Double-blind test of two CD players reveals audible difference
John Stalberg, Thomas Akerlund, and Mikael Sundman
Background
The purpose of this test was to investigate whether modern CD players sound different or not.
This is a controversial issue as there is little or no scientific evidence that CD players are
audibly different. A previous test have shown that the first CD player imported to the US,
the 14-bit Phillips CD player CD100, was audibly different to a Sony player
(
www.oakland.edu/?arlst/abx_cd.htm). The score was 67% correct guesses with a p lower
than 0.005. Other modern CD players did not show any audible difference.
These results were published in PSACS Sound Bytes Vol. 7 No. 1, 1996.
The Swedish Sound Audio Technical Society has previously been able to distinguish CD players
in blind tests, but the result has not been confirmed in any double blind test.
The result from the present study showed that there indeed is a significant and
audible difference between two specific CD players. Thus, the notion that all CD players
sound the same is not correct. Another conclusion is that changing CD players in an upgrade
path indeed may have a purpose Рi.e. to get a better sounding system.
However, there is little information of which CD players that sound poor and which sound good,
and for the consumer to know this, an extensive testing must be performed.
Material and methods
Reference system. The reference system consisted of Sentec SC9 pre-amplifier,
four NAD 208 Power amplifers (2 x 280 Watt/8 Ohm for each amplifier), INO Audio r 56 speakers
and 4 INO Audio profundus Z-4 subs, and Sentec DiAna DA converter.
Each item has been selected as be the most audible transparent for music and test signals
(i.e. uncoloured sound) using a Ңefore-afterӠtest(i.e. signals pass trough without
audible change of the pattern/waveform). The listening room is damped for early reflections,
and reproduce a frequency response of 20-20000 Hz within 1 dB limits (at listening position).
Max SPL is around 137 dB at 17 Hz.
Test objects. A Denon DCD 655 and a Harman Kardon HD7600 CD player with variable analogue output
were chosen as test objects. The output level of the Harman Kardon CDP was adjusted
to match the Denon at a dB level of ?0.01 dB using a oscilloscope and a 1 kHz Hz test signal,
as measured from the output of the pre-amplifier.
Listeners. Thomas Akerlund (1), Mikael Sundman (2) John Stalberg (3).
Both persons 2 and 3 are musicians and audio recording specialists and are very well trained
listeners, and familiar with both the reference system and the test objects.
Person 1 has limited or no previous experience of the reference system or the test objects.
Test procedure. All listeners were given a training session for about one to two hours.
First, the Harman Kardon CD player was auditioned and compared against the reference DA converter
and its character was noted. After a positive result in blind conditions with the
two different tests objects, they were chosen. Starting the DBT, one person (3)
tossed a coin ten times and wrote down the order on a paper, while the other two persons were
in another room. He made all the switches and was not visible or made any noise to the
test persons during the entire test. The CD players and the back of the preamplifier were
hidden with blankets. The test signal was two identical (verified by data analysis) copies
of a drum session. Each of these was set on repeat on the CD players, with exact matching of
length. For each test, one person made the switching and the other person made the listening
test. The switcher started to switch from CD (neutral) to either Aux 2 (CD player X) or
Aux 3 (CD player Y), without saying anything to the listener. Then the test was continued
with 5-10 seconds of listening at each input with about 1 second delay at neutral,
i.e. a test order of e.g. X - neutral РY Рnetrual РX etc. Since the sound of the CD players
were not synchronised the length of the listening period varied randomly during the test.
This procedure did cause some confusion, since it was not always obvious if a switching had
occurred and made it more difficult for the listener. Nevertheless, when the listener decided
which sound that related to which test object, the switcher noted which input he had chosen,
and the test result was written down. One person (1) conducted six listening trials,
and there was a switch between these two for the remaining four test, i.e. person 1 became
the switcher and person 2 the listener. After ten trials the result was compared to the written
scheme as made from person 3. After this, a second listening test with three trials was conducted
with person 3 as listener, 1 as switcher and 2 as the coin flipper.
Results and discussion
Training session. In the initial training session all listeners concluded that the Harman Kardon
CD player was very similar to the DiAna reference DA converter. A somewhat higher impact in drums
and a little brighter sound than the DiAna was noted.
When the Denon CD player was tested against the Harman Kardon, it became evident that the
Denon CD player had a brighter sound as noted on drums and background noise from the
recording(the noise produced by the CD players was never at audible levels).
Test person 1 was not as familiar with critical listening as person 2 and 3,
and had no previous listening experience with the test objects, although he also noted a
difference when these were tested. The listening session was ended with single blind guesses,
in which most cases were correct (Ÿ0%).
DBT session. The final result from the DBT tests was that 11 out 13 trials were correct.
This corresponds to 85% (p=0.01) correct choices, which is significantly different from
random choice. The two faulty choices came from person 1, who was not as familiar with the system
and had less training than the other two. Also, person 1 suffered from some listening fatigue
after the training session, and may have scored better with a different outlay of the experiment.
For example, listener 1 could have borrowed and listened to the CD players for a while,
before the test was done. The conclusion was however, that there is a 99% probability that the
CD players were audibly different.
och det är liksom omöjligt att bedöma om testerna görs vettigt eller ovettigt!
Jaja...
, när det jag skrev bara var att om man har att välja mellan antingen:
