2014 IEEE Biennial Congress of Argentina (ARGENCON)

Incomplete 2-Port Vector Network Analyzer

Calibration Methods
A. Henze, N. Tempone, G. Monasterios, H. Silva
RF Metrology Laboratory – Instituto Nacional de Tecnología Industrial (INTI)
Buenos Aires, Argentina
[email protected]

Abstract— Different types of incomplete 2-port vector

network analyzer (VNA) calibration methods are
explained. All of them are particular cases of the 12-
term error model and a comparison, including
advantages and disadvantages, between them and a Full
2-Port method, such as TOSM, is made.

Resumen— En el presente informe se explican los

distintos tipos de calibración incompleta de un VNA de 2
puertos. Todos ellos son casos particulares del modelo de
12 términos de error, y se realiza una comparación,
incluyendo ventajas y desventajas, entre ellos respecto a Fig. 1 Forward 12-term error model flow chart
un método Full 2-Port como el método TOSM.

I. INTRODUCTION In Fig. 1, the forward error terms are

When calibrating a 2-Port VNA, Full 2-Port calibration
is usually employed [1]. There are different types of
complete calibration methods depending on the error model e00 : Directivity (F)
to be considered. When measuring in coaxial systems, the e11 : Port-1 Source Match (F)
most common calibration method used is TOSM (also e10e01 : Reflection Tracking (F)
known as SOLT) which uses the 12-term error model [2]. e10e32 : Transmission Tracking (F)
However, this method requires an 8-step procedure to get e30 : Leakage (Crosstalk)(F)
both ports calibrated. e22 : Port-2 Load Match (F)
When it is not necessary to measure all four scattering
parameters (i.e. S11, S12, S21 and S22) of the Device Under
Test (DUT) or uncertainties are not necessary to be as small
as possible, most VNAs can be calibrated by using simpler
alternative methods. In fact, they are incomplete calibration
techniques and can be studied as cases derived from the
complete method. However, the implementation and details
of these incomplete calibration methods are seldom, if never
explained, in bibliography or instrument’s manuals, so the
user is usually not fully aware of the systematic errors that
they cause. This lack of information does not allow an
adequate consideration in order to determine which method Fig. 2 Reverse 12-term error model flow chart
will be the best option for each particular measurement
scenario.
This paper thoroughly analyses the different incomplete In Fig. 2, the reverse error terms are
calibration methods, introducing all correction error term
equations used for each case and calculating maximum
e′33 : Directivity (R)
deviation values with respect to the complete calibration
method. e′11 : Port-1 Load Match (R)
e′23e′32 : Reflection Tracking (R)
II. TOSM CALIBRATION e′23e′01 : Transmission Tracking (R)
Before measuring any DUT S-parameters, both VNA’s e′03 : Leakage (Crosstalk) (R)
ports must first be calibrated in order to calculate system e′22 : Port-2 Source Match (R)
errors. The most commonly employed method is TOSM. It
consists in calculating 6 forward (F) and 6 reverse (R) error
terms as shown in figures 1 and 2.

978−1−4799−4269−5/14/$31.00 2014
c IEEE 810
 Solving measured S-parameters from Fig.s 1 and 2 [2] TABLE I
PORT 1 CALIBRATION SUMMARY

b0 e10 e01 ⋅ (S11 − e22 Δ S ) (1) Error to be

S11M = = e00 + Reference
corrected
Description
a0 1 − e11 S11 − e22 S 22 + e11e22 Δ S
Open1 e11 Source Match (F)
b e10 e32 ⋅ S 21 (2) Short1 e10e01 Reflection Tracking (F)
S 21M = 3 = e30 + Match1 e00 Directivity (F)
a0 1 − e11 S11 − e22 S 22 + e11e22 Δ S
b3′ e'23 e'32 ⋅(S 22 − e'11 Δ S ) (3)
S 22M = = e'33 + 2) Port 2 Calibration: Making steps 4 to 6, an OSM
a3′ 1 − e'11 S11 − e'22 S 22 + e'11 e'22 Δ S
calibration to Port 2 is performed and the following reverse
b0′ e'23 e'01 ⋅S12 error terms are calculated from (3)
S12M = = e'03 + (4)
a3′ 1 − e'11 S11 − e'22 S 22 + e'11 e' 22 Δ S
e´33= S 22 M (match2 ) (13)
where
S 22M (open2 ) + S 22M ( short 2 ) − 2 ⋅ e´33 (14)
SxxM: Measured, i.e. uncorrected, S-parameters e´11=
S 22M (open2 ) − S 22M ( short 2 )
Sxx: Actual DUT S-parameters
− 2 ⋅ [S 22M (open2 ) − e´33 ]⋅ [S 22M ( short 2 ) − e´33 ] (15)
Δ S = S11S 22 − S12 S 21 (5) e´23 e´32 =
S 22M (open2 ) − S 22M ( short 2 )

