7958 IRE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES

Parak14kmpled Transmission~Line~Resonator Filters*

SEYMOUR

Summmy-This paper describes the synthesis of band-pass ripple. The accuracy of the design formulas has been
transmission-line titers consisting of series of half-wavelength
checked by exact computation, and they have been
resonant conductors such as strips. The design differs from the usual
found to give excellent results for bandwidths up to
end-coupled strip configuration in that successive strips are parallel
coup~ed along a distance of a quarter-wavelength. The resulting about 20 per cent in the case of maximally flat response
coupling between resonators is partly electric and partly magnetic. and 30 per cent in the case of equal ripple response. ‘1’he
Several important advantages are gained by this arrangement: 1) the method of analysis is basically that used in an earlier
length of the filter is approximately half that of the end-coupled type; paper3 for a number of other types of direct-coupled-
2) the gaps are larger and therefore less critical; and 3) the insertion-
resonator filters. In this method, the lumped-constant
Ioss curve is symmetrical on a frequency scale with the first spurious
response occurring at three times the center frequency of the pass low-pass prototype filter having the desired responso is
band. made equivalent to a set of either series- or paral [el -
Formulas are derived for the parallel-coupled-resonator trans- resonant LC arms interconnected by broa,d-band quar-
mission-line filter that permit accurate design for Tchebycheff,
ter-wavelength transformers. An approximate equiva-
maximally flat, or any other physically realizable response. The
lence then is established between the latter circuit and
formulas are theoretically exact in the limit of zero bandwidth, but
frequency-response calculations show them to give good results for the actual coupled-resonator structure.
band widths up to about 30 per cent. An experimental strip-line filter
of this typ,s has been constructed, and the data given in this paper
show that excellent performance has been obtained.
=Z2%E2Z2H22%Z+=
INTRODUCTION ~~ ‘a)
I [
I
A
S shown in Fig. 1(a) ,1–3 multiple-coupled-resonator 1

band-pass filters in strip line (or other TEM (b) ~

.1 a~~ ~~nu
transmission line) have been most comr$only
composed of half-wavelength strips coupled end to end. (c)
In this paper, an alternative arrangement is treated in Fig, l—Coupled-resonator strip-line filters: (a) end coupled;
which the half-wavelength strips are parallel-coupled, (b) and (c) parallel coupled.

as in Fig. 1 (b) and 1 (c). Parallel coupling offers a num-

ber of important advantages over end coupling: 1) the DESIGN PROCEDURE
length of the filter is reduced approximately by half;
The design formulas for an m-resonator paral”[el-
2) a symmetrical insertion-loss-vs-frequency response is
coupled filter are given in Table 1. They utilize the ele-
obtained with the first spurious response occurring at
ment values gl, @ . “ . , g. of the prototype low-pass
three times the center frequency, and 3) a much larger
filter, which may be computed for either maximally flat
gap between adjacent strips is permitted. The last ad-
or equal-ripple response by means of formulas given in
vantage is of particular importance, since it eases the
Table I I.3 The schematic diagram in Fig. 2 shows that
tolerance on the gaps for a given bandwidth, or permits
the filter is assembled from n + 1 sections, resulting in a
a broader bandwidth for a given tolerance. Furthermore,
structure containing n resonators. The sections are of
the larger gap permits a higher power rating of the
equal length (one-quarter wavelength at the center [re-
filter.
quency), and their electrical design is completely speci-
Formu Ias have been derived that allow the parallel-
fied by two characteristic impedances—Zo,, of the even-
coupled-resonator filter to be designed in a straight-
mode wave on the parallel conductors, and Z~” of the
forward manner, and to have any desired physically
odd-mode wave. These characteristic impedances have
realizable response, such as maximally flat or equal-
previously been defined, and formulas and graphs rellat-
ing them to the dimensions of the cross section are
* Manuscript received by the PGMTT, October 21, 195 ~. The available.4 The total filter structure always will be sym-
work described in this paper was supported by the U. S. Army Signal
metrical for maximally flat or equal-ripple response.
Eng. Labs under Contract No. D.\ 36-039-sc-64625.
+ Stanford Research Institute, Menlo Park, Calif. In the design of a parallel-coupled-resollator fil ker
1 E, G. Fubini, W. E. Fromm, and H. S. Keen, ‘{Microwave appli-
one should first select the type of response function and
cations of hi~h-O striu components, ” IRE CONVENTION RECORD, pt.
8, pp. 98–1 O:; Marcfi, 195~. the number of resonators that will yield the desired in-
2 E. A. Bradley, “Design and development of strip-line filters, ”
IRE TRANS., vol. MTT-4, pp. 86–93; April, 1956.
J S. B. Cohn. “Direct-couded-resonator filters, ” PROC. IRE, vol. 4 S. B. Cohn, “Shielded coupled-strip transmission line, ” IRE
45, pp 187-196 ~ February, 1$57. TRANS., vol. MTT-3, pp. 29-38, October, 1955.

