Jpn. J. Appl. Phys. Vol. 41 (2002) pp.

39873993
Part 1, No. 6A, June 2002
#2002 The Japan Society of Applied Physics

Double Force Compensation Method to Enhance the Performance of a Null Balance Force Sensor
In-Mook C HOI1 *, Dong-June C HOI2 and Soo Hyun K IM3
1

Korea Research Institute of Standards and Science, P.O. BOX 102, Yusong, Taejon 305-600, Republic of Korea
LG-Elite,16 Woomyeon-dong, Seocho-Gu, Seoul 137-724, Republic of Korea
3
Department of Mechanical Engineering, Korea Advanced Institute of Science and Technology
373-1 Kusong-dong, Yusong-gu, Taejon 305-701, Republic of Korea
2

(Received August 15, 2001; accepted for publication February 6, 2002)

Microforce measurement is becoming more essential as precision industries such as biomedicine, precision chemistry,
semiconductor manufacturing, and so forth develop. A null balance method has been introduced in order to improve on force
measurement performances involving a loadcell. The null-balance type force sensor is analyzed and designed for the
improvement of measurement performances. The measurement range and the resolution are dependent on the force generation
capacity and the various error sources. These characteristics are estimated and veried according to the mechanical sensitivity
and the force compensation sensitivity. Two dierent coil systems are designed and tested experimentally. Double force
compensation is proposed in order to obtain a large range and high resolution. The measurement range of the large coil system
and the resolution of the small one are fully realized by the double compensation method. After manufacturing, a range over
300 gf and resolution under
0:1 mgf were obtained by the double compensation method. [DOI: 10.1143/JJAP.41.3987]
KEYWORDS: force measurement, sensitivity, null balance, double compensation, electromagnetic force

1.

Introduction

Microforce measurement is becoming more important as

precision industries develop such as biomedicine, precision
chemistry, semiconductor manufacturing, and so forth.
Moreover, improved performances related to the measurement range and to resolution are required.
There are two basic methods for measuring microforce.
One involves a loadcell, which is relatively low cost and
can be manufactured easily. All systems that change
mechanical deformations under applied forces to electrical
signals are called loadcells. Since a loadcell has nonlinearity
with respect to the mechanical deformation itself and is
sensitive to temperature variation, it is dicult to obtain
high resolution through structural modications.
The other is a null balance method, which has been
introduced as an alternative to overcome the limitations of
the loadcell resolution. Lorentz force is used to prevent
mechanical deformation by an exterior force. The null
balance method, which uses magnetic force compensation,
has good linearity and is unrelated to zero-point shift under
overloading. But it cannot measure dynamic input force and
is sensitive to temperature variation, as is the case with other
measurement instruments.1,2)
In addition to these two methods, various force measurement methods have been introduced, but they are not
appropriate for measuring microforces under 10 mgf.35)
In the null balance method, mechanical characteristics are
very important since they are related to the improvement of
characteristics of the system performance such as measurement range, resolution, response time and so forth. The
system performance realized by the mechanism can be
degraded by the control error and noise. To recover the
decreased performance easily, double force compensation is
proposed in this paper. This method uses dierent
compensation characteristics provided by two electromagnetic systems.
Before the double force compensation method is proposed, the system characteristics particularly associated with

the position sensor, the electromagnetic system and the

control, are analyzed and designed.
In 2, the general principle and the null balance force
measurement system are introduced. In 3, each component
is analyzed and designed in order to improve the
characteristics. In 4, the double force compensation method
is introduced and designed. After the double coil system is
constructed, the performance is tested in order to verify the
design and analysis. Finally, conclusions are drawn in 5.
2.

Force Sensing System

Figure 1 shows the schematic diagram of a conventional

micro-weighing device of the null balance type.
If an object is loaded on the weighing pan, the parallel
spring moves vertically. The vertical displacement is
transmitted to the lever mechanism through the exure
hinges and the lever rotates counter clockwise. The position
sensor detects the displacement of the lever end. This
detected sensor signal is applied to the controller after a
simple computation and the current is applied to the coil
until the position reaches zero. The generated Lorentz force
prevents the lever from rotating. When the signal of the
position sensor converges to the xed error bound, the
applied coil current is proportional to the applied input force.
The system resolution has many relevancies to the
characteristics of the position control because the mechan-

E-mail address: mookin@kriss.re.kr

Fig. 1. Conventional micro-weighing device.

3987

3988

Jpn. J. Appl. Phys. Vol. 41 (2002) Pt. 1, No. 6A

ism is always led to the zero position.

