The Great Record-Club CD Conspiracy? Page 2
First I measured Mark's CDs. I loaded the first two minutes of track 1 from each of the eight CDs (four pairs) into a Sonic Solutions CD mastering system and proceeded with three increasingly demanding tests. If the CDs failed Test #1---a simple comparison of the peak levels of the two CD editions---it was all over. I was almost disappointed to discover that all the CDs passed the peak-level test (Columbia House and record-company editions of each title had identical peak levels). But a CD can be compressed in dynamics and still have the same peak level, just like those commercials on TV. In that case it can sound louder but measure the same on a digital meter, since the ear responds to average volume, not instantaneous peak level.
Test #2: Frequency Analysis
Test #2 was a frequency analysis of a section of each title, beginning and ending at the same sample number (to guarantee capturing exactly the same musical portion). The four titles passed this test as well, producing apparently identical frequency responses, to the limit of the resolution of the Discrete Fourier Transform within the Sonic Solutions program. But perhaps listeners' ears were perceiving differences we couldn't see on the DFT. I devised an even more critical test.
Test #3: Absolute Proof
Test #3 was a cancellation test. In theory, if two recordings are absolutely identical, then you can combine (mix) the two together, inverting the polarity of one vs the other, and there will be no audible output from the mix. Any audible output indicates incomplete cancellation, the result of a difference in level, frequency response, phase response, and/or time (delay) between the two.
For example, if the two recordings are matched in level and time, but if the residual of the mix is high-frequency-dominant, then this would confirm that a bass cut (or high boost) had been applied to one of the recordings. A test like this would have been impossible in the days of analog recording, for the timebase of two analog recordings drifts, resulting in the famous "Itchycoo Park" effect (footnote 1). This is the power of Test #3: If we can synchronize two identical CDs, invert the polarity of one, and mix the two signals together, there should be no sound for the entire duration of the CD---ie, complete cancellation. A skip in either CD, even as short as a sample, will produce an audible output. That is why the two CDs must be synchronized to the sample accuracy of 22 microseconds. I caution anyone trying to repeat this test that synchronizing two CDs takes some work; the waveforms must be magnified to show only a few samples on the screen.
All four of Mark's titles passed the cancellation test with flying colors. This is absolute proof that the compared CDs were made from identical data (identical master tapes or clones). I can only conclude that there is no "smoking gun," no conspiracy within the record-club industry. Of course, I had to correct the channel reversal of the Crooklyn CD before the test worked. Surprisingly, the two Crooklyn discs showed quite a bit of cancellation even before I corrected the channels, indicating that this commercial disc, like so many popular discs, has very little channel separation or stereo ambience information.
So, were Mark's ears stuffed? Has the Gotham City Audio Society gone batty? I don't think so. I believe the sonic differences between supposedly identical CDs are caused by a promiscuous evil known as jitter. A significant amount of jitter on a CD can produce a high-frequency edge (which Mark noted as a reduction in bass). Jitter can be caused by the mastering system at a particular plant. At this time, no plant has paid much attention to reducing the jitter of its cutting system (footnote 2), largely because, in general, the engineers at the plants are not audiophiles ("if it measures the same, it sounds the same..."). All of the Columbia House CDs were pressed by the same plant; the record-company CDs were pressed by four different plants; in fact, Tori Amos's store-bought CD was pressed by the same plant as Columbia House's, but with a different glass master.
Now I was really curious; I hadn't heard a big difference with the pop CDs, but could the drastic differences I heard with the Classical CDs really be caused by jitter? I loaded the Fennell and Stokowski into the Sonic Solutions and immediately proceeded with Test #3, which they both passed. Now I had six pairs of CDs reported as sounding different but made from identical masters. This also confirms my contention that stereo systems should be evaluated with the best recordings possible. The classical CDs were made from first-generation analog tape, while the pop CDs have gone through several generations---detail and resolution are more obscured, and additional jitter in playback doesn't have as noticeable an effect.
Footnote 1: The "Itchycoo Park" effect is due to comb filtering resulting from two recordings combining and canceling at various frequencies as their time delays vary.
Footnote 2: Anyone for building a better mousetrap?