Let us now consider the ideal situation whereby we increase the number of rings while reducing their widths until the delay becomes continuously variable along the radius of the membrane. Then we can isolate the effect of the delay profile from the discretization of the rings. If we treat the membrane as a pure pressure source with zero mass and stiffness, the far-field radiated sound pressure at a distance r and angle θ from its center is obtained by inserting Eqs. (13.70) and (13.121) into Eq. (13.124) to yield
p˜(r,θ)=jkcosθe−jkrr∫a0p˜+(w)J0(kwsinθ)wdw,
(15.1)
where J0 is the zero-order Bessel function, k=ω/c is the wave number, ω=2πf is the angular frequency, and p˜+(w) is the radial distribution of the electrostatic driving pressure. The tilde denotes a harmonically varying quantity where the term ejωt has been suppressed.
No Delay. If there is no delay, then the pressure everywhere on the surface of each side is just half the driving pressure
p˜+(w)=p˜02,
(15.2)
where
p˜0=ε0EPd2e˜in
(15.3)
and ε0 is the permittivity of air. The far-field pressure response then becomes that of Eq. (13.125)
p˜(r,θ)=jka2ε0EPd⋅e˜in2d⋅e−jkr2rD(θ),
(15.4)
where the directivity function is given by
D(θ)=2J1(kasinθ)kasinθcosθ.
(15.5)
The normalized directivity pattern 20 log10|D(θ)|−20 log10|D(0)| is plotted in Fig. 15.2.
Noting that D(0)=1, let the normalized on-axis response be
pnorm=p˜(r,0)ap˜0e−jkr/(4r)=jkaD(0)=jka,
(15.6)
which is plotted in Fig. 15.7 (black dashed). The input current I˜in≈jωCEe˜in is almost entirely due to the static capacitance CE=ε0πa2/(2d) so that the on-axis pressure simply becomes Walker's equation [7]
p˜(r,0)=EPd⋅e−jkr2πr⋅I˜inc.
(15.7)
Virtual point source. The geometry of a traditional “virtual point source” is shown in Fig. 15.3.
Because of its finite radius a, the membrane can only reproduce the part of the wavefront emanating from the source, which forms a spherical cap with half angle α and radius of curvature R, where
R=acotα.
(15.8)
To reproduce this, the delay must account for the time taken for the wave to travel the distance ΔR at each point w along the radius according to
ΔR=R2+w2−−−−−−−√−R.
(15.9)
Hence, the surface pressure distribution is given by
p˜+(w)=p˜02ejkΔR=p˜02e−jk(a2cot2α+w2√−acotα),
(15.10)
which leads to the directivity pattern
D(θ)=2cosθ∫10e−jka(cot2α+s2√−cotα)J0(kassinθ)sds,
(15.11)
where we have substituted s=w/a. The directivity pattern 20 log10|D(θ)|−20 log10|D(0)| is plotted in Fig. 15.4. The on-axis response is
Virtual oscillating sphere. Naively, we might insert the pressure produced by an oscillating sphere, given by Eq. (4.129), into Eq. (15.1), while setting P˜0=ρ0cU˜0/(πa2), to yield
pnorm=−2k2a22−k2a2+j2ka(e−jka−e−j2√ka2√),
(15.13)
where we have substituted r2=w2+a2 and cos θ=a/r. The first term in parentheses gives the true response of an oscillating sphere that would be obtained if the membrane were infinitely large. However, the second term, which is a “diffraction” term because of
the membrane's finite size, interferes with the first term to produce an irregular response as shown in Fig. 15.7 (light gray dashed). Hence, for our virtual oscillating sphere, we shall adopt the geometry shown in Fig. 15.5, where the axial distance between each point on the front surface of the virtual sphere and the membrane is a−a2−w2−−−−−−√.
The amount of delay required at each point along the radius of the membrane is the time taken for sound to travel this distance axially as a plane wave. Hence,
p˜+(w)=p˜02e−jk(a−a2−w2√),
(15.14)
which leads to the directivity pattern
D(θ)=2cosθ∫10e−jka(1−1−s2√)J0(kassinθ)sds,
(15.15)
where we have substituted s=w/a. The directivity pattern 20 log10|D(θ)|−20 log10|D(0)| is plotted in Fig. 15.6.
The on-axis response is
pnorm=kaD(0)=2jka(1−e−jka−jka),
(15.16)
which is plotted in Fig. 15.7 (black) along with the following first-order high-pass filter approximation (light gray)
pnorm≈2jka2+jka.
(15.17)
This approximation can be included as part of a crossover filter response, for example. The cut-off frequency is given by fC=c/(πa). Above fC, we have
pnorm≈2,f>cπa.
(15.18)
Then from Eqs. (15.4), (15.6), and (15.18), the voltage sensitivity is
p˜(r,0)=ε0aEPd⋅e˜in2d⋅e−jkrr,f>cπa
(15.19)
and at lower frequencies we have
p˜(r,0)=jε0ka2EPd⋅e˜in2d⋅e−jkr2r,f<cπa,
(15.20)
which is the same as Eq. (15.4) (with D(0)=1) for the voltage sensitivity of a massless flexible membrane at all frequencies. It is equivalent to Walker's Eq. (15.7) but given in terms of input voltage instead of current, so we shall call it Walker's voltage equation. The maximum field strength that we can realistically expect without breakdown is EP/d=2000V/mm. Similarly, the input voltage should not exceed 4000V peak across 2mm. If the radius a is 14cm and the permittivity of free space ε0 is 8.85pF/m, the maximum RMS sound pressure from Eq. (15.19) is 105dB SPL at 1m re 20μPa. This pressure increases by 6dB for every doubling of the diameter.
The on-axis plot of a virtual oscillating sphere shown in Fig. 15.7 tells us that, in theory, a continuously increasing delay in the driving pressure along the radius of the membrane produces a very smooth response, with just some very small ripples, and an almost constant figure-of-eight directivity pattern at all frequencies, as shown in Fig. 15.6. Although perfectly constant directivity is not achieved, the result is remarkably good considering the finite size of the membrane. Next, we will look at the discretization of the delay into rings of various widths.