I'm frustrated. Digital audio irritates me. It seems like there ought to be some novel way to losslessly store an analog signal in the digital domain. I suppose, though, there is some really simple reason that's not the case. Something especially annoying like the fact it would require a virtually unlimited number of bits... or some other such (perfectly sensible) nonsense.

The way digital audio works now is incredibly naive. It's like this: you hook up a volt meter up to your analog source and then periodically jot down the reading. It's that simple. Now imagine that your meter has a finite resolution. It can only give you a value from 0 to, say, 65535... (16-bit). Now imagine you're the flash and you can take a reading 44.1 thousand times a second (and that your meter is that reponsive). You could now make a CD.

Sony has come up with an alternative called DSD or, Direct Stream Digital. It turns out that this is exactly the same (more or less) as the previous example except your meter can only tell you whether the voltage increased or decreased (1 bit) and you've had a lot of coffee and decide to take a reading every 2 millionth of a second.

If you sit around and think about this for a bit Sony's method looks pretty good compared to the way CDs are encoded. It's simple and elegant. All's not right with the world, though. With traditional sampling methods you run afoul of something called the nyquist theorem, which says you can only represent frequencies that are less than half of your sampling rate. It turns out that the problem is actually worse than that. Nyquist was dealing with nice sine waves. In practical signals sampling can give rise to a nasty problem called aliasing. While Sony's method has its niceties you have a different (maybe) form of the same problem... there is only so fast your output voltage can change (think about slope) so you still can't represent high frequencies very well.

There are less naive approaches one can consider like taking a sample only when the input voltage does something interesting. The problem, of course, is that every little step you take towards smarter encoding the harder it gets to implement in hardware or software. Bleh.

I'm not giving up on the problem. In a sick way it's fun to think about. In the end it doesn't really matter. CDs sound pretty damn good. I'm just a freak.

...

I finished Janna Levin's How the Universe Got its Spots. It was a very nice read. I recommend it highly. She has a real knack for making complex concepts fairly accessible. That said, she still makes us lay-folk work for our lunches.

The book is odd in that it's a diary. The science is interspersed with accounts of her interesting interior life. It really breaks things up nicely.

Be warned: Levin seems to have a real fascination with the psychological dark-side of the science/math community. In fact, she almost seems to romanticize it. Maybe it's just me. I suppose, in all fainess, being faced day in a day out with things like the idea of a deterministic universe would be a bit challenging to morale. Also, when you consider how hard monomania, obsessive compulsiveness (redundant?), and the night-follows-day depression can be on personal relationships, it becomes a bit easier to appreciate how hard it must be for the real geniuses to keep their chins up.

Anyway, I'm rambling horribly...

My current reading is The Cultural Creatives (also in mediocre website form at www.cuturalcreatives.org). I've only finished the first chapter thus far but it looks fairly promising. The basic idea is that we may be in for a huge cultural shift because of the slow emergence of a new group of people (the cultural creatives) that espouse new types of views and ideals.

It's more interesting than that awful paragraph makes it sound.