IBIIE TRANSACTIONS MICROWAVE tTiEORY

AND

TECHNIQUES,

VOL.

WIT-27,

No.

12, DEc~~BR

1979

987

Thru-Reflect-Line: An Improved Technique for Calibrating the Dual Six-Port Automatic Network Analyzer
GLENN F. ENGEN,
SENIOR MEMBER, lEEE, AND CLETUS A. HOER MEMBER, IEEE

AfMtrvzct-In teelndque descrfbed. different either

an earlier

paper,

the use of a thru-short-delay automatic features network only a length of precision

(TSD) was
b4 f

I
COMPLEX RAT 10 OETECTOR

for calfbratiag Another

the dual six-port eircuk The better

analyzer

scheme required

transrnkdon for @

Ifrw and a caffbratlon approaches a known

of these two somewhat This paper will

L----J
t b3

have now been combined circuit

and the requirement eliminated.

shortj or a calibration

develop the theory for this new procedure.

REFLECTOMETER MEASUREMENT PLANE ! ;

I.
~ 1 HE APPLICATION

INTRODUCTION

of

digital

technology is perhaps analyzer however,

to the best In In

b4 ~.1 I 1 1
L--.

COMPLEX RATIO OETECTOR I . 1

field

of

microwave

measurement network features,

illustrated addition shift

by the automatic to the time-saving

(ANA).

a major
Fig. 1.

IOEAL
.-.

REFLECTOMETER
-- ----

; -1

in measurement

strategy

has been

introduced.

particular, the requirement for an ideal item of test equipment (e.g., reflectometer) has been replaced by a more complete theory in which deviations from the ideal are explicitly recognized, evaluated, and, in theory, eliminated from the final measurement result. The determination of these deviations as the ANA upon from the ideal is generally ANA, in referred to
WNT=

It is convenient to model a reflectometer as an ideal one in cascade with a two-port error box.

calibration. which Fig. is based [1] to the reflectometer, as shown it is convenient

In the case of the conventional the four-port visualize the calibration

1
i FICTITIOUS

2 OETECTOR PLANES

1. Here

nonideal reflectometer has been modeled in cascade with a two-port error box.

by an ideal one

4(------

The properties of the ideal reflectometer can be chosen in such a way that its sidearm wave amplitudes b3, b4 are, respectively, equal to the (emergent and incident waves at a fictitious detector plane which is the input to the two-port error box. The parameters of the error box are provided done, of the by the ANA ratio calibration. in combination detector, permit After with an this has been the reading exact dethese parameters, complex

------5

(------

(W:J

DELAY

PRECISION LINE

termination of the signals at the measurement plane. In order to make two-port measurements, it is convenient to, introduce a second (nonideal) reflectometer and complex ratio detector which is then modeled in the same way as the first one. This is shown in Fig. 2. The calibration requirement now calls for obtaining the parameters of error boxes A and B. One procedure for doing this is via the thru-short-delay (TSD) technique [2], [3]. In this technique, the error boxes A, B become the

Fig. 2. For two-port measurements a second reflectometer is added and modeled as the first. Parameters of the two error boxes are obtained from measurements with the connections shown, where for TSD a known short is generally used, while for TRL any unknown highly reflecting termination may be used.

key components

in three additional

(fictitious)

two-ports. of A and B measurement

The first of these is the cascade combination and is formed (Fig. 2) merely by connecting

planes 1 and 2 together (thru). The second two-port is a degenerate one and results from terminating the measurement planes with shorts (short). Finally, the substitution of an (unknown) length of (nonreflecting) line between the measurement planes 1, 2 yields the third two-port which is thus comprised of A, the length of line, and B in cascade (delay). Given the scattering parameters of these three two-ports, it is possible to solve for the individual scattering parameters of error two-ports A, B.

Manuscript received April 30, 1979; revised August 23, 1979. The authors are with the Electromagnetic Technology Division, tiona ! Bureau of Standards, Boulcler, CO 80303.

