You are on page 1of 6

IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 68, NO.

6, JUNE 2019 2041

Realization of Absolute Phase and AC Resistance


of Current Shunts by Ratio Measurements
Tobias Bergsten and Karl-Erik Rydler, Member, IEEE

Abstract— This paper presents a simplified method for


realizing absolute phase and ac resistance for current shunts
using only impedance ratio measurements. The method is based
on three geometrically identical current shunts with different
resistances, but with the same inductance, capacitance, and
ac–dc resistance change. We demonstrate how the inductance,
capacitance, and ac resistance can be calculated from the com-
plex impedance ratio measurements, therefore realizing absolute
current shunt impedance. The method gives competitive uncer-
tainties of around 200 μ/ for amplitude and 400 μrad for
phase at 1 MHz in the 1- range.
Index Terms— Capacitance, current shunt, impedance
measurement, inductance, measurement standards, phase
measurement, resistance.

I. I NTRODUCTION

W HEN a shunt is used for current measurement, its


impedance determines the relation between the input
current and its output voltage. For dc current, the impedance
is simply a resistance, but for ac currents, the impedance is
complex, Z = |Z |eiϕ (ϕ = arg Z ), and depends on frequency.
The common way of calibrating the impedance is to use
an impedance ratio bridge to measure the ratio of the shunt
impedance to a known reference impedance. This method is
relatively straightforward and gives accurate results, but it is Fig. 1. Top: three groups of shunts used in this paper. The shunts in each
group are identical in structure but with different resistance. The resistances
dependent on an absolute calibration of both the amplitude range from 0.16 to 4.0 . Bottom: four-terminal model used for the shunts in
|Z ref | and phase ϕref = arg Z ref of the reference impedance. this paper consists of a series inductance L, a dc resistance Rdc , a frequency-
The amplitude can be realized using thermal voltage convert- dependent resistance r which is zero at dc, and a parallel capacitance C. All
shunts in a group are assumed to have identical L, C, and r.
ers or calculable resistors [1]–[4], and for phase calibration,
a number of different methods can be used [4]–[11]. In this
paper, we present a method [12] to realize absolute phase and TABLE I
ac resistance using only an impedance ratio bridge and shunts OVERVIEW OF THE P RIMARY S HUNT G ROUPS U SED IN T HIS PAPER
calibrated for dc resistance, without the need for reference
shunt calibration using other measurement setups.

II. C URRENT S HUNT G ROUPS


In this paper, we used groups of three coaxial current shunts,
geometrically identical, but with different resistance values
Manuscript received July 6, 2018; revised October 17, 2018; accepted
November 6, 2018. Date of publication January 10, 2019; date of current
version May 10, 2019. This work was part of the TracePQM project. This
project has received funding from the EMPIR programme co-financed by the
Participating States and from the European Union’s Horizon 2020 research and
innovation programme. The Associate Editor coordinating the review process (see Fig. 1 and Table I). The assumption is that the three
was Djamel Allal. (Corresponding author: Tobias Bergsten.)
The authors are with the Department of Measurement Science and Technol- shunts in a group have the same inductance L, capacitance C,
ogy, RISE Research Institute of Sweden, SE-501 15 Borås, Sweden (e-mail: and ac–dc resistance change r = Rac − Rdc , where Rac is
tobias.bergsten@ri.se). the total shunt resistance at ac and Rdc is the calibrated
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org. dc value of the shunt resistance. This approach is a further
Digital Object Identifier 10.1109/TIM.2018.2882927 development of the methods in [7] and [8]. The capacitance
This work is licensed under a Creative Commons Attribution 3.0 License. For more information, see http://creativecommons.org/licenses/by/3.0/
2042 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 68, NO. 6, JUNE 2019

6) As a consistency check, the three measured ratios are


multiplied to verify that the product is close to 1.
The rms current was constant within each group (see Table I),
no level dependence was investigated in this paper. We, then,
defined the amplitude ratios and phase differences (i, j = 1..3,
ϕi = arg Z i ) as
Ai j = |Z i /Z j | (1)
φi j = ϕ i − ϕ j . (2)

IV. P HASE R EALIZATION


The phase ϕi of a coaxial shunt of dc resistance Ri with a
small phase error can be well approximated at a given angular
frequency ω as
ωL
ϕi = − ω Ri C. (3)
Fig. 2. Digital impedance ratio bridge used for the measurements in this Ri
paper. We can, then, write the phase differences as
   
