From: Patent Tactics, George Brock-Nannestad
I am beginning to feel old. The creation of standards for recording and
reproduction may have been a lovely ideal, but in practice if floundered, and
this was realized by the standardization bodies themselves. I wrote about
this in 2001:
Brock-Nannestad, George: "Pre- and De-Emphasis - A
Forgotten Necessity", AES Preprint No. 5360, 110th Convention 2001 May 12-15,
(an historical overview of standard compensations introduced in the
recording-reproducing chain and the arguments used over time, ending with
total resignation on the part of a standards body), and I quote the passage I
am referring to:
"Reproducing characteristic tolerances -- No tolerances are specified since
commercial reproducers normally contain a tone control which may vary their
frequency characteristic over a wide range" (IEC 1958)
This was after the following comment from the AES in 1951 concerning the
specification of the AES Standard Playback Curve:
"The decision to specify a standard playback response characteristic instead
of a recording characteristic was deliberate on the part of the Standards
Committee. This course was chosen because of the impossible task of achieving
a universal recorded characteristic compatible with all individual recording
conditions and systems"
The list of my publications that the above is taken from was posted on
ARSCLIST on 26 July 2004.
Recently, in discussions in the IASA Technical Committee over the upcoming
revision of TC 04 I stressed that there were quite broad tolerances in the
standards (also visible in Richard Hess' reproduction of the RIAA standards),
except at 1000 Hz. So in the IASA TC discussion I made the curve showing the
de-emphasis fat corresponding to the + and - tolerances, and you are now -
obviously - all over the place but - equally obviously - still within the
requirements of the standard. In other words, any curve that can be "hidden"
below the fat line is within the tolerances.
Furthermore, the blanket statement that the phase response follows
causatively from the amplitude response is only true for the class of minimum-
phase networks. This means that if the RIAA is made by passive components it
may be different from when it is generated as part of a feedback loop. In the
digital domain you are able to ensure that the transfer function you specify
is the correct one (well, the one you aim for!), within 0.001 dB (or better,
try to convert 24 bit into percentage precision and be ashamed!), and
furthermore that both stereo channels are absolutely equal and tracking. But
is the output from your pickup up to the challenge of interchannel
differences of 0.001 dB over the frequency range of interest? I think not!
Not for a mechanical pickup. And my pet peeve: 24 bits resolution. The bits
are there, they can be counted, but the step sizes corresponding to the last
5 binary digits are not the same all the time. For this to occur you need
analog stability, complete enclosure in an oven, after having burnt-in the
various components. Agilent does provide this - at a price and at a sample
rate of 1 Hz.
Equally in IASA TC work I have recommended the following book as a reference.
It is probably the last ever book to be written about analog circuitry for
phono pre-amplifiers. It comes at a price, but that should not frighten
'The Sound of Silence. Lowest-Noise RIAA Phono-Amps: Designer's Guide'
Springer-Verlag Berlin 2008
Theory, including noise basics 143pp.
Best practice, including MM, MC cartridges, RIAA Networks 127pp.
Noise measurement system (better than the object measured!) 21 pp.
The RIAA Phono-Amp Engine 22pp.