Na-

U.S. Government

work

not protected

by U.S. Copyright

988
The six-port reflectometer

mm TRANSACTIONS MICROWAVE THEORY AND TECHNIQUES, VOL. MTr-27, NO.

12, DECEMBER 1979

[4] differs

from

the four-port

MEASUREMENT PLANES

version in that it is based upon a six- rather than a four-port junction, and where the requirement for a complex ratio detector has been eliminated in exchange for four power meters or other detectors which yield ampli(nonthis
Fig. 3.
The

tude information that the six-port ideal) four-port

only. It has been shown may be modeled and complex may be applied

([5], [6]), however, Following

l
FICTITIOUS OETECTOR PLANES

by an equivalent as outlined

ratio detector.

the TSD procedure

above [7].

This is illustrated in Fig. 3. An alternative procedure for applying TSD to the dual six-port is found in [8]. Another calibration procedure [9] was based upon the use of a length of line as the only impedance standard. This method also required, however, the use of a calibration circuit which provided for terminating the two sixports with equal impedances and at the same power levels. The better features of these methods have now been combined in a technique known as thru-reflect-line (TRL). Here the calibration circuit is no longer required. As compared with TSD, the need for a short of known reflection has been eliminated. In its place a termination of unknown reflection (but different from zero !) is used to terminate, in turn, each of the six-ports. While a nominal short continues to be one of the more convenient point In Instead choices for the unknown as a by-product line termination, is no longer the key required. for is that a value for its reflection this is obtained the word

six-port reflectometer may also be reduced to an ideal four-port in cascade with an error box.

Let P3/Pd, P~/Pb, and P6/Pe represent a point in a three-dimensional P-space. Equation (l), which is of second degree, is thus represented by a quadric surface in P-space. However, the nine coefficients in (1) (i.e., a, b; 2, cp2, etc.), Moreover, described mediate are functions of the five parameters it has been shown [6] that the quadric by (1) is an elliptic task is to determine paraboloid which a. . . p. surface im-

is tangent

to the planes

P3/Pd = O, P5/P4 = O, PJ Pd = O. The

a. . . p, which, in turn, permits the six-port to be reduced to a four-port. In theory, it is only necessary to observe P3. . . P6 for five arbitrary and unknown terminations. These may then be inserted in (1) to obtain a set of simultaneous equations in s.. . p. Unfortunately, however, these are of third degree, and unless a good initial estimate is in hand, the iterations be lengthy obtain required and by standard may fail estimate, numerical methods tend to root. To to yield the desired

of the procedure. substituted

addition,

has been

delay as more descriptive of the technique. In common with TSD, the line length is arbitrary and unknown (other than being different from A/2!). Moreover, it need not be free of dissipation. There are three distinct parts to the associated theory: 1) the six-port to four-port reduction, 2) the determination of the scattering parameters of the two-ports which result from the thru, reflect, and line solution. connections, These will and 3) the application of the TRL in the order given. IL
THE SIX-PORT

the initial

it is useful to start with nine (or

more) arbitrary terminations, which in practice are provided by the phase shifter in conjunction with the thru, reflect, and line connections, etc. The corresponding sets of values for P3. . . Ph are substituted in the equation
A(P3/P4)2+ B(P5/P4)2+ + E(P3P6/P;)
+

c(p6/p4)2+~(~3p5/@

+ F(P~P6/P;) H(P5/P4)

G(P3/P4)

be discussed

+ I(P6P4)

+ 1=

(2)

