ELECTRICAL MEASUREMENT OF MECHANICAL QUANTITIES

VD 83003 e

Linearity and Sensitivity Error in the Use of Single Strain

Gages with Voltage-Fed and Current-Fed Circuits

by Dipl.-Ing. Manfred Kreuzer

HOTTINGER BALDWIN MESSTECHNIK

Linearity and Sensitivity Error in the Use of Single Strain
Gages with Voltage-Fed and Current-Fed Circuits
by Dipl.-lng. Manfred Kreuzer
Hottinger Baldwin Messtechnik GmbH, Darmstadt, West Germany

An important part of experimental stress analysis is the measurement of strain with single
strain gages and the type of circuitry employed influences the linearity and sensitivity of the
measurements. This article describes the basic principles of strain measurement by means
of strain gages, from the strain itself via the variation in resistance to the eventual electrical
signal, and compares the magnitude of error of the voltage-fed bridge circuit and the
current-fed circuit. The conclusion reached is that, for single strain gages, voltage-fed
bridge circuits give better linearity by several powers of ten and, when there is initial
detuning, a similarly more stable sensitivity than current-fed circuits.

1. Introduction In order to explain this self-compensation process in

more detail, it is necessary to examine the relationship
between the strain and the electrical signal produced.
Current-fed circuits possess a very linear characteristic
Fig. 2 shows the various stages of the process from
for the measurement of resistance and changes in
measurement of the actual strain to the eventual
resistance. However, this does not apply to the
production of an electrical signal. The strain in the
measurement of strain by means of single strain gages.
specimen causes a relative change in length of the strain
In such applications the current-fed circuit exhibits a
gage which, in turn, causes a relative change in
much greater linearity and sensitivity error than the
resistance which is converted into an electrical signal in
voltage-fed Wheatstone bridge circuit. Fig. 1 shows the
the relevant circuit. The relationships between the
circuit diagrams of the voltage-fed and current-fed circuits
various parameters in the individual steps of the process
for single strain gages. The cause of the different linearity
will be described individually later and then the relation-
and sensitivity characteristics is the not precisely linear
ships between the strain and the electrical signals from
relationship between the strain and the relative change in
the voltage-fed quarter bridge circuit and current-fed
resistance of the strain gage. In the case of the voltage-
circuit derived.
fed Wheatstone bridge circuit this non-linearity is
In the following section it will be explained why the
opposite to the non-linearity of the circuit, with the result
actual, i.e. effective, strain εe is only approximately equal
that the two tend to cancel each other out.
to the relative change in length ∆I/Io.

Fig. 1: Circuit diagrams for single strain gages Fig. 2: The steps from measurement of
a) voltage-fed bridge circuit the strain to the eventual
b) current-fed circuit electrical signal
2. The definition of strain Equation (6) represents an exact definition for strains of any
magnitude. According to this equation, measurements of
strain are independent of the prior history and part-strains can
According to VDI/VDE 2635 [1], technical strain e is the ratio be added or subtracted without incurring the discrepancies
described in the previous example.
of elongation or contraction Al to the original length Io
Comparing the effective strain εe with the strain ε defined in
(1) equation (1) gives the following relationships:

(7)

∆This definition leads to difficulties in the addition and sub-

traction of large values of strain, and this will be demonstrated (8)
by means of an example: When a wire 1000 mm long is
stretched by 10,000 µm/m it means that its length increases
For small values of strain ε ≈ εe as a first approximation, which
by 10 mm to 1010 mm. If the same wire is stretched by
is apparent if equation (8) is written as a series:
10,000 (µm/m again in a second, unconnected measurement,
this time its length increases by 10.1 mm because the original
length for the second measurement was 1010 mm. Therefore, (9)
the total elongation is 20.1 mm. If the original wire had been
stretched by 20,000 µm/m for the first measurement it would
only have increased in length by 20 mm according to the
definition given in equation (1). Therefore, for small values of strain both definitions give the
Consequently, the two independent stretchings of 10,000 same results. The relation between εe and (ε - εe ) is shown in
µm/m each do not correspond to a single stretching of 20,000 Fig. 3.
µm/m but to one of 20,100 µm/m, with the first accounting for
10,000 µm/m and the second for 10,100 µm/m. Naturally, this
is a totally unsatisfactory situation because it means that each
measurement is dependent on the prior measurement taken
with the device. It is essential, of course, for the same values
of strain to give rise to the same readings and this can only
be assured by restricting the definition of strain e in equation
(1) to very small values of e and, therefore, ∆l.
Equation (1) becomes totally correct if ∆l is made infinitely
small in relation to dl and divided by the total length I:

(2)

The precise equation for de is

Fig. 3: Deviation of strain e from the effective strainge in

(3) relation to the effective strain

Every value of effective strain ee can be regarded as the sum 3. Theoretical relationship between
of an infinite number of infinitely small values of strain: the relative change in length and
the relative change in resistance of
(4)
strain gages

Together with equation (2) this gives The following observations are only valid for strain gages
having metal grids and a k factor of approximately 2. How-
ever, since these are the most popular types of strain gage
(5)
the restriction is of no importance in the majority of practical
applications. The definition of k factor is