This force measurement system can be divided into a
mechanism part with a parallel spring and an amplication
mechanism, an electromagnetic part, and position sensor and
control components. The characteristics of each part can be
explained briey as follows.
The parallel spring is related to the position change of the
vertical input force, that is, the o center error. When a force
is applied to the center of the weighing pan or its edge, the
sensor outputs must always be the same. However, the
additional moment generated due to the force being o
center gives dierent results. The parallel spring prevents
the additional moment from transmitting into the lever
mechanism.
The amplication mechanism, that is, the lever mechanism, aects the characteristics of the measurement range.
When the force generation ability of an electromagnetic
system is uniform, the measurement range is dependant on
the lever ratio. If the lever ratio is 1 : 10 and the capability of
the electromagnetic force is over 1 N, the system can
measure forces over 10 N. As the lever ratio increases, the
measurement range increases.
However, since the parallel spring and the amplication
mechanism aect each other, these mechanisms must be
analyzed together.
In the electromagnetic part, the magnetic ux density
must be maximized while there is no saturation in the yokes.
The number of coil turns is related to the force compensation
sensitivity. Many coil turns can increase the compensation
force, but the resolution decreases because of the voltage
measurement limitation caused by noise.
Since the position control result is very important in the
null balance method, the sensitivity of the position sensor
must be increased. The measurement resolution of the lever
displacement is much more important than the range.
Generally, gap or optical sensors are used in the null
balance method.
In the control part, the steady state error must be
eliminated. Therefore, an integral action must be included
in the controller such as proportional-integral (PI). Moreover, in order to improve the readability of the system, lter
optimization design is required. A low-pass lter with a very
low cuto frequency is used because the measurement result
is meaningful only in the steady state.
The above-mentioned components aect the system
performance together and the mechanism part is particularly
important. If the mechanism design does not satisfy the
system requirements, even though other components may be
excellently constructed, the system performance cannot be
improved.
In this paper, the electrical part and the control part are
analyzed and designed with regard to the mechanical
sensitivity and the force sensitivity. The measurement range
and the resolution realized by the mechanism can degrade
the performance of the system. These performance degradations can be compensated by double force generation.
3.

System Characteristics

Before the control and the ltering are applied to the

system, the characteristics of each system component must
be analyzed experimentally.

I.-M. CHOI et al.

The resolution of the measurement system can be

estimated from mechanical sensitivity and the range from
force sensitivity. To increase the mechanical sensitivity, a
more sensitive position sensor is needed. Also, to increase
the force sensitivity, design of the electromagnetic component must be carried out.
After verication and estimation of the system characteristics, accurate measurement results can be obtained by
applying control and ltering.
3.1 Position sensor
The position sensor must be designed based not on the
measurement range, but on resolution. Also, the linearity of
the sensor is not important since the mechanism is always set
to the zero position. The ability to detect the variation of
input force can determine the resolution of the force
measurement. Gap or optical position sensors can be used
in the null balance method in consideration of these views.
In this paper, a two-element photodiode (PD) and a laser
diode (LD) are used, as shown in Fig. 2.
If the lever arm is rotated by input force, the slit moves up
or down. The displacement by force can be obtained from
the voltage dierence between the two photodiode cells
according to the slit movement. The optical position sensor
can yield about 0.1 m resolution in general. However, there
are a number of problems that must be solved in order to
obtain the displacement resolution. First, the variation of the
beam intensity from the laser diode brings about a drift
problem. The drift degrades the performance of the position
sensor. Second, the parasitic motion of the slit is detected by
the characteristic of the beam diusion. Third, there is the
phenomenon of the center movement of the beam.
The drift problem can be solved by normalizing and
amplifying the voltage signal, as in eq. (1):
VA  VB
Vs 10
:
1
VA VB
A nonlinear operation chip, such as a divider, is used for the
normalization.
The detection problem of the parasitic motion can be
solved by using a collimating lens. The parasitic motion is
generated by the o-center input force and can lead to an
inaccurate measurement result.
Figure 3 shows the modied null position sensor
construction.

Up & Down

A
B

Lever
Light

Photodiode

Slit

Laser diode

Fig. 2. Position sensor used in the force sensing system.

Jpn. J. Appl. Phys. Vol. 41 (2002) Pt. 1, No. 6A

I.-M. CHOI et al.

3989

9.0
8.8

Laser Diode

VB

B
Parasitic motion

8.6

LD Intensity Variation(V)

+
Vs = V A

8.4
8.2
8.0
7.8
7.6
7.4
7.2
7.0

Slit movement

10

12

Time (h)

(a) General position sensor

(a) Without normalization circuit

1.0
0.8
0.6

V A VB
VA + VB

Collimating
Lens

Divider

0.4

Sensor Output(V)

Vs = 10

0.2
0.0
-0.2
-0.4
-0.6

Parasitic motio

-0.8
-1.0

Slit movemen

10

12

Time (h)

(b) Modified position sensor

Fig. 3. Normalization and collimator.

