modules with different tubes?(oktavers)

[Dave Krooshof]

>Aahhh.... Nice try. Probably will not work.
>Problem... How do you decide the "waveform" crossing points.  I'm thinking of
>a Guitar wave... which is a bastard !  It may cross zero more times than you
>like. Tracking the peaks will not work either.
>
>DDL  pitch shifters do exactly this. If they have very short delays... the
>glitches in the waveform are severe. If you can wait a week... you could
>make a really long delay.

Dividing over peaks or zero's do have a lot of disavantages.
There are several things you could do.

1. Go for the zero's, but do it like this:
If the signal comes above a treshold, at say, 0.1 of the whole scale,
output an up trigger, and set the new treshold to -0.1.
Then wait for the signal to pass that -0.1.
If it does, output the a down trigger, and set the treshold back to .1.
(bis)
This should free you up from noisy triggers, and should free you
from overtones doing oktave jumps

2. go into Digital domain. Do a Fast Fourier analysis. (FFT)
(Fourier was actually concerned about heat distribution, but he did that
so well that he discoverd that complex functions can be sees as added
simpeler functions.)
So, in the FFT, you can find a peak, hopefully, which will be your frequency.
Or a load overtone. This is the disatvantage. You're overtones might be
loader in the spectrum. In guitars, they often are!

3. You can calculate the log of the FFT.  This will bring you in the
Capstrum domain, where everything's turned around. Quefrencies,
saphe, etc.
I'll need to go back into my math books again, but this is very interesting.
In this domain, periodicity is very obvious, and filtering out a tonalyty is
possible without to much difference in the rest, even withoutish phase
turnings!
I think I should refresh this knowledge, as it is very massive in
possibilities.*

4. Components analysis. I got the idea from sociology, been programming
on it with my DSP teacher in sonology. Use it, but give me credits.
The idea is: Go in to the spectral domain, FFT.
Then consider the peaks.
Roll up the frequency axis, one oktave per revolution.
Now, when periodicy happens, that is rich in harmonics, it's peaks will add up
on the oktaves and on the fifths, mostly!
So now the overtones are not a disatvantage, but it helps tracking the
fundamental, *even*when*it's*absent*!!!!
As happens in your ear/wetware! credits to the Creator.

The delay is mostly dependent on two things:
1. the window size that you pick to do your FFT on.
2. the lantncy of your DAC.
Your DSP should be fast enough to caculate is before the next
window's comming in. (an expensive spec)

You can state that all FFT pitchtrackes are as fast as one period
of the lowest freq you want to be able to measure.


Dave


*Capstrum is really the next step in filtering. I like LP HP and BP,
but I've heard them. Now lets go to these: Any thing that changes may pass,
Anything that stayes the same, is out.
This makes analogue synths and guitar solo's disappear from recordings,
and turn sweat singers into rough pitchless alcohol abusers/smokers.
It's more then fun. I called it the FROG EYE filter, as frogs only see
what moves (flies and partners do, mostly), the rest turns into a nice grey
blur (like reigns? (the birds that live on frogs I mean. 'Reigers' tend to
stand still)).
Another thing that's nice is to do the opposite: Only certain frequencies may
pass, if they are stable. Pick an odd scale, and every classical piece turns
into Stockhausen.


--------------------------------------------
Dave Krooshof http://www.xs4all.nl/~krooshof
geluidstechnicus @ http://www.ahk.nl/the/theatertechniek_ov.html
webmaster: http://www.popronde.nl
-------------------------------------------
no html mail please