You are on page 1of 4

An improved Multiline TRL method

J. E. ZUfiiga-Juarez, J. A. Reynoso-Hernandez, and M. C. Maya-Sanchez


Centro de Investigacion Cientifica y de Educacion Superior de Ensenada (CICESE)
Division de Fisica Aplicada, Depto. Electronica y Telecomunicaciones
Km. 107 Carretera Tijuana-Ensenada, 22860 Ensenada, B. C. Mexico;
E-mail: apolinar@cicese.mx.ezuniga@cices.mx.mcmaya@cicese.mx

b) the best propagation constant of the N-lines used in


the calibration is determined from N-1 pair of line
measurements by using the similar matrix properties
(equals eigenvalues, equal determinant and equal
traces) along with the Gauss-Markov method,
c) the best value of the error-box coefficients (ale, b,
etc) of the 8-term error model are calculated from
the previous knowledge of the best value of the
propagation constant of the line, used in the
calibration, by using the Gauss-Markov method.
In this paper, using the 8-terms error model along with
straightforward de-embedding method SFDM[4] and a reliable

Abstraet- This paper presents a multiline TRL (Thru-ReflectLine) method that is based on a straightforward de-embedding
method and a reliable method for determining a each
frequency the value of ale, b, and e -r 1 terms. The main
features of the proposed multiline TRL are: a) The
transmission line propagation constant is not used, therefore
the physical lengths of the lines are not needed, and b) The
Gauss-Markov method is used only to estimate the best values
of the constants ale and b. The S parameters of an offset
short, an offset open, and a load, measured in the frequency
range of 1-50GHz, corrected with the multiline TRL method
from NIST and the multiline TRL method from CICESE agree
each other.
Index
Terms-Multiline
TRL,De-embedding,
propagation constant, Straight de-em bedding method.

method[5,6,7] for determining value of the ale, b, and e- r1


terms, an improved multiline TRL method is presented. The
originality of the proposed multiline TRL method lies in the
methodology proposed to calculate the best value of the ale, b,

wave

I. INTRODUCTION

a, am (

Multiline TRL (Thru-Reflect-Line) from NIST [1,2,3] is


considered one of the best methods for errors correction of
network analyzers and error correction of test fixtures. One of
the key parameters used in the Multiline TRL from NIST [2,3]
is the determination of the wave propagation constant 'Y of the
lines. Its importance stems from the fact that the knowledge of
'Y allows determination the reference plane location and also
because the imaginary part of'Y is directly related to the phase
shift of the lines involved in the calibration. To summarize,
the main features of the multiline TRL method from NIST are:
a) the method uses N-non reflective lines of arbitrary
length and one Reflect,
This work was supported by CICESE and CONACYT, Project
No.39835-Y, Mexico. J.E. Zuniga-Juarez, J. A. ReynosoHernandez, and M. C. Maya-Sanchez are with the Centro de
Investigaci6n Cientifica y de Educaci6n Superior de Ensenada
(CICESE), Divisi6n de Fisica Aplicada, Km. 107 Carretera
Tijuana-Ensenada, 22860 Ensenada, B. C. Mexico; email:
apolinar@cicese.mx, ezuniga@cices.mx,mcmaya@cicese.mx

67th ARFTG Conference

a m = ae

-r 1

) and e -yl terms. While the Multiline TRL

method from NIST is based on the determination of the best


value of the wave propagation constant 'Y of the lines, the
Multiline TRL method from CICESE, proposed in this work,
does not need the previous knowledge of 'Y. Therefore the
physical lengths of the lines are not needed. The proposed
Multiline TRL method is based on: i)the straightforward deembedding method, ii) the reliable method for determining the
values of the ale, b, and e- y1 terms proposed by[4-7], and iii)
the Gauss Markov method for determining the best value of
ale, b, at each frequency. To demonstrate the usefulness of the
proposed multiline TRL method, the scattering parameters Sij
of an offset short, an offset open, and a load, measured in the
frequency range of 1-50GHz, are corrected with the multiline
TRL method from NIST [3] and compared with the scattering
parameters Sij corrected with the proposed multiline TRL from
CICESE.

