[LAU] 24-bit files and hisss.... GOT IT

Steve Fosdick lists at pelvoux.nildram.co.uk
Mon Feb 16 13:46:46 EST 2009

On Mon, 2009-02-16 at 10:50 -0700, Bearcat M. Şandor wrote:
> ok, so i'm confused.  If you're saying that upsampling is generally a bad 
> thing why does one pay extra for it in  a product like this: 
> http://www.meridian.co.uk/product-model/g-series/g082-upsampling-compact-disc-
> player.aspx ?

One probably gets upsampling whether he pays extra or not.

As I understand it there is a problem inherent issue with D/A conversion
in that you have to stop the clock used to run the D/A converter
appearing on converted analogue output.

When the sampling rate has been chosen to give a nyquist frequency only
slighly above the maximum frequency you intend to reproduce you the
designer has to implement a very steep filter in analogue components.

For CD this is the case - 44.1Khz sampling rate, 22.05Khz nyquist
frequency and you want to reproduce the whole human hearing range.

If instead the sampled data is upsampled to a higher frequency the clock
rate of the D/A converter is increased to match and the slope of the
analogue filter can be more gentle.

Upsampling does not enable you to hear anything not on the CD.  On the
other hand an insufficiently steep analogue filter could cut into the
audio band so an oversampling player may play all of the information on
the disk and not just most of it.

What it sounds like, of course, depends on the quality of the upsampling
algorithm - there are good and bad of those.

Back to the original issue with the sound card - most of the integrated
cards these days seem to be optimised for 48Khz rather than 44.1Khz.  It
may also be the case that the hardware upsamples everything a multiple,
for example 96Khz.  Perhaps the algorithm they implement in hardware is
bearable when the sampling rate a multiple (i.e. 2X for 48 to 96) and
pretty poor when not (i.e. 44.1 to 96) and the ALSA algorithm is a
better one.

More information about the Linux-audio-user mailing list