Sensors and Actuators A, 37-38 (1993) 449-452 449

Flicker noise and offset suppression in symmetric Hall plates

Zen0 Stoessel
La&s & Gyr Corp, Corporate Research and Deoelopment, Zug (Swrtzerland)

Markus Resch
KTH, Department of Apphed Phyxs, Stockholm (Sweden)

Abstract

For stable measurements of permanent or low-frequency magnetic fields, it LScmc~al to have a very good
signal-to-noise ratio and low offset voltage in Hall sensors Flicker noise and offset m Hall plates are suppressed by
orthogonal switching of a single symmetric Hall plate The supply current alternates at 0” and 90” through the Hail
plate The Hall voltage alternates its sign, while the fhcker-noise voltage and offset voltage do not With high-pass
filtermg, one gets suppression of the flicker noise and offset voltage A lower cut-off frequency for the flicker noise
of the order of 10% of the swtchmg frequency IS introduced

Introduction Thermal noise IS described as uth “, the ‘thermal noise

voltage’
A large vanety of Hall plates has been reahzed m With an external magnetic field perpendicular to the
different matenals, technologies and shapes [ 1,2] Ma- Hall plate, the Hall voltage IS calculated as follows
Jar goals were to nnprove the sensor’s sensltlvlty, stabd-
lty and offset voltage and to realize geometnc demands U=kZ(B+B,,+Bf.)+u,. (1)
for special apphcatlons Munter [3-51 describes how to
reduce offset voltage with the spmnmg-current tech-
nique on a special 16-contact slhcon Hall plate In this Symmetric Hall plates
paper, a slmphfied version of this spmmng technique 1s
desctlbed, to be used on commercially available sym- Symmetnc Hall plates are defined such that the sup-
metnc Hall plates made of any matenal A new aspect ply current contacts and Hall voltage contacts can be
of not only suppressmg the offset voltage but also exchanged as described below
suppressing the ticker noise 1s shown
Orthogonal swltchutg
We define two states A and B, see Fig 1
Hall plates State A the Hall current, Z, flows from contact (1) to
contact (3) and the Hall voltage, U,, 1s measured
The voltage of a Hall plate 1s proportlonal to its Hall between contacts (2) and (4)
constant, k, its supply current, Z, and an external mag-
netic field, B
Due to rmsahgnment m semiconductor technology
and imperfect unlformlty m the conductmty, there 1s an
addlttonal voltage, sometimes called the offset voltage,
that IS proportional to the supply current Therefore the
offset voltage can be described as kZBoff, where B,,E IS
called the ‘eqmvalent offset field strength
As m every physical system, there is noise m addition
to the Hall voltage In the low-frequency domam, tbs
noise 1s dominated by ficker or l/f noise This ticker-
noise voltage 1s proportional to the supply current Tlus A (3) B (3)

voltage can be described as kZB, “, where Bf n IS the Fig 1 D&&on of the two states A and B of a symmetnc Hall
‘eqmvalent flicker-noise field strength’ plate

0924-4247/93/$6 00 @ 1993 - Elsevzer Squola All nghts reserved

450

State B the Hall current, Z, flows from contact (4) to The temporal order of events 1s swltchmg to state A,
contact (2) and the Hall voltage, U,, 1s measured delay to allow setthng of the electromcs, measurement
between contacts (3) and (1) of signal A, s\lrltchmg to state B, delay to allow settling
In both states, the magnetic-field strength, B, 1s per- of the electronics, measurement of slgnal B, switching
pendicular to the Hall plate to state A, and so on Filtering is achieved by takmg
Switching between the two states A and B changes the difference of signal A and the following slgnal B
the sign of the Hall voltage but does not change the
u, - u, = kZ[B(t, _ ,) + B(tJ
sign of the offset voltage and ficker-notse voltage An
approptlate model for these effects 1s a resistor bridge, +kZ[&,(t,-1) -B,.(tJ
where the offset voltage is described by an Imbalance of
the bridge and flicker noise by fluctuations of the + 1% “(h - 1)- % “(bJ1 (5)
resistor values These resistors are replaced by ‘eqmva- The time of the kth sample 1s t, = k/fs As we can see,
lent field strengths’ we get two transfer functions, one for a sum and one
for a difference [6],
State A u, = k*Z(B + B,z,,rt-+ B,c, f .) + #A tt, ,, (2)

