Current audio systems do not reproduce the entire wavefront that
creates the listener's experience in the concert hall. At best, they
measure a few channels and reproduce those, inexactly, through a few
speakers.

Yet the playback experience can be enjoyable and thrilling. Obviously
something of the original sonic event is preserved. Something of the
original time-evolving spectrum of sonic energy is reproduced.

Ignoring for now the question of reproducing a wavefront, let's look at
just how the signal in one channel is handled. It can be quite
distorted and yet still recognizable. What aspects of a signal must be
preserved for it to be recognizable? What aspects must be preserved
for it to sound good, and to sound very much like the original signal?

Engineers have addressed this question in many ways, for example
designing compression algorithms. Some details of the original signal
can be thrown away without losing much, perceptually.

MP3's sound sorta like the orignal files. I'm interested in addressing
the question "what makes an accurate signal" at a higher level of
quality than that.

For example, I've always preferred analog sources to digital, finding
the former more lifelike. Does an analog recorder preserve some aspect
of the signal better than a digital recorder? I know that many of you
will say categorically not. Fine. Let's look anyway at one aspect of
the signal.

Intuitively, a musical signal is made of many "events"...for example
attacks of notes. Intuitively I hear even sustained notes as made of
events...little shifts of timbre, and so on. This idea is confirmed
when we look at an audio signal and see periodic spikes, and also
confirmed by the success of "granular synthesis" (a technique for
synthesizing sustained sounds by summing many individual wavelets).

Perhaps an important dimension of accurate sound reproduction is the
accurate reproduction of the *relative timing* of these events. To
clarify, perhaps we could conceive of each event as being recognized by
the neural machinery and triggering a neuron to fire. And something
about the pattern of this firing, the timing contained therein, is
important to defining the sound quality.

How does a particular recording/playback process affect the timing of
transients? Recording processes are sometimes characterized in terms
of frequency response. Digital has a very flat response in the region
audible to the ear, meaning it doesn't introduce much distortion.
However, it does introduce some distortion. And if we were somehow
able to examine the relative firing times of neurons in response to a
recorded/played-back signal, how much would a digital playback process
distort those times? How much would an analog process distort those
times?

This is not a question about jitter. Certainly jitter is one
distortion mechanism in digital (and analog) playback, but this more
about how even a linear playback system will distort transients because
it is band-limited. Changing the shape of the transient will likely
have a small effect on neural timing. Both digital and analog
recording processes distort the shape of the transient, but perhaps one
of them does so in a way that better preserves the relative timing of
neural events.

My *suspicion* is that analog in fact does better preserve the timing
of neural events. However, I would need to know more about
neuroscience and non-linear systems to have a good answer to this, but
perhaps someone reading is interested.

Best,
Mike