Spectral Analysis
What we could do is take a few real bells, hopefully similar to the target tele-
phone bell, and see what spectrum they give. One trick I learned from a jazz
drummer is that as with guitar strings you can selectively damp a cymbal or
bell by touching it in the right place. This lets you make recordings that sepa-
rate out some of the overtones and make the analysis clearer. Let’s take a look
at a spectrogram of a bell (actually a 10cm thin-walled brass bowl very similar
to the bells found in phones) and use these frequencies to help with the model.
The spectrums are thresholded, simplified so that only the strongest fre-
quencies come through. I have taken three snapshots, at the start, middle, and
end of the sound, so the evolution of frequencies can be seen. Neither these
diagrams nor the 3D plot given in figure 29.6 adequately show the moment-by-
moment growth and decay of the spectrum. It’s something you really need to
do for yourself with a good analysis package. As you step through a sound file
Model
371
Figure 29.2
Modes of a bell.
you will see the individual components move, and the way they move can give
clues about what is happening at a physical level.
Looking at figure 29.3 we see the initial burst of many frequencies. Like a
football game, you can see plenty from the initial positions, but can’t really
predict much about what will happen next. Anything could happen during
the game. The strongest partial is at 3
.
016kHz with two other strong ones at
1
.
219kHz and 5
.
316kHz. There’s also a spike of high frequency, which is proba-
bly the hit from the metal beater at 11
.
6kHz. Notice the grouping around the
bottom, between 400Hz and 1kHz. There are two strong peaks in this cluster
at 484Hz and 948Hz.
With two points of reference a lot more is revealed. The initial burst near
11kHz has completely gone, so we can assume this was the striker hitting the
bell. Much of the energy has now shifted into 483Hz, which seems a likely can-
didate for the fundamental, and a second mid-range peak has emerged near
1
.
229kHz at 1
.
439kHz. Notice that the harmonics move around slightly. The
current 1
.
229kHz is almost certainly the “same” modal resonance as 1
.
219kHz
(a small difference of 10Hz).
The question to ask now is, can we see any harmonic patterns or group-
ings? Well, 483
×
2 = 966, not far from 959Hz, so perhaps this peak is related
to the fundamental this way, and 483
×
3 = 1449, which might indicate a third
Do'stlaringiz bilan baham: |