Authorized licensed use limited to: INSTITUTO NAL. ASTROFISICA O. Y ELECT (INAOE). Downloaded on August 17,2020 at 03:44:43 UTC from IEEE Xplore. Restrictions apply.
 224 IRE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES April

TABLE I
FORMULAS FOR PARALLEL-COUPLED TRANSMISSIOWLINE-RESONATOR FILTER

(n

t-’-t-+-’ $ PROTOTYPE
~ LOW-PASS BAND-PASS
I I I
I [—,
I
I 1. [ I
2- f, f. f2
-
1
1
i
I I

n ~zo
n+l

(For either maximally-flat or equal-ripple response, the structure is symmetrical.)

1 = AO/4, where AO = wavelength in transmission line at ,fO

/0= (f, +,fJ/2

gl, gz, . . . , b = prototype elements in farads and henries from Table 11

r = right-hand load resistance in schematic of Table II.

W’ = pass band edge of prototype filter.
~,, .fz = corresponding pass band edges of transmission-line filter.
Z.,i = even-mode characteristic impedance with respect to ground of each conductor in ith section.
ZOO,= odd-mode characteristic impedance with respect to ground of each conductor in ith section.

(The dimensions for strip-line construction may be obtained as function of Zo.t and Zao, from Cohn:)

sertion-loss function in the pass and stop bands. This dimensions in each section should be designed to yield
may be done with the aid of the insertion-loss formulas these characteristic impedances. This ~ may be done
in Table I I and the following approximate relationship easily in the case of thin strip conductors, and with
between the frequency scales of the prototype filter and somewhat more difficulty in the case of thick strips, by
the parallel-coupled-resonator filter: means of the available graphs and formulas.1 It is noted
that the strip widths and spacings will, in general, differ

()
f–fo
J . Zall — (1) from section to section, and hence the width of the
f,–fl “ resonators will not be constant, as shown in Fig. 2. This
The exact response curves and the bandwidth-error variation in width is necessary to compensate for differ-
curve (which appear later) show how response errors are ences in the coupling gap. The amount of variation de-
introduced when the relative bandwidth exceeds 0.1. creases as the bandwidth is reduced and, except perhaps
The bandwidth error may be reduced by preadjusting for the first and last section, the variation is negligible
the design with the aid of these curves. in the case of bandwidths less than a few per cent.
Next, the element values gl, g2, . . . , g. may be com- One further important step in the filter design is to
puted, and in terms of these the Z.. and Zoo values of alter the length of the resonators to compensate for
each of the n+ 1 sections. Then, the transmission-line fringing capacitance at their ends. This may be done

Authorized licensed use limited to: INSTITUTO NAL. ASTROFISICA O. Y ELECT (INAOE). Downloaded on August 17,2020 at 03:44:43 UTC from IEEE Xplore. Restrictions apply.
 ! 95% Cohn: Parallel-coupled T’ransmission-Line-Resonafor Filfers 225

TABLE 11
PROTOTYPE LOW-PASS FILTER AND ITS DESIGN EQUATIONS FOR MAXIMALLY-FLAT AND TCHEBYCHEFF RESPONSE

‘@::%r=’ :Qr
n ODD n EVEN

Maximally-Flat Response

r=lforalltz Al
db
“=2 ‘in[%3 k=l,’,. ... n

A = 10 log,a (1 + co’z”)db

3db
L?14
o ,=
(J
cl) I
—— .—

Tchebycheff Response

r = 1 for n odd, r = tanh2 (6/4) for n even d! I

JnnnK.
g%= 2a~/y
‘i@d@
gb = ——— ,
k=2,3, ... n
bk-1zh-1

check: g. = glr

(L)’=1
krr
b~=y’+sinz — , k=l,2, . . ..n
() n

P = log, coth ~ , A,. in db

( )

A = 10 iog,o [1 + (lO%j’O – 1) COS2 (n COS-’ (d’) ]db,

A = 10 log,, [1 + (lOA#O – 1) cosh’ (n cosh-’ w’) ]db,

T ‘n+t ‘T
s ~1
n+l

Fig. 2—Basic dimensions of the strip-line parallel-coupled-resonator filter.