The movement of the beam center is an inherent

characteristic of the laser diode. To minimize the error
caused by the phenomenon, the distance between the LD and
the PD must be reduced.
Figure 4 shows the normalization circuit eect under the
null condition. The drift of the sensor output decreased
signicantly.
The sensitivity of the position sensor and the stiness of
the mechanism aect the force measurement resolution.
After construction of the mechanism and the position sensor,
the resolution characteristics can be estimated from the
mechanical sensitivity. When the input forces were applied
to the system, the sensor outputs were plotted without a
control. The result includes the nonlinearity of the position
sensor but if the mechanism deformation is controlled to the
null position, it will be removed.
The value of mechanical sensitivity was 3.395 V/gf. If the
position sensor can distinguish a 1 mV variation from ground
noise, the force measurement resolution can be estimated to
be less than 0.3 mgf.
As the mechanical sensitivity increases, the system
resolution increases. Thus, the position sensor must be
designed for high mechanical sensitivity.

(b) With normalization circuit

Fig. 4. Drift characteristics with no input force change.

3.2 Electromagnetic system

The electromagnetic part design is related to the force
compensation. The magnetic system must be designed for
large force generation and the number of coil turns is
considered in relation to force sensitivity.
Figure 5 shows the structure of the magnetic system.
The relation between the input force and the applied
voltage is expressed by eq. (2).

Fig. 5. Cylindrical magnetic structure.

3990

Jpn. J. Appl. Phys. Vol. 41 (2002) Pt. 1, No. 6A

Fi Lg Fc Lg Blic Lg

I.-M. CHOI et al.

Bl
Vc Kf Vc :
R

where Fc is the force generated by the coil, B is the magnetic

ux density, l is the eective coil length, Vc is the voltage
applied to the coil, Lg is the lever gain and Kf is the force
transformation factor.
The voltage applied to the coil is restricted to about 5 V
because of a heat generation problem, considering the
diameter of the coil. The lever gain or the magnetic ux
density must be increased in order to improve the capability
for force compensation. To increase the magnetic ux
density, B, the permeance method can be used.6)
The parameters that aect the magnetic ux density
dominantly can be obtained from the sensitivity analysis.
The sensitivity analysis of variables of the system shown in
Fig. 5 is carried out. As a result, the air gap size, thickness
and radius of the magnet were the dominant variables. Each
dimension was determined by considering the system size.
The magnetic ux density at the air gap was 0.25 T. To
verify the result by the permeance method, the electromagnetic analysis program, MAXWELL, was used and the
result was 0.24 T, a value similar to the experiment result.
Since the MAXWELL program can consider the ux
leakage omitted in the permeance method, it could give a
slightly smaller value.
Once the magnetic density is determined, the number of
coil turns can aect the measurement range and resolution.
The maximum voltage is set to be about 5 V because of the
heat generation problem. If the minimum voltage applied to
the coil through the low pass lter is 0.1 mV, the ratio of
range to resolution can be determined. In general, this ratio
is smaller than the ratio determined from the mechanical
characteristics because of a control error and a ltering limit.
If Kf is increased by many coil turns, the measurement
range is improved, but the resolution is degraded while the
ratio is preserved. On the contrary, if Kf is decreased by a
small number of coil turns, the measurement range is
decreased, but the resolution is increased while the ratio is
preserved. This force compensation sensitivity, Kf , can be
adjusted not only by the coil turn number, but by connecting
an exterior resistor the coil.
Figure 6 shows the force sensitivity for the input voltage.
If the input voltages are applied to the coil, the displacement

1500

3.3 Control and lter performance

For accurate measurement, the system must be controlled
and the obtained result must be ltered. The controller and
the lter used in the null balance method have previously
been introduced.710)
Figure 7 shows the block diagram used in a null balance
type force sensor. GM s is the transfer function, Ks is the
gain of the position sensor and Cs is the controller. The
ground vibration (df ), the drift of the position sensor (ds ) and
the noise () are considered as disturbances. The voltage
applied to the coil is given by eq. (3):
CKs GM
CKs
Vs
Fs
df
1 CKs Kf GM
1 CKs Kf GM
C
1
3

ds

1 CKs Kf GM
1 CKs Kf GM
Fs
1
1
1


df
ds
:

