1064 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQCJES, DECEMBER 1975

~rel
cidt to construct. A very good compromise can be found for C~
between O.014 <C, <0.04 .Thisresultsinl.04 < lE.1~~, <1.36.
Other important properties are the attenuation factor and the Q
1,3
factor. Taking into account finite conductivity of the waveguide
walls for a TM mode, the attenuation constant has the following
1>2
expression:

(C/2A )[ (fO/lJO) ]Ij2Rm ~c

1,1-
(15)
a = [1 (coo/& l)]/

1
where

circumference of the cross section;

area of the cross section;
surface resistance; I
0.01 0,02 Q03 Q04 c~
dimensionless number depending on the shape of the cross
Fig. 9. Reiative Q factor versus thepertmbation factor C*.
section and given bythe following relation:

u. permeability of the metal walls of the cavity;

6 skin depth;
The relative attenuation factor compares the attenuation in the d length of the cavity.
perturbed waveguide by the use of nonseparable solutions, with
In the same manner a relative Q factor can be defined in the form
that of the nonperturbed or rectangular waveguide and is given as
follows: 1 +~OCd/4A
Q, = (19)
(C/2A)& 1 +&Cd/4A
= [(c/2A) l&. (17)
a = (C/2A)~:
Fig. 9 shows Q, as an increasing function of the perturbation fac-
In Fig. 7weshow (C/2A ),and&, versus theperturbation factor Cz. tor C2.
It is obvious that only the factor .$=, has a considerable variation In conclusion we can say that nonseparable solutions of the Helm-
and is therefore a measure of the attenuation factor presented in holtz wave equation are suitable for describing TE and TM modes
Fig. 8. For cylindrical cavities, the Qfactoris given by: in waveguides and cavities with general cross sections; moreover,
they are exact wave functions for any deformed conventional rec-
(18) tangular or circular waveguide so that no approximate method of
solutions has to be taken. Further, it has been shown that wave-
where guides and cavities synthesized with nonseparable solutions have
better attenuation and Klgher Q factor than comparable conven-
tional waveguides shapes.
Further, it is clear that they have interesting properties for micro-
wave measurement and power applications. All calculations were
done on the computer of the Rekencentrum of the Catholic Univer-
sit y of Louvain.

REFERENCES

[1] P. M. Morse and H. Fesbach, Methods of Theoretical Phusics. New

York: McGraw-Hill, 1953, pp. 753757.
[2] D. S. Moseley, Nonseparable solutions of the Helmholtz wave
equation, Quart. APP1. Math., vol. 22, Pp. 356357, 1965.
[3] P. J. Luypaert and D. H. Schoonaert, in 1974 Proc. Microwave Power
Stimp., pp. 135-1/l-135-l/3 and B5-4/lB5-4/3.
[4] P. Lagasse and J. van Bladel, Square and rectangular waveguidas
I 0,0, 0,02
with rounded corners, IEEE Trans. Microwaue Theorfj Tech., vol.
0,03 ,,0, C2
MTT-20, pp. 331337, May 1972.
Nig. 7. Relative values of (C/2A)r and &, as a function of CL

Tapered Asymmetric Microstrip Magic Tee

M. H. ARAIN AND N. W. SPENCER, MEMBER, IEEE

AbsfractThe design, development, and construction of a very

compact decade-( 1-1 O-GHZ) bendwidth microstrip 8.34-dB cou-
pler are described. Calculations are given for the voltage coupling
coefficient and the low-frequency cutoff, and the method of deter-
~g the physical dimensions of the circuit is described. Also, the
feasibility of a decade-bandwidth microstrip magic tee by cascading
two 8.34-dB couplers is demonstrated by comparing the actual and
I theoretical results of a coupler.

