Analog Integrated Circuits and Signal Processing, 14,275-296 (1997)

© 1997 Kluwer Academic Publishers, Boston. Manufactured in The Netherlands.

A Smart Sensor System for Carbon Monoxide Detection

G. C. CARDINALI!, L. nORI!, M. FIORINI!, I. SAYAGO!,2, G. FAGLIA3 , C. PEREG03,

G. SBERVEGLIERI3, V. LIBERALI4 , F. MALOBERTI4 , AND n. TONIETT04
I CNR-IAMEL, Via Gobetti 101, 40129 Bologna, Italy; 20n leave of absence from: Laboratorio de Sensores, CSIC, Serrano 144, 28006 Madrid, Spain;
31NFM, Department of Chemistry and Physics for Materials, University of Brescia, VIII Valoni 9, 25133 Brescia, Italy; 4 Department of Electronics,
University of Pavia, Via Ferrata I, 27100 Pavia, Italy

Received January 26, 1996; Accepted January 30, 1997

Abstract. This paper illustrates a smart sensor system for carbon monoxide detection. An innovative technological
approach has been pursued to fabricate gas sensors on silicon substrate, compatible with IC fabrication. A mixed
analog-digital electronic interface processes the outputs of three sensors to compensate relative humidity and
interfering gases. Sigma-delta signal processing and low-frequency noise reduction techniques are used to minimize
silicon area and to meet the required performance in a standard CMOS technology.

Key Words: sensors, sensor interfaces, sigma-delta converters

1. Introduction full integration of the system), to reduce power

consumption to a hundred milliwatts;
It is well known that whenever combustion takes place iii) analog electronic interface for signal processing
with lack of oxygen, carbon monoxide (CO) gas is and for temperature pulsed measurement to max-
produced. This gas is extremely dangerous even at a imize sensor sensitivities.
low concentration and it must therefore be carefully The whole sensor system is outlined in Section 2.
monitored. Section 3 describes the microstructure of integrated gas
The working principles of solid-state devices used sensors, and Section 4 illustrates the electronic inter-
for CO detection are based on the conductance varia- face for control and measurement. Finally, Section 5
tion of the sensing layer or on the temperature increase presents the results from the experimental characteri-
in a heated element. Both effects are the results of zation of single blocks.
chemical reactions between the substances to be de-
tected and the material from which the sensor is made. 2. System Overview
However, CO sensors are intrinsically sensitive to in-
terfering gases like alcohol, relative humidity, methane The system described in this paper is designed to detect
and many others [1]. The influence of interfering gases the indoor presence of carbon monoxide. The target
on the output of CO sensors can be as great as the con- is to design a smart sensor system with the following
tribution of the CO itself or even greater. This limits characteristics: high reliability, low cost, reduced size
the use of conventional detection systems in most situ- and power consumption, high sensitivity to CO gas, re-
ations. To avoid this problem a multisensor system can duced cross-sensitivity (lower false alarm rate), and re-
be used. By combining signals generated by different duced long-term drift. The sensor system must provide
sensors, we can obtain a signal which is proportional an alarm signal when the CO concentration exceeds a
only to the CO concentration. pre-defined threshold value (around 150 ppm), as well
In this paper, we describe an innovative system for as a warning signal at half the alarm threshold. An ad-
CO detection, featuring: ditional output, proportional to the CO concentration,
i) an array of three sensors, to detect CO, ethyl al- can be directed to a microprocessor unit to monitor the
cohol (C 2 H 5 0H) and relative humidity (RH); variation of CO concentration overtime. To display the
ii) CO and alcohol sensing layers deposited on very CO concentration in percent units of the alarm thresh-
thin dielectric membranes obtained by microma- old, the resolution required of the sensor system is 7
chining the silicon substrate (in view of a future bits.

113
276 G. Cardinali et al.

Fig. 1. Block diagram of the smart sensor system.

The smart sensor system is made up by a full-custom which can be exploited to adapt calibration and mea-
integrated circuit implementing the electronic interface surement parameters to different sensors based on the
and an array of three sensors. The array contains a same detection principle (e. g. alcohol, methane).
commercial humidity sensor and two innovative micro-
sensors for CO and interfering gas detection. Fig. 1
3. Sensor Microstructure
shows a block diagram.
The gas sensors are based on Sn02 sensing lay- Fig. 2 shows the microstructure of one of the two gas
ers that show a variation in resistance when gases are sensors. The sensing film is placed on top of a layer
present. The Sn02 film is deposited on a thin mem- stack constituted (in order from bottom) by the mem-
brane realized on a silicon substrate using the bulk- brane, the heater element, the passivation layer and the
micromachining technique. To increase CO selectivity, metal contacts.
the two sensing layers are doped to enhance sensitivity Each of the above components is described in the
to CO in the first device and reduce it in the second following.
one.
As with other common interfering substances, air
relative humidity (RH) also perturbs sensor conduc- 3.1. Sensing Layers and Related Technology
tance. This requires constant monitoring and compen-
sation to guarantee the correct correlation between the Solid-state CO detectors chemically react with the
sensor output signal and the CO concentrations. A gases to be detected. The reaction produces a change in
low-cost capacitive sensor available on the market is the conductance of the sensing layer or a temperature
used to this end [2]. However, the system architecture increase in a heated element (catalytic sensor). Cat-
allows the discrete sensor to be easily replaced with an alytic sensors are low cost devices, but they have low
integrated humidity sensor. selectivity and need high gas concentrations to work
The electronic interface is integrated in 1.2-llm effectively.
CMOS technology. It carries out control and mea- The most widely used sensors in indoor applications
surement functions, spanning from reference and clock are based on the first of these two solutions and depend
generation to sensor temperature control and signal on the properties of wide-gap semi conductive mate-
processing with overs amp led techniques. The elec- rials, such as Sn02, deposited on suitable substrates.
tronic interface has a high degree of programmability When a Sn02 film is heated at 300 -7- 400°C in air, it

114
 A Smart Sensor System 277

Silicon Substrate

Fig. 2. Sensor device microstructure.