Kf
Kf GM
Ks Kf GM
CKs Kf GM
To eliminate the above disturbances, jKs j; jKf j and jCsj
must be increased.
Two coils were manufactured, as mentioned in 3.2, and
the PI controller was constructed.
Figure 8 shows the experiment result when the small coil
was controlled. The input force was 5 gf. The coil input
voltage was ltered by an average lter and by a Butterworth
lter with a 5 Hz cuto frequency. Figure 8(a) shows that the
sensor error converges to zero.
Figure 9 shows the experiment result when the large coil
was controlled.
As shown in Figs. 8 and 9, the coil input voltages are
dierent despite the fact that the same input force was
applied. If the ltered signals are the same, the two
performances are then determined by the transformation
factors, KfS and KfL . The large coil system has resolution and
measurement range about 10 times larger than the small coil

Measurement data for large coil

Measurement data for small coil
Linear fitting

1200

Output Force (mgf)

occurs and the result can be transformed to the force by

mechanical sensitivity. Two Kf s are selected by the change
of the coil turn number and the exterior resistor. The
measurement range and the resolution can be estimated from
these results. Since the system is not controlled, these results
are used only for the estimation of system performance.
Accurate system performance can be determined after
controlling the mechanism. Since the above results include
the nonlinearity of the position sensor and the inaccuracy of
the narrow measurement domain, they can give only
approximate estimations.

ds

df

FOL=76.824V C

900

E(s)

R(s) = 0

600

FOS =8.769V C

300

GM(s)

V(s)

Ks

C(s)

X(s)

0
0

10

15

20

Kf

Coil Input Voltage (mV)

Fig. 6. Force compensation sensitivity.

Fig. 7. Control block diagram.

I.-M. CHOI et al.

6

79.6

0.2

66.3

53.1

39.8

Measurement force (gf)

0.3

0.1

Sensor Error (V)

3991

0.0

-0.1

26.5

Butterworth (5Hz)
Average Filter
1

13.3

0.0

Coil Input Voltage(mV)

Jpn. J. Appl. Phys. Vol. 41 (2002) Pt. 1, No. 6A

-0.2

-0.3
0

Time (s)

Time (s)

(a) Coil i nput voltage

(a) Sensor error
6

745

621

497

373

4.96

248

Butterworth (5Hz)
Average Filter

Measurement force (gf)

Coil Input Voltage (mV)

Measurement Force (gf)

Average filter
Butterworth

+5 mgf
5.00

-5 mgf
5.02

5.04

124

4.98

0
0

Time (s)

Time (s)

(b) Amplified signal

(b) Coil input voltage

Fig. 9. Force compensation result with 200 coil turns.

Measurement Force (gf)

5.004

Average filter
Butterworth

Table I. Comparison of dierent coil systems.

Small coil

5.002

5 gf

5 gf

Coil input voltage

621 mV

66.3 mV

+0.5 mgf
5.000

Large coil

Input Force

-0.5 mgf
4.998

Kf

8.052 gf/V

75.414 gf/V

Range (5 V)

40.3 gf

377.1 gf

Resolution

0.5 mgf

5 mgf

4.996

Time (s)

(c) Amplified signal

Fig. 8. Force compensation result with 20 coil turns.

system. The characteristics of these two dierent coil

systems are shown in Table I.
By taking into account these dierent characteristics, a
double force compensation method can be proposed.

4.

Double Compensation

As mentioned in 3, the measurement range is larger

when using a large force compensation than when using a
small force compensation. But the resolution is degraded
because of the control error, the drift of the position sensor,
the limit of the lter performance, and so forth. On the other
hand, the small coil system cannot compensate for the large
input force. In this section, a double force compensation
method with the measurement range of the large coil system
and with the resolution of the small coil system is proposed.
Figure 10 shows a block diagram of the double force
compensation system. Two coils are piled on each other and
they use one magnetic system.
The principle of double force compensation can be
explained as follows: Assume the input force is 52.0013 gf.

3992

Jpn. J. Appl. Phys. Vol. 41 (2002) Pt. 1, No. 6A

I.-M. CHOI et al.

Fig. 10. Block diagram of the double force compensation.