TO FOUR-PORT

REDUCTION

As described in an earlier paper [6], the four power measurements associated with the six-port represent an overdetermined set in that three of them determine the fourth to the extent of a choice between two possible values. It is convenient to express this result by means of the constraining relationship, IJ2 + bJ2(P5/PJ2 +(ba + cp2(FJ c)p(P3PJPJ) PJ2

and the resulting set of linear equations solved for A. . I. In principle, (1) and (2) represent the same surface so that by equating the coefficients one has A = a/abc, B= b{2/abc,. . . . I= C(C a b)p/abc. The system of equations, which results from equating the coefficients, is next solved for a. . . p as functions of A. ..1. By substitution it can be confirmed thatl GH)/(2AH - GZ)/(2AZ G/A DG) - EG) (3) (4) (5) (6) (7)

a(P3/ +(c

b=(2D c=(2E a= b+c+

a b){(P3P5/P~)

+ (a b c){p(P#6/P~) + b(b a c){(P~/PJ+

+ a(a b C)(PJPO) {=+vzz C(C a b)p(PJPJ+ abc=O (1) p=+vzz

where a, b, c, J, and p are five real constants, whose values are intrinsic properties of the six-port. These five constants also characterize four-port. the reduction from a six-port to a

In addition, in the ideal case, there are constraining relations among the ,4... I such as DE= A HI, CDH = BEI, DF= BGI, EF= CGH, etc., so that a variety of alternative expressions for a. . . p is possible. However, since (3)(7) are only used to obtain an initial starting point, this additional information, in this context at least, is of no obvious practical value.

SNGSN ~

HOER ~u-RSFm~-~

9s9

where, by definition measurement conditions order will only

[6], J and p are positive. the paraboloid be approximately satisfied

Because of and tangency by (2). In p are

1<

DETECTORPLANES

error, however,

to improve

the accuracy,

these values of a..

used as the starting the multidimensional

point for an iteratative solution using Newton method, to the nonlinear

&-z-a
I .--.

system of equations based on (1). This also provides a convenient method for using all of the observations, although the system is now overdetermined. After determine noted, error that U=(P3 JP5 + cP4)/(2P4ti ) (8) (9) box. these improved with the complex values a.. ratio p have been obtained, P3 bJb4 Pb may be used to which, as already of the [6] at the input these together the observed

--

---.

-------

---

Fig, 4. The fictitious two-ports, from which the parameters of error boxes A,B are ultimately obtained, are formed from a thrw reflect; and line connection.

is also the reflection

coefficient

of three measurements and line connections,

(under only

different

excitation measurement

condiis re-

Let bJ bd = w = u +jo.

It has been shown

tions, i.e., values of a2/al)

are required a single

in each of the thru

quired with the reflect comection, Moreover, in this case

since here Slz = S2, = O.

v = (P3 pp~ + (b 2uu2)P4)/(2P&) where zz2=(b+c-a)/(2fi)

Sii = Wi
IV.
THE TRL

i=l,2.
SOLUTION

(18)

(lo)

and
?)2 = i

Given S1~, S22, S12S21 for the three fictitious two-ports which result from the thru, reflect, and line connections, bu; .

(11)
This

the final scattering

task

is to determine,

insofar

as possible,

the

(The signs of the radicals completes III. the six-port

DETERMINATION

are to be taken as positive.) reduction.

PARAMETERS

parameters

of the individual

error boxes, -4, B.

to four-port

First, it is useful to solve (12) and (13) for bl and al as functions of az and bp This gives

OF SCATTERING

The next task is to determine of the three fictitious two-ports the thru, reflect,

the scattering parameters which are obtained from as indicated by the b], matrix. wave amplitudes

and line comections

dashed box in Fig. 4. The emergent

(::)= +(=:2 :)(:) ()

a2 =Rb.
2

(19)

bz, (b~, bj) at terminals 1 and 2 are related to the incident waves al, a2, (bq, b~) by the well-known scattering equations b,= Sllal + S1za2 (12) (13)

In (19) the matrix two or more two-ports

R is known property

as the waue cascading that the R matrix the product two-ports for of

It has the important

in cascttde is merely of the error

the individual R matrices. Let the cascading matrices be denoted by R., their cascade thru

xl, B

bz = Szlal + Szzaz

where the S*. are the scattering coefficients. Dividing the first of these by al, the second by az, and then eliminating the ratio aJa2 between them yields (14)

R~, respectively, while connection. Then R,= RaR~

R, represents

(20)

while if Rd represents

the line

combination (21) has been inserted.

