Measurement 131 (2019) 1–6

Contents lists available at ScienceDirect

Measurement
journal homepage: www.elsevier.com/locate/measurement

Measurement of DC properties and relative humidity (RH) dependence

of wideband AC current shunts
Jure Konjevod a,⇑, Roman Malarić a, Martin Dadić a, Petar Mostarac a, Hrvoje Hegeduš a
a
Faculty of Electrical Engineering and Computing, University of Zagreb, Unska 3, 10000 Zagreb, Croatia

a r t i c l e i n f o a b s t r a c t

Article history: The paper describes the DC characterization of AC current shunts recently developed at the Faculty of
Received 20 March 2018 electrical engineering and computing, University of Zagreb (UNIZG-FER). The shunts of cage type design
Received in revised form 10 July 2018 include nine shunts that range from 1 mA up to 10 A. The following DC properties were inspected: long
Accepted 9 August 2018
time drift, temperature coefficients and power coefficients. The results have shown that shunts have low
Available online 24 August 2018
temperature (<7 mO/O/°C), and power coefficients (<10 mO/O/W), and small drift which makes them suit-
able for precise power measurement applications. Also, relative humidity dependence of currents shunts
Keywords:
is investigated.
Current shunts
Electrical resistance measurement
Ó 2018 Elsevier Ltd. All rights reserved.
Power coefficient
Temperature coefficient
Relative humidity (RH) coefficient

1. Introduction PCB material. They are equipped with N-type connectors, and con-
structed with three circular PCB elements, crossbars and resistors.
In recent years several National metrology institutes have The resistive element is Vishay S102C type. The current flows from
developed a new type of precise AC shunts that can provide mea- the input plate over the crossbars to the resistors connected in par-
surement of AC current and power up to 100 A and 100 kHz or allel on the output plate and returns over the other side of the
even 1 MHz with low uncertainties [1]. Those shunts can be crossbars [6]. Thus, seemingly large loop in the current path is bal-
divided into cage and foil type. In this paper emphasis is on the anced by an opposite loop to keep mutual inductance to a mini-
cage type shunt. For comparison, foil shunt construction is briefly mum [7]. The input plate is distanced from the two output plates
described in [2]. In addition to measuring amplitude and phase by the length of the crossbars. This design minimizes inductive
errors of such shunts [3], a proper DC characterization of shunts coupling between the input and the output circuits [8]. Details of
is also important for measurement uncertainty calculation of sam- the construction of the cage type shunt are presented in [9]. Prior
pling power standards and similar applications. The most impor- to the manufacturing, the extensive modelling of the cage and the
tant DC parameters are temperature coefficient, power coefficient dimensions of the particular elements of the cage (thickness of the
and drift, even though some recent experiments show that even PCB-s, the length and the width of the crossbars, dimensions of the
a humidity coefficient should be investigated as it may have con- disks) was performed to obtain optimal dimensions for shunt com-
siderable influence on the measurement uncertainty. ponents. Also, an in-house programmed numerical routine in
MATLAB, based on the method-of-moments (MoM) was developed
for the calculation of the capacitance and the inductance of the
2. Design and manufacturing of UNIZG-FER AC shunts
crossbars, while the dimensions of the disks were chosen as small
as possible. The capacitance of the crossbars is calculated using
At the Faculty of electrical engineering and computing, Univer-
MoM and 2-D approximation, neglecting the thickness of the cop-
sity of Zagreb, a set of cage type AC shunts have been produced [4].
per conductors. The conductor boundaries are discretized into 200
The shunts are based on SIQ design as it is most economic design
finite strips having the uniform surface charge densities, and the
offering reuse of most shunt components such as crossbars and
potentials are calculated using the exact solution of the midpoints
resistors of nominal resistance value to minimize the expenses
of the strips. Finally, the unknown charge densities are calculated
[5] and were fabricated using double-sided FR4 fiberglass-epoxy
using matrix inverse. It is supposed that the conductors are
immersed into a homogenous dielectric having the effective per-
⇑ Corresponding author.
mittivity calculated using the modified Schneider equation for
E-mail address: [email protected] (J. Konjevod).

