IMTC 2004 - Instrumenlation and Measurement

Technology Conference
corno, Inly. 18-20 May 2004

A Systematic Approach for Sensor Bridge Linearisation and Readout

G. de Graaf and R.F. Wolffenbuttel

Delft University of Technology, ITSEt, DIMES,
Mckelweg 4, 2628 CD Delft, The Netherlands
Phone +31 15 278 4707, Fax. +31 15 278 5755
e-mail: [email protected]

Abstract - A general concepr for the lineurirulion qf impedance conventional Wheatstone bridge is that impedances wi hoth
bridges wirh circuit components chon& in only one polariy is increasing and decreasing (e.g. Z+AZ and Z-AZ) values
presented. The method (Tun be applied for DC or AC bridges and the have to be available in opposite bridge arms. This means that,
linearisation can convenienrly be inregrural wirh the readoirt circuir
for example, in the case of planar capacitive transducers three
resulring in U highly linear senmr readour. A feedback-loop usins a
rruns-admittance umpliJ'ier properly designed and marched ro rhe stacked plates are needed for a differential device, also for a
nominul wlue of the bridge components reduces non-linearityfrom differential inductive bridge two stacked coils are needed in
9% to 0 . 4 6 over 4 0 % of bridge imbalunce. In uddirion io rhe each bridge arm.
lineurizatiori of rhe bridge transferfunction irsuK rhe method can be
applied for rhe lineurizarion of U range of trunsducers wirh U Fixed (dummy) o w c i t o m
hyperbolic rransferfunction.

Keywords- Transducer bridge. linearisurion, Sensor Readour,

OFA. bridge compensation, diflkrentiul amplifier.

INTRODUCTION
Fig. 2. Capncilive pressure sensor with only C+AC vnlues.
Impedance bridges are universally applied for precision
mezsurrment and the read-out of sensors. A general feature Therefore, for practical or cos1 reasons often only increasing
of a bridge configuration is its simplicity. In case the nominal or decreasing component values (Z, Z+AZ or Z, Z-AZ) are
values of the components are well matched, the bridge also available]. These types of bridges are inherently non-linear.
enables rejection of common-mode signals and improves Also in the case of capacitive transducers based on the
power supply rejection [I]. Moreover, parasitic transduction change in distance between the plates [2][3], as shown in Fig.
effects (notably temperature dependencies) can be 2, the relation between the plate distance and capacitance is
compensated for. However, a linear output is obtained only inherently nun-linear.
in the fully-symmetric bridge configuration shown in Fig. I . Many papers use modulation of the bridge excitation voltage
or current for linearisation or temperature compensation
[411.51[61.
This paper demonstrates that linear readout can also be
obtained by designing a readout circuit using a uniquely
determined trans-impedance.

BRIDGE NON-LINEARITY

Fig. 3 shows the non-symmetric half and full bridge

configuration with Z+AZ elements only. The transfer
function of the full-bridge can be derived as:
Fig. 1. Conventional fully-symmetric impedance bndge

The output voltage is given by: V, = -V,

Az = &Vb, provided
z
that the load impedance of the bridge (= the input impedance
of the readout circuit) is much highcr than the output If the input signal is a relative change in impedance, in two
impedance of [he bridge. The main problem of this of the bridge arms E = A D % and the load impedance ZI n 2 2
the bridge tnnslkr function can be simply written as:

0-7803-824X-X/04/$17.00 0 2 0 0 4 IEEE 1551

Authorized licensed use limited to: CSIC. Downloaded on May 28, 2009 at 12:18 from IEEE Xplore. Restrictions apply.
 e:
where Z denotes the nominal value of one of the four
bridge impedances. The output voltage V , caused by a

o2 $-
bridge unbalance E ( = m Z ) can be calculated as:
E
v, =
(-2+ 2g,"Z)+ &(-I+ 2g,Z)
V," (3)

A linearized Wheatstone bridge (Z=R) thus requires a

243~ voltage-to-current converter with a transconductance:

0.) bJ
Fig. 3. Full-(a)) and half~bridge(b.))circuit with onlyZ+AZvvalues vo =
This results in a linear bridge output: - --E (5)
Vr"
Generally the transfer function of these types of bridges is a
hyperbolic function. A relative change of the active bridge
components of 20% (E= 0.21 results in a non-linearity in the 1. Tnmsconductance.
output voltage of the bridge of about 9%. Equation (1)
implies that the bridge non-linearity can he influenced by the The value of the tmnScOndUctance should be set by two
load impedance Z, For a linear response Z, should be equal impedances that are well-matched with the nominal values of
to: ZI= -22. This can be electronically realised by combining the bridge components, which is relatively easily achieved,
the bridge read-out and feedback, as shown in the next especially in integrated sensors.
section. Capacitive and inductive bridges are also linearised if G,= -
jw(C+AC)/Z and G,=- I / (2joL) respectivily.
SYSTEMATIC DESIGN OF THE READOUT
AMPLIFIER For capacitive bridges, devices with change in permittivity
(A&,l&J or plate area ( M A ) and sensors with plate distance
variations (Add) as the measurand can be distinguished. The
Figure 4 shows the basic block diagram for bridge voltage relative plate distance change is related to the plate
readout with simultaneous compensation of the bridge non-
linearity. The general transfer function of the differential capacitance by: -= - Ac (5)
d C+AC
inputloutput amplifier is given by: This hyperbolic non-linearity can be compensated by the
inverse nonlinear transfer function of the bridge readout
(equation(3)). by selecting a proper value for the amplifier
gain G,. Combining equations (3) and (5) yields:

I I

Table 1 shows an overview of various bridge configurations

with the required amplifier transfer function for a linear
response.
z+un z h U Generally, it can be concluded that for a full bridge with a
Y Y linear output, the read-out amplifier should be realized using
a floating voltage-current converter with a trans-conductance
equal to twice the value of one of the bridge components.
Figure -1. Bndgr and read-out amplifier Note that E can be negative also, indicating decreasing
bridge impedances.
Since Vi = V, , the input impedance Zj and the current gain
Ai should be high for voltage rendout, the desired transfer
function is a trans-admittance with:

1552

Authorized licensed use limited to: CSIC. Downloaded on May 28, 2009 at 12:18 from IEEE Xplore. Restrictions apply.
 Bridge/Readout Required trans- Comparing equations ( 5 ) and (7) shows these offsct sources
Transfer IUJUJUI admittance [IJLI] can be neglected if these are small compared to the bridge
driving voltagc. Offsets introduce both a DC error and a non-
Full bridge linearity error. These errors can be compensated by trimming
Resistive ~ Volfsto a value equal to Y,,$ = -21,R
either 1 , or .
-jw(C+AC)/?
4. Amplifier input- und output impedunce
Cq.(Ad)
lnductivc Both the input and output impedance of lhe amplifier are in
parallel with the synthesized negative resistance, reducing the

Resistive
Cap (AE,AA)
I -714ARR
Half hridge

AC/4C , AN4A
I -2/(3R)
-(2/3) jw(C+AC)
I I
I
compensation. Increasing the trans-conductance of the
amplifier easily cancels the effects of input- and output
impedances:

Cap ( Ad) 2 Ad/d -(2/3) j

d
Inductive (314)ALJI. -2/(3ioLi

Table IOverview of various bndgr configurations

S. Amplifier frequency response

2. Stubilit, The poles of the closed loop compensated system follow

from (I+GmZ,)=O. Since the hridge impedance Zb is a
A small amount of positive feedback of the amplifier is function of the unbalance E, and assuming the amplifier bas a
required for synthesizing the negative trans-conductance. The dominant pole pI the pole of the closed loop system shifts
loopgain A,, is the product of the trans-conductance G , of with the bridge unbalance:
the read-out amplifier and the bridge output impedance 2,.
I
A*, = G- Za, wirh : C m = -- ond: z*= 2 ( Z / / ( Z C d Z ) )
22 As a result the bandwidth of the bridge-readout combination
is more than a factor 2 lower than the amplifier bandwidth.
follows: * =--
I+&
The positive feedback increases the gain also a factor 2 as
~' 2 + &
can be seen comparing eqn. (2) and (6). The bridge elements
From equation ( 6 ) can be seen that ideally the loopgain is a can add extra poles to the readout also. e.g. large parasitic
capacitances in a resistor bridge have to be taken in account,
function of the bridge unbalance &only, typically increasing
since these can cause instability at high frequencies.
from -0.5 to -0.565 (for E=O to 0.3) and decreaing if E is
negative. The absolute maximum is -2/3 for E=I(AZ=Z)
REALEATIONS
which is nevertheless smaller than the absolute stability limit
A,,= - 1 . However for a complete analysis the phase margin The differential voltage-to-current amplifier can be
and thus the frequency response of the amplifier also has to implemented in many ways, depending on: the type of
be taken in account, as will be shown in the next section. impedance to be mrdsured, required accuracy and frequency
range used for the bridge excitation signal. Simple voltage-
to-current converters can be designed, using extra bridge
READ-OUT AMPLIFIER NON-IDEALITIES elements to set the trans-impedance. However biasing of
simple-single stage amplifiers can cause problems, due the
3. AmpliJier o&t fact that the inputs and outputs are directly connected and the
differential voltages can be large. A common problem for the
For a DC-operated resistive bridge the intluence of amplifier
offset has to be considered. Figure 4 shows the read-out with amplifier design is that the input/output differential voltage
offset sources added. Since the output currents of the of the compensated amplifier, following from (3): V,=-
amplifier are much higher then the input currents the main (m) Vb can be in the order of several volts at a large
contribution to the equivalent offset current source is zff, bridge unbalance and bridge voltage, while only a few tens of
millivolts are allowed for sufficient linearity at the input
merely caused by mismatch of the output currents and Iu2
of the amplifier. The contributions of the input offset voltage stage of il MOS or bipolar differential pair. Therefore two-
V,,, and the output current mismatch lot,to the output can be amplifier instrumentation amplifier configurations have been
calculated as: further investigated for the read-out. Commercially available
v, = (l+&)(Yf17
+21",,R) (7)
instrumentation amplifiers do not have differential current
outputa and are thus not directly applicable here. The