Authorized licensed use limited to: INSTITUTO NAL. ASTROFISICA O. Y ELECT (INAOE). Downloaded on August 17,2020 at 03:44:43 UTC from IEEE Xplore. Restrictions apply.
 2’26 IRE TRANSACTIONS ON MfCROWAVE THEORY AND TECHNIQUES April

J++: $+d>

I
I
1, ++1, ~d

~d
-+d

(a) (b)
Fig. 3—(a) Strip-resonat?r design+ ne@ecting f:inging capacitance at ends. (b) Suggested
compensation for frmgmg capacitance at ends of resonators.

2.0 1.2
I I I [
I .8 — 1!

10
1.6 — f2 -f,
— = 0.05
fo 1.[
I .4 —

10
1.2” —
1.1
1 I I I I I I
1.0 ! I ,0
1.8 — Ii

[ ,0 I I
1.6 —
fg -f,
E
—= 0.1
f. :,, ,
1.4 —
>

I .0
I I
1.2 —
cc 12 —
I I
31.0
W
> 1.1

I .0 — 1 I
[.0 I

1.6 — I .4 — f2 -f,
—=050

I .4 — I .3 —

1.2 —
I .2 —

1.1

I I I
I .0
-1 ?. -10 -08 -06 -0.4 -0.2 0 02 04 06 08 10 12
1.8 —
(f -fo)/(f, -fJ

1.6 —
—= 0.3
Fig. 5—VSWR curves of six-resonator, parallel-coupled filter
designed for equal-ripple respons~vswr = 1.10.
I .4 —

I .2 —
bandwidth filters an error in d would be unimportant,

AA
while in narrow-bandwidth filters the error would be
I
I .0
-1.0 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1.0
cancelled by the tuning adjustments that are already
(f-fo)/(f2-fo) needed to overcome the effects of constructional toler-
Fig. 4—VSWR curves of six-resonator, parallel-coupled filter ances. This tuning may be done, for example, by means
designed for maximally-flat response. of screws or dielectric slugs in high-electric-field regions
of the resonators.

as shown in Fig. 3(b), where each end is cut back by the

average length, d. The dimension d would be expected
VERIFICATION OF DESIGN ACCUR.~CY
to be somewhat less than (Cf ‘/e) (b/2), where Cfr/e is a
quantity plotted elsewhere as a function of the strip- Because of the approximations necessary in the der-

thickness-to-plate-spacing ratio, t/b.5 Measurements on ivation of the design formulas, their accuracy was

the experimental filter described below indicate that d checked theoretically. This was done by exact computa-

should be approximately 0.75 (Cf ‘/c). (b/2). For very tion of the insertion-loss and vswr responses of the ac-

thin strips, this would be 0.75 (0.220b) = O. 165b. In wide- tual transmission-line filter networks specified by the
formulas for various bandwidths and for either maxi-
mally flat or Tchebycheff response. The computations
5 S. B. Cohn, ‘(Problems in strip transmission line, ” IRE TRANS.,
Vol. MTT-3, pp. 119-126; March, 1955. were performed on an electronic computer by a matrix-

Authorized licensed use limited to: INSTITUTO NAL. ASTROFISICA O. Y ELECT (INAOE). Downloaded on August 17,2020 at 03:44:43 UTC from IEEE Xplore. Restrictions apply.
 1958 Cohn: Parallel-Coupled Transmission-Line-Resonafor Filfers 227

60
‘0 --1- I I I I I / I 1 1 I I
I
I I I I I I I I I I I /1 I /~1

1 /

50 —

‘11-1~
40 — .
--i
:

8
3

~[,l,x,,,,
~ 30
F
.
%
~

20 —

10 — ‘.f :03
~=oz
f. <~,,

0 ~J--l--l.l
0 02 04 06 os IO 12 14 16 la 20 22 24 26 28 0 02 04 06 C
m 10 !,2 14 16 18 20 22 24 26 2.6
lf-fol/(f, -fo) lf-fol/(f2-fo)

