Hello to everyone who attended Analog Divide. For those of you who didn't get to attend, here's what happened: people tried to distinguish a vinyl record playing through a digitizer from the raw analog signal. People also drank a lot of wine. It was all part of the experiment, okay? You can read more about the setup (or lack thereof) here. Without further ado, here are the rather curious results, in time order:
apNOPE
szNOPE
mkNOPE
fmNOPE
abNOPE
oYUP
wbYUP
wbYUP
wbYUP
wbYUP
wbNOPE
wbYUP
wbNOPE
wbNOPE
wbYUP
agNOPE
cNOPE
bzNOPE
idYUP
aeNOPE
tNOPE
One enthusiastic participant took the test multiple times, so each line is the outcome of one guess. Only 3 out of the 13 participants correctly identified the analog and digital signals. If the participants really had no clue and were equally likely to guess correctly or incorrectly, the probability of getting such a lopsided result is a little less than 1 in 10. (See binomial distribution). So it looks like you guys had a clue. More likely is that most people really could distinguish the signals, but labelled them incorrectly. Before taking the test, several people asked "What's the difference--is analog supposed to sound better?" We told them "not necessarily, that's just what some people say" and tried to explain how the ADC <-> DAC circuit worked. Who knows if it made any sense. In the end, those people that weren't sure what they were listening for probably fell back on the labels "better" and "worse," mapping these to "analog" and "digital" respectively. Does this mean you guys actually think a digitized signal sounds better? That kind of freaks me out. So here's a theory: the high frequency pops and hisses coming from the record player were converted to lower-amplitude noise by the ADC <-> DAC circuit. I'm still reading up on this myself, but here's a starting point. Less popping and hissing = "better" = analog? This theory is supported by 2 of our 3 correct guessers, who said they were specifically listening for high frequencies and noise levels. One, wb, went from a string of correct guesses to a roughly 50/50 record after we started cleaning dust off the test albums. Our 3rd correct guesser, id, says she was listening for the "sound stage" which didn't come through in the digitized signal. In the end, we must concede that the experiment was fun but woefully inadequate as a scientific inquiry. Many suggestions were made to improve things: use the same, artifact-free analog test signal for all participants; get the survey script working so the test would be double-blind; drink less wine, etc. If you feel like taking an improved test, come on over, but I can't guarantee the less wine part just yet. ;) Thanks to everyone who attended. All the code for this party was open source.