State B Ug=kBZ(-B+Bg,oR+BBrn)+~gthn (3) [Z&&0)( = 2 cos f; (6)

1 ( 5)I
Important assumptron the coefficients k, BoR, B, n and
u,,, n are practically the same for the two states A and IITT~,~(w)
1= 2 sin G z
B, 1 e , the flicker-noise signals of the two states are I ( 0>I
correlated where o, IS the swltchrng frequency
If one switches peno&cally wth the frequencyf,, one The multimeter has an mtegratmg A/D converter,
can write which functions as an addItIona low-pass filter depen-
(4) dent on the mtegratlon time, t,
u = kZ[f(t)B + &, + Bf .I + %h n

where f(t) IS a square-wave function mth amplitude

f 1 and frequency f, In the frequency domain, the
spectrum of the measured magnetic-field signal 1s The swltchmg frequency has been chosen such that it 1s
shifted Us, + 3f, f Sf,, (convolution urlth spec- possible to measure tnth an mtegratlon tune of 20 ms
trum of the square wave) With this integration time, power-he noise 1s sup-
pressed Therefore, for the magnetic field, flicker and
thermal noise signals, we introduce an upper cut-off
Offset and flicker-noise suppression frequency,f,, of about 50 Hz For lower frequencies the
effect of the multimeter can be dlsregarded
Amphfymg the switched Hall voltage wth an a c The filter has the effect of a low-pass filter for the
amphfier and filtermg urlth a high-pass filter produces a magnetic-field slgnal, see eqn (6) For ticker noise and
sensor slgnal free of offset voltage Flicker nolSe 1s offset voltage, the filter acts as a high-pass filter, see
suppressed due to the titer charactenstlc eqn (7) The offset voltage IS fully suppressed (as long
as there are no thermal fluctuations and the above
assumptions are vahd) First we calculate the eqmvalent
Circuit and model used for the measurements power of the flicker noise A IS the equivalent power of
the flicker noise at 1 Hz
A set-up wth a symmetnc Hall plate, analog sv&ches
and amplticatlon electromcs was bmlt
Because we do not want to measure the thermal
fluctuation of the environment, the whole clrcmt 1s put
mto a box conslstmg of two layers The inner part 1s
made of alummmm to provide a large thermal mass fu
and the outer part 1s made of polystyrene foam to =A df
prowde good thermal msulatlon s f
The eqmpment outside the box 1s a power supply, a
square-wave generator and a tnggerable multimeter
The generator and the multnneter are connected to a
computer With the ad of the software package (Lab
View) the measurements were completed
The equivalent power of the ticker noise of the
smtched system 1s compared to that of an unsHrltched
system with a lower cut-off frequency, f, Because the
equivalent power of the flicker noise of the switched
system 1s taken from two signals, we have to compare it
wth four times the equivalent power of the unsmtched Fig 3 Calculated spectral power density of flicker and thermal
system (assuming that ticker-noise signals are corre- nose Solid hne, switched system, dashed Ime, unswltched system
lated)
Measurements
(10)
The measurements have been made mth Hall plates
fabncated at Landis & Gyr laboratory m Zug They
* 1; r (f,f,) ‘/* exp( - x2/4) z 0 085 x (f, f,) ‘P have a vertical structure, 1e , the four Hall contacts are
(11)
geometncally m one line on top of the chp It IS
Calculations of the spectral power density of the ticker sensitive to a magnetic field parallel to the surface and
noise of the switched and unswtched signals are shown 1s made of silicon Typical parameters are
in Fig 2 The switchmg frequency 1s 16 Hz and
A = 1 Vz The lower cut-off frequency, fi , IS 2 4 Hz Hall coe.f&ent, 300 V/AT
using fu = 50 Hz Figure 2 shows that the fhcker noise Equivalent offset field strength, f 10 mT
below 10 Hz is suppressed Resistance of Hall plate, 3 kQ
Because the thermal noise of the two subsequent Internal resistance of Hall plate as source, 20 kR
signals 1s not correlated, the thermal noise 1s not sup- The measurements have been done Hrlth the followmg
pressed The spectral power density of the ticker and settings
thermal noise of the switched and unswtched signals 1s
found m Fig 3, where (h,, ,,)’ = 0 1 V2/Hz Swltchmg frequency, 16 Hz
Hall current, 1 mA
Integration tnne of multimeter, 20 ms
Delay to allow setthng of the electromcs, 7 ms
A typical result of the spectral power density of the
signals of the switched and unswtched system IS shown
m Fig 4 The spectral power densities of the signals A
and A mmus B were calculated and dlsplayed 1024
samples were measured, 1e , 512 samples of state A and
512 samples of state B As one can see, the fhcker ncnse
is suppressed and the measurement corresponds well
with Fig 3
The equivalent offset field strength of the Hall plates
IS attenuated about 1000 ties and 1s typically + 10 pT