W2S11+ W1S22 A= W1W2 where A= S1,S22 S12S21 and WI= bl/al Wz= bJa2 Given the appropriate power =bJb4 = V3/b& meter readings obtained

Rd = R.RIR~ where R, represents the line which

(15)

Solving

(20) for Rb gives Rb = R.- lR, (22) (22) is (23)

(16) (17) and a p

so that Rb may be obtained from R. and Rt. Next, used to eliminate R~ from (21). This yields TR~ = RaRl where T= RdRt and which may thus be found thru and line two-ports. from the parameters

for each six-port,

the Wi maybe

from (8) and (9).

Excitation of the junction under three different conditions provides a system of linear equations based on (14) which may be solved for S11, S22, and A [101.Following this, the product S12S21 may be obtained by use of (15). The technique does not permit the individual terms, Slz or S21 to be obtained, but this is not required. While a minimum

(24) of the

If y and 1 represent respectively the propagation constant and length of the delay line, then assuming the line

990 is nonreflecting,

DD3ETRANSACTIONSMICROWAVE YHSORY AND TECHNIQUES, VOL. MTr-27,

NO. 12 DECEMBER 1979

=(:9
Finally, the elements respectively. and tti, Expansion tllrll+tlzrzl=rlle

(25) by rti

To continue, the reflection coefficient WI which obtains at the fictitious detector plane for error two-port A is related to the reflection coefficient of the load Tl by

of Ra and T will be represented of (23) gives yl

(34)
where3 a
= r1Jr22

(26) (27) (28) (29)

t21rll + tzzrzl = r21e - yl

(35)

t,1r12+ t12r2z= r12e yl

t21r12+ t2zrz2 = r22e 71 . Next, gives taking the ratio

b= r12/r22 c=r21/r22.

(36)
(37)

of (26) to (27), and of (28) to (29)

Comparing these definitions for a, b, c with (30) and (31) indicates that b and a/ c have already been determined by the solution to the quadratic equations. To complete the determination of the parameters of error box A and thus effect the calibration of six-port 1, it is sufficient to determine a. Rearranging (34) one has wlb a= If J7, is known (38) by

t21(r1,/r2J2+ (t22lll)(rll/r21)
t21(r12/r22)2 + (t22 tl J(r1Jr22)

t12=0 t12= O.

(30) (31)

The ratios (rll/r21) and (r12/r2J are thus both given by a solution of the same quadratic equation, where the coefficients are parameters of the T-matrix. The problem of which root represents which ratio will be deferred to the following (rll/r2J section. For the moment have been it will be assumed that determined. Following and (r12/r2J the ratio eh! =

r~(l wlc/a)
from

and different

zero, a is determined

(38) since the remaining parameters are also known. This represents the TSD solution although the formulation is somewhat different of error from what has been previously solution obtained pubby unlished. To complete the TSD box B procedure, the corresponding

this, by taking

of (29) to (27) one has t21(r12/ r22) + t22 t12(r21/rl J + tll equations (26)(29)

(32)
can be

parameters use of (20). For

may be readily

Ordinarily,

a system of four

the TRL

171 is, by hypothesis,

solved to yield four unknowns. At this point, three have been obtained: (r, ,/rJ, (rJ rz), and e271.BY t~ing the determinant of (23), and noting that the determinant of a product is equal to the product easy to show that of the determinants, it is

known. Before proceeding, it may be useful to recap what has been accomplished thus far. Apart from the determination of the line parameters, which has no bearing on the immediate problem, one has obtained b and c/a. In addition a and J71are connected by (38) but both are as yet unknown. Returning to (20), this may be written

tlltzz t,~tzl = 1
thus there are only three independent matrix. It follows that there in are only (26)(29), experimental observations parameters three and