1 1
φ12 = ω L − − C(R1 − R2 )
R R2
comes mostly from the printed circuit boards (PCBs), while   1  
the inductance is an effect of the resistor legs and resistive 1 1
φ23 = ω L − − C(R2 − R3 ) (4)
foil forming inductive half-loops above the circuit boards. The R2 R3
ac–dc resistance change is mainly due to the skin effect in the and correspondingly for φ31 . Solving these equations for L
resistor leads and, therefore, does not depend on the shunt and C gives us
resistance. Since the resistive foil in the resistor body is much  
R1 R2 R3 φ12 φ23
thinner than the skin depth, it does not contribute to the ac–dc ωL = − (5)
R1 − R3 R1 − R2 R2 − R3
resistance change.  
−1 R1 φ12 R3 φ23
ωC = − (6)
R1 − R3 R1 − R2 R2 − R3
III. D IGITAL I MPEDANCE B RIDGE where we use the calibrated dc values for R1 , R2 , and R3 .
We, then, repeat the calculation of (5) and (6) using the
The complex impedance ratios Z 1 /Z 2 , Z 2 /Z 3 , and Z 3 /Z 1
combination φ23 , φ31 and φ31 , φ12 , respectively, to get three
of Shunts 1–3 (arbitrary numbering) in each group were
different values of L and C and then take the average to get
measured using our digital impedance ratio bridge [8]
the final values [see Fig. 3 (top and middle graphs)]. Using (3),
(see Fig. 2). This bridge consists of a current source, a current
we get the phase of the three shunts. Fig. 4 shows the time
T-connector for connecting the two shunts in series, and a
constants τi = ϕi /ω for the three shunts of Group 1.
four-channel digitizer that measures the output voltages from
the shunts differentially. Several procedures are performed to V. AC R ESISTANCE R EALIZATION
ensure the accuracy of the measured impedance ratios.
The amplitude of the shunt impedance can be written as
1) The linearity of one digitizer channel is calibrated (assuming a small phase)
using a thermal transfer standard, as described in [13].    
Remaining channels are calibrated relative to the first 1 L 2
|Z i | = Ri + r + Ri ω LC +
2
− (Ri C) 2
. (7)
one by connecting all channels in parallel. 2 Ri
2) The relative gain and the phase difference are calibrated
We can, then, write the amplitude ratios as
at frequencies of interest by again connecting all chan-  
nels in parallel. R1 R2 − R1
A12 = 1+r +ε (8)
3) In order to eliminate the influence of the digitizer input R2 R1 R2
impedance (including measurement cables), we measure where ε is the correction for inductance and capacitance
each ratio twice, turning the current T-connector in the  
opposite way between the measurements [8]. Therefore, 1 2 2 L2
ε = ω (R2 − R1 ) 2
+C .
2
(9)
we effectively measure the unloaded condition of the 2 R12 R22
shunts (except for the unavoidable contact capacitance). This term is second order in ω and is generally negligible, but
4) We switch the digitizer channels (the two channels for the highest frequencies, it has a small influence. Solving
measuring Z 1 are switched to Z 2 and vice versa) to the equation for r gives us
reduce gain errors caused by short-term drift.  
R1 R2 R2
5) Measurements are repeated at least three times and r= A12 −1−ε (10)
averaged to reduce noise and verify reproducibility. R2 − R1 R1
BERGSTEN AND RYDLER: REALIZATION OF ABSOLUTE PHASE AND AC RESISTANCE OF CURRENT SHUNTS 2043

Fig. 5. Measured values of ac–dc resistance change for the three groups of
shunts using the new method. The inductance and capacitance values are only
weakly dependent on frequency and are presented in Table II.

TABLE II
A PPROXIMATE VALUES OF I NDUCTANCE , C APACITANCE , AND AC–DC
R ESISTANCE C HANGE FOR THE T HREE G ROUPS OF S HUNTS