https://doi.org/10.1016/j.measurement.2018.08.016
0263-2241/Ó 2018 Elsevier Ltd. All rights reserved.
2 J. Konjevod et al. / Measurement 131 (2019) 1–6

the effective permittivity of the microstrip line. For a target charac- Also, the dissipated power on each S102C resistor in each shunt
teristic impedance of the equivalent line, the crossbar width is cal- does not exceed 30% of ambient power rating at nominal voltage
culated using the MoM routine and bisection method, based on the output to minimise the effect of resistor self-heating due to the fact
capacitance calculation. Finally, the inductance of crossbars is cal- that resistors are placed close to each other. Fig. 1 shows double
culated from the capacitance using the lossless transmission line row of Vishay resistors in 10 A shunt during manufacturing. It is
approach. For a target characteristic impedance of the line equal the only resistor that has two resistors per crossbar, all other
to Z0 = 10 X (to match the resistance of the resistors for shunts shunts have one resistor per crossbar. The completed set of shunts
with nominal currents 1 A, 5 A, and 10 A), the width of the crossbar is shown in Fig. 1b. The photo also shows two standards to easily
is calculated as 15.48 mm and its capacitance as C = 721.22 pF/m. connect shunts with PMJTC for AC/DC comparison purpose [12].
The numerical results have been compared with the professional
software Infolytica ElecNet used as benchmark, where the capaci- 3. Determination of temperature coefficients
tance of the crossbar calculated with ElecNet is 699.14 pF/m, thus
giving a good agreement with the difference 3.2%. The model in To begin with analysis the shunts have been photographed with
ElecNet applied the accurate modelling of the electrodes using thermal image camera to determine the temperature distribution
their real thickness. The difference between results can be attribu- and actual temperature at nominal current (Fig. 2) for two shunt
ted mostly to the limited accuracy of the Schneider equation in this types (14 crossbars and 70 crossbars). It was determined that
particular application, since it assumes an infinite dielectric plane. shunts are heated up to 31 °C at nominal current, significantly from
Using a more accurate effective permittivity, calculated in ElectNet 23 °C usual laboratory temperature.
for the obtained crossbar width equal to 15.48 mm, the MoM rou- Next, to determine the actual temperature coefficients (TCR),
tine yields capacitance 697.09 pF/m, which is within 0.2% as com- the shunts were compared against L&N set of resistance standards
pared to FEM. The inductance is obtained from the calculated (1 O–10 kO) with low TCR using two 6,5 digital multimeters in
capacitance as L ¼ C  Z 20 and it equals 72.12 nH/m. This numerical usual DVM type resistance comparison method as is discussed in
result has been also compared with the professional software literature for the similar measurements [13]. The method is cap-
Infolytica MagNet used as benchmark, where the inductance of able to compare two resistors with needed precision with simulta-
the crossbar calculated with MagNet is 72.25 nH/m, thus giving a neous triggering. Each shunt was matched with resistance
good agreement with the difference 0.18%. The actual width of standard with closest nominal resistance to tested shunt. The used
the crossbars was finally, due to the manufacturing reasons DVMs have 24-bit analog to digital converters (ADS1256) and pre-
rounded to 15 mm. cise voltage reference type ADR421; 2.5 V that can replace usual
The permittivity of the PCB material was measured over the Keysight 3458A 8,5 digitsmultimeters in DVM method for this pur-
wide frequency range using a LCR-bridge and two rectangular sam- pose [14]. The system to measure shunt TCR consists of two tem-
ples. For that purpose, two parallel-plate capacitors were cut from perature regulated air baths capable to set up temperature in
FR4, one of them having a 1 mm thickness and the other one hav- 10–40 °C range independently, DVM resistance method, current
ing a 2 mm thickness. For the crossbars (their thickness is 1 mm), a reversal module using two SPDT OMRON G5LA-14 5 V power
