[linux-audio-dev] Re: applying RIAA curves in software

[fons adriaensen]

On Tue, Oct 25, 2005 at 01:51:43PM -0500, Cornell III, Howard M wrote:
 
> Maybe use a crystal cartridge that puts out approximately one volt instead
> of a magnetic cartridge that puts out tens of millivolts.  That attacks the
> signal to noise problem from the signal side.

A crystal pickup requires a very high-impedance, low-capacitance input. 
Wiring this, or a magnetic one directly to an input that is not designed
for the purpose will just generate trouble.

> The RIAA record curve reduces bass and increases treble, and the reverse
> RIAA curve for playback does the opposite. 

Sorry, but this is plain wrong. The RIAA filter used when cutting a disk
master will boost bass (below 50 Hz), and reduce high frequencies. This
actually leads to a worse S/N ratio on playback. It looks like this:

   \
    \______       (1)
           \
            \________


The high frequency shelving is (was) required to adapt to the limitations
of disk cutting technology, and the bass boost allows to reduce rumble
(mecahnical VLF noise) on playback. The inverse (playback) filter is of
course:

             ________ 
            /
    _______/      (2)
   /
  /

Combine this with the -6dB / octave required for a magnetic transducer,
(and which has nothing to do with RIAA equalisation) and you get the
familiar curve:

 ___
    \
     \_____       (3)
           \
            \
             \ 

For a crystal transducer (assuming it has a flat frequency response),
you would actually need (2).

-- 
FA