traps oxygen molecules on the sensor surface in the position of very thin catalyst layers (gold or platinum)
form of ions (0-,02"), lowering its overall electrical on the surface of the sensing material. In addition, the
conductivity. If CO molecules are present in the sur- use of an activated carbon filter has been considered.
roundings, they react at the surface of the sensor lead- Indeed, experimental measurements have shown that a
ing to a charge injection that produces a conductivity carbon filter drastically reduces the response to ethyl
variation, in agreement with the following reactions: alcohol with concentrations in the range between 250
and 2000 ppm, while the magnitude of the sensor re-
Sn02 + e- + i02 -+ Sn02(0-)ad sponse to CO remains almost unaffected even at low
Sn02(0-)ad + CO -+ (Sn02 + e-) C02 concentration levels.
To reduce the effect of RH on the sensor response
The well-established Reotaxial Growth and Thermal and to increase CO sensitivity, the sensor operating pro-
Oxidation (RGTO) technique is ideal for the fabrica- cedure has been optimized. The device is operated in a
tion of this type of sensing layers [3--6]. RGTO sen- pulsed temperature mode (rather than at constant tem-
sors exhibit good sensitivity and time stability in their perature) and thus the temperature is cycled between
response, but they always exhibit some sensitivity to 120 -;- 150°C and 375 -;- 450°C every 60 sand 90 s, re-
interfering gases, which could be present in ambient in spectively. The conductance value is measured when
high concentrations. It therefore becomes vital to dis- the sensor temperature is in the lower range of the heat-
criminate between the response due to CO and to other ing cycle. Experimental data collected with the sensor
interfering substances. To overcome this problem, a operating in this manner have shown that the effect of
solution based on two sensitive layers has been con- humidity is reduced and selectivity to CO increased in
sidered. The first layer has been designed to be highly comparison with the standard operating mode. In ad-
selective of carbon monoxide while the second is sen- dition, power consumption (depending on the pulsed
sitive to ethyl alcohol, which is the most commonly duty-cycle) is lower.
found interfering gas. Finally, to minimize the response drift due to aging
To optimize the response of the sensor specifically (1 -;- 2 years), the sensors were stabilized by thermal
designed to detect the CO, several solutions have been treatment at 500°C in the open air for two weeks.
considered and incorporated into the thin film fabrica- Figs. 3-6 present experimental data showing the
tion technology [7, 8]. A first approach entails the de- response of Sn02 sensors under different conditions,

115
278 G. Cardinali et al.

4,510- 6 r - - - - - - - - - - - - - - - - - - - - - - - - - - ,
150 ppm CO

410- 6
75 ppm CO

~ 3,510- 6

Air
310- 6

2,5 10-6
o 20 40 60 80 100
RH (%)

Fig. 3. Sn02 Pt-RGTO film current (at I V bias) vs. RH at 0, 75 ppm, 150 ppm of CO in air. (The sensor case contained an activated carbon
filter.)

710- 6
R.H.=40% [C01=150ppm
T=20°C
610- 6 Tsensor=400°C

-----
510- 6 IF

410- 6

310- 6

210- 6

1 10-6

o
o 200 400 600 800 1000 1200 1400 1600

ALCOHOL (ppm)

Fig. 4. Response ofthe sensor in Fig. 3 at fixed CO concentration (150 ppm) vs. C2HSOH concentration. (The sensor case contained an activated
carbon filter.)

used to evaluate sensor behavior. The reported data an activated carbon filter on top. Measurements were
and trends were employed in designing the electronic carried out with CO concentrations set to 0, 75 ppm and
interface (see Section 4). For these measurements, the 150 ppm, respectively. We can see that the output cur-
sensors were biased at 1 V, to obtain sensor conduc- rent increases as the RH increases, and that the slope of
tance figures which are identical to the output current. the current as a function of the RH does not depend on
Fig. 3 shows how the relative humidity affects the CO concentration (at least for the concentration range
output current of a Sn02 Pt-doped RGTO sensor with considered).

116
 A Smart Sensor System 279

7 W- S
R.H.=40% [C01=150ppm
T=20°C IF
61O- s T sensor=400°C

5 IO- s

41O- s
$
3 10- 5

21O- s

1 IO- s

0
0 200 400 600 800 1000 1200 1400 1600

ALCOHOL (ppm)

Fig. 5. Sn02 Au-RGTO film conductance vs. C2HSOH concentration at 150 ppm of CO in air. (The sensor case did not have any activated
carbon filter.)

61O· s
R.H.=40% [C2HPH1=500ppm
T=20°C
5 IO- s T =400°C
sensor - IF

41O· s

$ 310- 5

21O- s

1 IO- s

o
o 20 40 60 80 100 120 140 160

CO (ppm)

Fig. 6. Sn02 Au-RGTO film conductance vs. CO concentration at RH = 40 % and [C2HSOH] = 500 ppm in air. (The sensor case did not have
any activated carbon filter.)

117
280 G. Cardinali et al.

Fig. 4 shows the conductance variation of the same pattern of a positive photoresist layer reproduces mask
sensor, as a function of the ethyl alcohol content in the apertures corresponding to the membrane in the silicon
air. The CO concentration is set to 150 ppm. This fig- nitride deposited on the back side of a double polished
ure shows that the Sn02 Pt-RGTO sensor with a carbon silicon wafer (500 or 300 /Lm thick). After this litho-
filter is almost insensitive to the alcohol at concentra- graphic step, the wafer is put in a KOH solution at
tions below 500 -;- 600 ppm and slightly dependent on T = 80°C and bulk silicon is anisotropically removed
alcohol concentrations above that value. through the silicon nitride openings. The process stops
Without the carbon filter, the sensor's response to when the silicon nitride deposited on the front of the
a gas mixture similar to the one in Fig. 4 changes wafer is reached.
completely and exhibits a pronounced sensitivity to To optimize the mechanical properties of the mi-
C2H5 0H. Fig. 5 shows the output current of a Sn02 cromachined membranes, their dimensions and fabri-
Au-doped RGTO sensor, which exhibits a quite linear cation processes have been considered with care. A
dependence for ethyl alcohol concentrations above 500 number of membranes, with sizes ranging from 1 x 1
ppm. to 2 x 2 mm 2, were realized varying the parameters
Fig. 6 shows the Sn02 Au-doped RGTO sensor re- of the silicon nitride deposition process: temperature
sponse as a function of the CO concentration, at fixed and gas ratio (diclorosilaneiammonia). A test was per-
amounts ofRH (40 %) and C 2H5 0H (500 ppm). We formed to investigate the mechanical stability of the
can see that for CO concentrations from 0 to 150 ppm, membrane by applying a differential pressure. The
the output current value of this sensor does not depend membrane deflects in a blister-like shape and deflec-
on the presence of CO in the environment. tion is a function of membrane size and intrinsic film
The reported set of experimental data contains suf- stress. Test results show that an increase in membrane
ficient technical information to engineer and optimize area or in the gas ratio leads to an increase in deflection.
the two sensors and ensure selective detection of both On the other hand, the deposition temperature does not
CO and ethyl alcohol. affect the deflection magnitude. Experimental charac-
terization of fabricated devices shows that a 1.5 x 1.5
mm 2 membrane, 250-nm thick, breaks at a maximum
3.2. Thin Film Membrane on Silicon overpressure of 100 kPa, while it deflects about 35 /Lm
when a pressure of 40 kPa above the atmospheric pres-
A thin film applied as mechanical support is an im- sure is applied.
portant feature in microsensor technologies. Essential
membrane requirements can be summarized as follows:
i) mechanical stability; 3.3. Heater Resistor and Temperature Sensor
ii) low stress material;
iii) low thermal conductivity; To set and measure the operating temperature of the
iv) high chemical inertness under the KOH etchant gas sensitive layers, a heating resistor and a tempera-
(the same film must work either as etch-stop al- ture sensor must be integrated on micromachined mem-
lowing the membrane to form or act as masking branes. Both components can be fabricated by selec-
layer during the silicon bulk micromachining pro- tively etching the same conductive film deposited on
cess); top of the surface of the nitride film (the membrane).
v) film deposition process and material compatible Sputtered platinum is the most widely used and charac-
with IC technology. terized material for this type of application. It exhibits
Many solutions have been proposed to realize thin high stability in temperature cycles and a quite high
film membranes [9-13]. They span from a sin- temperature coefficient that ensures intrinsic temper-
gle layer to a sandwich of dielectric layers like ature uniformity of the heater and provides a simple
Si02/Si 3 N4 /PSG, SiC. To fulfil these requirements way to obtain a temperature measurement by moni-
without increasing fabrication process complexity, we toring resistance variations. The temperature sensor
modified the silicon nitride low pressure chemical va- could be the heater itself, but design characteristics of
por deposition (LPCVD) process to obtain a low stress the temperature control electronics suggest separate re-
membrane film. Membrane fabrication relies on a sin- alization of these components would be more suitable.
gle lithographic step. Through reactive ion etching, the Indeed, at the operating temperature T = 400°C, the