(33)
in the T since they independent

have been solved for three unknowns, there is nothing more that can be learned from this system of equations. It is perhaps desirable to stop and make certain observations on the practical application of the theory developed thus far. By means of (30), (31), and (32), it is possible to obtain (rl l/rJ, (rJrzJ, and e271from rati~~ among the elements of T. This matrix, in turn, is given by (24). At this point, it appears that since a complete determination for the complete set of scattering from A and B in cascade, while there maybe a problem of Ri, for example, calls parameters which results only S1l, S22, and snszl

where4 a, & y, P22 and d, e,

f, g

correspond,

respectively, (39) by R=- 1

to a, b, c, r22 in R~ and Rt. Premultiplying and expanding it is easy to show that f = de/a

lcc/a

(40) (41)

and dbf aa= lcc/a (42)

have been measured. Examination of (19) shows, however, that Rt can be determined from this partial information except for an unknown constant multiplier for each of its elements (in this case 1/ S21). The same observation holds for Rd. Comparing this with (24), the experimental procedure does yield the ratios between the elements fortunately this is all that is required.2 ZIf de~ked, the remtig of T, and

Since b and c/a are already known, and d, e, andj can be obtained from ratios among the elements of Rt, then y, ~/cY, and aa may be obtained from (40)-(42).

3The motivation

in the choice

of notation,

as in a preceding

section,

factor in T could be obtained, a Part from a

has been to provide continuity with prior work. However, this creates a double use of the symbols a, b, c. These symbols refer to (34)(37) for the rest of the paper. 4This definition of y is not to be confused with its prior use in the exponential associated with the delay line!

& sign, by the use of (33). However, there is no immediate use for this information.

EN13EhAND HOE?R: 1fIItU.RFJFL.JjCX.LINE

991

For written

error

box

the

counterpart w2+y

of

(38)

may

be

WI

a=~(l+w2&LX) where the reversed signs and interchange between

(43) the
Fig. 5.

ro Ies of ~ and y are due to the reversed direction associated with error box B. Eliminating 17, between (38) and (43) one has

Itis convenient to redefine certain parameters associated with

error box A as shown here.

~/a_

(W, -b)(l (W2 + y)(l

+ w,p/a) - w,c/a)

(44)

where, with allowance be obtained from

for the change in notation,

~2 may

(47), while
,b 2 1 2 =

by use of (48) and (49) (51)1

so that by combining

(42) and (44) -t w2fl/a)(dbf) - cc/a)

la,(a - CW,)12 /r**(a - bc)l [5] that

= a? F and

(w, - b)(l (w,+

y)I(l - wlc/a)(l

(45)

Substituting /all*= lb~12sP,, P=

(47) and (51) into yields Pdlacw112 lr2J21a - bc12 (-

(50), and noting

(d-bf) a= ;(lec/a) Apart from the choice

(46)

2,

52) at this

of sign in (45), the requirement

The only system parameter

which

is undetermined

that F1 be known matter, a nominal

has been eliminated. As a practical short continues to be a convenient

point is [r221. It may be found by observing the system response Pd and WI with P indicated by a standard power meter. After this has been done, the power delivered to other terminations may be found by use of (52). Moreover, by a simple extension, laz12 may also be obtained. As explained in [10] the evaluation of a nonreciprocal two-port calls for measurement of the ratio of the incident waves at the two-port, as well as the reflection coefficients. reversal), for error Since for the thru planes the ratio B taking connection the powers at for a sign expression the measurement is two-port are the same (except of (52) to a similar the ratio yields

chclice for the unknown reflector, although for some applications it appears that an open may provide better repeatability. In either case, a nominal value for the argument of the unknown reflection is available and with the help of (38) this permits be made in (45). At this point, sufficient coefficient, Returning refkction given by detail the system the proper has been sign choice to calibrated of reflection two-ports. the plane (38) that in

to permit

the measurement of reciprocal from follows

and the S-parameters to Fig. coefficient 2, it easily

171at the left measurement wlb cwl+a

Ir221/jp221. This

r,=

(47)

