Th-PS2 128752 1

Resistive Dividers for Wide-Band Voltage and

Power Meters
D. Slomovitz, Senior Member, IEEE, L. Trigo, Senior Member, IEEE, G. Aristoy, Member, IEEE,
M. Brehm, Member, IEEE, C. Faverio, G. A. Kyriazis, Senior Member, IEEE

Abstract—This paper describe the design and construction of TABLE I. Characteristic details of the set of the dividers.
resistor based voltage dividers for using at the input of electronic
Input Input
power quality meters. They are intended for working with Power
voltages up to 1000 V and harmonic frequencies up to 100 kHz. voltage Ratio resistance
(W)
As not only voltage, but also power must be measured, the (V) (kΩ)
accuracy requirements involve ratio and phase shift. Special
1024 1280 256 4.10
resistors are used with very low dielectric losses. Electrostatic
shields avoid the influence of parasitic capacitances extending the 512 640 128 2.05
bandwidth. 256 320 64 1.02
128 160 32 0.51
Index Terms— distortion measurement, phase shift error, 64 80 16 0.26
power, ratio error, resistor, voltage divider.
32 40 8 0.13
16 20 4 0.06
I. INTRODUCTION 8 10 2 0.03

D UE to development of power electronic, the need to

measure current, voltage an power under very distorted
waveforms has increased. PWM modulation and other similar
4 5 1 0.02

For all cases, the output resistance is 200 Ω and the nominal
techniques generate high frequency harmonics up to tens of output voltage is 0.8 V, so that the current is 4 mA. The ratios
kilohertz. There are power meters in the market for measuring were selected with binary ratios to allow the calibration
this kind of waveforms, but National Standards are necessary against a standard binary inductive voltage divider [2] at low
to calibrate them. A project for designing and constructing a frequencies, and for covering the whole voltage range with the
standard meter for measuring voltage, current and power, has same step-ratio between each adjacent pair. The input arms are
been jointly developed by the National Metrology Institutes of constructed with multiple individual resistors in parallel-series
Argentina (INTI), Brazil (INMETRO) and Uruguay (UTE) connection to limit the power of each one at 100 mW.
[1]. The project was divided in different blocks, one of them is Precision requirements are: ratio error lower than 30×10-6 in
the divider described in this paper. It is required to cover the frequency range 50 Hz to 100 kHz and phase displacement
voltage ranges from 4 V to 1024 V. For that, nine resistive lower than 60 µrad/kHz. To achieve these values, non-linear
voltage dividers are used, one at a time, with binary nominal losses and stray capacitances must be considered as main error
ratios. Table I shows details. The input resistance and the sources. Both effects are disused in following sections.
dissipated power depend on the input voltage. Previous proposals for wideband dividers involve shields
using partial or total screens and guarding dividers. Although
these proposals obtain good results in limited bandwidth
This work was partly supported by the Conselho Nacional de applications, they have backwards for the application
Desenvolvimento Científico e Tecnológico (CNPq), of the Ministry of
Science, Technology and Innovation of Brazil, under Grant CNPq/Prosul described in this work. Partial shield proposal consists in
Processo Nº 490271/2011-1. cover the part of the resistor more close to the input terminal
D. Slomovitz is with UTE Laboratory, Montevideo, Uruguay (e-mail: with a cylindrical shield connected to the input. An external
[email protected]).
L. Trigo is with UTE Laboratory, Montevideo, Uruguay (e-mail: coaxial screen connected to ground covers the entire device.
[email protected]). Fig. 1 shows an example of this technique. The external
G. Aristoy is with UTE Laboratory, Montevideo, Uruguay (e-mail: cylindrical screen was removed to view the internal shield. It
[email protected]).
M. Brehm is with UTE Laboratory, Montevideo, Uruguay (e-mail: is placed near the input terminal and connected to it. In this
[email protected]). way, the capacitances between the resistor and the input are
C. Faverio is with UTE Laboratory, Montevideo, Uruguay (e-mail: increased, compensating the stray capacitances to the external
[email protected]).
G.A. Kyriazis is with Instituto Nacional de Metrologia, Qualidade e
cover, at ground potential. The divider can be modeled,
Tecnologia, Brazil (e.mail: [email protected]). approximately, by a two time constants circuit; and the length
 Th-PS2 128752 2

of the shield can be adjusted to get equal values for these time
constants. In this way, from a theoretical point of view, the
frequency response would be ideally flat. However, although
this system works well up to 100 kHz with ratio differences
below 50×10-6, it has large errors in phase displacements. A
better modeling must use distributes parameters instead of Fig. 2. Resistors with and without isolating cover.
concentrated ones. This technique does not reduce radial stray
currents, but compensates them. Anyway, it is used in ac-dc They were tested using ac-dc thermal converter against a
thermal converters and other applications where phase standard resistor. For this test, the resistors were mounting
displacements do not have any influence on the result. with no PBC, directly fixed to the measuring structure. Table
II shows the ac-dc differences. The resistor with cover reduces
its impedance 48 µΩ/Ω when the frequency goes from 53 Hz
to 100 kHz, whereas the resistor without cover increases its
impedance 19 µΩ/Ω. This last variation is due to a small stray
capacitance to ground. Part of the input current was shunted to
ground without passing through the thermal converter. In case
of the covered resistor, the increase of the dielectric losses is
much greater than the effect of the stray ground capacitance
Fig. 1. Partial shielded resistive voltage divider. resulting in a reduction of the apparent impedance.