118
 A Smart Sensor System 281

optimum resistance value for the heater is 100 Q, while Table 1. Resistance of passivation layers
the value for the temperature sensor is around 1 kQ.
Passivation thickness resistance
layer [nm] range
3.4. Passivation lAyer Si3N4 100 fewkQ

Following the scheme reported in Fig. 2, a passivation SbN4 210 fewMQ

layer is needed to electrically insulate the heater from LTO 300 fewkQ
the sensing film. The resistance of the passivation layer
must be high enough to ensure that leakage currents SbN4 + TiN 235 » GQ
are not affecting the measurement of the sensible film SOG+LTO 1500 »GQ
conductance at working temperature (150°C < T <
430°C) and at the applied bias (few volts). Moreover, SOG + LTO + TiN 1750 »GQ
the dielectric must have long term stability, to avoid
lifetime reduction of the device due to Sn or Pt diffusion
across the layer stack.
In standard processes, silicon nitride (SbN4) and low
temperature silicon oxide (LTO) films can be deposited
using the LPCVD technique. Other possible passiva-
tion layers are: stress-free spin-on-glass (SOG) and show that a reasonable trade-off between membrane
titanium nitride (TiN). size and conduction losses is obtained when this ra-
Insulation properties of passivation layers have been tio ranges from 2 to 3. At the same time, convection
thoroughly investigated by measuring the resistance and radiation losses depend on the size of the active
between the sensing layer and the Pt resistors in several area itself: the smaller the active area, the smaller the
dielectric configurations. Experimental results show heating power required. On the other hand, the size of
that tin diffusion through the passivation layer during the active area depends on the shapes needed to opti-
the tin oxidation process (30 hours at T = 600°C) sig- mize the response of the sensing film. An active area
nificantly degrades the insulating properties of a single with a 700 p,m side results from fabrication constraints.
passivation layer, making its resistance comparable to Starting from this active area and considering a 1500
that of the sensor. From data reported in Table 1, the p,m membrane side, the heating power needed to raise
best choice is 1 p,m of LTO on two SOG layers. Tita- the temperature of the active area to T = 400°C is
nium nitride also offers very high electrical resistance, calculated at about 100 mW.
but its use is not recommended because it adds a tensile To define heater geometry, the temperature distribu-
stress to the entire film stack and its realization process tion inside the active area was considered. Good uni-
is more complex. formity is required to optimize the sensing layer detec-
tion mechanism. The performance of variously shaped
heaters was evaluated by numerical simulation and by
3.5. Microstructure Design and Optimization means of experimental measurements. The meander-
shaped heater, widely proposed in the literature, ex-
As shown in Fig. 2, a platinum heater is placed at the hibits poor temperature uniformity. This is probably
center of a membrane, beneath the gas sensitive layer. due to the fact that the heating power per unit area is
The region where the sensing layer conductance is mea- almost uniform, while heat exchange follows a quasi-
sured by means of an interdigitized shaped contact, is radial distribution. An average gradient of 0.4°C; p,m
commonly referred to as the active area. The ratio be- was obtained. Better results have been achieved by the
tween the edge of the square membrane and the edge double spiral shaped resistor shown in Fig. 7(a). The
of the active area is a key figure of the device design. spiral layout has a variable pitch, to account for the
The greater the ratio, the smaller the losses due to con- distribution of heat losses inside the device active area.
duction heat transfer. Simulation results, carried out by Fig. 7(b) shows the simulated temperature distribution,
means of the three-dimensional finite-element analysis obtained with SOLIDIS. A minimum gradient of about
program SOLIDIS [14], and experimental observations 0.2°C/ p,m has been obtained.

119
282 G. Cardinali et at.

(a)

Temp
[C)

555
495
435
375
325
265
205
145
85
25

(b)

Fig. 7. Simulation results of a double spiral shaped heater: (a) geometry; (b) temperature distribution.

4. Design of the Electronic Interface easily conform to sensor characteristics and the require-
ments of the external alarm system.
The main target in the design of control and measure- The electronic driving circuitry will perform the fol-
ment circuitry was achieving the best trade-off between lowing four main functions.
performance and costs. To this end, the whole elec- i) Control logic. Flexibility and programmability
tronic interface was realized using a 1.2-/Lm CMOS during control and measurement steps must be
technology and it was made as flexible as possible to ensured.