A. VNA Calibration
TABLE II
The calibration procedure consists in measuring 7 PORT 2 CALIBRATION SUMMARY
different reference standards (2 Opens, 2 Shorts, 2 Matches
and a Thru) with known reflection and/or transmission Reference
Error to be
Description
values from a TOSM calibration kit. In this paper reference corrected
standards are considered to have ideal values as follows Open2 e′11 Source Match (R)
Short2 e′23e′32 Reflection Tracking (R)
ΓOPEN = 1 (6)
Match2 e′33 Directivity (R)
ΓSHORT = −1 (7)

ΓMATCH = 0 (8) 3) Isolation Ports Calibration: Step 7 is optionally made

only when very low transmission parameters must be
⎡0 1 ⎤ measured. In most cases this error term is neglected.
STHRU = ⎢ ⎥ (9)
⎣1 0 ⎦
To perform a complete 2-Port calibration, an 8-step e30 = S 21M (match1, 2 ) (16)
procedure must be done as follows
′ = S12 M (match1,2 )
e03 (17)
Step 1: Connect Open1 to Port 1
Step 2: Connect Short1 to Port 1
Step 3: Connect Match1 to Port 1 TABLE III
Step 4: Connect Open2 to Port 2 ISOLATION PORTS CALIBRATION SUMMARY
Step 5: Connect Short2 to Port 2 Error to be
Step 6: Connect Match2 to Port 2 Reference Description
corrected
Step 7: Connect Match1 Port 1 / Match2 Port 2 Match1 e30 Crosstalk (F)
Step 8: Connect Thru between Port 1 and Port 2 Match2 e’03 Crosstalk (R)

1) Port 1 Calibration: Making steps 1 to 3, an OSM 4) Calibration between Ports: When making step 8 both
calibration [2] to Port 1 is performed and the following Load Match and Transmission Tracking error terms are
forward error terms are calculated from (1) calculated from (1), (2), (3) and (4) as follows

e00 = S11M (match1 ) (10) S11M (Thru ) − e00

e22 = (18)
S (open1 ) + S11M ( short1 ) − 2 ⋅ e00 S11M (Thru ) ⋅ e11 − Δe
e11 = 11M (11)
S11M (open1 ) − S11M ( short1 ) e10 e32 = [S 21M (Thru ) − e30 ]⋅ (1 − e11e22 ) (19)
− 2 ⋅ [S11M (open1 ) − e00 ]⋅ [S11M ( short1 ) − e00 ] (12) ′
S 22M (Thru ) − e33
e10 e01 = ′ =
e11 (20)
S11M (open1 ) − S11M ( short1 ) S 22M (Thru ) − Δe′
′ = [S12 M (Thru ) − e03
′ e32
e23 ′ ]⋅ (1 − e33
′ e11
′ ) (21)

811
where III. INCOMPLETE 2-PORT VNA CALIBRATION
Δe = e00 .e11 − e01e10 (22) In the past, VNAs had only a transmission/reflection
(T/R) test set. This allowed only forward parameters to be
Δe´= e´33 .e´22 −e´23 e´32 (23) measured, since Port 1 acted as a source and Port 2 as a load.
Then, calibration methods used were:

TABLE IV • Transmission Response (TR)