while a similar expression for 172 may be obtained from (43). This capability now permits the evaluation of twoports in the manner described in the preceding section for the fictitious ones by simply replacing w ~ and WJ by r ~ and rz in (14).
III

suffices for a determination of the ratio of the incident wave amplitudes, which in turn yields IS121 and ]S121 of the nonreciprocal two-port. If the individual phases are also required, the method may be further developed along the lines described in [10]. V. Comparison that the two Another of the roots
THE PROBLEM OF ROOT CHOICE

of (30) and (31) with (35)(37) indicates roots of either quadratic are a/c and b. which must be ccmsidered, a/c and is that one latter, is equal to both b. The

order

to measure and power,

the properties

of nonreciprocal theory is required. planes at the

possibility,

two.ports, Referring

some additional

to Fig. 5, for clarity,

the reference

1, 2 as input and output of error Ibex A will be relabeled shown. The wave amplitudes az, bz at measurement plane 2 are expressed in terms of al, bl at the detector plane by (48) where a- The net power = r,,(al-from bc)

however, requires a= bc. This, in turn, is only possible if for error box,4, S,2S21 = O which obviously cannot be true for a practical measurement system. Thus the two roots a/c and b are distinct. Next, it is useful to consider values of the two roots, lb\/]a/cl [bcl/lal <1, then Ibl < la/cl for root choice. TO demonstrate the ratio of the absolute = lbcl/lal. Provided that serve as a basis
the

and this will

(19),

that this is ordinarily

(1- c
lr212)

case, it is convenient

to expand

~ b)

(49) bl = r12 ~ a2+ b2 /) r12 (50) al = r22 a2+b2 ( r22

r21 \ )

output

error box ,4 is given by

(53)

P= Ib,f(1

(54)

992

IEEE TRANSACTIONS MICROWAVE THBORY AND TECHNTQU13S, VOL. MIT-27,

NO. 12, DECEMRER 1979

In (53), rlz may be considered a scale factor while rll/r12 is a measure of how tightly b ~ is coupled to az as compared to b2. A similar observation holds for (54). Now, port, under portional four-port remaining requires in most measurement to the wave amplitude reflectometer, it systems, including bz incident the sixproit is desirable to have one response nominally

Instead

of a collection

of offset

shorts,

or the single

short in conjunction with sliding terminations, which have characterized the prior art, the only items now required to calibrate the measurement system are a section of the transmission termination of the While mediately line or waveguide a nominal The in which the measurements are to be made (which for which applicable the method reflection is known. implicitly value defines ZO), plus a for the argument is thus line. applicable to immethod equally

upon the item that Ir, ~/rlJ the -

test. This calls for

Ir21/ rzzl <<1. In the case of the is further required however, proportional to az which

response be nominally

to the solid dielectric is, in principle,

Ir, ~/ r121 >>1. For the six-port,

1 is more typical. In any event, except for gross departures from the usual design objectives, one has (55) which requires lbl<la/cl. (56)

a dual four-port reflectometer, the latter has found but limited application to date because of the requirement, for duplicating the complex ratio detector, which is usually a costly item, or alternatively, the need for an elaborate switching scheme and its associated errors. The advantages of the TRL procedure are thus more readily realized in the six-port environment; this represents another advantage of the six-port technique. For a report on the performance of a measuring system in which this calibration technique plays a key role, the reader is referred to a companion paper [12]. While the effect of residual imperfections, such as line reflections, remains to be evaluated, the cited results [12] leave little room for doubt that the method is capable of high accuracy ! This calibration of adapter technique evaluation. may also be applied of adapters. to the If the problem let error Returning a pair again to Fig. 4,

