Thanks for the explanation - I think I understand.
Hmm. Well in that case, what would stop you from setting up such a
bandpass filter for each note and using it to trigger a tone generator
instead of an LED? This would serve the purpose of simple pitch
tracking, no? It would only be capable of tracking notes in half step
intervals (no string bends or anything) but that would probably serve
my purpose. Would that work?
This is one of those things that sounds good at first, but when you
get down to details there are some problems (besides the cost, as
another post mentions). The main problem is that you will need
very sharp tuning to isolate each note. If you build a simple
bandpass for each note, the "passband" must be very narrow
in order to reject adjacent notes... but then it will reject the
desired note if it is not dead-on. So you need a multi-pole
bandpass that has a flatter top and still has steep slopes.
These get complex and fussy to tune.
But suppose that is not a problem. (Actually, there are
digitally tuned single-chip filters that could handle this,
where the fitler frequency is controlled by a clock. This
could be made very accurate using a crystal reference.)
Then you will discover the hidden "gotcha": The response
time of a filter is proportional to the inverse of its bandwidth,
which means there will be a lag between the time you pluck
the string and the time the peak comes out of the bandpass.
Then, if that wasn't bad enough, remember than the bandpass
output is still AC, and for a guitar most of the notes of interest
are in the 100 Hz or less range. If you are going to convert this
to a DC value you will need further smoothing after you (full-wave)
rectify it, further adding to the delay.
OK, maybe with one filter per note you don't really need to convert
to DC, just some sort of a threshold comparator. But you will have
to have some way to keep the subsequent synth oscillator going
between comparator peaks, which means a time delay on the
shut-off... probably much better than delaying the onset.
Anyway, as a rule of thumb, figure that an overall 10 msec onset
delay will probably be the upper limit of tolerability. You might be
able to tolerate more if you use the synth as a backup instrument
that echoes your leads with a slight reverb-like lag. A 10 msec
lag is roughly what you'd get from a 100 Hz passband, but the
actual passbands you need are much narrower because each
semitone is only 6% from its neighbors. So at 110 Hz you need
passbands that are only 17 Hz wide, which means a lag of
roughly 59 msec... pretty bad. And it gets worse for lower notes.
Still, despite all this gloom and doom, I strongly encourage you
to play around with this idea. Start with a simple single-note
bandpass and see what you can do. You might come up with
some clever workaround for the above problems that other
folks haven't thought of yet, or you might not... but you'll have
a fun and educational experience that will almost certainly lead
to other insights and maybe to other inventions.
Best regards,
Bob Masta
D A Q A R T A
Data AcQuisition And Real-Time Analysis
www.daqarta.com
Scope, Spectrum, Spectrogram, Signal Generator
Science with your sound card!
