[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Decoding the CD318 Genome

Hi, All:

Thought I would forward this to the List, in case others might find
the subject interesting.  Jim, I hope you don't mind a larger audience
for this one...


Jim Morrison wrote:

> You're in the recording business, so I thought you'd be the person to go to
> with this question.
> Is there any way to make a "sonic" fingerprint of CD 318 and use that to
> test that against the other recordings?
> The same question about the Goldberg Yamaha.

Hey, Jim:

The easy answer is "yes".  You could sample a long tone from a specific key/
pitch on the instrument (say, C4) and using a digital audio workstation (DAW)
create something known as a "waterfall" plot, which maps amplitude against
frequency over time.  This display would give you a nice indication of which
harmonics are gaining energy over time after a note has been struck.  You
could also do this for short, non-sustained tones.

For a really good "fingerprint" you'd need to sample every key on the
instrument (all 88) at various dynamic levels and for various durations of
time.  Yes, this would generate quite a bit of data, but would give a pretty
accurate "picture" of CD318's output.  You can't represent the instrument
adequately with just one "snapshot" because what's going on acoustically is
too complex for that.

The waterfall plots of CD318 would look different from the Yamaha or,
for that matter, any other piano.  Things get a bit tricky, however, in
that, because each piano note causes three strings to vibrate in unison,
there is a question of "voicing" to also be considered (ie: the tunings of
the three strings per note).  Since CD318 was being "tweeked" almost
continuously, it would be likely that sampling two of the same CD318 notes
at different recording sessions (or from different albums) would result
in differences in the waterfall plots.  Perhaps not gigantic differences;
it would be interesting to see.

So, it's not as simple as a traditional fingerprint or (here in the age
of the Internet) each person having a unique iris composition/configuration
that can be represented by a single digital image.  Isn't it amazing that the
human ear/brain can integrate all of that data so easily and determine
(with pretty fair accuracy) the unique qualities of certain instruments?

Recording engineers can develop the same type of timbral identification
skills with microphones.  An experienced engineer can often identify
which mics have been used to record certain instruments just by listening
to the final mix of something.

But back to CD318 and the waterfall plots.  That's a nice research idea
Jim, if you have a year or two.  If you need a thesis advisor, just let me