120
 A Smart Sensor System 283

ii) Reference voltage generation. Accurate voltage perature swings in the control circuitry.
references must be generated for control and mea- ii) high rejection to supply variation, because a high
surement. voltage swing in the power supply may occur be-
iii) Heating control. During heating cycles, maxi- tween low and high temperature cycles.
mum accuracy and stability of the temperature is These two requirements are met by the band-gap volt-
required. age generator shown in Fig. 9. This circuit uses the
iv) Measurement and compensation circuitry. This base-emitter voltage VBE of a vertical bipolar transis-
function is based on bDo NO conversion and on tor (parasitic PNP BJT in N-well CMOS technology)
bitstream signal processing. This kind of elabo- to generate a reference voltage [15, 16]. It is well
ration requires a minimum amount of electronic known that the VBE voltage has a negative tempera-
circuitry to compensate spurious signals gener- ture coefficient. To compensate such a variation, the
ated by interfering gases. circuit relies on the positive temperature coefficient of
Fig. 8 shows the block diagram of the electronic inter- Do Vbe. The ratio RJ/ R3 is selected to cancel the first
face IC. order temperature effects. Low resistance values are
used to minimize the silicon area required. The nomi-
nal reference voltage is 1.205 V. The layout of passive
4.1. Control Logic components has been carefully designed, to minimize
mismatch effects. The variation of the generated volt-
Given the strong dependence of sensor behavior on the age is less than 1 mV within a 70°C temperature range,
operating temperature, measurement and control re- and is less than 0.2 mV for a 1 V supply variation.
quire a high degree of flexibility to optimize system
performance for different sensors. Therefore the con-
trol logic must allow the timing and duty cycles of
heating steps to be changed, as well as the number and 4.3. Heating Control
timing of measurements.
The timing section employs a 32 kHz quartz oscil- To operate in temperature pulsed mode, temperature
lator to generate a master clock signal with maximum stabilization is needed during heating cycles. As ex-
stability. As shown in Fig. 8, the 32-kHz crystal is the plained before, the heating system is constituted by two
only external component of the timing section. Two sputtered platinum resistors: the first is the heating re-
binary numbers corresponding to the length of each sistor R h , while the second is the measuring resistor
cycle are stored in the EEPROM (Fig. 1) to control Rm. Their nominal resistance values are 100 Q and
the programmability of the heating cycles. Since the 1000 Q, respectively. The operating temperatures are
time interval of each cycle will not exceed 128 s, and 400°C and 150°C. To reach 400°C during the heating
a time resolution of 0.5 s is considered satisfactory, cycle, at least 100 mW must be supplied to the heating
the master clock is divided by 2 16 , and the output of a resistor, that means that a current of 33 mA must be
synchronous 8-bit counter is compared with the mem- supplied by the control system. To furnish the heat-
ory contents. A toggle flip-flop switches between two ing current, a special output stage is added to a con-
states and generates the control waveform with the re- ventional single stage mirrored amplifier. The output
quested duty-cycle. Measurement steps are timed and stage consists of an open-drain PMOS transistor with
controlled in the same way. W / L = 1000. The heater is connected to the drain to
optimize the voltage swing.
Temperature is controlled through an "on/off'
4.2. Reference Voltage Generation switching circuit. The basic idea is to have, for a given
temperature, a reference resistor Rref which has the
A voltage reference generator is needed both to mea- same resistance as that expected for the measuring re-
sure the conductance of the sensors and in heating con- sistor.
trol. Such a voltage reference must have two important Hence Rm is compared with the reference resistor
key features: and the power is switched on (off) if Rm is smaller
i) high temperature stability, because the pulsed (higher) than Rref • The designed circuit implementing
mode operation of the heater may introduce tem- this control scheme is shown in Fig. 10. The tem-

121
284 G. Cardinali et al.

ext. trimming resi,filors

o==
ASIC ~
00
o
u

Bandgap
Voltage
Generator

~A v/"..,...,..",,,J..f.j v"'i
~ ~
~ ~
~
~
~
~
] E E E '" '"
~ ~ "'-
.., =
ESo '~" ~ '~"
~""'.J""'///'/"'..I'..I'''''''''''''/.z S-6 ~ ~ ~
.::: .::: .:::
QUARTZ ~

"" """ """ """

Fig. 8. Block diagram of the electronic interface circuit.

VREF2
(1.2 V)

VREF1
(1.1 V)

Fig. 9. Band-gap reference voltage generator.

122
 A Smart Sensor System 285

VDD

V REF1 V REF2
(1.1 V) (1.2 V)

temp. Rheater
cycle

R ref(150)

Fig. 10. Scheme of the "on/off' temperature control loop.

perature cycle is selected by switching between two 4.4. Measurement and Compensation Circuitry
reference resistors, according to the control logic sig-
nal. Let us suppose that the system has to perform the To obtain an electrical output which is only propor-
cycle at T = 400°C. Since the temperature is lower, tional to the CO concentration, the electronic interface
the value of the sensing resistance Rm is smaller than must compensate effects produced by alcohol and by
its reference value, R ref (400)' The output of the com- humidity on the Sn02 sensor. Before considering how
parator assumes the high state, and power is switched the conductance variation is converted into a readable
onto the heater. When the temperature exceeds 400°C, signal, it is worthwhile to analyze how such compensa-
the value of Rm becomes greater than the reference tion can be executed. The compensation scheme must
temperature and the comparator switches off the heat- take into account how interactions between interfering
ing element. In this control scheme the output PMOS gases and the sensing layer influence the response of
transistor behaves like a switch. A quite large output the sensor itself.
dynamic range is achieved, and the supplying power Let us denote the output of a CO sensor with an
capability is considerably increased. activated carbon filter on top with Sea, the output of a
The temperature accuracy of the proposed control sensor without a filter (specifically designed to reveal
circuit depends on the matching between the reference the presence ofC 2H5 0H) with SA/coho/, and the output
and the measuring resistors. To account for deviations of the humidity sensor with SRH.
of real parameters from design values, the two refer- By considering only signals from the CO and
ence resistors are discrete components (e. g. two multi- C2H5 0H sensors, we can write the following first-order
twist trimmers) external to the ASIC. The system can equation set:
be set to cycle between two given temperatures simply
by adjusting the values of these resistors.
Computer simulations confirmed that the design Sea = al . [CO] + b l . [C2H5 0H] (1)
{
specifications are also met in worst case conditions. A SA/coho/ = a2 • [CO] + b2 • [C 2H5 0H]
temperature control accuracy of about 2°C is achiev-
able if a tight thermal link exists between the heater and
the temperature sensor. This condition is satisfied in where square brackets [] are used to denote the gas
the device considered, due to the reduced dimensions concentrations.
of sensor microstructures. The equation set (1) can easily be solved to ex-
press [CO] as a linear combination of signals Sea and

123
286 G. Cardinali et al.