Fig. 3. Top and middle graphs: inductance L and capacitance C of the


current shunts of Group 1 from 10 kHz to 1 MHz. The symbols correspond components, and geometry. Group 1 shunts are the largest,
to results when using different combinations of φi j in (5) and (6), and the containing 48 parallel through-hole resistors in a coaxial cage
solid line is the average. Bottom graph: ac–dc resistance change r of the same
current shunts. Symbols correspond to different values of Ai j in (10), and the design. Group 2 shunts are smaller with 12 parallel through-
solid line is the average. Note that the symbols do not indicate L, C, and r hole resistors, while Group 3 shunts are the smallest, with
for the individual shunts as they are assumed to be the same. 12 parallel surface mounted resistors in a radial pattern,
enclosed in a metal box. The purpose was to perform the phase
and ac resistance realization using three quite different shunt
groups and verify that the end result was the same. The results
for the three groups are shown in Fig. 5 and Table II.
The L, C, and r parameters are quite different, and the
three groups can be considered fairly independent. By using
these three quite distinct groups of shunts rather than using
a single group with more than three shunts, we can assume
that the covariance of the uncertainty is small and treat the
three groups as uncorrelated realizations of impedance, and,
thereby, reduce the uncertainty by averaging.
We, then, measured another current shunt (0.6 /1 A, made
by CMI) by comparing it to one shunt from each primary
group (see Fig. 6), using the digital bridge. Ideally, the three
results should be identical, but due to some uncertainties in
the realizations, the results are slightly different. In Fig. 7,
the results are shown together with the mean value. The
difference between the time constants, up to 150 ps, corre-
Fig. 4. Time constant τi = ϕi /ω for the three current shunts of Group sponds to about 1000 μrad at 1 MHz. Standard deviation is
1 from 10 kHz to 1 MHz. There is a slight decrease at higher frequencies,
which can be attributed to the skin effect in the resistor legs, which decreases
about 60 ps/380 μrad and 150 μ/ at 1 MHz, respectively.
the self-inductance. The CMI shunt was used in an informal trilateral impedance
comparison between national metrology institutes CMI (Czech
Republic), RISE (Sweden), and BEV (Austria) [14].
and correspondingly for A23 and A31 . As can be seen in Fig. 3
(bottom graph), the three ratios give consistent results for r .
VII. U NCERTAINTY A NALYSIS
VI. C OMPARING S HUNT G ROUPS A. Monte Carlo Simulation
In order to verify the method, we have constructed three The uncertainty of the phase realization depends mostly on
different groups of current shunts, with different dimensions, the assumption that the inductance and the capacitance are the
2044 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 68, NO. 6, JUNE 2019

Fig. 6. Overview of the measurement scheme for the current shunts. The
arrows represent relative measurements using the digital impedance bridge.
Each group of three shunts is used to realize independent phase and ac
resistance references. One shunt from each group is used as a reference for
calibrating the test shunt (0.6 /1 A, made by CMI).

TABLE III
S ENSITIVITY OF T IME C ONSTANT TO I NDUCTANCE ,
Fig. 7. Results of phase and amplitude calibration of a 0.6- current shunt
C APACITANCE , AND P HASE U NCERTAINTY
using one shunt from each of the three primary groups as reference. Top graph:
time constant of the shunt τ = ϕ/ω. Bottom graph: relative ac–dc resistance
change. Solid lines are the arithmetic means of the three curves and the shaded
areas represent the uncertainty of the means.

The sensitivity is different for each group of shunts.


Group 1, with the lowest resistances, has the highest sensitivity
to inductance; Group 3, with the highest resistances, has
the highest sensitivity to capacitance, while Group 2, which
has the smallest resistance ratio between the shunts (ratio 2,
compared with 2.5 for Groups 1 and 3), is the most sensitive
same, and only the resistance differs between the shunts in a to phase measurement errors. Table III summarizes the results
group. The uncertainty of the phase difference measurement for the three groups.
in the digital bridge also gives a contribution. The uncertainty
is furthermore affected by the resistance values of the shunts. B. Inductance Uncertainty
While it is possible to use (5) and (6) to express the uncertainty The shunt inductance of Groups 1 and 2 comes mainly
analytically, the expression would be very complex and not from the half-loops formed by the leads and the resistive foil
very intuitive. of the resistor components. For the most part, the inductance
We, therefore, took another approach, using a Monte Carlo is the same for all the shunts because they are geometrically
simulation to estimate the sensitivity of the time constant identical. We compared two identical shunts of 0.4  and
to uncertainties in inductance L, capacitance C, and phase found that the difference in phase was 10 μrad, corresponding
difference measurement φi j . The procedure was to simulate to 0.6 pH in inductance difference. We, therefore, conclude
the bridge measurements using simulated shunts with the same that the uncertainty caused by the manual assembly of the
parameters as the actual shunts, but with small individual shunts is less than 1 pH. However, the current distribution
values added to L, C, or φi j . These small added values were in the resistive foil might be different for different resistor
generated randomly from a normal distribution with a specified values, and therefore the inductance would differ between the
width, δ. We, then, calculated the time constants τi = ϕi /ω shunts. Since the foil is a small part of the inductive loop,
using (3), (5), and (6), and recorded the mean error τ of perhaps 5% of the loop area, we estimate the uncertainty in
the time constants compared to the actual time constants of inductance to 1%.
the simulated shunts. Repeating this procedure for 105 times The inductance of Group 3 shunts comes mostly from the
generated a distribution of τ , and we could calculate the resistive foil and the solder joints, and possibly from mutual
standard deviation s( τ ). The sensitivity factor could, then, inductance between the current and voltage loops. We estimate
be calculated as the uncertainty to 10%.
∂τ s( τ ) C. Capacitance Uncertainty
= (11)
∂X δ Most of the capacitance of the shunts comes from the
where X is L, C, or φi j . PCBs. The resistive components contribute only about 0.5 pF
BERGSTEN AND RYDLER: REALIZATION OF ABSOLUTE PHASE AND AC RESISTANCE OF CURRENT SHUNTS 2045