CALIBRATION BETWEEN PORTS SUMMARY • 1-Port + Normalization (1-P+N)
Error to be • Enhanced Response (ER)
Reference Description
corrected • One-Path 2-Port (1-P 2-P)
e22 Load Match (F)
e’11 Load Match (R) Nowadays, most VNAs have a full S-parameter test set.
Thru
e10e32 Transmission Tracking (F) This allows the source to be switched to both ports, hence it
e’23e’01 Transmission Tracking (R) is able to measure all four S-parameters and a TOSM
calibration can be done.
Equations (10) to (21) represent the 12 error terms to be However, when it is not necessary (or convenient for
calculated. some reason) to measure all four DUT S-parameters or,
uncertainties are not necessary to be as small as possible,
above mentioned incomplete calibration methods can be
B. DUT S-Parameters Measurement used [5]. All of them are partial calibrations based on
Solving equations (1) to (4), corrected S-parameters of TOSM method described in section II.
the DUT can be expressed as follows [1]−[4] To simplify mathematical expressions, crosstalk error
terms will be considered null valued for all cases
N11 ⋅ (1 + N 22 ⋅ e′22 ) − e22 ⋅ N 21 ⋅ N12 (24)
S11 =
D e30 = 0 (33)
N 21 ⋅ [1 + N 22 ⋅ (e22
′ − e22 )]
(25) ′ =0
e03 (34)
S 21 =
D
N 22 ⋅ (1 + N11 ⋅ e11 ) − e11
′ ⋅ N 21 ⋅ N12
(26)
S 22 = IV. TRANSMISSION RESPONSE
D
It is the simplest 2-Port calibration method and is used
N12 ⋅ [1 + N11 ⋅ (e11 − e11
′ )]
(27) when only S21 (or S12) is of interest. A Thru reference
S12 =
D element is connected between ports for the calibration, so
where only transmission tracking error term is partially calculated.
This causes the highest uncertainties in transmission S-
S11 M − e00 (28) parameter measurements.
N11 =
e10 e01
A. VNA Calibration (between Ports)
′
S12M −e03 (29)
N12 = From (2) and (33):
′ e01
e23 ′
S − e30 (30) e10 e32 ⋅ S 21 (Thru ) (35)
N 21 = 21M S 21M =
e10 e32 1 − e11 S11 (Thru ) − e22 S 22 (Thru ) + e11e22 Δ S (Thru )
′
S 22 M − e33 (31)
N 22 =
′ e32
e23 ′ Applying (9) in (30)
D = (1 + N11 ⋅ e11 ) ⋅ (1 + N 22 ⋅ e′22 ) − N 21 ⋅ N12 ⋅ e22 ⋅ e11
′ (32)
1 (36)
S 21M = e10 e32 ⋅
1 − e11e22
where Nxx are normalized S-parameters [4].
As neither e11 nor e22 is calculated, the correction term
C. TOSM: Advantages and Disadvantages related to them is considered null valued.
1) Advantages: Provides low uncertainties as all 12 error
terms are calculated and all four DUT S-parameters are e11e22 = 0 (37)
measured.
Replacing (37) in (36)
2) Disadvantages: Needs an 8-step procedure calibration.
It is always necessary to measure all four DUT S- (38)
e10 e32 = S 21M
parameters even if only one is of interest.

812
 Similar considerations are applied for the reverse
transmission tracking term Similar considerations are applied for reverse parameters

′ e′23 = S12M
e01 (39) ′
S 22M ( DUT ) − e33 (44)
S 22 =
′ − Δe ′
S 22M ( DUT ) ⋅ e22

B. DUT S21 (or S12) Measurement

S12 M ( DUT ) (45)
As only S21 parameter is measured and e10e32 error term S12 =
′ e01
e23 ′
is calculated, equation (25) is reduced to

S 21M ( DUT ) (40) C. 1-Port + Normalization: Advantages and Disadvantages

S 21 =
e10 e32
1) Advantages: Corrects Directivity, Reflection Tracking
and Source Match of Port 1 (or Port 2).
Similar considerations can be applied for S12
2) Disadvantages: The same as in Transmission
S ( DUT ) Response method.
S12 = 12M (41)
e'23 e'01