Fig. 6–-Insertiomloss curves of six-resonator, parallel-coupled Fig. 7—Insertion-loss curves of six-resonator, parallel-coupled
filter designed for maximally-flat response. filter designed for equal-ripple response—vswr = 1.10.

multiplication method. eBecause the filter is symmetrical,

it is sufficient to compute the matrix of half the filter.
The insertion loss and vswr then were computed as
functions of the elements of this matrix.
The pass band vswr curves for six-resonator maxi-
mally flat filters of various bandwidths are shown in
Fig. 4. The relative bandwidth is defined as

#2-fl
(2)
jo ‘
_20Li-L_LLl_.l
0 005 010 0[5 020 025 030
where f o= (fl +fJ /2, and fI and .fz are the $db points Of f2 - f,
f.
the response curve. The response is seen to be truly
maximally flat for relative bandwidths up to 0.1, while Fig. 8—Bandwidth error vs relative bandwidth for six-resonator
at 0.2 the deviation from maximally flat is slight. At parallel-coupled filter. Error is expressed as a percejltage of the
bcmdwidth; e.g., at (.fz–.fJ /fo = 0.10, corrected bandwidth = ~.098.

0.3, however, the deviation would be serious in the more

critical applications. The corresponding curves are
shown in Fig. 5 for six-resonator filters designed to have frequency, only half of each pass band is shown. In both
an equal ripple vswr of 1.10 in the pass band. In this figures, the insertion loss is virtually identical with that
case, fl and fz are the pass band limits for the equal of the prototype function for relative bandwidths u p to
ripple-level vswr of 1.10. The desired response is ob- 0.1, and even for greater bandwidths, the deviation is
tained very accurately for relative bandwidths up to quite small,
0.05, and gradually deteriorates as the bandwidth is in- In addition to the errors in response-curve shape that
creased further. In spite of this deterioration, the pass develop as the bandwidth is increased, the predicted
band vswr limit of 1.10 is not exceeded for relative band- bandwidth is slightly in error. In Fig. 8 the bandwidth
widths up to 0.3, and even at 0.4 or 0.5 the response is error in per cent is plotted vs relative bandwidth for the
adequate for many applications. But, it is important to cases considered above. It is seen that the discrepancy
note that these conclusions are drawn from specific cases does not exceed 2 per cent of the bandwidth for rela tive
considered, and may vary somewhat with other numbers bandwidths up to 0.1, and is only about 6 per cent at
of resonators or with other equal ripple levels. 0.2. The actual bandwidth of the filter appears to be al-
The insertion-loss curves for the above cases are ways less than the value assumed in the design, and
plotted against a normalized frequency scale in Fig. 6 therefore, in the case of the wider bandwidths, it would
and Fig. 7. Because the response is symmetrical with be desirable in calculating the parameters of a given
filter to use a somewhat larger bandwidth than actually
is required. Fig. 8 mav be used as an approximate guide
8 P. 1, Richards, “Applications of matrix algebra to filter theory, ”
PROC. IRE, vol. 34, pp. 145P-150P; March, 1946. in selecting the bandwidth design value.

Authorized licensed use limited to: INSTITUTO NAL. ASTROFISICA O. Y ELECT (INAOE). Downloaded on August 17,2020 at 03:44:43 UTC from IEEE Xplore. Restrictions apply.
 228 IRE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES April

l--’-d

7
Fig. 9—Layout of parallel-coupled-resonator filter.

TABLE III
DESIGN PARAMETERS FOR EXPERIMENTAL PARALLEL-COUPLED STRIP-LINE-RESONATOR FILTER

z.
‘i Zo.i zoo, Wi si di
KL-Li
.
0.449 82.5 ohms 37.6 ohms 0.236 inch 0.021 inch 0.073 inch
: 0.1529 58.8 ohms 43.5 ohms 0.346 inch 0.110 inch 0.084 inch
0.1038 55.7 ohms 45.3 ohms 0.360 inch 0.158 inch- 0.085 inch
: 0.0976 55.4 ohms 45.6 ohms 0.361 inch 0.163 inch 0.085 inch