1553

Authorized licensed use limited to: CSIC. Downloaded on May 28, 2009 at 12:18 from IEEE Xplore. Restrictions apply.
MAX135 [7] wideband amplifier is a 275 M R r differential AC Bridge CornpenJatlon
input- and output wideband transconductance amplifier
- a-
( W A ) that can he applied without additional active
components. The device allows large input differential
voltages and has sufficient low noise and DC precision for a
low-impedance Wheatstone DC or AC bridge read-out,
shown in Fig. 5.
Bridge

I I
I I
U, ..
Wdge unbalance nWR

Fig. 7. Measured non-linearity of Ihe bcidgc-amplifier combination of figure

5 using AC signals

A capacitive bridge using the MAX435 transconductance

amplifier from Fig.5, has also also been tested. Considering
the very high bandwidth, the high power dissipation, high
bias currents of the amplifier and low impedances involved
the readout is not optimal for D C or LF bridges hut has good
performance at working frequencies in the MHz range using
capacitive, inductive or resistive bridges.

Another implementation that is now being investigated for

DC or low-frequency AC bridges is shown in Fig. 8.

. -.~"="Om-nn-nnm,
_-'m = 2 , : H ; - 1,"?

-B
r
%

-c
.-
c-
0
z -4 \

<.
Figure 8. Resistive bridge readout using two operational floating amplifiers
- a i . . , .. . , . ... . . . , . . . . . .1 (OFA's).

This figure shows a full-floating voltage-to-current converter,

using two differential input-output amplifiers (OFA's) 181 for
the readout. The circuit has been designed and fabricated in a
CMOS process with well-matched on-chip resistors and
capacitors that can he switched on and off by MOS
transistors for testing. The lower part of the chip photograph
in Fig. 9 shows the resistive bridge read-out and the upper
part the capacitive read-out. The circuit can operate at bridge
excitation voltages between DC and a few MHz. The design
predicts a non-linearity smaller than H.4% over +20%
bridge imbalance, which is an improvement by a factor 22 as
compared to a conventional readout.

Authorized licensed use limited to: CSIC. Downloaded on May 28, 2009 at 12:18 from IEEE Xplore. Restrictions apply.
 [7] MhX43x Wideband Transconductnnce Amplifier Family
Datashcct, Maxim Integrated Products, 120 San Gabriel Drive,
Sunnyvale, CA 940x6, USA, http:Nwww.m;vtin-ic.com.
[S] Operational Amplifiers Theory and Dcsign , J.H.Huijsing,
Kluwer. Ac.Publishers, 101 Philip Drivc Norwell, MA02060, USA.

Fig. 9. Chip photograph of the CMOS bridge readout with on-chiphridge

C0mpO"e"tS.

CONCLUSIONS

A systematic approach to the linearisation of impedance

bridges is presented. The concept presented can be
implemented in conventional half- o r full bridges and
combined with a basic readout amplifier. In addition a
hyperbolic transfer function of a sensor can be compensated
also by selecting a proper value for the amplifier gain. The
method has been verified using a commercially available
amplifier and a full-custom CMOS amplifier design. Good
matching of bridge components. amplifiers and amplifier
gain setting is imponant to minimize offset and non-linearity.
Therefore, accurate, low-power, floating voltage-to-current
converters with integrated resistors and capacitors have been
designed and fabricated in CMOS, and are now under test.

REFERENCES
111 Designers Guide to Bridge Circuits, J. Williams, Reprint from
EDN magazine, Cahncrs (Elsevier Business Information), Rue des
Palais 100. 8-1030 Brussels, Belgium.
[Z] K. Puers, U. D. Bruyker and A. Cozma, A novel combined
redundant pressure sensor with self-test function, Sensors and
Actuators A: Physical, Volunie 60, Issues 1-3, May 19Y7, Pages 68-
71.
[3] E. Kalvesten, L. LRfdahl and G. Samme, Sniall piezoresislive
silicon microphones specially designed for the characterization of
turbulent gas flows. Sens. Actuators A Phys. A46 (IYYj), pp. 151-
155.
[4] A.J. Lopez-Mattin, J.I. Osa, M. Zuza and A. Carlosena, Analysis
of a Negative lmpcdance Converter as a Temperature Compensator
for Bridge Sensors, IEEE Transactions on IM, Vol.52. No.4, August
2003, pp. 106X-1072.
[SI P. Kopsytynski and E. Ohermeier, An interchangeable silicon
pressure sensor with on-chip compensation circuitry, Sensors and
Actuators, vol.lX, no.;, 19XY, pp.23Y-245.
161 C.D. .lohnson and C. Chen. Bridge-to-computer Data Acqui-
sition System with feedback nulling. IEEE Transactions on IM ,
Volume: 39 Issue: 3 , June 1990, pp. 531 -534.

I555

Authorized licensed use limited to: CSIC. Downloaded on May 28, 2009 at 12:18 from IEEE Xplore. Restrictions apply.