137 mm  148 mm sample was cut, and its capacitance was mea- relays controlled by National Instruments 6008 DAQ card, Rigol
sured using a Rohde&Schwarz (Hameg) HM8118 LCR bridge, being DP1116A DC power supply with a switchable output up to
890.5 pF at 1 kHz, with the accuracy as defined by manufacturer 16 V/10 A or 32 V/5 A and computer that serves as a controller to
equal to 0.11% at this frequency and impedance. Using the fully automate the measurement process as shown in Fig. 3. The
parallel-plate capacitor approximation, the permittivity of cross- measurement procedure was programmed using LabVIEW pro-
bars material is 4.96. To determine the possible error caused by gramming language.
the parallel-plate approximation, the calculation of the capacitance The thermal air baths are specially developed for this purpose
of the sample model was run in the finite-element-method (FEM) based on Ezetil E21 Electric Cool Boxes with capacity of 20 L. These
commercial package Maxwell. For the model with the permittivity boxes have 12 V Peltier elements and their basic purpose is cooling
equal to 4.96, the FEM-based calculated capacitance was 896.42 pF, only. As Peltier elements have ability to cool and heat, depending
thus giving the difference 0.67%. on polarity of DC voltage, the boxes were modified to get the ability
For the disks (thickness is 2 mm), a 157 mm  168 mm sample of cooling and heating of air inside air box. Peltier elements are
was cut. Its measured capacitance is 565.21 pF at 1 kHz, with the controlled separately from the fan that works on constant DC
accuracy as defined by manufacturer equal to 0.12% at this fre- 12 V, and they are driven by two double H-bridges (two L298 ICs
quency and impedance. Using the parallel-plate approximation, connected in parallel) which are controlled by Arduino Uno micro-
its permittivity is 4.84. The FEM based calculated capacitance controller boards. Each air bath has own microcontroller that mea-
was 571.39 pF, thus giving the difference 1.09%. sures temperature inside box with NTC temperature sensor and
In total, nine AC shunts have been manufactured: 1 mA, 10 mA, control Peltier element depending on measured temperature and
30 mA, 100 mA, 300 mA, 1 A – two shunts, 5 A and 10 A. The values desired temperature that is defined by user. Temperature of air
of shunts were chosen so it is possible to calibrate them with usual bath is controlled in range of +10 °C and +40 °C with resolution
step up method starting from 10 mA Planar Multijunction Thermal of 0,01 °C by PID algorithm that runs on Arduino microcontroller
converter (PMJTC) current standard obtained from Leibniz- board. This range is possible with room temperature set to 23 °C.
Institute for Photonic Technology Jena (Leibniz-IPHT) up to 10 A. Arduino microcontroller accepts commands over USB cable
Resistance values of developed shunts are well suited for use with (RS232 protocol over USB) from application in LabVIEW that run-
these PMJTCs which have 90-O heaters with a rated current of son PC. The whole TCR measurement system is shown in Fig. 4.
10 mA [10].The high precision Vishay S102C foil resistor have tem- The fully automated procedure allows measurement to be per-
perature coefficient of ±2.0 ppm/°C, and it also features low induc- formed when temperature stabilizes to predefined temperatures of
tance, low capacitance, low power coefficient, stability and low both baths. The TCR measurements have been performed in 18 dif-
noise [11]. The resistor is also available in wide range of nonstan- ferent temperatures ranging from 18 °C to 39 °C to cover tempera-
dard values. The shunts were designed to provide peak voltage out- ture range of shunts. The whole procedure lasts approximately four
put of approximately 1 V at nominal current, to suite most modern hours, and each measurement was repeated at least twice to check
precision DMMs, AC measurement and AC/DC transfer standards. the repeatability of measurements. To reduce the self-heating
 J. Konjevod et al. / Measurement 131 (2019) 1–6 3

Fig. 1. Double row of Vishay resistors in 10 A shunt on a single plate during assembly on the left and complete set of shunts on the right.