TABLE IV measurements or specially designed time constant standards,


U NCERTAINTY OF THE T IME C ONSTANT W ITH THE this method can be used with existing coaxial current shunt
E STIMATES OF THE I NDIVIDUAL T ERMS
designs and an impedance ratio bridge. It is most suitable for
shunts, where RC and L/R are small (compared with 1/ω)
and of the same order. The phase uncertainty is comparable
to calculable time constant standards, while the amplitude
uncertainty is a bit higher than what is achieved with ac–dc
transfer methods. Perhaps the amplitude uncertainty can be
improved by including a model for parallel conductance.
The phase uncertainty may be improved by better control
of the inductance, e.g., by using machine assembly of the
each, which means about 2% of the total capacitance in circuit boards and selecting components with low internal
Groups 1 and 2 and 20% in Group 3. We estimate that the inductance such as microwave components. This method was
PCB capacitance is identical between the shunts within 0.5% used by RISE in an informal trilateral comparison of phase and
for Groups 1 and 2 and within 5% for Group 3, and the amplitude characterization of current shunts [14], and shows
resistor capacitance is identical within 10%. This gives a total good agreement with the other two labs (CMI and BEV).
uncertainty of 0.7% for Groups 1 and 2 and 6% for Group 3.

D. Phase Difference Uncertainty ACKNOWLEDGMENT


We can make a decent estimate of the phase difference The authors would like to thank Stanislav Mašláň at CMI
uncertainty by calculating the sum of the three measured phase for providing the test shunt used for this work. This work was
differences in a group. Ideally, the sum should be 0. In our a part of the TracePQM Project.
measurements, this sum is maximum 10 μrad at 1 MHz,
corresponding to 1.6 ps. We, therefore, estimate the uncertainty R EFERENCES
to 3 ps. [1] J. R. Kinard, T. E. Lipe, and C. B. Childers, “AC-DC difference
relationships for current shunt and thermal converter combinations,”
E. Uncertainty of Time Constant IEEE Trans. Instrum. Meas., vol. 40, no. 2, pp. 352–355, Apr. 1991.
[2] M. Klonz, H. Laiz, T. Spiegel, and P. Bittel, “AC-DC current transfer
Combining the main uncertainty contributions, we arrive at step-up and step-down calibration and uncertainty calculation,” IEEE
the numbers in Table IV. The results range from 90 to 180 ps. Trans. Instrum. Meas., vol. 51, no. 5, pp. 1027–1034, Oct. 2002.
By combining all three groups and taking the average as we [3] K. Lind, T. Sørsdal, and H. Slinde, “Design, modeling, and verifi-
cation of high-performance AC-DC current shunts from inexpensive
did with the 0.6- CMI shunt, we can reduce the uncertainty components,” IEEE Trans. Instrum. Meas., vol. 57, no. 1, pp. 176–181,
to 65 ps, corresponding to 400 μrad at 1 MHz. Jan. 2008.
[4] S. Mašláň, M. Šíra, and T. Skalická, “Progress on simple resistance stan-
dard with calculable time constant,” in Proc. Conf. Precis. Electromagn.
F. AC Resistance Uncertainty Meas. (CPEM), Jul. 2018, pp. 1–2.
The assumption for the ac–dc resistance change is that [5] S. Svensson, K.-E. Rydler, and V. Tarasso, “Improved model and phase-
the skin effect in the resistor leads and solder joints is the angle verification of current shunts for AC And power measurements,”
in Proc. Conf. Precis. Electromagn. Meas. (CPEM), Jun./Jul. 2004,
main contribution. The resistance change in absolute terms, pp. 82–83.
therefore, does not depend on the resistor value. However, [6] I. Budovsky, “Measurement of phase angle errors of precision current
there may also be a contribution from frequency-dependent shunts in the frequency range from 40 Hz to 200 kHz,” IEEE Trans.
Instrum. Meas., vol. 56, no. 2, pp. 284–288, Apr. 2007.
leak conductance through the substrate and encapsulation of [7] G. C. Bosco et al., “Phase comparison of high-current shunts up to