Fig 2 Calculated spectral power density of flicker nose Sohd Orthogonal s\Nltchmg of a symmetnc Hall plate,
he, swtched system, dashed Ime, unswltched system measurmg the two states and taking the difference of
452

ments were made by Markus Resch as his diploma

work as a physicist He was a student of Professor
J A Tellefsen, KTH, The Royal Institute of Tech-
nology, Department of Apphed Physics, Stockholm,
Sweden

References

1 H P Baltes and R S Popovq Integrated semlconductor

magnetic field sensors, Proc IEEE, 74 (1986) 1107-l 132
2 R S Popovlc, Hall Efict Devrces, Adam Hdger Senes on
Sensors, Bnstol, 1991
3 P J A Munter, Electronic cmxttry for smart spmnmg-current
Fig 4 Measured spectral power density of fhcker and thermal Hall plate ~lth low offset, Sensors and Actuators A, 27 (1991)
747-751
4 P J A Munter, A low-offset spmnmg-current Hall plate,
Sensors and Actuators, A2ILA23 (1990) 743-746
the two signals, attenuates the offset voltage about 1000 5 P J A Muter and S Mtddelhoek, The electromc arcmtry for
times and suppresses ficker noise the low-offset spmnmg-current Hall plate, Proc i7wd Int
The assumption that the Hall parameters are the same Forum ASIC Transducer Technology, Banff, Alta, Canada, May
for both states 1s nearly correct Offset of the Hall plate 20-23, 1990, pp 109-112
1s not fully ehmmated because of second-order effects, 6 A Papouhs, Szgnul Analysis, McGraw-Hdl, New York, 1984
e g , the electnc-field &strlbutlon m the plate 1s not the
same for the two states (Imperfect unrformlty of dopmg)
Offset voltages of the amphficatlon and filtermg stages
and of the A/D converter are elnnmated except for the
thermal fluctuations between the two states Zeno Stoessel was born m Interlaken, Surltzerland, on
The assumption that flicker noise IS correlated for the 14 Aprd, 1953 He completed an apprenticeship as a
two states 1s vahd m the range 0 03-10 Hz Thermal physics techmclan He studied electrical engmeermg at
noise masks fhcker noise m the domam where thermal- the Ingemeurschule Burgdorf He contmued his educa-
noise IS larger If there 1s flicker noise, it 1s measurable tion at the Umverslty of Bern and m 1988 he received
wth a sufficiently low frequency (sufficiently long mea- a hc phJ nat degree m physics Smcc 1988 he has been
suring time) working m the sohd-state physics group at Corporate
Research and Development, Landis k Gyr, Switzerland
He is responsible for sensor interface electronics Stoes-
Acknowledgements se1 is a member of the IEEE He also collaborates v&h
Professor H Baltes, SWSS Federal Institute of Technol-
The measurmg set-up was built and the measure- ogy, Zurich, Surltzerland