev] Re: Looking for fast integer resampling code
Alfons Adriaensen wrote:
On Mon, Jun 13, 2005 at 10:49:38PM +1000, Erik de
Castro Lopo wrote:
Yes, but if you meaure the SNR the way I do, you are also
ensuring the absence of bugs in the implementation of the
algorithm. As you are probably aware, looking at code is
not the best way to find bugs in the implementation but
testing actually works quite well.
Erik
--
+-----------------------------------------------------------+
Erik de Castro Lopo nospam(a)mega-nerd.com (Yes it's valid)
+-----------------------------------------------------------+
"C++ is like jamming a helicopter inside a Miata and expecting
some sort of improvement." -- Drew Olbrich
The SNR and bandwith cannot be determined by
reading code.
Measurement is the only option.
It's perfectly possible to calculate this, and it isn't even
very difficult. In the case of a sinc filter, everything is
determined by the window on the complete sinc function that
defines your filter, and the ratio of the two sample rates.
In fact the 'measurement' you describe in the presentation
you referred to is not really a measurement. It is 100%
numerical and there's nothing physical involved. So it's in
fact a long winded way to do the calculation. 
7:34 p.m.
New subject: [linux-audio-dev] Re: Looking for fast integer resampling code
On Tue, Jun 14, 2005 at 06:46:03AM +1000, Erik de Castro Lopo wrote:
Yes, but if you meaure the SNR the way I do, you are
also
ensuring the absence of bugs in the implementation of the
algorithm. As you are probably aware, looking at code is
not the best way to find bugs in the implementation but
testing actually works quite well.
Absolutely, and testing with simple signals such as sines
often reveals problems you wouldn't notice otherwise. But
when performing such a test, I also want to know if the
result is what I expected to see, i.e. to verify it against
a calculation.
--
FA
7047
days inactive
7047
days old
linux-audio-dev@lists.linuxaudio.org
4 comments
4 participants
participants (4)
-
Alfons Adriaensen -
Erik de Castro Lopo -
fons adriaensen -
Kjetil Svalastog Matheussen