I
0,01 0D2 0,03 0,04 rz Manuscript received April 21, 1975; revised August 26, 1975.
The authors are with the Autonetics Group, Rockwell International,
Fig. S. Relative attenuation factor versus the perturbation factor Cz. Auaheim, Calif. 92803.
 SHORT PAPERS 1065

INTRODUCTION It is assumed that the four ports of the coupler are terminated in a
characteristic impedance 20, that the relationship dejined by (2)
Thk short paper describes the design of an asymmetric broad- holds along the coupling section, and that the even- and odd-mode
band magic tee. The high-pass nature of this coupler makes it velocities are equal. Hence the characteristic impedance of the cou-
inherently broad band, and such a component is applicable for use plingregion ZOis given by (2) as
in many broad-band microwave assemblies includlng balanced mixers
and beam forming networks. Alternate magic tees which have been zo@(z/l).zoo(z/t) = 202. (2)
realized in microstrip have not had the inherent bandwidth capa-
The voltage coupling coefficient k and the physical ~ength of the
bility. Both the slotline/microstrip magic tee [1] and the magic
coupling section tforthe coupler are determined from (3) and,(i).
teeusing cascade-tandem connection of directional couplers [2] are
limited to 2:1 frequency bandwidths. ho m..
1 = (l/A). (3)
To validate the design approach, a test circuit was made which (,,)/
required the development of a fine-line delineation process and
special packaging techniques tosupprew spurious modes. The actual
/c(Z/l) = ~KJZ/l)m (41)
results correspond with the theory except for the phase shtit below ~=~
the minimum frequency of operation. Thk effect has not been ex-
where ~, is the relative effective permittivitv of the media and
plained. While the results obtained at this time indicate-further
kr~a. isthedesired maximum free-s~ace wavelength. (i/X), and K~
work is required, the feasibility of such a network has been demon-
arethelow-end cutoff frequency andthecoupling facto, coefficients,
strated.
respectively. For a given mean coupling value and voltage ripple &
The theory for asymmetric couplers has been described by Arndt
K and (l/h), can be obtained from Arndts table [3]. For a 8.34-dB
[3] and Tresselt [4]. Also, Duhamel and Armstrong [5] demon-
coupler with a 0.5-~ercent voltage ripple, the low cutcff frequency
strated this magic tee in low dielectric stripline transmission media.
(l/A). is to be 0.409 and the coupling factor coefficients are 0.6687,
However, prior to this work the feasibdity of producing this coupler
1.0507, 0.7719,2.3164, 1.0692, 0.3929, and O.31)44.
in a microstrip configuration had not been demonstrated. Thk was
To calculate the impedances and dimensions, the total length of
due to several limitations. First, the planar structure with edge
the coupler is divided into 200 equal-length sections, and the volt-
coupling made it dMicult to obtain the required tight coupling.
age coupling coefficient k(z) at the midpoint of each section is deter-
Second, the fabrication techniques for achieving repeatable 0.0002-in
mined using (4). Then the even- and odd-mode impedances that
gaps werenot available forathick metallization layer (=0.0005 in).
correspond to these k(z) values are calculated from (l) and (2)
Third, theinhomogeneous transmission medlawith ahlghdlelectric
and the physical circuit dimensions are determined using the com-
constant substrate degraded the dlrectivity of such coupler+ because
puter program subroutine provided by Smith [8] for :ornputation
the even- and odd-mode velocities of propagation were not equal.
of even- and odd-mode fringing capacitances of coupled microstrip
The potential solutions to these problems are described.
lines insuspended substrates. Alumina (e, = 10) waschosen as the
Theasymmetric microstrip magic tee consists of two 8.34-dB
high-dielectric substrate material, and the substrate thickness was
couplers cascaded intandemto forma 3-dBcoupler [6]. Reference
chosen to be 0.025 in. The effective dielectric constant e.~f with the
lines are added to each side of the coupler to compensate for the
close proximity ground-plane spacing was equal to 5.5, Thenfrom
frequency-dependent length and to give a transmission phase cliff er-
(3) thelength of thecoupler wascalculated to be 2.05 in for a low-
enceof 180 and OOat theoutput ports when fed from the main port
frequency cutoff of 1 (lHz.
and isolated port, respectively. The 8.34-dB coupling was selected
The circuit was fabricated on a 0.025-in-thick alumila substrate
because the edge coupling required at the tight coupling region was
with a top ground plane separated from the circuit by an air space
considered within reason; subsequent processing development re-
of 0.025 in. The gap width tapers down from 0.070 in at the loosely
sulted in an additive process which achieved the required coupling.
coupled endto 0.00018in atthetightly coupled end. Tlke 8.34-dlB
Improved directivity was achieved by positioning the top ground
coupler was tested initially on an automatic network armlyzer. Satir~
plane very close to the dielectric substrate; from the literature [7]
factory results were obtained over a 5:1 frequency bandwidth. .How-
0.025 in was selected as the spacing. However, it was determined
ever, better performance had been anticipated. Therefor: the 8.34k
empirically that 0.015 in was superior for our coupler dimensions.
dB coupler was tested on a time-domain reflectometer which meas-
ures the coupling amplitude along the coupled lines. Ctoupling wa~s
ANALYSIS OF 8.34-dB COUPLER observed beyond the actual coupled section. It is belie~red that this
undesirable coupling was caused by the surface waves and trapped
Theschematics of theasymmetric hgh-pass 8.34-dBdkectional
waveguide modes, which can occur in the rectangular package. To
coupler are shown in Fig. 1. They consistof two transmission lines
correct this undesirable condition, the width of the cir,xrit package
which are coupled nonuniformly along the coupling section 1. The
was reduced from 1 to 0.15 in. After thk change had been made,
voltage coupling coefficient k(z) changes continuously along the
the measurements were repeated and it was observed tha,t the overall
longitudinal direction Z. It is defined by even- and odd-mode im-
response of the coupler was improved. The direction:d coupler is
pedances ZO. and ZOo in (1) as
shown in Fig. 2 with top cover removed. Fig. 3shows the theoretical
zoe(z/l) 20.(2/1) and the experimental results of the coupled and transrlitted signal
,%(2/1) = (1)
208(2/1) +200(2/1)