VI. ENHANCED RESPONSE

C. Transmission Response: Advantages and Disadvantages It is an improvement of the 1-P+N method for
1) Advantages: Very fast one-step calibration procedure. measuring Forward (or Reverse) S-parameters. It needs the
Only S21 needs to be measured in order to get its corrected same four steps as before to calibrate the VNA but, in this
value, so it is a good option for unidirectional devices. case, it also calculates Load Match error term. This allows
Transmission Tracking error term to be correctly calculated.
2) Disadvantages: Only for transmission (S21 or S12)
parameters. As Transmission Tracking error term is not A. VNA Calibration
calculated correctly, this method is not very accurate with
1) Port 1 (or Port 2) Calibration: Procedure is applied in
lossy DUTs.
the same manner as in Section II A.1 (or section II A.2).
2) Calibration between Ports: Procedure is applied in
V. 1-PORT + NORMALIZATION the same manner as in Section II A.4.
This method performs a 1-Port calibration (at Port 1 or B. DUT Forward (or Reverse) Parameters Measurement
Port 2) and, separately, a transmission response. This is
Although e22 is calculated in this case, DUT reverse
usually employed when only forward parameters (S11 and
parameters are not measured and none of the reverse error
S21) or reverse parameters (S22 and S12) are required.
terms is calculated. Hence, all correction terms related to
A. VNA Calibration them in (24) and (25) are considered null valued and S11 and
S21 can be derived as follows
1) Port 1 (or Port 2) Calibration: Procedure is applied
in the same manner as in Section II A.1 (or Section II A.2). S11M ( DUT ) − e00
S11= (46)
2) Calibration between Ports: Procedure is applied in S11M ( DUT ) ⋅ e11 − Δe
the same manner as in Section IV A.
B. DUT Forward (or Reverse) Parameters Measurement S 21M ( DUT ) ⎡ e01e10 ⎤
S 21 = ⋅⎢ ⎥ (47)
As DUT reverse parameters are not measured, e22 and all e10 e32 e S
⎣ 11 11M ( DUT ) − Δe ⎦
reverse error terms are not calculated, hence all correction
terms related to them in (24) and (25) are considered null
Similar considerations are applied for reverse parameters
valued. S11 can be expressed as follows
′
S 22 M ( DUT ) − e33 (48)
S 22 =
S11M ( DUT ) − e00 (42) ′ − Δe ′
S 22M ( DUT ) ⋅ e22
S11=
S11M ( DUT ) ⋅ e11 − Δe

S12M ( DUT ) ⎡ ′ e32

e23 ′ ⎤
(49)
As OSM and transmission normalization calibrations are S12 = ⋅⎢ ⎥
′ e01
e23 ′ e′ S
⎣ 22 22M ( DUT ) − Δ e′ ⎦
performed separately, S21 corrected value remains the same
as in section IV B.
C. Enhanced Response: Advantages and Disadvantages
S 21M ( DUT ) (43)
S 21 = 1) Advantages: Calculates Transmission Tracking error
e10 e32
term correctly.

813
 2) Disadvantages: As only Forward (or Reverse) S- A. S11 Deviation Results
parameters are measured, Load Match value can not be used Maximum deviations of |S11| for 1-P+N and ER methods
for correcting DUT S-parameters. respect to TOSM are shown in Fig. 3 for different values of
|S21|. Since both of these incomplete methods calculate S11
in the same way (see equations (42) and (46)), such
VII. ONE-PATH 2-PORT deviations with respect to TOSM are the same. These
deviations depend on Port 2 Load Match, i.e. e22, and the
Originally named One-Path Full 2-Port, this method was DUT´s attenuation, i.e. S21.
introduced to T/R VNAs in order to measure all four S- For example, if the DUT is a 6-dB attenuator and |e22| =
parameters. However, the DUT must be manually reversed 0.1, then the maximum deviation with respect to TOSM will
to measure Reverse S-parameters . be 0.026.
At present most VNAs have this calibration option, but If now the DUT consists of a coaxial cable with nominal
special care must be taken as some manufactures call the 0 dB attenuation, and |e22| remains in 0.1, then maximum
ER method as 1-P 2-P. deviation respect to TOSM method rises to 0.100.
A. VNA Calibration
This method considers Forward and Reverse models the
same as follows

′
e00 =e33 (50)
e11 = e′22 (51)
e10 e01 = e′23 e32
′ (52)
′ e′23
e10 e32 = e01 (53)
′
e22 =e11 (54)
′
e30 = e03 (55)