[139]

II. THE DE-EMBEDDING METHOD FOR NON COAXIAL DEVICES

resulting WCM when Ii and lj are measured with the


uncalibrated system are given by

For de-embedding non coaxial devices, the expressions for


computing the scattering parameters using the straightforward
de-embedding method SFDM are given by [4]

M L =TA TL
j

TB

(3)

M L =TA TL TB

(4)

where

~).

TA =r22 ( :
a
--b
c

S = Lim
12
a
a
-m22 -b(m11-- m21)-m12
c
c

e-r (li-21)

(6)

(1 )
and

_____
-;_-_b

(5)

e-y( li-21)

(-Yin
e

(7)

21

22

The constants b and ale are derived from (3), (4) as follows.
First, we have that combining (3) and (4) one has
ij
ij
(8)
MLTA = TATL '
where

a a
a
-m22 -b(ml1-- m21)-m12
c c
c

ij
;
) -1
ML=ML[M ]
L

and
where

am = ae- yl;

(2)

The equation I will be named the two port de-embedding


equation because it allow the calculation of the DDT Sparameters embedded in any two port network. Furthermore,
depending on whether or not the OUT is embedded in a test
fixture, equation 1 can be simplified. It is very important to
comment that the right hand side of (1) depends on the
measured mij parameters, the a, ale and b elements
(unknowns) of the transition A, the wave propagation constant
y of the lines, the mechanical dimensions of line where the
device is embedded I, and the mechanical dimension of the
reference line Ii. The right hand side of (I) can be simplified
for de-embedding noncoaxial devices and can also be
simplified utilizing a reference li=2/. An example of the use of
the SFDM when li=21 along with the improved multiline TRL
for de-embedding noncoaxial devices mounted in fixture
devices is presented in this work.

i}
TL

(2;)0

12

122

l.-

(9)

'

;Aij

=e

y(lj-li)

(10)

Using (5) and substituting (9-10) in (8) yields the following


equations

+ et]2 = aAi}'
at2] + e122 = CAi}'

(II)

at]]

(12)

bl]] +112 = - ,
Ai}

(13)

b12 ]+122 = - ,
Ai}

(14)

Finally, the ale term is expressed as

a=

112

Aij

Aij - 122

-til

(15)

t 2]

and the b by

--1

b=

The method for computing b and ale terms using the L-L
method [5] use two nonreflecting lines. The shorter and
longer lines will be referred to as Ii and lj respectively. The

67th ARFTG Conference

11

121

Ai}

III. PROCEDURE FOR CALCULATING THE BEST Alc AND B TERMS


WITHOUT THE PREVIOUS KNOWLEDGE OF THE PROPAGATION
CONSTANT

j -1

= TL [TL ] =

(1 1)

t/ 2

--1

A..
y

[140]

1]

A..

22

IJ

12 ]

(16)

It is obvious that for the calculation of alc and busing (15)


and (16) the term er(U-li) have to be known. Is in this
parameter where the method developed by [1] lies in. Using
the similar matrix properties of equation (8), it was explained

in [1, 2], how to determine erUj - ). The calculation of


.
1mow Ie dge 0 f the physIcal
.
e r(U-Ii) needs th e precIse
li

dimensions of the lines Ii,

Ii

and the complex propagation

constant. Furthermore, because the calculation of erUj - )


starts with the calculation of the eigenvalues of (8) a
complicated algorithm to resolve the sign ambiguity problem
and to achieve a continuous traveling wave vector most be
used [2]. An alternate method for determining ale, b and
li

er(U-li) without the previous knowledge of Ii, lj was


proposed in [6,7]. Is this method that we use as initial values
,.1
b and er(U-li) ~lor nnp

IementIng
. the Gauss-Markov
o f wC,
method leading to the best estimation of these parameters. The
Gauss-Markov method was implemented according to [11]
using equation (15) and (16) as follow.

a)
(

_
= ;t..U - t i22j
ij
cIJ
t 21

::1 "~~-~~~~1
~ t):?I'
[~.'-'.