SAlcoho/. as follows: To conclude, the following set of linear equations

constitutes a good model of actual sensor behavior:
[CO] = b 2 Sea - bl SAlCOhOl. Sea = al . [CO] + b l . [C 2H s OH] + CI . [RH]
b2al - b l a2 { SAlcohol = b2 . [C 2H sOH] + C2 . [RH] (2)
SRH = C3 . [RH]
It is well known that this kind of operation can be
executed using ~!l techniques. Mathematical opera- By solving the equation set (2), we can express
tions are performed on high-frequency bitstreams gen- [CO] as a linear combination ofthe three signals Sea,
erated by second-order ~!l modulators from the sen- SAlcohol, and SRH as follows:
sor outputs [17, 18]. The resulting output bitstream
is integrated to obtain low frequency samples of the [CO] = 2. (Sea - bbl. SAlcohol- (b 2CI -b I C2) .SRH)
CO concentration evaluated. A diagram of signal pro- al 2 b2 C3
cessing is illustrated in Fig. 11. The~!l modulator This operation can be performed directly on bit-
architecture is described in [19]. stream signals. However, it is worth noting that the
Now let us consider how the results of prototype multiplication in time domain corresponds to a con-
sensing layer characterization fit the first-order equa- volution in z-domain. This means that the result of a
tion set (1). multiplication of two noise-shaped bitstreams in time
By comparing Figs. 5 and 6 we can see that the con- domain is a non noise-shaped waveform [18]. Conse-
ductance of the alcohol sensor is much more dependent quently multiplications must be performed before bit-
on C2H s OH concentration than on CO concentration. stream generation. The simplest method is to mUltiply
This means that the conductance dependence on the the analog signal by a bitstream corresponding to the
CO concentration can be neglected, and we can as- multiplication coefficient [20, 21]. Conversely, two
sume that this sensor is sensitive only to ethyl alcohol. bitstreams can be added using very simple digital cir-
Curves of conductivity as a function of alcohol con- cuitry without losing the noise shaping benefits [17, 18,
centration (Fig. 5) exhibit a logarithmic behavior that 22-24].
becomes linear when the ethyl alcohol concentrations The block diagram of the measurement interface is il-
are larger that 500 ppm. Hence in the second equation lustrated in Fig. 12. Three 8-bit digital coefficients K I ,
the coefficient a2 is set to zero. K 2 and K 3, defined on the basis of sensor characteriza-
Another point to be emphasized is the following. tion, are converted into bitstreams by three digital-to-
From Fig. 4, we can see that the conductivity of the bitstream ~!l converters. The impedance variations
sensor with carbon filter is practically insensitive to of the three sensors are sampled according to compen-
ethyl alcohol below 500 ppm, while above the 500 ppm sation coefficients and converted into three bitstreams
value it shows a weak linear dependence. Hence, the by analog-to-bitstream second-order 1:!l converters.
coefficient b l in the first equation has to be set to zero The resulting bitstreams are added to obtain a single
when the C 2H 5 0H concentration is below 500 ppm bitstream that can be low-pass filtered using a simple
and set to the appropriate value for higher alcohol con- digital accumulator. Bitstreams are added using the
centrations. This means that the effects of alcohol on interleaving technique [23]. Alternatively, three input
the sensor conductance need only be compensated with stages could be used in a single 1:!l modulator as pro-
concentrations exceeding 500 ppm. Moreover above posed in [21], but this approach limits the flexibility of
this threshold value the conductance can be linearly the whole system.
compensated. Resistive sensors are directly connected to the first
The conductance variation as a function of the rela- stage of the ~!l converter, as illustrated in Fig. 13.
tive humidity can be processed in a similar way. From A precise voltage reference is applied to the sensor,
Fig. 3 we can see that the CO sensor behaves linearly thus obtaining a current signal which is proportional to
with RH values in the range 25 -;- 90 %. The same the sensor conductance. The sensor signal is chopped
trend has been observed for the interfering gas sensor. by the digital generated bitstream and integrated into
This means that humidity effects can be easily compen- the first stage of the ~!l modulator. In this manner
sated in both sensors by using the procedure described the compensation coefficients are multiplied in mixed
above. The capacitive sensor used to measure the RH analog-digital domain, maintaining the noise-shaping
is insensitive to CO as well as to C2Hs OH. characteristics of modulator outputs [18, 20].

124
 A Smart Sensor System 287

sigma delta4

Dithering
Pulse
sigma dellaS Generator

Fig. 11. Scheme of alcohol compensation.

8 bit
compensation ,.....-_ _ _-,
coefficent
KI
,--_L-_--, digital
bitstream
[CO]

8 bit
compensation ,--_ _ _-,
coefficent
K2

[ALCOHOL]---_..I

8 bit
compensation ,--_ _ _-,
coefficent
K3
digital
bitstream
[RH]------1~1

Fig. 12. Block diagram of the measurement interface circuit.

Conductance variations must be converted into read- capacitor Caz' During the successive clock phase the
able signals by taking into account problems due to circuit acts as an integrator and the voltage stored on C z ,
offset and flicker noise. Flicker noise and offset were applied to the auxiliary input, cancels low-frequency
cancelled using a sampled data auto-zero technique. flicker noise and offset components.
During the auto-zero phase <I> AZ, the inputs of the first With the RH capacitive sensor we have a problem
integrator stage are connected to analog ground, the due to the large value of the sensor capacitance (around
integrating capacitor is disconnected, and the integra- 100 pF). Such capacitance is not compatible with the
tor output is connected to the negative auxiliary input. integrating capacitor, since it would saturate the inte-
The output voltage (equal to the input-referred offset grator output because of the over-large ratio Csens / Cj •
and flicker noise) is sampled and held on the auto-zero We overcame this problem using the circuit shown in

125
288 G. Cardinali et al.

Ci ~AZ
.-----I~

Vrefo---j sensor r----~-+----l

to second
in + stage

aux in+

feeedback path
L -_ _ _ _ _ _ _ _

.... T
--~

<PAZ

<PAZ

bitstream

Fig. 13. Scheme of conductance measurement interface with auto-zero compensation.

Fig. 14. An additional capacitance (the reference ca- High-level simulations of the measurement interface
pacitor C ref ) is switched like the RH sensor, but to- have been performed using MATLAB [25] and TOSCA
wards a voltage opposite to the sensor bias voltage. [26].
Consequently the charge injected into the integrating Fig. 15 illustrates the results of the TOSCA simu-
capacitor is: lation of the multiplication of a sinusoidal waveform
by a constant coefficient. From the power spectral
Q = I::!C· Vref density of the resulting bitstream, we see the typ-
ical noise-shaping characteristics exhibiting a low-
where frequency noise level 60 dB below the signal. Fig. 16
illustrates the resulting spectrum of a chopped resistor
measurement, obtained by circuit level simulation in
time domain and fast Fourier transform (FFT). The two
This operation prevents integrator saturation.

126
 A Smart Sensor System 289

m
to second
stage

feeedback path

Fig. 14. Scheme of capacitance measurement interface with auto-zero compensation.

,,
I I I I
I I I I I I I
-20 _ ___ ____ .1. ________ . ' . ________ .1. ________ . ' . ________ .1. ________ .1. ________ .J. ___ __ _
I I I I I , ,
I I I I I
I I I I I

-40

·60 , , ,,
-----~---, ,
-----~---------,-----

,
I I I I I I
·80 _.1.. _____ __ _ .1. _ _ _ _ _ _ _ _ . , . _ _ _ _ _ _ _ _ ..I. ________ ..I. ________ .J. ______ _
I I I
,,
I
,,
I I

-100 ~~~noTTTrrrnnTT.rnnnnTTTTrrnnTTTTrrnnTTTTrrrnnnTTTTrnnnTTrrrrinnTTTTrrrl

o lk 2k 3k 4k 5k 6k 7k 8k

Frequency [Hz]

Fig. 15. Spectrum of the bitstream signal for a single sensor output (TOSCA simulation).

127
290 G. Cardinali et at.