An alternative, and in principle more definitive, test may be obtained from (32). Solving (30) and (31) for and substituting in r,, / r21, and rlJ rzz, in terms of the ty, (32) gives, (57) where R=~~ and where assignment for e27~,as other. Thus (58)

boxes A, B represent

the choice of sign depends upon how the root has been made. The two corresponding values given by (57), however, are reciprocals of each as an alternative criterion, the root assignment

should be made such that Ie2y1<1. As noted, this is in theory a more definitive test, however, since the difference between Ie2y[l and unity is usually small, it is possible for measurement error to mask this effect unless good quality Finally, the argument power detectors are used. from (57), of ezy~, as obtained

entire assembly of Fig. 4 is inserted between the measurement planes of Fig. 3, then al, bl, a2, b z (Fig. 4) may be measured by the six-ports where it is assumed the calibration has already been completed. The procedure thus calls for a repeat of the TRL technique with the adapters in place and yields the adapter parameters in place of those of a fictitious Finally, ered namely which, six-port an error box. noted that while TRL of TSD, another maybe of considTSD, however, problem in the variant it maybe extension super for TSD,

depends upon the root assignment. If the line length differs in a known direction from the design center of a quarter wavelength, root choice. this could also serve as a basis for is more than

also exists [13]. The latter, account purposes, of a crosstalk is nonexistent

was developed

to take

all practical

environment. VII.
REFERENCES analyzer system: Micm-

Ordinarily, however, the first criterion adequate in a practical application. VI. The foregoing provides
SUMMARY

[1]

R. A. Hackborn, An automatic network waoe J., vol. 11., pp. 4552, May 1968.

[2]

the

theoretical

basis

for

a
[3]

calibration procedure which impact upon the microwave

promises to have a major art. In particular, the long-

standing role of the short or open as an impedance standard, if not obsolete, at least needs to be reexamined. This is in agreement with the observation by Allred [11] that open and short circuits are not standards in the true sense. As brought out in a companion paper [12], the TRL technique, quite possibly, represents the best experimental provide method for evaluating Moreover, a short or an open which has yet been devised. the values thus obtained

[4]

[5]

[6]

a useful monitor

of the system performance.

N. R. Franzen and R. A. Speciale, A new procedure for system calibration and error removal in automated S-parameter measurements, in Proc. 5 [h European Microwave Conf. (Hamburg, Germany, Sept. 14, 1975). Sevenoaks, Kent; Engtand: Microwave Exhibitions and Publishers, pp. 6973. N. R. Franzen and R. A. Speciale, Accurate scattering parameter measurements on nonconnectable microwave networks, in Proc. 6fh European Microwaoe Conf. (Rome, Itaty, Sept. 14-17, 1976.) Sevenoaks, Kent, England: Microwave Exhibitions and Publishers, pp. 210-214. G. F. Engen, The six-port reflectometer: An alternative network analyzerj IEEE Trans. Microwave Theoty Tech., vol. MfT-25, pp. 1075 1080, Dec. 1977. , Calibration of an arbitrary six-port junction for measurement of active and passive circuit parameters, IEEE Trans. Irr.rtrum. Mess., vol. IM-22, pp. 295299, Dec. 1973. Calibrating the six-port reflectometer by means of stiding terminations, IEEE Trans. Microwaue Theory Tech., vol. MTT-

D3EI! TRANSACTIONS MICROWAVE TNEORY AND TECHNIQUES, VOL. MsT-27, NO. 12, DECEMBER 1979

993

[7]

[8]

[9]

26, pp. 951-957, Dec. 1978. G. F. Engen, C. A. Hoer, and R. A. Speciate, The application of Tbru-Short-Delay to the calibration of the dust six-port, in 1978 IEEE MTT-S Int. Symp. Dig., pp. 184-185. L. Susman, A new technique for calibrating dual six-port networks with application to S-parameter measurement: in 1978 IEEE MTT-S Int. Symp. Dig., pp. 179 181. C. A. Hoer, Calibrating two six-port reflectometers with an unknown length of precision transmission lineu in 1978 IEEE MTTS Int.

