You are quite correct - for a full ABX test there should be two or more reference samples for A and B.
(I am just trying to do a simple "screening" to determine "whether there's anything to test or discuss".)
For that matter, there should also be some sort of randomization protocol. After all, one of those "unconscious cues" I was talking about is the order in which samples are arranged. For example, when offering three glasses of soda to taste, you must mix up the order to rule out the possibility that right handed tasters tend to prefer the glass on the right - or the opposite, or that most people prefer the first sample they taste - or the last. In this case it's quite possible that both the person setting up the test and the person taking it have an unconscious preference for early tracks, or late tracks, or tracks with even numbers, or tracks with numbers evenly divisible by 3 (by letting a machine randomize the order you rule out all these possibilities). I'm assuming that the tracks will be at lest more or less mixed up and not, for example, arranged as 1-5=mat and 6-10=no-mat.
I base my expectations - and tests that I propose - on a few simple basic facts. There are only two places in "a digital audio
stream" where information is contained. The
values recorded in the individual digital bits are information, and the
timing with which they are presented is also information (the clock). In the case of a digital audio
file, the timing is not stored on the disc or other storage medium, and so the
only information is contained in the bits themselves. (This is much simpler than analog audio, where it is quite possible that there is some specific distortion or error that you don't happen to be measuring. It is also simpler than considering an entire CD player, where the mat could influence other parts of the player that are involved in the conversion process besides the data itself.) With a digital file by itself, if the bits are the same, then they
are the same - and, if the files are different, then the bits
must be different.
Honestly, with all the discussion of test protocols, I seem to have glossed over an important specific question:
HAVE WE DONE THESE RIPS WITH ACCURATERIP OR OTHERWISE CONFIRMED THAT WE DO IN FACT HAVE THE SAME BITS WITH AND WITHOUT THE MAT?
(If not, then the most obvious assumption would be that the mat is in fact altering the data - which might reasonably be audible. And, if so, it should be pretty easy to determine which is correct - with or without.)
However, we must be careful to separate things about the digital data file from things about the file
while its playing. (For example, let's assume that I put ten copies of the
same file on a CD. It's possible that various instances of that track might sound different because the CD drive has more jitter on early tracks than on later tracks, or even because one area on the disc is more subject to bit errors. If I'm playing that CD player into a DAC, that extra jitter could make the early tracks sound audibly different. And, if I'm ripping that CD, it could result in more uncorrected data errors on early tracks, which might also make the ripped tracks sound different. However, if the jitter results in data errors, then the ripped tracks will contain different data, which will be obvious when we compare them. Likewise, if the jitter does not result in data errors, the I can easily demonstrate that as well - by doing a bit compare on the files. Most modern CD ripping programs do this automatically using AccurateRIP or another similar data confirmation service. (In fact, CD players read their data through a buffer, so the data itself should have any jitter that's a result of the read process itself removed. However, it's possible, for example, that vibration from the spinning CD could introduce jitter or other distortion at the output by causing a wire somewhere to vibrate. If so, then a mat could possibly help that - or make it worse.)
A lot of the claims for "cd mats" and "magic liquids" are based on some facts and assumptions that are somewhat correct - but not entirely. For example, let's say we're playing a slightly scuffed CD on a CD player. Because it's worn, the data stored on the CD actually has some bad or unreadable bits. Luckily for us, when the player reads that CD, there are two levels of digital error correction that will do their best to repair those problems. These use extra information, and some rather fancy math, to "reconstruct" the missing or garbled data. (Gaps up to about 1/8" on a single track should be perfectly corrected.) If they are successful, then the data is "new again" -
it is not in any way different than the original data (this is known as "a recoverable error"). If the damage is too extensive, then most CD players have the option of "patching over the gap" with "similar data". When this happens, you may hear a tick or a pop, but, whether you notice it or not, the data is
NOT the same as it was, and a simple bit compare test will show that it has changed. (Most computer CD drives lack this final correction step and will simple report a failure if they encounter an unrecoverable error - which is one reason why "audio players" will sometimes play discs that computers won't.)
Most mats and treatments I've read about claim to reduce the number of uncorrectable errors and/or reduce read jitter. Remember that uncorrectable errors are missing data that is replaced with interpolated "data that's sort of close" - and so we would expect them to be audible. However, you should also be aware that it is very unlikely that reducing jitter during the actual read process will affect the jitter at the output of the CD player. Virtually all CD players read the data from the disc into a buffer. Therefore, any jitter present in the read process will not be passed on to the output anyway. (However, it's theoretically possible that a high enough level of jitter in the mechanism could cause additional read errors.) The only flaw in this argument is that uncorrectable read errors are in practice
VERY RARE. Of the 500+ CDs that I've ripped using my current software, which confirms every bit of every rip against the AccrateRIP database, I have seen exactly three uncorrectable errors on three individual tracks. One turned out to be a damaged disc, one went away when I reripped the disc - and so was probably dust, and one turned out to be a manufacturing flaw on the disc itself - and was present on two additional copies. Therefore, uncorrected errors would seem to be "a problem that doesn't really exist". (Of course, it's also possible that the $19 CD drive in my computer is really much better, and so has lower error rates, than the expensive audiophile players that the mats and liquids claim to improve. Likewise, it's quite possible that a solution that makes scratches optically less visible could reduce the number of unrecoverable on some few badly damaged discs. However, and this is where audiophiles seem to get confused, the fact that a given "fix" actually helps a few badly damaged discs
DOES NOT suggest that it will improve ones that aren't damaged.)
At this point, I do want to mention one possibility that is obvious to me - but we haven't even considered yet...
If I rip a CD without the mat and the data from the disc is perfect (which we can tell by our AccurateRIP report), and it does sound different when we use the mat, then the most likely cause would seem to be that the mat is changing the data (and, since it was perfect to begin with, the only possibility would be that the mat is adding more errors). Luckily, if that's the case, we would be able to tell easily enough by doing a bit compare.
The other possibility is that, perhaps by reducing vibration or something similar, the mat makes the analog output of certain CD players sound better for some reason
other than because it improves the data. However, if that's the case, then we would expect no difference in the bits of digital file rips done with it, and no difference on the sound of those files.... unless they were played on that CD player (with the mat).
Quote:
I like the way you think.
If there is no real difference, the probability of getting 9 or 10 correct "by chance" is approx 1.07%. (Per the calculator at http://stattrek.com/online-calculator/binomial.aspx . )
For a full ABX, shouldn't there also be one labeled reference copy for "no matt" and one labeled reference copy for "fancy matt" ?