Professional Documents
Culture Documents
7.uncertainty of VNA S-Parameter Measurement Due To Nonideal TRL Calibration Items
7.uncertainty of VNA S-Parameter Measurement Due To Nonideal TRL Calibration Items
7.uncertainty of VNA S-Parameter Measurement Due To Nonideal TRL Calibration Items
2, APRIL 2005
Abstract—For the 7-term TRL calibration of a four-sampler The noncorrected S-parameters measured by the VNA
vector network analyzer (VNA), expressions for the deviations of (raw values) are then [1]
the S-parameters of two-port test objects are presented as func-
tions of the deviations from the ideal values of the S-parameters
of the TRL standards used. The sensitivity coefficients obtained
are suitable for establishing the Type-B uncertainty budget for the
S-parameter measurement.
Index Terms—7-term TRL calibration, sensitivity coefficients,
S-parameters, uncertainty of calibration, uncertainty of measure-
ment, vector network analyzer
.
(1)
I. INTRODUCTION
where the , , , are the S-parameters of a calibra-
disturbed
Manuscript received July 2, 2004; revised September 29, 2004.
The author is with the Physikalisch-Technische Bundesanstalt, 38023 Braun-
schweig, Germany. ideal
Digital Object Identifier 10.1109/TIM.2005.843521
where is the ideal reflection coefficient of the loads. instead of the exact expressions given in [4], as a good approx-
Through: imation
disturbed
Line:
(5)
disturbed
(6)
with , line length, propagation coeffi-
cient and , , the deviations associated with
Reflect, Through, and Line, respectively.
At first, the deviation is determined. By the VNA
(7)
firmware, is calculated from raw values obtained for Through
and Line (see [3]), thereby ideal scattering matrices and
are assumed. To calculate a disturbed , we insert the (8)
entries of the disturbed or , respectively, into (1) for ,
, , and , and calculate disturbed raw values, which To get corresponding expressions for and , index 1 is
we then insert into the equation used for the calculation of replaced by 2 and vice versa in (5)–(7).
[3]. Finally, we obtain . For the nonideal Through and Line, the sensitivity coeffi-
When the seven error terms are calculated by the firmware cients may become large if approaches 1, i.e., if approaches
from the raw values obtained for Reflect, Through and Line, also (where is the wavelength). All reflection deviations ,
ideal scattering matrices , , and are assumed. To obtain may become large when using loads of small instead of
disturbed error terms (ideal plus deviation), we insert the entries short-circuits as Reflects, while all transmission deviations are
of the disturbed or , or , respectively, into (1) for , not affected by . Further on we observe a direct dependence of
, , and , and again calculate disturbed raw values. and on and , respectively, unaffected by
We insert these (and the expression for ) into the appropriate any other parameter [cf., (6)], as in case of a nonideal TMSO cal-
equations (similar as in [3]) used for the calculation of the error ibration (cf. (23) in [2]). A small phase shift in the transmission
terms. Finally, we obtain deviations of the error of the nonideal Through will generate an equal phase deviation
terms as functions of , , associated with Reflect, of the transmission coefficients of the test object.
Through, and Line. For low-reflective test objects ( and near 0), in case of
the nonideal Line, there is no significant influence on the trans-
III. RESULTING SENSITIVITY COEFFICIENTS mission. The sensitivity coefficients dependent on the deviations
associated to Reflect become very small.
Combining the results from the first and second step, we ob-
tain sensitivity coefficients for the four S-parameters of a test
IV. EXPERIMENT AND CALCULATIONS
object which are separately given for the deviations , ,
associated with Reflect, Through, and Line. The expressions (3)–(8) have been experimentally verified
using a set of coaxial (7 mm/PC-7) high- and low-reflective
A. Deviations Associated With Reflect two-port test objects of (nominal) attenuation 0 to 50 dB, like
If the reflection coefficient of one of the terminations is the set described in [2]. The high-reflective devices were ei-
while the other has a deviating reflection , there is influ- ther a short line or attenuators sandwiched between the side
ence only on the reflection coefficients, without any influence of arms of two T-junctions. The feeding arms of the junctions were
shorted, hence the S-parameters varied between about 0.8 and
0.8. The reflection of the low-reflective devices was maximum
(3) 0.05.
