The wrap-around you are referring to does not happen because the lowpass
filter in the converter prevents any frequencies above 22.05 khz from making
it to the converter. The lowpass filter is a "brick wall filter." It
typically starts to rolloff at around 20khz and by the time it gets to 22kHz
should be at -96 db or lower. The steep slope of the filter means phase
shifts happen along with the rolloff. In cheap filters these phase
artifacts often happen in the audible range. This is what causes the loss
of "air." In fact some cheap filters start to rolloff before they get to
20khz. The higher quality filters are designed to minimize the phase shift
and to preserve the sound quality. It's also possible to oversample. You
can for instance sample at 88.2 khz set the filter to rolloff at 20kHz but
only store every other sample. This could sound better because the filter
has more room to roll off and therefore does not need to be so drastic.
There are even 1-bit converters that sample in the Mhz range. These would
take a little more time to explain, but the bottom line is that any
artifacts or loss of sound quality you notice in a recording sampled at
44.1kHz is due to the filter preceeding the A to D converter and possibly
the lowpass filter following the D to A converter to a lessser extent.
My studio uses an Apogee AD8000 SE converter and yes these are very
expensive. But I cannot hear the difference between using this converter at
44.1 kHz and a ProTools HD converter running at 192kHz. I also use Apogee's
dither which is very good and does add apparent resolution when converting a
24 bit recording to 16 bits. I record from the MOTM straight into the
converter, after passing the signal through a very good transformer to make
it Lo-Z.
> I'm aware of the 24 bit technique, and the dithering. But
> I'm surprised that no one suggests the need for high sample
> rates. Afaik, sampling at 44.1 or 48k is asking for
> aliasing caused by signals living around the 20k range (I'm
> told they 'wrap' around the freq roof of an AD. Afaik, if
> I'd use a '300 saw waveform, I get lots of partials in that
> frequency range, and maybe even higher.
>
> Mind you I've never tried it (no 96k option here). But I got
> this knowlegde from web articles written by a mastering
> engineer (can't remember the source right now). What I
> concluded from those articles is that you have to use the
> highest possible sample rate to approach that analog audio
> feel (apart from a jittter-free clock, etc, etc.).
>
>
>
>
>
> Yahoo! Groups Links
>
>
>
>
>