mark@mips.COM (Mark G. Johnson) (12/17/90)
A long while back I remember reading an article that said (roughly) the
following.
Just because measurements on a piece of equipment show low Total
Harmonic Distortion or low IM distortion or flat frequency response
within +/- 0.05 dB, doesn't guarantee it will sound good. A stunning
demonstration and Proof of this hypothesis was presented by
Mr. xxxx [ don't remember name ] at the recent yyyy meeting.
Mr. xxxx exhibited a small aluminum box which was placed in series
with the tape loop of a $nn,000 preamplifier. THD, IM, and frequency
response were measured with the box in the loop and with the box
removed from the loop. The preamp's measurements were identical,
box-in vs. box-out. Measurement resolution was 0.01% for distortion,
+/- 0.05 dB for frequency response. Then, after the measurements,
the audience was asked to listen to the preamp, with and without
the aluminum box. Everyone agreed that the box's presence
changed the sound from good to horrible even though the
measurements predicted identical sound. So the aluminum box
"Measured Perfect, Sounded Awful."
Question: if *you* had to present such a demo, what would *you* put
in the little metal box?
(I _think_ I remember what the article claimed Mr. xxxx put in his box,
but [i] my recollection could be wrong; [ii] netters' ideas for
sound-lousy-measure-great contraptions might prove even more clever.)
--
-- Mark Johnson
MIPS Computer Systems, 930 E. Arques M/S 2-02, Sunnyvale, CA 94086
(408) 524-8308 mark@mips.com {or ...!decwrl!mips!mark}noamb@over.cs.caltech.edu (Noam Bernstein) (12/26/90)
One such box I've heard measured perfectly in every measurement anyone (at some audio engineering society meeting) could come up with, except listening to music. With actual music, the sound sucked. It was a digital circuit which looked for symmetrical waveforms. When the input was symmetrical, it shorted the input to the output. When it wasn't symmetrical, it didn't conduct. noamb@over.cs.caltech.edu
sethb@Morgan.COM (Seth Breidbart) (12/28/90)
In article <8551@uwm.edu> noamb@over.cs.caltech.edu (Noam Bernstein) writes: >One such box I've heard measured perfectly in every measurement anyone (at >some audio engineering society meeting) could come up with, except listening >to music. With actual music, the sound sucked. It was a digital circuit >which looked for symmetrical waveforms. When the input was symmetrical, >it shorted the input to the output. When it wasn't symmetrical, it didn't >conduct. That's an interesting idea. How did it handle the following test: Play music through it. Subtract the output from the input. Take the ratio of the power of the result to the power of the input (that is, the ratio of the integral of the square of the voltages). In some sense, that's a measure of distortion. 0 is shorting the input to the output, 1 is an open circuit (output = 0). You might want to "normalize" the output by multiplying by a constant before subtracting, using the constant that minimizes the resulting measure. This lets a perfect amplifier score 0. It should also be fairly easy to design a box that has wonderful specs when played into a resistor, but which has terrible specs (and sounds terrible) when played into a "real" speaker. Seth sethb@fid.morgan.com
mark@mips.COM (Mark G. Johnson) (01/03/91)
> Everyone agreed that the box's presence > changed the sound from good to horrible even though the > measurements predicted identical sound. > I _think_ I remember that the article said the box contained: A timer and a DPDT mercury-wetted-relay. The relay shorts the signal output to ground for 1.0 milliseconds. This occurs once every N seconds (N=4??). Enough to drive listeners batty but not enough to perturb the measurement gear. -- -- Mark Johnson MIPS Computer Systems, 930 E. Arques M/S 2-02, Sunnyvale, CA 94086 (408) 524-8308 mark@mips.com {or ...!decwrl!mips!mark}