wjm@whuxk.UUCP (10/17/83)
Some more thoughts based on my message of this morning and a message about same
from Steve Bellovin (who did not note the loudness differences
mentioned in the earlier message - Andy Koenig did - sorry for the mixup
to both Steve and Andy)
raised a point which I thoroughly agree with - any A/B comparisons should be
DOUBLE-BLIND - that is, neither the experimenter nor the subject(s) know which
is A and which is B until all the trials are completed. I have seen schematics
for building double-blind switch boxes in several electronics magazines - where
the controller (a small microprocessor or microcomputer like the 8048 would
do nicely) selects A or B based on some pseudo-random algorithm and writes
the selection on either magnetic or paper media sealed in such a way that it
cannot be read until the trials are concluded. As my psycholgist friends tell
me, it is virtually impossible for the experimenter to avoid giving subconscious
clus to the subject(s) as to which is A and which is B.
For the now infamous speaker wire tests - I'd suggest the following - get a
double blind switch box (I don't think the contact resistance of some suitable
power relays that can carry 30 amp or so (you could get them out of a junked
motor controller) would affect the results - and if you really wanted to get
fancy, you could use sealed mercury contact relays or those with gold plated
contacts to keep the contact resistance down), a good power amp with LOW
output impedance (a.k.a. high damping factor) - a Hafler or the Perreaux 2150
would do nicely, and a high end pair of speakers (KEF 105.2's seem particulary
appropriate). Arrange the setup as follows:
AMP -----0---0---------------cable A----------------------0
\ \
\ \
0---------------cable B----------------------0---0----Speakers
Two notes - you need TWO relays controlled from the switch box and if one
wants to isolate the effects of other things than DC resistance, the
cables should have the same DC resistance. (or for that matter resistance
and some standard test frequency , e.g. 1KHz)
Happy Testing,
Bill Mitchell
Bell Laboratories
Whippany, NJ
(whuxk!wjm)
P.S. When you post your results to the net, it would be a good idea to state
the test conditions as well