DESIGN OF EXPERIMENTAL FILTER mode characteristic impedances ZO. i and ZOO{are com
A strip-line parallel-coupled-resona tor filter has been puted from the formulas in Table I. In this calculation
constructed and tested in order to demonstrate the v2–~1)/J”o is set equal to 0.1, and ZO to 50 ohms. Then
feasibility of this type of design. As shown in Fig. 9, the strip widths w; and separations s, of the various sec-
the filter has six resonators. The two ground planes were tions are obtained from the nomograms of a previous
spaced 0.5 inch apart by a polystyrene dielectric. The paper,4 with E, equal to 2.55, L equal to 0.5 inch, and
strips are of copper foil, 0.0017 inch thick. The filter was strip thickness assumed to be zero. Finally, the resona-
designed for 10-per cent bandwidth, centered at 1200 tors are shortened at their ends by di to compensate for
mc, and an equal ripple pass band vswr of 1.10. The fringing capacitance. The various quantities referred to
exact theoretical vswr and the insertion-loss curves of above are contained in Table 111.7
the actual transmission-line filter network, as calculated A further quantity required is the width w~ of the
by an electronic computer, are shown in Fig. 5 and Fig. terminating 50-ohm strips. This may be obtained from a
7. In each figure the appropriate curve is labeled graph of characteristic impedance vs strip widths,8 The
result is wT/b = 0.744 or wT = 0.372 inch. The section
(~z :~1)/~ = 0.1. The details of the design procedure and
a discussion of the experimental results are given below. length 1 is a quarter-wavelength in the dielectric, and
Fig. 9 shows that the filter has seven sections, but be- hence equals 1.540 inch.
cause of the symmetry of the filter, it is necessary to Photographs of the completed filter are shown in Fig.
compute the parameters of only the first four sections. 10. The structure is a sandwich of copper foil separated
The first step in the design procedure is to compute the by a pair of polystyrene plates each machined to have
element values gl, ga, ga, and gl of the prototype seven- flat surfaces and a thickness of 0.250 inch. Thus, the

element, low-pass filter. This is done by means of the total ground-plane spacing is one-half inch. The copper

formulas for Tchebycheff response in Table II, with foil was cemented to the polystyrene plates with Dow

n = 7 and Am = 0.00986 db (which corresponds to an in-

put vswr of 1.10). The resulting values are 7 The values of d~in Table I I I were computed from an approximate
formula, and are quite close to the value O.165~ mentioned above.
Because the formula is rather complex, and its assumptions are as
gl = 0:77968, gz = 1.35921, g~ = 1.68800, gl = 1.53454.
yet unproved, it is not given here.
8 S. B. Cohn, “Characteristic impedance of the shielded-strip
Next, the quantity Z./K;. I,, and the even- and odd- transmission line, ” IRE TRANS., vol. MTT-2, pp. 52-57; July, 1954.

Authorized licensed use limited to: INSTITUTO NAL. ASTROFISICA O. Y ELECT (INAOE). Downloaded on August 17,2020 at 03:44:43 UTC from IEEE Xplore. Restrictions apply.
 1958 Cohn: Parallel-Coupled Transmission-Line-Resonator Filters j!29

40, , 1 1 1 / 1
I
\l I
THEORETICAL FOR 1. = 1207..
.. ..”. ( EXPER, MENTAL

35 \

.- —
30 —- -
!

>. —
I
:
I
~

; 20 -—
0
~
.
2
II +/ ‘--”--

,5 - \ .

40 -. --—

I // “-”
5

. ....... .....
“m.b 1100 II ,,00 ,2s0
>,.,0”,,.”- ,,00 II 0
.,
(a)
Fig. 1l—Insertion loss of parallel-coupled-resonator filter.
I
$8