,,,
, ,
---- --- - -:- - - - - - - - - -:- - - - - - - - - -:- - - - - - - - - -:- - - - - - - - - ---- ---- -:- - - - - - - - - ------ ---
I I I I I t I
-20 ~- ~
,
I
,,
I I
,,
I

-40

, , ,
-60 I I I I t I
- - - - -,- - - - - - - - - -,- - - - - - - - - -,- - - - - - - - - -1- - - - - - - - - -.- - - - - - - - - , - - - - - - - - -

, ,,
I I I I I I ,
___ _ .1.. ________ .'- ________ .1. _ _ _ _ _ _ _ _ .1. _ _ _ _ _ _ _ _ ..I. ________ ... _________ .J _____ ___ _
-80
I I I I I I I

,,
I I I
,
I

o 1k 2k 3k 4k 5k 6k 7k 8k

Frequency [Hz]

Fig. 16. Spectrum of the bitstream signal for a single sensor output (circuit-level simulation).

simulation results are in good agreement, and demon- temperature mode, with cycles of 150 s. To evaluate
strate the validity of the proposed approach for gas the magnitude of the stress applied, we point out that
sensor signal processing. the membrane, when heated at T = 450°C, deflects
1 -;- 2 f.1.m while during the differential pressure test
5. Experimental Results it was deflected by tens of microns. All the tested de-
vices exhibited excellent mechanical stability and none
To evaluate sensor performance under different oper- broke during the tests, proving the long-term reliability
ating conditions and to optimize the whole system, the of the membrane.
basic components have been fabricated and character-
ized separately. This section describes some experi-
mental results allowing estimation of system behavior.
600

500

5.1. Characterization of the Microstructure 400

L\ TH
I'C] 300
The blister test was also used to establish the maximum
pressure that the ShN4 membrane can sustain without 200

breaking. This is useful in estimating the long-term 100

membrane stability under operating conditions. The
o~--~--~~--~--~----~--~--~
stress was applied by ramping a differential pressure o 50 100 150

up to one quarter of the breaking pressure, at a fre- P HlmW]

quency of 100 cycle/min. The number of applied cy-

cles was about 200000, equivalent to 1 year in pulsed Fig. 17. Temperature increase in active area versus heating power.

128
 A Smart Sensor System 291

U 1,6 U 1,4
0
N 0
N
:Ii 1,4 :Ii
1,3
Ii Ii 1,2
~ ~
0 0
\0 \0
1,2 1,1 30% R.H. 20°C
0
U 0
u '\
S S
8: '\ 0..
0
0 "-
0..
0
0
0..
0,9
b ,
N N Ct,
.:z.0% R.H. 20°C
~ 0,8
" 0..
().. ~ 0,8 "().

'"
OJ)
IX
--0--0-_
'"
OJ)
IX 0,7 70% R.H. 20°C
0,6
°
0,6
100 200 300 400
ppm CO
500 600 700 800
° 100 200 300 400
ppm eo
500 600 700 800

Fig. 18. Sensor response to CO concentration in constant temperature Fig. 19. Sensor response to CO concentration in pulsed temperature
mode. mode.

Further mechanical and electrical investigations has been thoroughly investigated by means of elec-
were carried out on the complete structure (sensing trical and thermographic measurements. The depen-
layer included). In particular, an accelerated test was dence of the active area temperature on heating power
performed where the device was stressed pulsing the has been determined (Fig. 17). A total power of about
temperature between 200°C and 500°C each 2 s for 100 mW is required to heat the active area at 450°C
10 days. The results of the accelerated test have con- and 35 ms is the time interval needed to switch from
firmed both the excellent mechanical stability of the 200°C to 500°C. The thermal variation of the platinum
whole system and the considerable endurance of the resistance has been investigated by electrical measure-
passivation layer. ments performed on the real device and on test struc-
The thermal behavior of the sensor microstructure tures, maintained at a uniform and well-controlled tem-

RH =40%, V_heater =7 V
10 -s *---+ 180 ppm Alcohol + 100 ppm CO

~,

180 ppm Alcohol + 100 ppm CO

100 ppm CO ",
4---.a "
Air ...
!:=.
--------=: 100 ppm CC ~
---. .~ir ,
'"
,~

" ,
,
~
"

without filter with filter

10-6
60 65 70 75 80 85 90 95 100
Days

Fig. 20. Comparison of sensor responses to CO and alcohol (with filter and without filter).

129
292 G. Cardinali et at.

Tek 1111111150.0k5/s perature (±O.I°q. A positive temperature coefficient

I
ex = 2.71 ± 0.06°C- 1 has been evaluated.
(2 RMS
1.166 V

(2 Max
5.2. Characterization of the Sensing Films
2.67 V

, C2 Period
Experimental results on the stationary response of the
2.544ms sensing layers to different gas concentrations, are sum-
marized on section 3.1. Here some aspects of the tech-
niques used to reduce the sensitivity of the CO sensor
to interfering gases, such as relative humidity and ethyl
alcohol, are considered.
25Jun 1996 Figs. 18 and 19 show the responses of the same sen-
10:H:25
sor at two different levels of RH using the constant and
(a) the pulsed temperature procedures. The sensor output
is normalized to a reference condition of 200 ppm of
Tek 1111111150.0kS/s 2 Acqs CO and 60% of relative humidity. The pulsed tem-
f---------l---·-·-·----j
perature mode decreases the RH effect on the sensor
response, without affecting CO sensitivity.
C2 RMS
1.192 V Ethyl alcohol interference can be minimized by us-
ing an activated charcoal filter (by SUPELCO, Inc.),
C2 Max
3.61 V positioned inside the sensor housing, in front of the
sensing layer. Experimental results (Fig. 20) show that
(2 Period
4.136ms
Unstable
filtering is very effective, especially at low alcohol con-
histogram
centrations, without any appreciable efficiency reduc-
tion after several months of activity.

m . ms 25 Jun 1996 5.3. Characterization of the Electronic Interface

10:21:17

(b) The blocks described in Section 4 have been character-

ized separately.
The on-off regulator which controls the membrane
temperature has been proven to work properly, with a
maximum output current equal to 40 mAo The maxi-
C2 RMS
1.214 V mum supplied heating power is 160 mW. The wave-
t
,
'r
forms at the regulator output, with three different bias
+ (2 Max
voltages: VDD = 3 V, 4 V and 5 V, are shown in
·f 4.34V
T
Figs. 21(a), (b) and (c) respectively. From the figures,
>+++H"H+>++'~f+++4 . +i-" i , , , , i ' , , C2 period
4.S60ms it is clear that the average heating power is independent
Unstable
histogram of the bias voltage.
The three 1:: i"!. modulators have been tested sepa-

~l T
rately. They have been proven to work properly and
in agreement with the project specifications. Each 1:: i"!.
• Lo09d M{Obms eti2'7 2.169'251un Ig96 modulator output is a bit stream, whose low-frequency
10:25:21
analog content is proportional to the sensor conduc-
tance multiplied by the sampling coefficient.
(c)
Fig. 22 shows a single bit stream together with its
Fig. 21. Wavefonns at the temperature control output for differ.. power spectrum, obtained at the oscilloscope through
ent values of supply voltage: (a) VDD = 3 V; (b) VDD = 4 V;
(c) VDD = 5 V. Hanning window and 1024-point FFr. Noise shaping

130
 A Smart Sensor System 293

Fig. 22. A single output bitstream and its power spectrum (obtained at the oscilloscope through FFT).