1.6

1.2

0.4

0.0

Compensation
by large coil

Compensation
by small coil

-0.8
0

Time (s)

(a) Sensor position error

0.8

760 mV
750 mV

KfS and KfL are given in 3.2. The resolution of VfL is almost
the same as that of VfS if each coil system is controlled
independently. However, in the double force compensation
system, the resolution of VfL is higher than that of VfS .
The large coil system can be used as a force reference and
the small coil system as the null balance type force sensor.
Hence, the resolution of the total system is dependent on that
of the small coil system.
Figure 11 shows the experiment result of a double force
compensation system. The input force was 57.000 gf.
In the compensation by the large coil system, after the
error of the position sensor converges to a denite band, the
input voltage applied to the large coil is altered to a known
voltage divided into 10 mV. The input force corresponding

0.8

-0.4

0.6

Coil Input Voltage (V)

Fi FL FS KfL VfL KfS VfS :

2.0

Sensor Error (V)

If the large coil system is controlled in a state of no control

of the small coil system, it will yield 52.00 gf as the force
measurement result. If the small coil is controlled, it will be
saturated. A dierent coil system with a medium transformation factor should be constructed in order to obtain a
better control result.
The large coil system is not controlled in the double force
compensation system. Only the known force, for example,
50.0000 gf, is compensated for. The remaining input force is
compensated for by the small coil system. The obtainable
result of the small coil system would be 2.0015 gf in this
case.
The control algorithm is as follows: The large coil system
is controlled until the sensor error converges to a denite
band. After this time, the coil input voltage is altered to a
discrete and known preset voltage. Then, the small coil
system is controlled until the sensor error becomes zero. The
preset voltage resolution of the large coil system is
determined only by the resolution of a digital-to-analog
converter. The force measurement result of the large coil
system does not include the control error or the position
sensor drift and so forth, but is determined only by the preset
voltage resolution.
The force measurement result is given by eq. (4):

0.4

Small coil
Large coil
0.2

0.0

Time (s)

(b) Coil input voltages

Fig. 11. Control results of double force compensation.

Jpn. J. Appl. Phys. Vol. 41 (2002) Pt. 1, No. 6A

I.-M. CHOI et al.

Measurement Force (gf)

57.004

57.002

57.000

56.998

Average Filter
56.996

Time(s)

Fig. 12. Double force compensation result.

3993

The measurement range and the resolution are dependent

on the force generation capacity and the various error
sources. These characteristics could be estimated and
veried from the mechanical sensitivity and the force
compensation sensitivity. Two dierent coil systems were
designed and double force compensation was proposed. The
large coil system was used as a force reference and the small
one was controlled for null balance. The measurement range
of the large coil system and the resolution of the small one
were fully realized by the double compensation method.
The measurement range, over 300 gf, and the resolution,
less than
1 mgf, were obtained according to the experiment
results.
The eects of temperature variation and lter design on
fast response will be studied in the future.
Acknowledgement

to the division step must be compensated for by the small

coil system. This step was divided to be smaller than the
compensation range of the small coil system designed in
3.2
The ratio of the two force generations must be about 10 :
1 and the discrete step of the large coil system must be
divided to be much smaller than the force generation
capability of the small coil system.
Each input voltage obtained as shown in Fig. 11 can be
transformed by using eq. (4) and KfS and KfL . The force
measurement result was scaled up and is shown in Fig. 12.
The obtained resolution is less than
1 mgf.
5.

Conclusions

In this study, a force sensor system of null balance type

was introduced and the characteristics of each component
were analyzed and designed with particular emphasis on the
force generation component. The position sensor and the
electromagnetic system were designed and tested experimentally.

This work was supported in part by the Brain Korea 21

Project.
1) I. R. Peterson: Rev. Sci. Instrum. 68 (1997) 1130.
2) T. J. Quinn: Meas. Sci. Technol. 3 (1992) 141.
3) A. Cheshmehdoost and B. E. Jones: The 8th Int. Conf. Solid-state
Sensors and Actuators and Eurosensors IX. Stockholm, Sweden 1995,
June 2529, 1995, p. 608.
4) S. Hara, S. Kasai, T. Nishitani and A. Nagashima: Rev. Sci. Instrum.
70 (1999) 435.
5) S. Kurosu: Proc. Measurement of Force and Mass, 1996, p. 129.
6) A. Kusko and T. Wrovlewski: Computer-Aided Design of Magnetic
Circuits (MIT Press, Massachusetts, 1989) Chaps. 1 and 2.
7) R. K. Butler, V. S. Rao and Leslie R. Koval: Proc. SPIE 2190 (1994)
358.
8) F. Gronlund: J. Phys. E 22 (1989) 899.
9) W. Balachandran, M. Halimic, M. Hodzic, M. Tariq, Y. Enab and F.
Cecelja: IEEE Instrumentation/Measurement Technology Conf.
IMTC/95, Waltham, Boston, MA, U.S.A., April 2426, 1995, p. 293.
10) R. Maier and G. Schmidt: IEEE Trans. Instrum. & Meas. 38 (1989)
998.