Equation (5) describes the "logarithmic" or effective strain

which has been known in this form since 1909 [2].
When a wire, e.g. a strain gage, of length I0 is stretched to (10)
length I0 + ∆l, the effective strain εe becomes

Therefore, k = 2 means that the relative change in resistance

(6)
is twice the strain e. The reason for specifying strain gages
with a k factor of 2 is because the k factor is always 2 in the
plastic range of every material [3].
For metal strain gages with a k factor of 2, change 4. Relationship between measured
in form accounts for approximately 80% of the k
factor in the elastic range and a change in specific
voltage and change in resistance
resistance for only about 20%.
When a strain gage is connected as shown in Fig. 1
In the plastic range of a material the relative change a with three equal resistors R0 in a Wheatstone
in resistance is caused exclusively by the change in bridge, and the resistance of the strain gage
form with simultaneous constancy of volume and changes from R0 to R0 + ∆R due to the strain e, the
specific resistance. Therefore, strain gages with a k measured voltage Um can be calculated from
factor of 2 possess constant sensitivity in both the
elastic and plastic ranges, which is a very important
requirement when measuring large values of strain,
in order to obtain reliable readings. Another
advantage is that the readings are therefore also
almost entirely independent of the prior treatment The voltage US/2 is that across the resistors R0 and,
and use of the strain gage. therefore, also across the strain gage when it is
The resistance of a conductor is calculated with the unstrained; for the purposes of this description it will
formula be denoted by U0:

where I is the length, A the cross section and ϱ the It is normally necessary to know the magnitude of
specific resistance of the conductor. If the volume V the measured voltage Um (change in voltage) in
and specific resistance ϱ are assumed to remain relation to the voltage U0 across the strain gage
constant (plastic strain), the following relationship when it is unstressed. Dividing the two voltages
between the resistance and the length can be gives
derived:

the formula for the voltage-fed bridge circuit. The

relationship between measured signal and change
in resistance is non-linear.
In the case of the constant-current circuit as in Fig.
1b is

Which gives the formula for the constant-current

circuit

This gives the relative change in resistance

Therefore, the measured signal varies proportionally

with ∆R.

With equation (1) this becomes

5. Relationship between the effec-
tive strain and the electrical signal
and, as a first approximation, The next step is to determine the overall
transmissibility of the measured value "effective
strain εe " to the electrical measured signal "Um/U0"
Therefore, the derivation gives a k factor of 2 as a for voltage-fed bridge circuits and current-fed
first approximation. When considering the linearity, circuits. Substituting equation (8) in equation (19)
2
however, the term ε in equation (19) must not be gives
neglected.
For metal strain gages with a k factor of 2, equation (26) re-
presents the theoretical relationship between the effective strain
εe and the relative change in resistance R/R0. The series
expansion of equation (26) gives
(27)

which is a good approximate solution. Practical measurements

have given the same result in the past [4].
In order to obtain the relationship between the effective strain ee
and the electrical signal Um/U0 for voltage-fed bridge circuits,
equation (26) is substituted in equation (23), giving

(28)
Fig. 4: Small linearity error of the voltage-fed bridge
circuit in relation to the effective strain
Multiplying the numerator and denominator of equation (28) by
exp (- εe) gives

(29)

Therefore, the theoretical relationship between the electrical

signal Um/U0 and the effective strain εe for an ideal metal strain
gage with a k factor of 2 is determined by the hyperbolic
tangent.
The third approximation of equation (29) is given by

(30)

Fig. 5: Large linearity error of the current-fed bridge circuit

in relation to the effective strain
The error of this approximation compared to the exact solution is
extremely slight; even with a very large strain of 100,000 µm/m it
is only - 0.0013% of the actual value. 6. The effect of extreme initial detuning
From equation (30) it can be seen that the measured signal on measuring sensitivity
increases linearly with the strain at small values of strain. Only
3
with very large values of strain does the term ε e/3 produce a So far, this article has dealt mainly with the linearity error of
measurable linearity error. Even then, it is only 0.3 µm/m for a voltage-fed bridge circuits and current-fed circuits at very large
strain of 10,000 µm/m. Fig. 4 shows the linearity error as a values of strain. However, for most strain measurements, errors
function of the effective strain ee for the voltage-fed bridge in linearity are not of overwhelming interest because the strains
circuit. to be measured are often relatively small and the linearity error
can be neglected within the range of accuracy required.
In the case of current-fed circuits the electrical signal is equal to
Nevertheless, the relationships which have been derived are of
the relative change in resistance of the strain gage so that, by
considerable significance when considering the effect on
substituting equation (26) in equation (25), the relationship
measuring sensitivity of prestrain and the resistance tolerances
between the effective strain and the electrical signal is obtained
of the strain gages, i.e. initial detuning in general.
directly:
The measuring sensitivity E for very small strains is given by the
(31) slope of the transmission curve at the particular working point,
i.e. at the particular value of effective strain εe:

Expanding the series to the third term gives the third ap- (33)
proximation of equation (31):

