As a rule of thumb, I do find that higher sampling rates of 88.2 kHz or
more make a considerable difference when it comes to removing impulse
noise on mechanical carriers (records with pops, clicks, and crackle).

However, do higher sampling rates help remove broadband noise on
magnetic media?

For the most part, I use 24-bits for archiving analog material that
has been converted to digital.

But I will use 32-bits for processing (digital speed or EQ changes,
noise reduction) because audio editing software like Wavelab and
Cube-Tec will automatically increase the word length to more
accurately represent new digitally interpolated values.  Once
any processing has been done, I will continue to work in 32-bit
(or 64-bit in Cube-Tec) and save intermediate files with the
longer wordlengths, and as the very last step dither down to 24-bit.

Eric Jacobs
The Audio Archive

At 10:33 AM 6/24/2005, Richard Hess wrote:

> May I respectfully submit that there needs to be some sanity check as
> to what you're doing...If you had a thesis that was typewritten on
> one side of the page, would you scan the backs of the pages and save
> them? Well, that's what you're doing  when you record a cassette or a
> 3.75 in/s reel at 88.2/24 (or 88.2/32). I wouldn't save anything
> archival at 88.2/32 unless it was born digital that way.