Shortly after we published "Thursday Afternoon Ripping Experiment," fellow writer S. Andrea Sundaram wrote me and posed a good question: Do you think byte size is a reliable way to determine file differences? My answer to him was mixed.
On the one hand, I do think byte size is a good method to determine differences with losslessly compressed files, as any change in the data will likely result in a change in the resulting file size because the compression will differ. But I also conceded that there could be an instance where the file size remains the same even if the data are different. And I also conceded that with non-compressed files, it's not a good way to assess differences because the size is determined by the bit depth, sampling frequency, and track length. As a result, we both thought that I should look for other methods to validate my results.
We both agreed that Audacity, which is a sound editor that you can download for free, would be a good tool to assess any differences. Audacity allows you to input multiple files and, through various tools, compare them.
The way I chose to do the comparison was by importing two song files at a time and then inverting a channel in one song file and mixing it with the same channel from the other song file. If the channels are identical, the result is a "null." If there is some difference between the channels, the result will show up as a waveform in Audacity, meaning that they're not identical. It's also possible with Audacity to sum the left and right channels of a song to make it mono, which I also did, and compare that result with a mono summation from the other song. I'm pleased to say that with every song file I compared -- compressed or not, stereo or mono -- the result was a null, meaning identical.
After that, I used a program called ExamDiff Pro to compare the first FLAC file that was created from the ripped CD to the FLAC file that was derived by taking that original file and first converting to WAV, then to ALAC, then to Monkey's Audio, and then back to FLAC. ExamDiff Pro is a program that allows you to compare the contents of pretty much any file, bit by bit if you have to, which is exactly how I did it. There were small differences in the header information between the two files, which didn't surprise me given all the formats that the one FLAC file had been converted to, but the sections that contained the music data were bit-for-bit identical.
With Audacity and ExamDiff Pro both confirming that the files I tested were identical, we probably could've stopped examining them. But S. Andrea and I figured that it wouldn't hurt to try one more thing, so S. Andrea took the same two files I compared in ExamDiff Pro and imported both into a program called Mathematica, in which he was able to run a process that, not surprisingly, confirmed them to be the same.
What this all means is that what I said in "Thursday Afternoon Ripping Experiment" holds true: "It's possible to convert back and forth between lossless compression formats (e.g., FLAC, ALAC, or Monkey's Audio), or non-compressed ones (e.g., WAV or AIFF), as I did, and not lose a thing." And if someone tells you otherwise, you can confidently tell them that they're full of it.
The only thing I wonder now is whether rips done on other computers will be identical to the ones done on my own computer. I also wonder whether different ripping software will create files that are the same as the ones I created with JRiver Media Center. You'd think they should all be the same, but you never know -- so those sound like interesting experiments for another time.
Doug Schneider
Publisher, SoundStage!