Complete screens dividers are used by other proposals. TABLE II. Variation of impedance versus frequency of resistors with
and without cover.
To compensate the capacitive currents between the resistor
and the screen the screen potential is adjusted by an Frequency With cover Without cover
auxiliary capacitive divider [3], or using fix capacitors [4]. (Hz) (µΩ/Ω) (µΩ/Ω)
Other approach to eliminate the influence of stray 53 -2 3
capacitances is to add a guarding divider close to the main
1000 -1 1
one in order to reduce the radial electric field [5]. This
auxiliary divider drives many partial screens at potentials 10000 -7 5
close to the main resistor potentials. In this way the radial 50000 -28 15
electric field is reduced as well as the currents though the
100000 -50 22
radial stray capacitances. This technique works well up to
medium bandwidth but at the frequencies needed in this
work the non-ideal-behavior and parasitic elements
A similar test was done with the non-covered resistors
introduced by the shielding divider degrade the results.
mounted in different PCBs. In this case, PTFE PCB which has
In [6] a new technique is proposed that uses two
low dissipation factor at high frequency was used. It was
symmetrical shields nulling the radial electric field on the
compared with the behavior of the resistor mounted without
whole surface of the main resistor divider. PCB (see Table III). The differences are below the
measurement uncertainty, denoting that the PTFE PCB does
not increase the non linear losses.
II. NON-LINEAR EFFECTS
In high value resistors most non-linear effects are due to TABLE III. Variation of impedance versus frequency of resistors
dielectric power losses in their isolating cover. Other non- mounted on PTFE PCB and without any PCB (air mounted).
linear sources, as skin effect are negligible. As that losses Mounted on
depends on the frequency, the net effect is the variation of the Air mounted
f (Hz) PTFE PCB
(µΩ/Ω)
impedance when the frequency changes. At higher (µΩ/Ω)
frequencies, the apparent impedance value decreases due to
53 3 2
extra losses. To reduce this effect, special resistors without
cover were used [7]. Tests on very stable resistors with 1000 1 1
temperature coefficients lower than 10-6 K-1of type Vishay Z- 10000 5 2
Foil type, 10 kΩ, 0.6 W, were done measuring the variation of 50000 15 8
their impedance against frequency. One is a conventional 100000 22 23
resistor, and the other is the same type but without cover (see
Fig. 2).
III. SHIELDING
Once nonlinear effects have been reduced, stray
capacitances must be controlled to get frequency plane
response. Capacitance to ground is generally larger than the
 Th-PS2 128752 3

resistors parallel capacitance. For compensation, one simple

technique consists in adding a parallel capacitor to the input
arm in order to have the same time constant in both arms.
Although this works well in low precision dividers, it is not
easy to use it in high precision dividers. That capacitor has
dielectric losses and stray inductances that affect the ratio at
high frequencies. Other proposals were described in pervious
section.
In this work we propose a different shielding technique with
the aim of nulling radial electric fields. Two symmetric cone-
shaped electrostatic shields are mounted at each end of the Fig. 5. Interior view of the input resistor arm showing the
resistor to null radial electric fields at the whole surface of the cone shields.
resistor. The cone at the input is connected at the applied
voltage and the cone at the output is connected to ground. Fig.
3 shows a schematic draw and a vector representation of the
IV. EXPERIMENTAL RESULTS
electric field made by simulation software. The directions of
the vectors around the central resistor (tube shape) are parallel Errors and thermal effects were experimental determined.
to the resistor. The dimensions of the shields were determined Two methods were used to test the ratio error and phase shift
using the simulation software. of the dividers. The first one compares the divider under test
against a standard ac-dc thermal converter [8]. It only tests the
modulus of the ratio error, as it does not give information on
phase displacements. The other method is based on the step
response and gives the ratio error as well as phase angle error.
A. AC-DC Method
The variation of the modulus of the voltage ratio, when the
frequency varies between 50 Hz and 100 kHz, was measured
against a standard voltage thermal converter. Table IV shows
Fig. 3. Electric field simulation for the proposed shield.
the ac-dc differences for the prototype. The ac-dc error varies
between -15 and 35 parts in 106 for the whole frequency
range, which is nearly close to the requirements. The
A more detailed simulation was done using the actual uncertainties, k=2, are detailed in the third column.
geometry of the input resistor of the divider. Fig. 4 shows the
TABLE IV. Ratio errors by ac-dc testing method
resistor corresponding to 512 V. It has 30 individual units of
Ratio
38.4 kΩ in three parallel arms of 10 resistors each. Fig. 4 f (Hz) error
Uncertainty
shows this set mounted in a PTFE PCB, and Fig. 5 the cone -6 (×10-6)
(×10 )
shields.
50 -8 17
1000 -15 16
10000 -7 16
50000 18 18
100000 35 23

Fig. 4. Input arm of the 512 V divider.

