Subject: Re: [linux-audio-user] Resampling help
From: Fernando Pablo Lopez-Lezcano (nando_AT_ccrma.Stanford.EDU)
Date: Thu Jul 25 2002 - 06:09:23 EEST
> >Anyway, I am really sort of confused as to whether this is
> >the best quality way to resample because I just resampled a
> >900 meg file in about a minute! That seems odd to me because
> >in soundforge this takes about 30+ minutes. Can anyone suggest
> >a sure fire method of getting the best 48kHz > 44.1kHz resampling
> >results in linux?
I just got back an answer back from Julius Smith (the author of resample)
and he says (regarding quality - I was asking about an old command line
option called -aaa that I recalled seeing in old resample versions):
> Hi Fernando,
>
> The default is pretty darn good (80 dB quality given a 20% oversampling
> factor). The -aaa option was replaced by the -filterFile option, and the
> old 'aaa quality' filter file is still in the distribution, and as I
> recall, it allowed only 10% oversampling for comparable quality. However,
> I'm not pushing it because its cut-off frequency is set a little too low,
> which means it is working a little harder than it should for the quality it
> gives. The default filter is nicely balanced.
>
> I suppose I should come up with a well balanced 10%-oversampling filter
> sometime this summer, since that's what we normally get with CD sampling
> rates. Then perhaps I could make options "-dat" and "-cd" (with perhaps
> -cd becoming the default, since computers are really fast nowadays, after all).
>
> Since my filter-design program is distributed with resample, anyone can
> design new filters. It's kind of strange to me that nobody has ever sent
> me any new filters to include in the distribution. The resample home page
> actually asks for help on that topic.
>
> Thanks for the ping,
> Julius
So, I'd say it is pretty good _and_ fast.
In the source distribution of resample look at largefilter.h, it further
documents what Julius says above. The "-expensiveFilter" options is still
there in resample, albeit not documented. Apparently the gain in quality
is not too much related to the run time performance hit. I guess you could
use that option and slow things down a bit :-) And then compare with the
default filter to see if you can hear any difference.
-- Fernando
This archive was generated by hypermail 2b28 : Thu Jul 25 2002 - 06:07:22 EEST