Fig. 2. Thermal image of two types of shunts at nominal current (a) Shunt with 70 crossbars, (b) Shunt with 14 crossbars.

Fig. 3. TCR measurement procedure.

Fig. 4. TCR measurement system.

during TCR measurements the shunt is powered to less than 10% of All reference standard resistors (L&N Rosa (NBS) type produced
nominal current and the reference resistor is powered to maxi- from 1963 to 1966) used for comparison have temperature coeffi-
mum of 100 mW. The reference resistance standard is kept in sep- cient a23 less than 10 mO/O/°C. The results of TCR measurement for
arate regulated air bath at constant temperature of 23 °C ± 30 mK. shunts have shown that TCRs for all shunts are in the range from
4 J. Konjevod et al. / Measurement 131 (2019) 1–6

Fig. 5. Measurement of 0,1 A TCR (temperature has been changed from 15 °C to 39 °C with standard deviation of each measurement).

Table 1
Temperature coefficients.

Shunt TCR [lX/X/°C]

1 mA 0,6
10 mA 0,7
30 mA 4,3
100 mA 1,3
300 mA 1,5
1A 2,8
5A 6,5

1,5 mO/O/°C to 6,2 mO/O/°C. The Fig. 5 shows typical TCR measure- Fig. 6. Maximum 3458A integration error on DC range.
ment for 0,1 A shunt, which is similar to all other shunts.
The total uncertainty contribution of TCR depends on the data fit
uncertainty calculated with Gauss-Newton fitting program [15], The maximum integration error (1 V RMS superimposed on 0,1
non-homogenous air bath temperature and temperature measure- V DC) can be calculated as:
ment accuracy itself and it is estimated to be 0,5 lX/X/°C for all 1V max 1 1
shunts. Calculated TCRs of shunts are presented in Table 1. Pmax ¼ :  ð1Þ
1; 11 2fT 0; 1V
With the AC frequency f = 10 kHz, and the integration time of
4. Determination of power coefficients
T = 1 s, the maximum error is pmax = 450 lV/V which is significant
for high precision measurements. However, with double RC filter
Power coefficient measurement was conducted in the same
connected to the input of voltmeter, the influence of AC current is
setup as described above using one reference resistance standard
lowered according to expression:
and one shunt, both enclosed in separate air baths and kept at con-
stant temperature of 23 °C. The main challenge in this setup is to U out 1
¼ ð2Þ
increase the power dissipation of shunt while keeping the power U in ð2pfRCÞ
dissipation of standard resistor constant. This was solved by sup-
plying AC current (10 kHz frequency) to the shunt using arbitrary With the selection of R = 10 kX and C = 1lF, the ratio Uout/
generator (Rigol DG4062) and power amplifier (Rigol PA1011), Uin = 2,53*106. Accordingly, the maximum integration error is then
while keeping DC current passing through standard resistor and pmax = 0,00114 lV/V which is practically negligible. Thus, the
shunt constant at 10% of nominal current of measured shunt. In shunt/standard resistor ratio will change only because of shunt
this way it was possible to power shunts (I = 1 A; I = 5 A and resistor change due to added power dissipation in each step of
I = 10 A) with additional AC current to obtain several measuring the measurement process. To enable extra precision, DVMs used
points: 50%, 60%, 75%, 90% and 100% of nominal AC shunt current for this measurement are HP3458A. PCR value determined for 1 A
ratings. The reference resistors are powered to maximum of shunt is (8,6 ± 0,5) mO/O/W. The whole PCR measurement system
10 mW during measurement to minimize the heating of reference is shown in Fig. 7.
resistors. In addition, the capacitor is connected on the power The calculated temperature coefficients of both reference resis-
amplifier to avoid shunting the resistor shunt with the function tors and shunts were taken into account to obtain more precise
generator. Further, voltmeters on reference resistor and shunt are values of shunt power coefficients. The PCR measurement for
connected through double RC filter used to cancel AC component shunts below 0,3 A rated current have not been measured, as the
during DC voltage measurement. The voltmeter measuring the PCR for these shunts are negligible due to very low power dissipa-
standard resistor does not need to have RC filter as there is no tion on Vishay resistors that make the resistive elements of the
AC current passing through standard resistor but it was added to shunts.
equalize the input characteristics of both voltmeters [16]. The used
voltmeter is cancelling only full periods on DC range, so the max- 5. Drift measurement
imum error during voltmeter integration time T will be when inte-
gration time lasts for n + 1/2 periods of AC current with frequency f Drift investigation was conducted in the same setup as for TCR
(Fig. 6). and PCR using one shunt (I = 1 mA, R = 714,29 O) and one reference
 J. Konjevod et al. / Measurement 131 (2019) 1–6 5