(17)

ijf

.0'1

j::

..0,

ij

t
I

W~~~~E~~ ::~~l
f~"t~~t;:tl

-., - t "

'''~<rl~1

a)

..

r----_-'~-c::-~
~ ?'f
J ;.l.:'

5:.:,

L~;i~.J I

~~"t')v~~,"""",:v,"'o)o<f'-,"""-,,",,."t'.';''-':---~'.~.''''~~~''~~'''''''''''r'''''''''"'''''N1!t~'>oIy"""~

(18)

ij ij

t 21

N'.

1::.

1
:. .

I ,:

~--!-:-~~~s-'f.:'~~~3:~1S~~~-!6---"'.;;. . -~. _
. . -J4

I
~

'~~!>~"in~!'S -.~1O-2!!i~
. ~:j1""".~36~ -60"'-:1-...~-...J~.
~~~'1

$1:1

Figure 1. (a) Magnitude and (b) Phase of S II and S22 of an


offset short corrected with multiline TRL method from NIST
and multiline TRL method from CICESE.

(19)

where C is the number of combinations ofN-lines.


Once the best alc and b are determined the am term are used
according to the procedure reported in[ 10]

V. CONCLUSIONS

IV. EXPERIMENTAL RESULTS


The multiline TRL method was implemented as follows.
Because the 8-term error does not remove the switching errors
due to the non ideal source and load match, the switching
errors are removed first using the method proposed in [8]. The
scattering parameters were measured by using a HP851 OC in
the configuration that allows the removal of the switching
errors due to non ideal source and load match. The multiline
TRL calibration technique was implemented using the
calibration standards CM05 provided by JmicroTechology
[9]. Five lines, Thru included, were used in order to cover the
1-50GHz frequency range. According to the multilines TRL
method proposed in this work, the number of combinations of

67th ARFTG Conference

l~;i!jf

rot
1

~~.M;, f~'i

(N -2)1 21'

~""'.-""'Y._"""-':>'",,,!.... '~.o:o."'''-~'''liCo;l''a.~-''lo''T''~''t'''~~~'*'''~'''''''''

j ""

""f

An important feature of the proposed method is how the


combinations of N-lines measurements are used. While the
multiline TRL method from NIST uses N-l combination, the
proposed method uses

CN -

::,

(b )ij =

five lines from an array of two will generate ten lines in total,
which is not easy to implement in commercial network
analyzers. Then, using the free switching errors measurement
data of the S parameters of the calibration standards used for
multiline TRL, the best value of the constants alc, b, and am
are estimated using the Gauss-Markov method. Next, by using
the expressions of the straightforward de-embedding method
[4] the corrected S parameters of an offset short, an offset
open, and a load are computed. The S parameters of an offset
short, an offset open and a load were corrected using the
multiline TRL software from NIST [3] and compared with the
S-parameters corrected with the multiline TRL method
proposed in this work. Fig. 1 and Fig.2 show, respectively, the
magnitude and phase of an offset short and offset open. The
real and imaginary parts of the load are presented in Fig.3. It is
notable that both multiline TRL methods provide similar
results. Therefore the usefulness of the multiline TRL method
from CICESE is demonstrated.

The S parameters of an offset short, an offset open, and a load


are corrected with the multiline TRL from CICESE and with
the multiline TRL software from NIST[3]. Both methods
provide similar results. Due to the difficulty and degree of
complexity of the method proposed in [1-3] to compute the
propagation constant of the line, the multiline TRL method
from CICESE is easier to implement than the multiline TRL