!6 I

,.315 ,/50 ,,;5 I I I I

,.00 ,2,6 1250 ,27!> ,300
,REOUENC Y-MC

Fig. 12—Measmed vswr of parallel-coupled-resonator filter,

(b) insertion loss of the filter and the exact theoretical curve
Fig. 10-–Photographic views of parallel-coupled resonator filter:
for the transmission-line network computed for a center
(a) with upper plate removed; (b) completely assembled. frequency of 1207 mc and neglecting dissipation. Above
2 db, the agreement is excellent. The bandwidth error is
Corning XC-271 adhesive, which was applied to the seen to be only about 1.0 per cent of the bandwicith.
copper and air dried for 45 minutes before pressing the In the pass band, however, disagreement is inevitable
foil ontcl the polystyrene. The strip circuit then was cut because of the finite Q of the strip-line resonators.
on one surface with a sharp knife and the unwanted foil Nevertheless, the measured pass band insertion loss is
peeled off. quite uniform at 0.5 to 0.7 db.
The input vswr of the filter measured with a 50-ohm
DATA FOR EXPERIMENTAL FILTER
termination on the output is shown in Fig 12. Although
A preliminary model of the parallel-coupled resonator the vswr exceeds the theoretical ripple level of 1.10, it
filter was tested first with the resonator ends uncompen- would be quite acceptable for most applications. The
sated; Le., with di = O. The center frequency was lower highest vswr peak in the pass band occurs near the low
than the design value by 4.4 per cent. Then, the open- frequency end, where it reaches a value of 1.33, while
circuited strip ends all were cut back uniformly a dis- the vswr elsewhere in the pass band is under 1.20. l[t is
tance d = 0.220b, the value of (Cf’/e)/(b/2) for very thin believed that the vswr could be made to conform more
strips, and the center frequency was higher than the closely to theory if tuning adjustments were provided
design value by 1.7 per cent. Finally, the filter was re- for the resonators. These adjustments also could serve
constructed with the values of di given in Table III. The to set the center frequency of the filter exactly on the
center frequency of this model is 1207 mc, rather than desired value.
the design value of 1200 me—an error of only 0.6 per
DERIVATION OF DESIGN FORMULAS
cent. In the three cases the bandwidth and response
curves ~are quite similar. Although the best pass band The analysis of the parallel-coupled-resonator filter- is
respons~ was obtained with the last case, it may have based upon the characteristics of the individual section
been due to more accurate construction. of Fig. 13(a). The image impedance ZI and image phase
Fig. 11 shows a comparison between the measured shift p of this section are given in a paper by Jones and

Authorized licensed use limited to: INSTITUTO NAL. ASTROFISICA O. Y ELECT (INAOE). Downloaded on August 17,2020 at 03:44:43 UTC from IEEE Xplore. Restrictions apply.
 230 IRE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES April

=
(b)

--- lpi t—-++ , , -.

INPUT K
Z. Kol Z. Z. 12 Z. Z. ’23
LINE Z.
—-- -90” ,. “ ~, -90” ,- __

(c)
X=-z. cot #

*
“at (d)
zE2iii=-yO (e)

——
1 1 K
Kol 1 K12 1 ’23 n,n+l
TC TCr
-90” ‘r -90° ‘r -90° -90”
j ~ —-
7 :H

(f)
Fig. 13—Development of equivalent circuit of the parallel-coupled-resonator filter.

Bolljahng as follows The matrix of the ideal inverter is obtained from this by
substituting @ = — 90° and Zo = K, as follows
[(2.. – zoo)’ – (Zoe + zoo)’ Cos’ +]’/’ (3)
Zr =
2 sin +
o –jK

Cos /8 = (4) –.j

(::3’0s” L -. o“ 1
LA -1
where O is the electrical length of the coupled transmis-
sion lines and Zo. and Zoo are the characteristic imped-
Therefore, the A B CD matrix of the complete filter sec-
ances of each conductor with respect to ground for the
tion of Fig. 13(b) is
even and odd modes, respectively.
It is now shown that the filter section of Fig. 13(b) is
approximately equivalent to that of Fig. 13(a). The box
cos @ ~ZO sin+ o –jK-
AB
represents an ideal impedance
stant image impedance,
inverter having a con-
K, and a phase shift of — 90°
[1[ CD—
= ~sin+

20–
Cos cp
1[
x –j
7.
o
at all frequencies. The image impedance, Zr, and image
phase shift, /3, of the section will be derived by means of
the A B CD matrix parameters.b ,10
X ~sin~
The A B CD matrix of a transmission line of length @ Cos +
and characteristic impedance 20 is 1 20

[+’:::]- , (K )
(K
~ ~~ sinz ~ Cos’ d
g E. M. T. Jones and J. T. Bolljahn, “Coupled-strip-transmission- o K E’
line filters and directional couplers, ” IRE TRANS., VOI. MTT.4, pp,
75-81; April, 1956.
10Radio Research Laboratory Staff, “Very High Frequency Tech-
niques, ” McGraw-Hill BoolI Co., Inc., New York, NT. 1’,, vol. 2, ch.
The image parameters are related to the matrix ele-
26; 1947. ments by ZT = \/B/c and cos ,fl = A and, hence,