[11]

[12] [13]

[10]

Symp. Dig., pp. 176--178. A network analyzer incorporating

two six-port

reflectome-

ters, IEEE Trans. Microwave Theory Tech., vol. M7T-25, pp. 1070-1075, Dec. 1977. C. M. Alfred and C. H. Manney, IIN calibration and use of directional couplers without stan&rd$: IEEE Trans. In.rtt-um. Meas., vol. IM-25, pp. 8489, Mar. 1976. C. A. Hoer, Performance of a dust six-port automatic network analyzer; this issue, pp. 993998. R. A. Speciale, A generalization of the TSD network-analyzer calibration procedure, covering n-port scattering parameter measurements, affected by leakage errors, IEEE Trans. Micrawaoe Theoy Tech., vol. MIT-25, pp. 110( 1115, Dec. 1977.

Performance

of a Dual Six-Port Network Analyzer

CLETUS A. HOER,
MEMBER, IEEE

Automatic

Ahwet-lrritfal

results network

of the performance ssmdymr operating are given.

of an experimental in tbe 2-lS-GHZ

dual

sfx-port reflection two-prt vafues tematic 0,0004. coostmts

automatfc

range with in measurfng of

thermistor-type

power detectors of one-port

The imprecision repeatability. from The

coefficients

devices, or the scattering eosmector varies

parameters

/= ,,!,,s,, cm,,

FLEXIBLE CABL[ .% \

devices is 4 X 10-s, in measuring to of attenuation attenuation

excluding dll

At 3 Gl@ dB at low error Ilw in sys-

the imprecision
measuring

attmmation

0.0003 systematic

0.15

at 60 dB. coefficient

appears

to be Iesa than the imprecision. appears errors

error

in measuring systematic

reflection

to be less than

Additional

caused by changes in the calibration to be less than 0.003 dB in coefficient.

over a 20-week period

were observed

attenuation

and less than 0.002 in reflection

I.

INTRODUCTION

A
GHz.

Fig. 1. Block diagram of a dual six-port automatic network analyzer, where P indicates power detector. When a termination is connected to either measurement plane, p, or p2 becmmes the usual reflection coefficient T1 or r~ of the termination.

N EXPERIMENTAlL ~ (ANA) incorporating at NBS of one-port

automatic two

network

analyzer The accuracy function standard of a six-port measurement is primarily quality a

six-port

reflectometers the network from 2-18 of the ANA so far are

has been constructed parameters The precision,

for measuring devices obtained and stability

and two-port Results

of the quality transmission stability,

of the connectors, and linearity

of the and of sidearm

accuri~cy,

line used in the calibration, of the four

are now being summarized

investigatecl.

the resolution,

in this paper. II.

SYSTEM DESCRIPTION

A block diagram of the dual six-port ANA is shown in Fig. 1. Measurements of the reflection coefficient r of one-port devices are made by connecting reference device are measured the termination parameters the twoThe theory to eil,her six-port of a two-port port between plane. The network by inserting reflectometers.

power detectors. Greatest accuracy has been obtained with NBS Type IV power meters [2] using thermistor type power detectors. The thermistor detectors are housed in an aluminum block whose temperature is held constant to 0.01 C. The present system has a phase-locked source whose output power is externally leveled. Connectors at the measurement trolled planes are GPC-7. calculator.
TECmQIQUES

The system is con-

by a programmable III.

the two six-port

of operation and a description already been published [1].

of the basic system have

CALIBRATION

Manuscript reeeived May 8, 1979; revised September 5, 1979. This work was supported in part by the Calibration Coordination Group, Department of Defense. The author is with the Electromagnetic Technology Divisio~ National Bureau of Standards, Boulder, CO 80303.

The technique used to calibrate the dual six-port ANA is the thru-reflect-line (TRL) technique [4] augmented by including a nominal 10-dB pad in the set of measurements. The steps in the calibration are shown in Fig. 2 and outlined below. by U.S. Copyright

U.S. Government

work

not protected