the resistive films, which would vary for different resistor 100 kHz,” IEEE Trans. Instrum. Meas., vol. 60, no. 7, pp. 2359–2365,
values. Furthermore, the current level will influence the self- Jul. 2011.
[8] K.-E. Rydler, T. Bergsten, and V. Tarasso, “Determination of phase angle
heating of the resistive components, and this could affect both errors of current shunts for wideband power measurement,” in Proc.
the leak conductance and the skin effect. We do not have a Conf. Precis. Electromagn. Meas., Jul. 2012, pp. 284–285.
model to take this into account, and therefore, the results will [9] X. Pan et al., “A coaxial time constant standard for the determination
of phase angle errors of current shunts,” IEEE Trans. Instrum. Meas.,
have some additional uncertainty. vol. 62, no. 1, pp. 199–204, Jan. 2013.
We can get an idea of the uncertainty by looking at the [10] L. Lai, J. Feng, L. Shi, H. Shi, Y. Pan, and B. Zhou, “Measurement
differences between the primary groups in Fig. 7. The standard of the time constant of four-terminal resistor based on dissipation
factor of capacitor,” in Proc. Conf. Precis. Electromagn. Meas. (CPEM),
deviation at 1 MHz is 150 μ/. By multiplying the three Jul. 2016, pp. 1–2.
measured impedance ratios within a primary group, we get [11] K.-E. Rydler, T. Bergsten, and G. Eklund, “A method for realisation
an error of less than 10−5 at 1 MHz, so the bridge contri- of inductance and quality factor to 1 MHz,” in Proc. Conf. Precis.
Electromagn. Meas. (CPEM), Jul. 2018, pp. 1–2.
bution to the uncertainty is negligible. We set the uncertainty [12] T. Bergsten and K.-E. Rydler, “Realisation of absolute phase and AC
conservatively to 200 μ/ (at 1 MHz). resistance of current shunts by ratio measurements,” in Proc. Conf.
Precis. Electromagn. Meas. (CPEM), Jul. 2018, pp. 1–2.
[13] T. Bergsten, V. Tarasso, and K.-E. Rydler, “An electrical power ref-
VIII. C ONCLUSION AND D ISCUSSION erence system up to 1 MHz,” in Proc. Conf. Precis. Electromagn.
This method of impedance realization for coaxial current Meas. (CPEM), Jul. 2016, pp. 1–2.
[14] S. Mašláň, M. Šíra, and T. Skalická, “Four terminal pair digital sampling
shunts offers a simplified procedure compared with alterna- impedance bridge up to 1 MHz,” IEEE Trans. Instrum. Meas., submitted
tive methods. Rather than time consuming ac–dc difference for publication.
2046 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 68, NO. 6, JUNE 2019

Tobias Bergsten was born in Lund, Sweden, Karl-Erik Rydler (M’08) received the M.Sc. degree
in 1970. He received the M.Sc. degree in engineering in electrical engineering from the Chalmers Univer-
physics and the Ph.D. degree in physics, with a sity of Technology, Gothenburg, Sweden, in 1979.
focus on Coulomb blockade thermometry, from the In 1979, he joined the Electrical Metrology Labo-
Chalmers University of Technology, Gothenburg, ratory, RISE Research Institutes of Sweden, Gothen-
Sweden, in 1996 and 2001, respectively. burg. During the years, he has been involved in the
In 2004, he joined the Spintronics Group, NTT calibration and development of measuring methods
Basic Research Laboratories, Atsugi, Japan, where for electrical dc and low-frequency quantities. His
he worked on Rashba spin orbit interference devices. current research interests include the development of
In 2006, he moved to the Semiconductor Physics measuring methods and standards for ac quantities.
Group, University of Cambridge, Cambridge, U.K., Mr. Rydler was a convener of the LF subcommittee
where he worked with AFM measurements on graphene. He joined the Electri- of EURAMET TC-EM from 2004 to 2010.
cal Metrology Laboratory, RISE Research Institutes of Sweden, Gothenburg,
in 2009, where he is developing electrical measurement methods.

You might also like