WUPLING sEcrION
1 1 OUTPU7 &Jrl-

w l*L#aB fwr-

Fig. 1. High-pass transmission-line directional coupler. Fig. 2. Tapered asymmetric directional coupler.
1066 IEEE TRANSACTIONS ON MICROWAVE TFIEORY AND TECHNIQUES, DECEMBER 1!)75

---------- -- *------- -- ..--.:-. ---------- .* -----

8,8 I 1: t:
:: 1: *: 1, 1
s 1, 1, 1:
, :
+-.::--+.- ..--- ;--. -.---- F----W---=l-----

,---L-
1
I
-------? ~6i%&%n~c
---------fy;p= *Qfur :
w uw~r --l--- ------ ------ --- .!-- --- :

is i~ i~ it i
--A----:.-.- 4---- +-.--b----~-...b--.<-- -. &---- :
s , , .,
i i ii ii i i i i
is i
--- L-- ----- ---- . ------------- .---- d- ---. ----.

wig. 5. Tapered asymmetric magic tee.

20 .--- r--- ---- -.$ ----- ----,---- ~ ----,---- ----- 1---- .

1: * 1,, ,*
1, 1 :

: : :
----r---- L -y ----
I a I :
t 1 t
1 a 1:
----- .-- -!---
1 *
-_- :1::1%%-i
,---- i---
1-------------- I I

--_I ~ ~ ! J!-;
:

:----:
---- ::

I
I 1
J,----
t:
~
+--_+
!
LL_--+----L
::
I
1:~
1
:
s
FREQUENCY (GHz) t
11 VI
I
1 1,1:,
1 8

---.: ----;- .-- L-. --~--- .&-... L.-..,. -.-J.-.-j ----,

8
Fig. 3. Coupling and direct loss of S.34-dB coupler.
----t-----
1:11
#
la
b---a----
1:
I
.4----
~!!~:i
* ---.
1
-&-----
II
.-_d----j__ :
1: 1: i: i: i:
I If 1 I 1
amplitudes as a function of frequency. The theoretical analysis is 1: IJ
---.lr-_-,- _-+-- -:----
performed by evaluating the cascade ABCD matrix of the line
segments as a function of frequency The isolation and the return ISla,
loss of $he coupler (no~ shown here) were greater than 15 dB over
1:18
the entire band.
----
-.-
-- if-.
-- ..-- ~---+---- L .--+.---i ,----:
+. , , , I
8: t:i:
MAGIC TEE CONSTRUCTION ,. ,
.1: 1: I 1 I
1 Ieta1 1
+--A.----L_- &--- J__-- JL_--L---- L_---1---- i----,
1;1: ,1:1 s
The magic tee was to be constructed by cascading two 8.34-dB : :,
I
i: :1 t: 8. *I*
la