is in good agreement with simulation results in Figs. 15 Acknowledgement

and 16. The low-frequency noise level flattens at -57
dB, due to the analog noise superimposed to the digital This work has been supported by the European Union
bit stream. The resolution required (7 bits) can easily under ESPRIT Projects 7500-MEPI (Demonstrator D-
be achieved by digitally filtering the bit stream with a 114) and 8795-AMFIS.
narrowband low-pass filter. By using a simple comb
filter that integrates the output bitstream over 0.5 s,
References
the resulting signal-to-noise ratio is better than 40 dB,
corresponding to a maximum error which is less than 1. G. Williams and G. V. Coles, "NOx response of tin dioxide
1% of full scale. based sensor," Sensors and Actuators B 15-16, pp. 349-353,
1993.
2. Philips, Technical Note 134 for Capacitive Humidity Sensor
(catalogue number 2322 691 90001).
3. G. Sberveglieri, G. Faglia, S. Groppelli, and P. Nelli, "A new
technique for preparing Sn02 sputtered thin film gas sensors,"
6. Conclusion in Proc. Transducers '91, 6th Int. Con! on Sensors and Actu-
ators, San Francisco, CA, USA, 1991, pp. 165-168.
This paper presented a multisensor architecture for car- 4. G. Sberveglieri, "Classical and novel techniques for the prepa-
ration of Sn02 thin film gas sensor," Sensors and Actuators B
bon monoxide detection. Experimental characteriza- 6,pp.239-247,1992.
tion activities allowed us to define and, in some cases, 5. C. Di Natale, A. D'Amico, F. Davide, G. Faglia, P. Nelli, and
to optimize the different technological steps involved G. Sberveglieri, "Performance evaluation of an Sn02-based
in the realization of a sensor prototype. An electronic sensor array for the quantitative measurement of mixtures of
H2S and N02," Sensors and Actuators B 20, pp. 217-224,
interface has been designed, both to generate the con- 1994.
trol signals and to process sensor outputs. By using 6. G. Sberveglieri, "Recent developments in semiconducting
signal processing in the bitstream domain, hardware thin-film gas sensor," Sensors and Actuators B 23, pp. 103-
109, 1995.
complexity is greatly reduced whitout sacrificing sys-
7. P. Bonzi, L. E. Depero, F. Parmigiani, C. Perego, and
tem flexibility, while a satisfactory signal-to-noise ratio G. Sberveglieri, "Formation and structure of tin iron oxide thin
is maintained at low frequencies. film CO sensors," J. Material Res. 9, pp. 1250-1256, 1994.

131
294 G. Cardinali et al.

8. L. E. Depero, F. Parmigiani, C. Perego, G. Sberveglieri, and

A. Taroni, "Sn(l-x)CoxO y : a mixed oxide for carbon monox-
ide detection," in Tech. Digest Micro System Technologies '94,
Berlin, Gennany, 1994, pp. 439--444.
9. M. Sekimoto, H. Yoshihara, and T. Ohkubo, "Silicon ni-
tride single-layer x-ray mask," 1. Vac. Sci. Technology 21(4),
pp. 1017-1021, 1982.
10. E.!. Bromley, 1. N. Randall, D. C. Flanders, and R. W. Moun-
tain, "A technique for the determination of stress in thin
films,"J. Vac. Sci. Technology B I, pp. 1364-1366, 1983.
II. M. G. Allen, M. Meheregany, R. T. Howe, and S. D. Sen- Gian Carlo Cardinali was born in Lugo (Ravenna),
turia, "Microfabricated structures for the in situ measurement Italy, in 1953. He received a degree in electronic engi-
of residual stress, Young's modulus, and ultimate strain of thin neering from the University of Bologna in 1979. Since
films," Appl. Phys. Letter 51, pp. 241-243, 1987.
October 1979 he has been working at the Lamel Insti-
12. I. George, P. Carmeli, B. Bonvalot, M. Wagener, A. Girard,
A. Zarudiansky, and 1. Suski, "Thin membranes with optimized tute of the Italian National research council in Bologna.
thermomechanical properties for microsystem applications," His scientific interests are in the area of design, fabri-
in Proc. EUROSENSOR VIlI Conf., Toulouse, France, 1994, cation, and testing of semiconductor devices and in-
vol. 2, pp. 38--42.
tegrated circuits. He is currently involved in research
13. S. Moller, 1. Lin, and E. Obenneier, "Material and design con-
siderations for low power micro heater modules for gas sensor projects concerning the applicability of microelectron-
applications," in Proc. Fifth Int. Meeting on Chemical Sensors, ics design methodologies and fabrication technologies
Rome, Italy, 1994, vol. I, pp. 433--436. to the realization of smart microsensors. He also dealt
14. 1. G. Korvink, The SOLIDIS User Manual: Version 1.0. PEL-
with physics of photovoltaic conversion and its applica-
ETH Zurich, 1995.
15. E. A. Vittoz, "MOS transistors operated in the lateral bipolar tions. Now he is head of the Department of Electronics
mode and their application in CMOS technology," IEEE J. at Lamel Institute.
Solid-State Circ. 18, pp. 273-279,1983.
16. P. E. Allen and D. R. Holberg, CMOS Analog Circuit Design.
Holt, Rinehart and Winston: New York, NY, USA, 1987.
17. F. Maloberti, "Non conventional signal processing by the use
of sigma delta technique: a tutorial introduction;' in Proc.
IEEE Int. Symp. on Circ. and Syst., San Diego, CA, USA,
1992, pp. 2645-2648.
18. V. F. Dias, "Signal processing in the sigma-delta domain,"
Microelectronics Journal 26, pp. 543-562,1995.
19. V. Liberali, F. Maloberti, and D. Tonietto, "Sigma-delta pro-
cessing in multisensor systems for carbon monoxide detec-
tion," in Proc. IEEE Int. Symp. Circ. and Syst., Atlanta, GA,
USA, 1996, vol. IV, pp. 376-379. Leonello Dorijoined the CNR-LAMEL Institute in
20. F. Op't Eynde, "A power metering ASIC with a sigma-delta-
1969. In 1980, he received the master degree in Physics
based multiplying ADC," in IEEE Int. Solid-State Circ. Con!
Dig. of Tech. Papers, San Francisco, CA, USA, 1994, pp. 186- at the University of Bologna. With the IBM visiting
187. scientist program, he was 4 years at the IBM-Thomas
21. G. van derHorn andJ. H. Huijsing, "Smart temperature sensor J. Watson Research Laboratory in Yorktown Heights
with integrated bitstream calibration," in Proc. Eur. Solid-State
(NY, USA) working on gate dielectric structures of
Circ. Conf., Lille, France, 1995, pp. 234--237.
22. N. Kouvaras, "Operations on delta-modulated signals and their DRAM or Flash memory devices. At the present time
application in the realization of digital filters," Radio and Elec- he is working on sensor and micro systems R&D pro-
tronic Engineer 48, pp. 431-438, 1978. gram of the LAMEL Institute.
23. P. O'Leary and F. Maloberti, "A bitstream adder for over-
sampling coded data," Electronics Letters 26, pp. 1708-1709,
1990.
24. P. Ma1covati, C. Azeredo Leme, P. O'Leary, F. Maloberti,
and H. Baltes, "Smart sensor interface with AiD conversion
and programmable calibration," IEEE 1. Solid-State Circ. 29,
pp.963-966,1994.
25. The MathWorks, Inc., MATLAB user's guide. 1993.
26. V. Liberali, V. F. Dias, M. Ciapponi, and F. Maloberti,
"TOSCA: a simulator for switched-capacitor noise-shaping
AiD converters," IEEE Trans. Computer-Aided Design of In-
tegr. Circ. and Syst. 12, pp. 1376-1386, 1993.