Fig. 7. PCR measurement system (photo and electric diagram).

resistance standard (R = 1000 O), both enclosed in each air bath 23 °C and shunt with nominal current of 1 A is kept in second ther-
and kept at constant temperature of 23 °C. The resistance value mal chamber with equal temperature of 23 °C where relative
of the shunts was measured over a three months period in order humidity has been regulated from 29% up to 85%. Fig. 8 shows that
to determine the long-term drift. The measurement procedure DC resistance is negligibly changed when RH is increased from 29%
was repeated several times and consists of approximately three to 35%. Mean values are overlapped inside standard deviation lim-
weeks of intensive measurement performed each hour to discern its for both RH coefficients. Measurements are carried out eighteen
any day-night pattern from the drift itself. Drift is only measured times where each measurement last three minutes to satisfy
for shunt resistor with nominal current 1 mA. The calculated drift repeatability. Rosa type standard have resistance element sealed
value is 0,238 mO/O/year. in a can filled with mineral oil, and thus very low humidity
coefficient.
6. Humidity coefficient measurement Further, to investigate behaviour in high humidity environment,
RH coefficient is increased to 65% and 85%. Fig. 9 shows change of
Recently it was reported in [17] that the shunts have a relative DC resistance for three different RH coefficients. Samples are
humidity (RH) dependence in the order of one mO/O per % with a approximated with regression lines and for RH coefficient values
time constant in the order of ten days so to investigate that claim, of 30 and 65 results are practically overlapped. But, significant
the humidity coefficient was measured using the similar setup for deviation is obtained at humidity level of 85. In that case, dc resis-
determining PCR and TCR coefficients. In this setup, standard resis- tance has changed in the order of about 1 mO/O/RH% which is
tor is kept in enclosed regulated chamber with temperature at shown in Fig. 9.

Fig. 8. Relative humidity dependence for RH 29% and 35%.

6 J. Konjevod et al. / Measurement 131 (2019) 1–6

Fig. 9. Relative humidity dependence for RH 30; 65 and 85.