B. Step response
The external cylindrical aluminum cover was taken out to The step response method consists in applying a very fast
see the interior shields. A small adjust in the position of the step voltage at the input of the divider, recording the output
resistor inside the shield is possible to compensate for stray voltage waveform g(t). This is a well-known technique for
capacitances in the low voltage output arm. It is done moving linear devices. From this method ratio error ε as well as phase
the two screws that support the PCB. shift δ can be computed [9-10]. The equations to get those
values are the following.

  (1)

where ω is the angular frequency and T is defined by the

following equation

T   1  gt dt (2)
 Th-PS2 128752 4

V. CONCLUSION
The ratio error is calculated according to
ACKNOWLEDGMENT
ε = -Ts2ω2/2 (3) This work was partly supported by the Conselho Nacional
where
de Desenvolvimento Científico e Tecnológico (CNPq), of the
 Ministry of Science, Technology and Innovation of Brazil,
   2  1       (4) under Grant CNPq/Prosul Processo Nº 490271/2011-1.
For generating the step voltage, a dc current source is
connected to the input of the divider and a mercury-wetted-
relay is used to short-circuit this point to ground [11]. The REFERENCES
transition to zero has a very short time, lower than 1 ns. As the
recorder, a 10 bits 400 MHz digital oscilloscope was used, and [1] Kyriazis G A, Di Lillo L, Slomovitz D, Iuzzolino R, Yasuda E, Trigo L,
its data was processing by software that computes T, Ts2, ε and Laiz H, Debatin R, Franco A M R, Afonso E, “Trilateral South
American project: a reference system for measuring electric power up to
δ. Fig. 6 shows the response of the tested divider. In this case,
100 kHz –– progress report,” IOP Journal of Physics: Conference Series
T=-7.9 ns and Ts2=-1.6×10-18 s2, which leads to ratio error (JPCS), to be published.
ε=31×10-6 with uncertainty of 51×10-6 k=2, and phase [2] Slomovitz D, Faverio C, Izquierdo D, Trigo L, Castet C, “An
displacement δ=5000 µrad with uncertainty of 2000 µrad k=2, electronically assisted binary inductive voltage divider,” 2006,
Conference on Precision Electromagnetic Measurement, CPEM 2006
both at 100 kHz.
DIGEST, 274, Torino, Italy.
Ratio errors calculated by both methods are compatible, [3] G. A. Kyriazis; A. M. Ribeiro Franco; R. M. Debatin; W. G. Kürten
taking into account their uncertainties. Ihlenfeld, “Precise resistive voltage divider for harmonic
measurements,” Conference on Precision Electromagnetic
Measurements (CPEM), pp. 140 - 141, 2012.
[4] M. Grubmülle, B. Schweighofer, H. Wegleiter, “Characterization of a
resistive voltage divider design for wideband power measurements,”
IEEE SENSORS 2014 Proceedings, pp. 1332 – 1335, 2014.
[5] Rydler K E, Svensson S, Tarasso V, “Voltage dividers with low phase
angle errors for a wideband power measuring system,” 2002, Conference
on Precision Electromagnetic Measurements, CPEM 2002 Digest, 382.
[6] D. Slomovitz, L. Trigo, G. Aristoy, M. Brehm, “Resistor Dividers for
Power Quality Meters,”Conference on Precision Electromagnetic
Measurements (CPEM), 2016.
[7] http://www.vishaypg.com/docs/63140/var.pdf
[8] Hermach F, Williams E, "Thermal voltage converters for accurate
voltage measurements to 30 megacycles per second," 1960, AIEE
Trans., pt. 1, 79, 200.
Fig. 6. Step response of the 512 V divider. [9] Rungis J, Schon K, "The evaluation of impulse divider response
parameters," IEEE Trans. Pow. Del., 3, No 1, pp. 88, 1988.
[10] Slomovitz D, Bergalli P, Daher J F, “Evaluation of low-frequency
C. Thermal effects voltage dividers, using the step response,” 1998, Conference on
Precision Electromagnetic Measurements, CPEM 1998 Digest, 319.
The power dissipated for the resistors increases their [11] Fernandez J, Slomovitz D, "Desarrollo de un generador de step rápido
temperature changing their values. But, as the thermal para pruebas de divisores de tensión de impulso,", Electrotécnica, 81,
coefficients of this type of resistor is lower than 1 (µΩ/Ω)/K, pp. 233, 1995
the resistance variation in under the maximum project
tolerance. For the 512 V divider, it was measured a ratio
variation of 13×10-6 in 10 min with a time constant of 3.5 min.
These values can be reduced adding force air with a small fan.
Although it is not necessary for the tested divider, it must be
for the 1024 V divider. The fan will be installed in the output
side of the external cylinder and the air flux will be conducted
to the resistor though small holes in both cones.