132
 A Smart Sensor System 295

Photo
not available
at time
of print

Cesare Perego received his M.S. degree in physics

Mauro Fiorini was born in Bologna in 1966. He
from the University of Milan (Italy) in 1990 discussing
graduated with Laurea degree in Phisycs from the Uni-
a thesis on FTIR and Raman measurements. From
versity of Bologna in 1995. Now he is working at
1990 to 1996 he was with the Thin Film Laboratory of
C.N.R. LAMEL institute (Bologna, Italy) with a schol-
the University of Brescia working on the development
arship from Philips Automation to develop a benzene
of solid state sensors. His main interests are in the
sensor with Sigma Delta NO interface.
field of thin layer deposition, and development of gas
sensors devices based on mixed oxide. In 1996 he
joined Steel Tecnica.

Isabel Sayago received the Master degree in Physics

from the Universidad Complutense de Madrid, in 1993.
Since then, she has been working on chemical sen-
sors at the Laboratorio de Sensores (CSIC). She has
worked on microsensor at the C.N.R. LAMEL Insti-
tute (Bologna, Italy) under a Spanish MEC post-doc
program. Giorgio Sberveglieri was born in Reggio Emilia,
Italy in 1947. In 1968 he received his Diploma de-
gree in Electrical Engineering. In 1970, he joined
the Montedison Corporation at the Donegani Research
Centre in Novara. In 1987 he was appointed Associate
Professor in General Physics at the University of Bres-
cia (Faculty of Engeneering). Since 1987, G. Sberveg-
lieri is leading the Gas Sensor Lab. at the Department
of Chemistry and Physics of Material at the same Uni-
versity. In 1994 he became full professor in General
Guido Faglia has received an M.S. degree from the
Physics and Director of the INFM (National Institute
Polytechnic of Milan in 1991 with a thesis on gas sen-
for the Physics of Matter) research unit (18 scientists)
sors. Since then he has been studying to obtain the PhD
at the University of Brescia. G. Sberveglieri's main
in electronics by the University of Brescia. In 1992 he
research interests are on the deposition and electrical
has been appointed as a Researcher by the Thin Film
characterisation of thin films of gas-sensitive materi-
Lab at the University of Brescia. he is involved in the
als using PVD techniques. He has published more
study of the interactions between gases and the tin ox-
than 120 papers in international scientific magazines
ide surface and in sensor electrical characterization. In
and congress proceedings on many subjects of mate-
1996 he has received the PhD degree by discussing a
rial science, particularly on electronic devices based on
thesis on semiconductor gas sensors.
semiconducting thin films.

133
296 G. Cardinali et al.

for telecommunication and sensor systems, CAD for

analogue and mixed A-D design. Dr. Maloberti has
written more than 180 published papers, 2 books and
holds 14 patents (2 pending). He was the recipient of
the XII Pedriali Prize (1992) for his technical and scien-
tific contributions to national industrial production. He
was co-recipient of the lEE Fleming Premium (1996)
for the paper "CMOS Triode Transistor Transconduc-
Valentino Liberali was born in Broni in 1959. He tor for high-frequency continuous-time filters". He re-
graduated with the Laurea degree in Electronic En- ceived the Dr. Honoris Causa in Electronics awarded
gineering from the University of Pavia in 1986. In by Inaoe (lnstituto Nacional de Astrofisica, Optic a y
the same year he received a one-year scholarship from Electronica), Puebla, Mexico in November 1996. He
SGS (now SGS-Thomson Microelectronics), within has been responsible, at both technical and manage-
the frame of the Piano Nazionale Microelettronica. ment levels, for many research programmes including
From 1987 to 1990 he was with Italian Nuclear Physics 10 ESPRIT projects. He organised the first Special Ac-
Institute (INFN) working on the development and char- tion for Microelectronics in Italy (I-SMILE, ESPRIT
acterisation of low-noise electronics for particle detec- frame); he was the CEC representative in the advisory
board of Special Actions I-SMILE, MEPI and SUMIS;
tors. In 1990 he joined the Department ofElectronics of
the University of Pavia, where he worked on the devel- he was in the Steering Board of the European Actions
opment of the simulator TOSCA for sigma-delta NO EUROCHIP and Combined Action; he was member of
converters. His main research interests are the design the Executive Board of the ESPRIT Project MEDCHIP.
of analog/digital interfaces, CAD for analog and mixed Dr. Maloberti also served the European Commission as
integrated circuits and mixed signal testing. V. Liber- ESPRIT projects' evaluator and reviewer. He served
ali is a member of AEI (Associazione Elettrotecnica ed the Academy of Finland (1996) for the evaluation of the
Elettronica Italiana). He has been elected to the Aca- electronic research in Finland. Dr. Maloberti is a mem-
demic Senate of the University of Pavia for the period ber of the AEI (Italian Electrothecnical and Electronic
1997-1999. Society), a Fellow of the IEEE (1996), Vice-President
Region 8 IEEE-CAS (1995-97), a member of the Ed-
itorial Board of Analog Integrated Circuits and Signal
Processing, a member of ESSDERCIESSCIRC Steer-
ing Committee, and he was a member of the Steering
Committee of the CEC Network NEAR (Network of
European Analogue Research).

Franco Maloberti received the Laurea Degree in

Physics (Summa cum Laude) from the University of
Parma in 1968. He joined the University of L' Aquila,
then the University of Pavia. For the years 1975-79, he
was technical co-ordinator of the Engineering School
at the University of Mogadishu, Somalia. He is cur-
Davide Tonietto was born in Roma, Italy, in 1970.
rently Professor of Microelectronics at the University
He received the Laurea Degree in Electronic Engineer-
of Pavia, and is also head of the Micro Integrated Sys-
ing from the University of Pavia, Italy, in 1995. In
tems Group. His professional expertise is in the design,
the same year he received a scholarship from INFM,
analysis and characterisation of integrated circuits and
Italy, and he worked on development of an innovative
analogue digital applications, mainly in the areas of
carbon-monoxide smart sensor interface into ESPRIT
switched capacitor circuits, data converters, interfaces
Project 7500 MEP!.

134