(32)
The values of valid, effective strain ee are obtained by con-
verting the resistance errors ∆R/R0 due to the resistance
For an effective strain of 10,000 µm/m the current-fed circuit has
a linearity error of 100.7 µm/m. Fig. 5 shows the linearity error in
relation to the effective strain for the current-fed circuit.
tolerances and prestrain of the strain gages according to
equation (34);

The sensitivity EB of the voltage-fed bridge circuit is then

and therefore

Equation (36) can be written in the following approximate

form:
Fig. 6: Small variation in sensitivity in relation to
effective strain for the voltage-fed bridge circuit
The sensitivity ES of current-fed circuit can be calculated
similarly:

and written in the approximate form:

A point of interest is how the measuring sensitivity E

varies in relation to the effective strain εe. If the sensitivity
for εe =0 is denoted E0 and the sensitivity at the
appropriate working point on the curve E, the relative
change in sensitivity ∆E/E0 can be written:
Fig. 7: Large variation in sensitivity in relation to
effective strain for the current-fed bridge circuit
Substituting the values from equation (36) for the
voltage-fed bridge circuit in equation (40) gives the Substituting equation (34) in equations (41) and (43)
relative change in sensitivity: gives the relative change in sensitivity in relation to the
relative change in resistance ∆R/R0 of the active strain
gage.
which can be written approximately: For voltage-fed circuits the following equation can be
written:

Treating equation (40) for the current-fed circuit similarly

with equation (38) gives the relative change in sensitivity
of the current-fed circuit:
and in the approximate form:

which can be written approximately:

For current-fed circuits the equation is:

Fig. 6 shows the relative variation in sensitivity with effec-

tive strain for the voltage-fed bridge circuit and Fig. 7 the
Figs. 8 and 9 show the relative variation in sensitivity in
same for the current-fed circuit. The relative variation in
relation to the relative change in resistance ∆R/R0 of the
sensitivity of voltage-fed bridge circuits in the range from
active strain gage. It can be seen that, with voltage-fed
- 20,000 (µm/m ≦ εe ≦ 20.000 µm is less than 0.04% bridge circuits, relative changes in resistance of ± 6%
whereas that of the current-fed circuit in the same range cause changes in measuring sensitivity of less than
is approximately ±4%. 0.1%.
 In the case of the current-fed circuit the variation in measuring their effect cannot be calculated either for the voltage-fed circuit
sensitivity is proportional to the variation in resistance and is 6% or the current-fed circuit and this is the reason for comparing the
for a resistance of 6%. This disadvantage of the current-fed linearity and sensitivity error of the two types of circuit on the
circuit has already been referred to in [5]. basis of systematic, calculable relationships.

Fig. 8: Small variation in sensitivity in relation to change in

resistance of the strain gage for a voltage-fed bridge
circuit

As a result of this fact it is necessary to match the constant Fig. 9: Large variation in sensitivity in relation to change in
current for the strain gage to the actual value of strain gage resistance of the strain gage for a current-fed
resistance if the sensitivity error of the current-fed circuit is to be bridge circuit
eliminated. It means that zero balancing of the hardware must
be employed, which involves a considerable amount of
The general conclusion can be drawn that, for the use of strain
additional circuitry in the case of multiple-gage systems.
gages in quarter bridge connection, the voltage-fed bridge
With voltage-fed bridge circuits, on the other hand, a simple circuit is superior to the current-fed circuit in many ways,
computing zero balance is sufficient up to very large values of primarily because of better linearity by several powers of ten
initial detuning. and similarly better sensitivity stability under conditions of initial
detuning.

7. Summary 8. References
The purpose of this article is to draw attention to the fact [1]VDI/VDE 2653 Blatt 1: Dehnungsmeßstreifen mit
that the linearity and sensitivity of strain gage measure- metallischem Meßgitter, Kenngrößen und Prüfbedingungen :
ments cannot be assessed solely from the relation
August 1974
between change in resistance and change in electrical
signal; the transmissibility between strain and change in [2] Ludwik, P.: Elemente der Technologischen Mechanik;
resistance must also be taken into account. However, Verlag von Julius Springer. Berlin 1909
one important proviso to the relationships developed in [3]Rohrbach, Chr.: Handbuch für elektrisches Messen
the article is that they are only precisely valid for ideal mechanischer Größen (pp. 126 and 130); VDI-Verlag GmbH,
strain gages applied ideally. In reality the transmissibility Düsseldorf 1967
of "strain/change in resistance" is adversely affected by [4] Hoffmann, K.: Zum Linearitätsfehler bei Dehnungs-
many factors such as less-than-ideal backing materials messungen mit Dehnungsmeßstreifen im Hochdehnungs-
and adhesives and multiple-axial stress fields in the grid bereich; MTB 12 (1976), pp. 53-57
of the strain gages. Such factors are usually [5] Knapp, H.: Vorschläge zur Vereinfachung und Verbesserung
unsystematic and, therefore, cannot be dealt with in a der Schaltungstechnik bei rechnergesteuerten Vielstellen-
universally-applicable form. Moreover, Dehnungsmeßgeräten;
tm 1981, pp. 265-271