asymetric couplers in tandem and adding phase reference lines on 1 I* I l~la

each side of the device. However, the output port and the isolated
~,4
2.4 4.4 6.4 8.4 10.4
port of the 8.34-dB coupler had to be reversed in order to solve FREQUENCY (Giiz)
the topological layout problem. This was achieved by having the
Fig. 6. Coupling and transmission loss of magic tee.
coupled lines cross over at the tightly coupled region by using a gold
ribbon ~ridge 1 mil thick and 7 mil wide (see Fig. 4 which gives a
detailed view of the crossover region). In makln~ this crossover the -.-.C-.--
----=---- -----r---- -----, ---- s----t----
.
1:11, ,,,
bri,dging ribbon was kept high enough to prevent coupling between 1 I* I ::, , ,
1+11, 1 ,1,,
ad] scent circuit lines. Also, a small channel w& cut in the top ground -- ;--- --&-- ~---+-_- :._--lr---&--- ;-_-<
plane in order to provide adequate space between the ribbon and
the top ground plane. Thk was necessary because it was observed Ill! ! i ! ! ! ! !
that if the bridging ribbon is too close to the coupled lines or to the

KsFERENCS PLANE
TO WHICH TSE
LENGTH SHO\VN IN
* TABLE 4RE REFERENCED

--11- ..
ii
.oV2-

i rl
!
COWLED LINES
I
III*,
L
--.-,----
1, ;----:----
i---- 1,1
i:
!}I-,
vL.1
;----%-----
,,
I

9-----;
----
t :,
i: i,
1!
f
#
7 MIL WIDE X 1 NIL WIDE
GOLD RIBBON CROSSOVER

----
11( 1:141 1
1, 1,1 1:1111, ;v,.,
G---- L ..---,---- J --------- L ----L---J-- .. -J----,
,!, , ,,, , .
i :,
i
:,
i
:,,
; * ;
, ,
I

.Zk 0.4 2.4 4.4 6.4 8.4 10.4

FREQUENCY (GHz)
--&J-- Fig. 7. Isolation of magic tee.
Nig. 4. Detailed view of crossover at the tightly coupled region,
SHORT

---------
i ~.
.--,..-
1,I
1,s ....
:11.
10
PAPERS

1II
-_-p-_-:-
_= II
_--&_
&-...,
It
1#
88----------
_:_-_:.----,
:
11:
--&_-;_
-.;
106,7

1,1 11
1st : :11, II :, :.
50 ACKNOWLEDGMENT

40
I 1,11
--
#81
-}----:-----:---- j.--- ;----:. ---} ----:---- j----j
The
their
authors
efforts
wish
in fabricating
to thank F. A. Plzzarello
the circuits.
and D, E. Stiegler fc(r

.1[ ~ --
:,
+.--_LJ-
, 8,
t,
19
19

.1 ---;----}----;----;---J----;----;
II
11 :,

-:
#

:
[1] l?. C. De Ronde, A new class of microztrip
in IEEE
[2] E. Carpenter,
G-MTT Sfimp.
REFERENCES

Diq., May 1970, PP. 184189.

The virtue of mixing tandem
directional couplers,

and cascade coupler

* connections, in IEEE G-M TTSgmp. Dig, May 1971, I)P. 8-9.
I
I [3] F. Arndt, Tables for asymmetric Clhebyshev high-pm; TEM-mod~e
-, directional couplers, IEEE Trans. Microwave Theory Tech., VOL
, MTT-18, pp.633638, Sept. 1970.
,
I [4] C. P. Tresselt, Design and computed theoretical performance of
-: three classes of equal-ripple nonuniform line couplers, U3EE Trans.
, M$crowave Theory Tech., vol. MTT-17, pp. 218-230, Alw. 1969.
[.5] R. H. Duhamel and M. E. Armstrong, The tapered-line magic-T:
- .1
A wide-band monopulse antenna, in Abstracts of tht 15th Annu.
Symp. USAF Antennu Research und Detielopment Procram (Monti-
.; cello, Ill.), Oct. 1214, 1965.
[6] H. J. Hindin and A. Rosenzweig, 3-dB couplers constructed from
1, 1 1 i i ;1 two tandem connected 8.34-dB asymmetric couplers, .rEEE