Hence, it needs the same four steps as in the previous

methods to calibrate the VNA and only 6 forward error
terms need to be calculated using (10), (11), (12), (16), (18)
and (19).
Fig. 3 |S11| maximum deviation for 1-Port + Normalization and Enhanced
B. DUT Parameters Measurement Response with respect to TOSM method.
When measuring the DUT, forward parameters are first
measured, then the DUT is connected backwards and B. S21 Deviation Results
reverse parameters are measured. This allows equations (24) Different maximum deviations of |S21| for TR and ER
to (27) to be used with all correction terms correctly methods with respect to TOSM are shown in Fig. 4 and 5
calculated. respectively for a DUT having a nominal attenuation value
C. One Path 2-Port: Advantages and Disadvantages of 0 dB, for different values of |S11| and |S22|. For the ER
case, the deviation of |S21| is independent of its nominal
1) Advantages: All four DUT S-parameters can be value.
measured. As Forward and Reverse error terms have same
values, only a four-step procedure is needed to calibrate the
VNA.
2) Disadvantages: Not recommended for VNAs using
any combination of coaxial sexed port connectors due to the
necessity of adapters. The DUT must be manually reversed
in order to measure all four S-parameters. In practice,
uncertainties may be higher that TOSM due to connector
mechanical repeatability or cable movement.

VIII. SIMULATIONS
A series of comparisons between incomplete calibration
methods with respect to TOSM were carried out. S11 and S21
measurements were simulated and maximum deviation
results are shown in Fig. 3 to 6. In these simulations, the
TOSM case represents an ideal measurement of the DUT, Fig. 4 |S21|dB maximum deviation for Transmission Response with respect
to TOSM method when measuring a S21 value of 0 dB.
since all the error terms are perfectly determined and used
to correct the S-parameters.

814
 be used to estimate the maximum systematic errors for all
the cases.
In practice, if VNA´s error terms and/or DUT´s
mismatches are quite low, there will be no significant
deviation between incomplete and complete calibration
methods when measuring S11 or S21. However, if VNA´s
source and load match error terms are high enough and the
DUT is lossy, then incomplete methods may not be suitable
since significant deviations may occur.

REFERENCES

[1] AN 1287-3, “Applying error correction for VNAs”, 2nd. Ed., Santa
Clara, CA: Agilent Tech., 2002.
[2] D. Rytting, “Network analyzer error models and calibration
methods”, Palo Alto, CA, Hewlett Packard Inc., 1998.
Fig. 5 |S21|dB maximum deviation for Enhanced Response with respect to [3] B. Hall, “VNA error models: Comments on EURAMET/cg-2/v.01”,
TOSM method (in this case the deviation does not depend on the S21 value). ANAMET Report 051, Measurement Standards Laboratory of New
Zealand Lower Hutt, New Zealand, 2010.
For example, if |S11| = |S22| = 0.1, and |e11| = |e22| = 0.1, [4] J. Dunsmore, “Handbook of microwave component measurement”,
then the maximum deviation with respect to TOSM will be Agilent Tech., John Wiley & Sons, Ltd., UK, 2012.
0.17 dB for TR method and 0.09 dB for ER method. [5] M. Hiebel, “Fundamentals of vector network analysis”, Rohde &
Schwarz GmbH & Co. KG, 5th Ed., 2011.
According to Fig. 5 and 6, if now the same DUT has an
[6] G. Wübbeler, Clemens Elster, Thomas Reichel and Rolf Judaschke,
attenuation value of 6 dB and all other values remain the “Determination of the complex residual error parameters of a
same, then maximum deviation respect to TOSM method calibrated one-port vector network analyzer”, IEEE Transactions,
rises to 0.24 dB for TR method and remains in 0.09 dB for Vol. 58, No. 9, 2009.
ER method. [7] AN 1287-11, “Specifying calibration standards and kits for Agilent
vector network analyzers”, Agilent Tech., 2011.

Fig. 6 |S21|dB maximum deviation for Transmission Response with respect

to TOSM method when measuring a S21 value of -6 dB.

IX. CONCLUSIONS

Different types of incomplete 2-Port VNA calibrations

methods are explained in this paper. Each one of them has
its own advantages and disadvantages respect to a complete
2-Port method as TOSM. The VNA user must be aware of
how each method works and the systematic errors that can
and cannot be corrected, so a judicious choice can be made
as to which method should be used for any given
application.
The approach in this article of studying each incomplete
method as a simplified case of the 12-term error model,
allows the reader to highlight the conceptual differences
between them. The expressions given through this work can

815