Richard Rauch wrote:
> On Mon, Nov 29, 2004 at 02:53:42PM +0000, Simon Truss wrote:
> 
>>Richard Rauch wrote:
>>
>>>Thanks (and thanks to other contributors) for interesting insights.
>>>
>>>I guess I'll record at 44.1KHz then for convenience at burning CDs,
>>>unless I find myself in possession of a much better audio source
>>>than the tape players I have.


>>I usually capture at the highest resolution possible and then convert
>>down after post processing. while I agree that 44.1/16 sounds the same
>>as 48/24 I believe that any significant amount of processing will reduce
>>the effective resolution of the signal.
> 
> 
> I have some issues here that get in the way:
> 
>  * Honestly, most of the tapes that I'm working from are not that
>    good.  The ones that are very clean are also the ones where
>    high fidelity matters the least (interviews---as long as the
>    interview is clear, I'm going to move on).

Sounds like 16/44 will do the job for your purpose then. gwc may prove
useful. be careful with it, too much denoising blindly applied can
introduce strange squeals and squelches so listen out and back off if
you do use too much. I've been very happy with the results I've
obtained so far. If the end use is to please the senses then that is the
criteria to measure by. My job requires more from the filters
we use however.

>  * I do not think that any of my sound hardware can sample at more
>    than 16 bits per sample.  Not with NetBSD drivers, anyway.
>    If I wanted to play with this, what are some rough guides to
>    the cost of such sound hardware?

I patched a  £6 cmi8738 to sample 24bit(spdif) for the linux drivers.
It was just a matter of setting the 'use 24bits' in a register and
ensuring the dma buffers were larger. The 'just' took me a while to
discover :-/

>  * At least one of my systems appears to have some internal buzz on
>    the motherboard audio.  I assume that it is interference from
>    the fields inside the box.  That limits the quality of recording
>    that that machine can do.

Thats not so good. I use an outboard DAC/ADC and spdif to transfer data.
Most sound cards I've played with (on the analogue side) have all been
reasonable to good except the mic inputs. I have personally never had
any luck with creative and I stopped trying when I saw a top of the line
creative Audigy something underperform compared to my £6 card :-)

If you choose to spend money then an outboard ADC may be worth looking
at. Either USB (unknown quantity for me) or an analogue ->
spdif/TOSlink. Both can be obtained at reasonable prices. Im afraid
I cannot offer any really useful advice here.

>>The reason pro systems use 48KHz is they have a wider stop band, thus
>>lower pass band ripple. This enables cheaper or higher quality filters
> 
> 
> My familiarity with signal processing and audio engineering terminology
> is bounded by a sphere of radius epsilon, where epsilon << 0.  (^&
> 
> Low-pass/high-pass I gather has to do with passing frequences below/above
> threshholds.  What is "ripple" in this context?

The design criteria of practical filters means that there is some
distortion either in band or out. A perfect filter would pass 100% of
the signal in band and 0% out of band, thus the response is flat wrt
frequency. Ripple refers to the variations in response in the pass and
stop bands. Nice pictures of this on JOS' page.

http://www-ccrma.stanford.edu/~jos/resample/Theory_Practice.html

There is lots more such as group delay and phase responses but thats for
another day :-)

Simon