What is hidden in Aphex Twin’s WAV files?

WAV files (or WAVE as they would be called if the Quick and Dirty Operating System hadn’t had a three-character limit on file extensions), probably the most common kind of files for uncompressed PCM audio, follow a loosely defined format with ancient roots, where sections have four-character labels. Each file consists of a single “RIFF” chunk of type “WAVE”, containing a “fmt ” bloc that specifies metadata (sample frequency etc.) followed by a “data” bloc containing the actual PCM audio samples. Usually there is nothing else, but occasionally files contain other kinds of blocs.

The Java standard library can decode WAV files, but in Klipspringer v4 I added my own decoder to read WAV files using NIO in the simplest cases, and delegated the files to AudioSystem when files contained something strange. But when I noted that my decoder couldn’t play the files of Aphex Twin’s most recent release, I changed it so that it could, by just ignoring irrelevant blocs. I also took a look at what those blocs were.

The WAV files of the Blackbox Life Recorder 21f / in a room7 F760 release turned out to contain some surprising junk. For instance, the four originally uploaded files all contained a “pad ” block with 130 kB of just zeros. But two of the differently mastered versions that Richard (as we old Aphex Twin fans like to call him) uploaded a few days after the original release contain something that might be more interesting.

It’s the two title tracks that each have a “LIST” bloc of type “INFO”, and inside are a number of subsections that I can’t find documented anywhere. The files don’t contain exactly the same segments, but both have “NITR”, “NTKB”, “data”, “DMRT”, “RDH””, “SKHC”, “DOMF”, “NSRV”, “ATAD”, “BDNA”, “DIUA”, “RTIB”, “QMPB”, “PEUC”, “SGLF”, “MPBH”, “TDPI”, “YEKM”, “BDCP”, “BDKP”, “CNYS”, “YTAM”, “2TIT” “NELT”, “3NRT”, and “INAM”. Most of these sections just a few bytes long. A couple are comprihensible to me: 2TIT contains the track title in a 16-bit text encoding and INAM the same in normal 8-bit text.

The interesting part is the 3NRT section. It’s 28212 bytes long in blackbox life recorder 21f 760 omc and 20476 bytes in in a room7 F760 omc s and the data looks like something that should be possible to interpret.

Is this a little Easter egg? It wouldn’t be surprising, since Richard has a history of hiding things in various places. Most famously, the 1999 Aphex Twin release Windowlicker had track titled \(\Delta M^{-1}_{i}=-\alpha \sum^{N}_{n=1}D_i[n]\left[\sum_{j\in C} F_{ji}[n-1]+Fext_{i}[n^{-1}] \right]\) that turned out (when people listened to it in a computer player that displayed a spectrogram) to contain an encoded picture of Richard’s grinning face:

I have no idea if what the WAV files contain is anything like that, but it definitely has regularities that make it seem like it could be some sort of image data. But of course, it could also just be something output by the software Richard used, without intent and without any interesting interpretation.

Anyway, my attempts at interpreting it have failed, so I’m uploading it here to let others have a go:

• blackbox life recorder 21f 760 omc: this is full LIST bloc and this is the extracted 3NRT segment that presumably is something interesting.
• in a room7 F760 omc s: this is full LIST bloc and this is the 3NRT segment.

If anybody out there has an idea what this is, please leave a comment below! Also, I would be interested if anybody knows the origin of all those section labels (“NITR”, “NTKB”, etc.).