Only one of the calibration items (Reflect, Through, and Line)
(4) at a time was considered nonideal. For a nonideal Reflect, one of
the two short-circuits of the calibration kit, designed for test port
depending on the position of the nonideal termination: either 1 or 2, was loaded with Teflon disks (thickness 0.06 mm, diam-
at port 1 or port 2 , eter 5 mm). For a nonideal Through, a thin copper foil (thickness
respectively. 0.03 mm, diameter 3.2 mm) was introduced between the inner
conductors with the Through connection (cf. Fig. 1 in [4]). A
B. Deviations Associated With Through and Line nonideal Line was realized either by a Teflon ring (length 3.41
For Through and Line, the sensitivity coefficients are free mm, diameter 5 mm) around the inner Line conductor 6.95 mm
of all error terms but the “match” term (as with the sen- in length, or by replacing the “ideal” inner conductor (diam-
sitivity coefficients for a nonideal TMSO calibration [2], where eter 3.040 mm) by an inner conductor of the same length but
a Through, Matched loads, Short and Open circuits are used as of diameter increased by an amount of .
calibration standards). Since is negligibly small, we obtain, The measured S-parameters of the last-mentioned nonideal Line
678 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 54, NO. 2, APRIL 2005
Fig. 1. Flow graph of the 8-term error model [1]. For a TRL calibration, the
test object is replaced by the TRL standards (Through Connection, Line and
Reflective Loads). a , b refer to test port 1 and a , b to test port 2 of the
VNA.
Fig. 3. Calculated deviations j S j in dB and phase deviations arg( S )
in for the high-reflective line and the low-reflective 20 dB attenuator, using for
the TRL calibration a nonideal Line with the “ideal” inner conductor diameter
increased by an amount of 1D :
= 0 002 mm.
are about maximum 0.08 (for ) and 0.006 (for ). Fig. 4. Calculated deviations j S j in dB and phase deviations
arg( S )
in for the high-reflective line and the low-reflective 20-dB attenuator as test
Their effect on the calculated and measured deviations objects, with a hair 0.1-mm in thickness in the Through connection, by use of
and of a high-reflective line as test object is shown in the values ofT taken from Fig. 5 in [5] for the calculations.
Fig. 2. No significant difference of the results was observed
when using the exact sensitivity coefficients in [4] instead of 0.1-mm thick, will cause S-parameter deviations of about
(5)–(8) for the calculations. maximum 0.01 of the Through connection (cf. Fig. 5 in [5]). The
Some additional calculations were carried out to show the in- deviations of modulus and phase of , calculated by use of
fluence of small values of , , or as they may occur these values are shown in Fig. 4 of this paper, for the high-
during the calibration or with real calibration items from com- reflective line and the low-reflective 20-dB attenuator. Even for
mercial calibration kits. The experimentally obtained values low-reflective test objects, the deviations are considerably
of the set of test objects were used in these calculations. large, as in case of a nonideal TMSO calibration (cf. Fig. 6 in
For a Line taken from a precise calibration kit, the diame- [5]). The deviations in reflection may rise to 0.01.
ters of the conductors will deviate not more than some from Connector imperfections such as gaps in the inner conductor
the ideal values. For , the effect on for the may also give rise to deviations. For a 0.025-mm-long gap, re-
high-reflective line and the low-reflective 20-dB attenuator is flections were measured [6] which vary linearly with fre-
shown in Fig. 3. Even with this high Line precision, the devia- quency from 0 to about 0.0055 at 18 GHz for N connectors.
tion in attenuation could be close to 0.01 dB for high-re- Assuming equal gaps of this kind at the Line connections to test
flective test objects, whereas, for example, the deviations of the ports 1 and 2, then we have , where
S-parameters of the low-reflective 20-dB attenuator are negli- the real part of is set equal to . For low-reflective test ob-
gibly small. jects the effect is very small, but for high-reflective test objects
Small particles which get in between the end planes of the deviations of about maximum 0.07 dB may occur. Devi-
connectors with the Line or Through connections, may cause ations of similar magnitude are caused by a gap in the inner con-
considerable deviations of the S-parameters of high-reflective ductor with the Through connection, where
but also low-reflective test objects. For example, a human hair, is assumed.
STUMPER: UNCERTAINTY OF VNA S-PARAMETER MEASUREMENT DUE TO NONIDEAL TRL CALIBRATION ITEMS 679