Authorized licensed use limited to: INSTITUTO NAL. ASTROFISICA O. Y ELECT (INAOE). Downloaded on August 17,2020 at 03:44:43 UTC from IEEE Xplore. Restrictions apply.
 1958 Cohn: Parallel-Coupled Transmission-Line-Resonator Filters

(7

(Z. “’” w K ‘“” ‘J

42.3

— –1 1 K
—
4z0’(:+:)c0s2+
—
]“2

!
2 sin+ K Z.
~+:–
0 K sin’ +

zoo
Cos /3 =
() ~+~
o
sin@ cosq5. (6) —.
z.
1–;+:. (14)

A comparison between (3) and (5), and between (4) and Jt may now bc noted that, for narrow bandwidth,
(6), shows that the sections of Fig. 13(a) and 13(b) Z./K<< 1 and, therefore, (12) is satisfied approximately
would be electrically equivalent if the following condi- by (13) and (14). Thus the effect of the approximations
tions were met, involved in relating the sections of Fig. 13(a) and 13(b)
is negligible for narrow bandwidth, but increases in im-
z.. + zoo
()
portance as the bandwidth is increased.
$+: sin+= (7)
0 z., – zoo When the individual sections of the filter are as-
42.3 sembled, it is seen that the original parallel-coupled-
resonator filter of Fig. 1 (b) or 1(c) is approximately
K equivalent to the circuit of Fig. 13(c), in which trans-
= (Zo. – zoo)’ (8)
Z. mission lines approximately &/2 long are separated by
:+$– inverters. A lumped-constant equivalent of a line of
0 K sinz @
length O is shown in Fig. 13(d). This exact equivalent cir-

()
42.2 : + 5 cuit is particularly convenient for 6’ near 180°. In Fig.
K 13(e) the equivalent circuit has been simplified by
- = (Zo, + zoo)’, (9) omitting the phase-reversing transformer which has, no
z.
~+~.—– effect on the insertion-loss response. Also, the series re-
actance of Fig. 13(f) are small near 6 = 180° and are
negligible in comparison with the high characteristic
The presence of sin@ in the left side of each of the above
impedances of the inverting elements. Thus, the circuit
makes a frequency-independent equality impossible.
of Fig. 13(f) is approximately equivalent to that of
However, sin @ is stationary in the vicinity of @ = 90°,
Fig. 13(c), and hence to the original parallel-coupled-
anc[ hence may be replaced by unity with negligible error
resonator filter.
over a moderate bandwidth. Thus,
The analysis will be shortened at this point by com-

z.. + zoo paring the filter circuit of Fig. 13(f) with that shown in
:+:= (lo) Fig. 11(c) of a previous paper.’ They are seen to be the
0 z“. – zoo same, except that the former contains parallel-resonant
22.2
arms while the latter contains series-resonant arms.
— = Zo, – zoo (11)
K Hence, the circuits are duals, and the formulas obtained
for the latter may be used in the present analysis if -K is
4z03 K za
(12) replaced by I/A”, Zo by 1/20, and if L and C are inter-
y ~+ ~ = (z”. + zoo)’.
() 0 changed. Thus,

From these equations, formulas for Zo,/Z~ and Zoo/Zo as z. 7rw

. i=lton —1, (15)
functions of Z~/K may be found. Only two of the three Ki, i+ I 2w’dgEg%+l ‘
equations are needed for this purpose, and since the
where W is the relative bandwidth defined by (2). For
three do not form a consistent set, a judicious choice
the first and last section of the filter,
must be made among them. Examination of (3) to (6)
indicates that the third relation is the least important —.—
z. . w z. 1 3rwr

“d —
T
of the three. Therefore, (10) and (11) are solved simul- ——
taneously to yield KOI – 2WI’gl ‘ K.,n+I – ( —
2wl’gn
. (16)

z.. The formulas in Table I follow directly from (13)

.I+; +g (13)
z through (16).

Authorized licensed use limited to: INSTITUTO NAL. ASTROFISICA O. Y ELECT (INAOE). Downloaded on August 17,2020 at 03:44:43 UTC from IEEE Xplore. Restrictions apply.