method from NIST. Since the proposed multiline TRL method
does not use the physical lengths of the lines, it has the
following advantages:
a) it is compatible with any commercial available
calibration substrate,
b) it is less prone to errors due to the position of the
probes in the lines,

[141]

c)

it is very useful when the physical lengths of the lines


are unknown to the users,
probably the software developed in CICESE to implement the
proposed multiline TRL method is faster than multiCal from
NIST[3].

work was supported by CICESE and CONACYT, Project No.


39835-Y, Mexico.
REFERENCES

[1]

1-1Ql I
--~

~-~

,(ll

>02

ie),)

I~'
J~.$

.06
.Q.~

..0'6

-",
~

..

__.-""-_,.

y>.",~_""

_._~,,

008 --~ho~~-~~-~~-~-~-'~"-'Io

_~,,.

"~r,,;i-t:1

'~1l6k;.~

a) 3.1),.........-

---,

lW

,;m

16

tS

:s

Xl

4)

4&-

,,~,.fl.-l

,3D

til)

:Ie

2S

:I>

JO

.:

If)

~.,~

b)

Figure 2. (a) Magnitude and (b) Phase ofSll and S22 of an


offset open corrected with multiline TRL method from NIST
and multiline TRL method from CICESE.
("~.

~f.~

~r___._-----_-..----.-----.---,

1==..:1

f)1

h .....-oIllGH"

10

15

16
~
''''''''1tGm~

ta

R. B. Marks, "Multiline Method of Network Analyzer


Calibration," IEEE Trans. Microwave Theory Tech.,
Vol.MIT'-39, No.7, pp. 1205-1215, July 1991.
[2] Donald C. DeGroot, Jeffrey A. Jargon, and Roger. B.
Marks, "Multiline TRL Revealed," 60th ARTFG
Conference Digest,Fall 2002,pp131-155.
[3]
NIST De-embedding software
program multiCal
version 1.04a
[4]
J.A. Reynoso-Hernandez, Everardo Inzuza Gonzalez,
"A Straightforward De-embedding Method for Devices
Embedded in Test Fixtures" 57th ARTFG Conference,
Phoenix, Arizona, USA, may 2001.
[5]
G.F.Engen and C.A.Hoer,"Thru-Reflect-Line: An
improved Technique for Calibrating the Dual Six Port
Automatic Network Analyzer," IEEE Trans.
Microwave Theory Tech., Vol.MIT' 27, No.12, pp.987993, December 1979.
[6]
J.A.Reynoso-Hernandez," Reliable Method for
Computing the phase Shift of Multiline LRL
Calibration Technique", IEEE Microwave and Wireless
Components Letters, vol. 12,pp 395-397,October 2002.
[7]
J.A Reynoso- Hernandez, "Unified method for
determining the complex propagation constant of
reflecting and non -reflecting transmission lines," IEEE
Microwave and Guided Wave Lett, vol 13, No.
8pp.351-353, August 2003.
[8]
ARFTG Short Course on Measurements
and
Metrology for a Wireless World, November 27-28,
2001, San Diego CA.
[9]
A. Fraser, and J.Schapper, "Test adapter substrates ease
the task of measuring PHEMT FETs," Microwave
Journal, vol. 38, No. 3, pp. 120-122, March 1995,.
[10] J.A.
Reynoso-Hernandez,
J.R.Loo-Yau,
Hugo
Ascencio-Ramirez, Juan Alberto Saldivar, J.E.ZunigaJuarez, Maria del Carmen Maya-Sanchez, "LineAttenuator-Line: An alternative method for In-fixture
Calibration" 65 th ARTFG Conference, Long Beach,
California, USA, June 2005.
[11] Henry Stark, John W. Woods, "Probability, Random
Processes, and Estimation Theory for Engineers" ed.
Prentice hall, 1986

b)
Figure 3. (a) Real and (b) Imaginary parts of ZlJ and Z22
parameters of the load corrected with multiline TRL method
from NIST and multiline TRL method from CICESE.
ACKNOWLEDGMENT

The authors wish to thank Jose de Jesus Ibarra Villasenor


and Benjamin Ramirez Duran for device measurements. This

67th ARFTG Conference

[142]

You might also like