>Q: How can you represent it as a sound? Sound is supposed to be a wave that travels through air, and there was no air in the early stages of the Big Bang?
>A: The Big Bang Sound in the simulation is derived from the sound propagating as compression waves through the plasma/hydrogen medium of the early universe some 100 to 700 thousand years after the initial Big Bang. The density of this medium was changing as the universe expanded, but should have been considerably more dense than air on our little planet. One does NOT need air to have sound, only some medium in which compression/rarefaction waves can propagate. The sound waves were very low in frequency and had wavelengths comparable to some fraction of the size of the universe. For the convenience of humans, who could not hear such low frequencies, I have increased them to the audio range of the human ear.
Or... One could simply say it's a rendering of J.R.R. Tolkien's Ainulindalë. ;)
Interesting speculation in light of the article I read earlier that the universe may have started out with one time and one space dimension. No "sound" then ... and no gravity.
I haven't looked at his source yet, but my guess is that the power spectrum was inverted directly back to real space/"time" domain, without randomizing the phase.
Keep a grain of salt at the ready as you analyze the problem. It's a tricky thing to do "right", and even when you're done, it's not clear that "right" will be well-defined in this case.
Edit: Since the code is a *.nb form, I can't read it directly (insufficient space/desire to install the Mathematica Reader), but from what I can glean reading it as ASCII text, the sound is synthesized from in-phase cosines. Each signal at every frequency starts together, leading to a spike in amplitude. You can see each pop very clearly if you look at the wav file over a short duration in the time domain.
The relative phases of the spectra, following the analysis that Cramer's used (inverting the angular power spectrum into sound) actually are in the raw WMAP/Planck data, and may be in print. If you have the raw temperature map and are comfortable computing the angular power spectrum, you'll have access to the phase too.
I can't find a perfect reference at the moment, but to emphasize the importance of the phase in reconstructing real-space data, check out the pictures in [1].
P.S. I fully admit this is a stupid joke but it makes me laugh so there ya go.