Field Cricket
563
Resonating surface (amplifier)
Flexible material
MOVEMENT
Harp (wing)
Plectrum
FIle
Figure 50.6
A cricket wing.
Figure 50.7
Field cricket synchronous AM
method.
Subtracting 0
.
5 puts the line around zero
with half its original amplitude, and squar-
ing produces two curves, one for the positive
and one for the negative part, that meet
at zero. Flipping the sign, recentering and
adding an offset makes this a circular hump
that sits on zero. Now that we have a slow
modulation, we need to obtain the fast
source that makes the little clicks in each
chirp. If each burst is 0
.
12s and contains 7
segments then the frequency of modulation
is 58Hz, so we can derive this from the base
by multiplying by 40
.
6. Taking the square
of the cosine of this value produces posi-
tive going pulses with which to modulate
the main frequencies. On the right we derive
these frequencies from the base phasor in a
similar fashion, 1
.
43
×
3147 = 4
.
5kHz, plus a second harmonic at twice the fun-
damental and about one-third the amplitude. This seems good enough without
a third harmonic, which is very quiet anyway.
564
Insects
Field Cricket 2
Figure 50.8
Field cricket pulse and
band-pass method.
As an alternative, let’s look at a different way to
achieve a similar result. We can calculate that each
click is 17ms apart, so let’s begin with a timebase
that produces clicks continually at that rate. This
is done with the
object. How many pulses will
there be in one period of 0
.
7s? Dividing the chirp
period by the pulse period gives 700ms
/
17ms = 41,
so let’s now add a counter and
operator to con-
strain counting over this range. To split this number
stream into groups of 7 and 34 (the remaining silent
pulses),
is used. So that the pulses grow in ampli-
tude the number is scaled into a range 0
.
2 to 1
.
0 by
adding 2 and dividing by 9, then substituting that
value as the upper limit of a 0
.
2ms pulse obtained
through
. Because these pulses are a little bit too
sharp, a low-pass filter softens them before we feed
them to some high-resonance band-pass filters which
produce the correct tone. Unfortunately this leaves a
residual low frequency at the pulse rate, so a further
high pass removes this. The result sounds much too
mechanical and unnatural with only two bands, so I’ve
added an extra two peaks very close to the fundamen-
tal to provide the effect of two wings at close frequencies. Notice the high gain
needed to recover the ringing signal from the filters when excited by such a
short pulse. This leads to a potentially troublesome patch that can make loud
clicks on startup if the filters are not initialised carefully. It works nicely in
Pure Data, but beware if you are translating this patch, and be sure to zero
any signal prior to the filters and trap any initial DC components.
Cicada
An approach is summarised in figure 50.9. We start with some harsh precondi-
tioning of the noise source to keep it well inside the 5
,
000Hz to 8
,
000Hz band.
Extraneous frequencies in the lower region seem to badly affect the sound by
bleeding though as modulation artifacts, so we kill everything we don’t want
in the spectrum first.
Two narrow peaks are split off at 5
.
5kHz and 7
.
5kHz and subjected to
a modulation at 500Hz, which widens up the sidebands and gives us a time
domain texture like that observed in the analysis. A pulse wave is derived from
a cosine oscillator using the 1
/
(1 +
x
2
) method to modulate the noise bands.
The width and frequency of this are controllable to set the singing call. Three
examples heard from the analysis recordings are switched between at random.
Modulating the noise after such high-resonance filters works to keep the sound
Do'stlaringiz bilan baham: |