I
1 I :U Trans.
L_L-_J-_J__ L_L-_._--L Microwave Theory Tech. (Corresp. ), vol. MTT-16, pp. 125-126, Feb.
1,1 1,1 ----+-i--v,
0,1 :1, Ill, 1968.
1,8,1 ,,,,.
[7] E. G. Crietal and S. Frankel, Design of hairpin-lim and hybrid
ol!!~! i # i I
hairpin-parallel-coupled-line filters, in IEEE G-MTT Symp. Dig.,
0.4 2.4 4.4 6.4 8.4 10.4 May 1971, pp. 1213.
FREQUSIJCY(GHZ) [8] J. I. Smith, The even- and odd-mode capacitnace pa ,rameters for
coupled lines in suspended substrate, IEEE Trans Mtcrowaw
Fig. 8. Return loss of magic tee.
Tlzeorv Tech., vol. MTT-19, pp. 424431, May 1971.

E Evaluation of the Equivalent Circuit Parameters of

: .
200 r -R/ ? Microstrip Discontiuuities Through Perturb/ltion of
g
.
z .
~ 180 -- . ----- --*-- .--2- ._: a Resonant Ring
~
U) 150
~ WOLFGANG J. R. HOEFER, MEMBER, IEEE, AND
i!
k ASOKNATH CHATTOPADHYAY, STUDENT MEMIIER, IEEE
g 100

# AbstractA resonaut technique for evaluating the equivalent

50 circuit of reciprocal microstrip discontinuities is des:ribed. Th{e
complex Z parameters of a discontinuity are related to the change in
0 resonant frequencies and Q factors of a microstrip ring it perturbsi.
04 14 2.4 3.4 ,., 34 6.4 7.4 84 94 10.4
FSEQUENCY(Ctk)
As an example, measurements made on inductive posts are presented
and compared with theoretical values.
Fig. 9. Phase performcmce of tapered asymmetric magic tee.

I. INTRODUCTION
top ground plane, undesirable ripples are induced in the coupling.
Also, each coupler section and reference line was made on individual
The main difficulty in measuring the circuit parameters of rnicro-
substrates and metal walls separated the substrates in order to
strip dkcontinuities resides in the elimination of systematic errors
reduce spurious modes.
introduced by coaxial--tomicrostrip transitions. This problem can
The magic tee circuit designed for a low-frequency cutoff of 1 GHz
be avoided by testing discontinuities in a resonant microstrip ring
is shown in Fig. 5 with the top ground plane removed. The actual
which may be coupled very loosely to the test equipmel It.
characteristics of this device areshown in Figs. 69. Fig. 6 comparea
Stephenson and Easter [1] and Douville and James [2] have
power at the two equal-power output ports. The average power
demonstrated the resonant technique as applied to the characteriza-
deviation is less than 0.5 dB over the 19-GHz frequency band.
tion of rectangular bends. Groll and Weidmann [3] ha~e evaluated
Fig. 7 shows theieolation and Fig. 8 shows the return loss. Fig. 9
impedance steps in a resonant ring.
depicts the measured and the computed phase dlfferencebetween
The present short paper gives a comprehensive genre-al analysis
the output ports of the magic tee.
of a microstrip ring containing a reciprocal discontinuity y. First it is
shown how the Z parameters of a dkcontinuity are re Iated to the
CONCLUSION change in resonant frequencies and Q factors of a micrc strip ring it
pert nrbs. Then the measurement technique is described, the accuracy
This work has demonstrated the feasibility of designing a decade-
of the method is dkcussed, and some measurements made on induc-
wide magic tee, ueing microstrip and a top ground plane. The cou-
tive metallic posts are presented.
pling factor k(.), which changes continuously along the length of
the coupler, wascalculated fortbis device using thecouplingcoeffi-
cient prepared by Arndt for equal-ripple high-pass directional cou- Manuscript received April 14, 1975; revised August 25, 1975. This
plers. Theline and gap widths were determined by using the com- work was supported in part by the National Research Coum :il of Canada
puter subroutine of Smith. The coupled aections and the reference under Grant A 7620, and in part by the Communicati( ms Research
Centre, Ottawa, Ont., Canada.
lines were separated by metal walls in order to reduce the standing- The authors are with the Department of Electrical Engineering,
wave modesin the box. University of Ottawa, Ottawa, Ont., Canada.