7. Conclusion [4] G. Rietveld, J.H.N. van der Beek, E. Houtzager, DC characterization of AC current
shunts for wideband power applications, IEEE Trans. Instr. Meas. 60 (2011)
2191–2194.
The DC characterization of FER AC current shunts with current [5] R. Malaric, M. Dadic, P. Mostarac, D2.2, Report on designing, developing and
range from 1 mA to 10 A are presented. The shunts have been testing current shunts, Deliverable D.2.2. of Croatian Science Foundation
project SMAGRIMET (2016) 1–20 [in Croatian].
developed for precise power measurement applications, and the [6] B. Voljc, M. Lindic, R. Lapuh, Direct measurement of AC current by measuring
results have shown that the shunts have low drift, low TCR and the voltage drop on the coaxial current shunt, IEEE Trans. Instr. Meas. 58
PCR which make them suitable for the intended purpose. It was (2009) 863–867.
[7] S. Svensson, A wattmeter standard for the audio frequency range, IEEE Trans.
also shown that the heating of the shunts by the measurement cur-
Instr. Meas. 48 (1999) 431–433.
rent is distributed very homogenous over the circle plate where the [8] B. Voljc, M. Lindic, B. Pinter, M. Kokalj, Z. Svetik, R. Lapuh, Evaluation of a 100 A
Vishay resistors are placed. According to this fact, there is direct current shunt for the direct measurement of AC current, IEEE Trans. Instr.
correlation between the power coefficient and the temperature Meas. 62 (2013) 1675–1680.
[9] Piotr S. Filipski, M. Boecker, AC–DC current, shunts and system for extended
coefficient of the shunts. Considering further investigations on rel- current and frequency ranges, IEEE Trans. Instr. Meas. 55 (2006) 1222–1227.
ative humidity dependence, improvements can be achieved using [10] K. Lind, T. Sørsdal, H. Slinde, Design, modeling, and verification of high-
digital voltmeters with higher resolution. For this purpose, we performance AC–DC current shunts from inexpensive components, IEEE Trans.
Instr. Meas. 57 (2008) 176–181.
can use HP3458A as in the PCR measurement. Besides, current [11] Vishay Precision Group. High Precision Foil Resistor with TCR of ±2.0 ppm/°C,
shunt can be placed in humidity and temperature chamber which Tolerance of ±0.005% and Load Life Stability of ±0.005%, URL:
has finer regulation of RH coefficient. <https://datasheet.octopart.com/S102C-10K-0.01%25-Vishay-datasheet-
8852715.pdf/>.
[12] M. Klonz, L. Laiz, T. Spiegel, P. Bittel, Ac-dc current transfer step-up calibration
Acknowledgment and uncertainty calculation, IEEE Instr. Meas. Technol. Conf. (2001) 1435–
1440.
[13] V.N. Zachovalova, M. Sira, J. Streit, L. Indra Daejeon, Korea, in: Measurement
This paper is fully supported by Croatian Science Foundation
System for High Current Shunts DC Characterization at CMI, Precision
under the project Metrological infrastructure for smart grid Electromagnetic Measurements (CPEM 2010), 2010, pp. 607–608.
IP-2014-09-8826. [14] H. Hegedus, J. Konjevod, P. Mostarac, R. Malaric, Comparison of DMMs based
on RMS noise and standard deviation noise Smolenice, Slovakia, in: J. Manka,
M. Tysler, V. Witkovsky, I. Frollo (Eds.), Proceedings of the11th International
References Conferenceon Measurement, 2017, pp. 111–115.
[15] K. Malaric, R. Malaric, H. Hegedus, A nonlinear least-squares graphical tool
[1] T. Bergsten, V. Tarasso, K.E. Rydler, A calibration system for electric power at (‘Gaussfit’) for educational purposes, Int. J. Electr. Eng. Educ. 47 (1) (2010) 11–
frequencies up to 1 MHz, 17, Int. Cong. Metrol. (2015). 22.
[2] M. Garcocz, G. Heine Scheibenreiter, W. Waldmann Rio de Janeiro, Brazil, in: [16] R. Malaric, I. Lenicek, M. Kresic, Measurement of Resistor Load Coeficients, XVI
Progress on BEV Foil Shunts for Ac-dc Current Transfer, Precision IMEKO World Congress, Vienna, AUSTRIA, 2000, September 25-28; pp. 211–
Electromagnetic Measurements (CPEM 2014), 2014, pp. 272–273. 214.
[3] I. Budovsky, Measurement of phase angle errors of precision current shunts in [17] S. Svensson, K.E. Rydler, V. Tarasso Ottawa, Canada, in: Upgrade of Accuracy
the frequency range from 40 Hz to 200 kHz, IEEE Trans. Instr. Meas. 56 (2007) and Traceability the SP Sampling Wattmeter, Conference on Precision
284–288. Electromagnetic Measurements (CPEM 2016), 2016, p. 2.