Delay Path Length. We saw in the last section that discretization of the delay into rings of finite width produces irregularities in the response at higher frequencies. However, the delay was an ideal delay like that produced by a DSP, whereas in practice, the delay is more likely to take the form of an analog delay line on the high-voltage side
of the transformer, such as that shown in Fig. 15.12. Otherwise, a separate stepping up transformer would be needed to feed each ring. Unless complicated inductors with center taps or a very large number of inductors are used, analog delay lines tend to introduce a degree of attenuation at the higher frequencies. In this case, this turns out to have a smoothing effect on the response.
Table 15.1
Dimensions of rings.
Ring
Equal delay (mm)
Equal area (mm)
Equal width (mm)
a0
72
53
20
a1−a0
26
22
20
a2−a1
17
17
20
a3−a2
12
14
20
a4−a3
7.7
12
20
a5−a4
4.4
11
20
a6−a5
1.4
10
20
From Fig. 15.5, we see that the total delay path length zT at each point along the radial ordinate w is given by
zT=a−a2−w2−−−−−−√=a(1−1−w2a2−−−−−−√).
(15.26)
For discretized rings, the delay section zn required for the nth ring is the difference between the total delay zTn at that ring and the sum of all the previous delay sections
zn=zTn−∑n−1m=1zm,
(15.27)
where
zTn=a(1−1−(an−1+an2a)2−−−−−−−−−−−−√).
(15.28)
Notice that we have taken the delay path to the mid-point (an−1+an)/2 of each ring. The nth path length zn is related to the time delay Tn of the nth section by
zn=cTn,
(15.29)
where c is the speed of sound.
Delay Line ignoring Stray Capacitance. In Fig. 15.12, CRn are the ring capacitances that are given by
CRn=ε0π(a2n−a2n−1)2d
(15.30)
CR0=ε0πa20/(2d)
(15.31)
while Cn are shunt capacitors used to make up the required capacitance for the correct delay and impedance. The delay line comprises inductors Ln together with the total capacitances of each section
CTn=CRn+Cn/2
(15.32)
where RT is the termination resistance on the far right-hand side of Fig. 15.12. The same resistance RT is also connected across each inductor to create a series of Zobel networks such that the impedance presented to the preceding section is always 2RT. The voltage transfer function of each delay section is given by
e˜ne˜n−1=ωns+ωn
(15.33)
provided that the inductor values are set to
Ln=2R2TCTn,
(15.34)
where s=jω and the angular turnover frequency is given by
ωn=12RTCTn.
(15.35)
The time delay Tn per section is defined by
Tn=znc=1ωn=2RTCTn=LnRT.
(15.36)
so that the total capacitance per section is given by
CTn=zn2cRT.
(15.37)
We can now furnish each section of the delay with its respective component values