2030
IEEE JOURNAL OF SOLID-STATE CIRCUITS, VOL. 36, NO. 12, DECEMBER 2001
A CMOS Multiparameter Biochemical Microsensor With Temperature Control and Signal Interfacing Erik Lauwers, Jan Suls, Walter Gumbrecht, David Maes, Georges Gielen, and Willy Sansen, Fellow, IEEE
Abstract—A fully integrated multiparameter microsensor chip is presented for continuous monitoring of concentrations of different blood gases (e.g., pH, pO2 , pCO2 ), ions, and biomolecules, and for conductometric measurements. The chip can monitor up to seven different chemical substances depending on the membranes deposited on the sensor units (on-chip ion-sensitive fieldeffect transistors (ISFETs), amperometric and conductometric cell). The sensors, which are positioned in a flow channel, are surrounded by on-chip interfacing and processing electronics so that external readout goes via a simple data acquisition card. In addition, temperature control of the measured fluids and a onetime-use security check have been provided for proper operation. Fabrication was done in a standard 1.2- m CMOS process to which extra postprocessing steps have been added for the chemical sensors and membranes. The chip operates at 5 V and the total die area is 25.7 mm2 . Full integration is obtained including the ISFETs and ISFET buffers, as well as a reference electrode structure, all integrated on the same chip in the same technology. Index Terms—Amperometric and conductometric cell, biochemical, CMOS microsensor, EPROM, ISFET, signal interfacing, temperature control.
I. INTRODUCTION
M
ASS production, high yields, and low manufacturing costs are concepts readily associated with electronics integrated on a single piece of silicon. Other considerations such as low power, high speed, etc., are also motivations for silicon integration. Apart from these factors, small dimensions and portability are of great value in the field of medical healthcare. Analysis of blood gases in intensive-care units is common practice nowadays. For this analysis, many samples can be required every day, consuming a lot of blood, time, and resources. Continuous monitoring of blood gases is therefore a major improvement for critical-care patients and reduces the amount of blood samples needed. For many other applications, as for example, in bioreactors, the possibility to perform continuous measurements is also advantageous. This paper presents the implementation of a complete microsensor system for the continuous monitoring of ions, dissolved gases, and biomolecules. Even more functionality has been integrated on chip, such as a conductometric sensor, an on-chip absolute temperature control, and a one-bit EPROM Manuscript received March 22, 2001; revised July 20, 2001. This work was supported by the Brite-Euram COMMONSENS project. E. Lauwers, J. Suls, G. Gielen, and W. Sansen are with the Katholieke Universiteit Leuven, ESAT-MICAS, 3001 Leuven-Heverlee, Belgium (e-mail:
[email protected]). W. Gumbrecht is with Siemens AG, Erlangen, Germany. D. Maes is with IMEC, 3001 Heverlee, Belgium. Publisher Item Identifier S 0018-9200(01)09334-9.
for medical security reasons. The full system is processed in a 1.2- m single-metal single-poly CMOS technology and operates at 5 V. Extra postprocessing steps have been added to manufacture the sensor (interface) structures and a calibration system. The total chip area is 4.11 mm 6.25 mm. The process technology used was explained in detail in [1]. In this paper, the overall system and the electronic subblocks are explained in detail and measurements are presented. Minor additional details can be found in the visuals supplement of [2]. The paper is organized as follows. Section II presents an application for the microsensor chip and explains the global system setup. In Section III, the sensors are described, and in Section IV, the sensor interfacing electronics. Section V covers the heat regulation system, and Section VI explains the control electronics and the EPROM. In Section VII, chemical measurements are presented, and finally, in Section VIII, conclusions are drawn. II. TOTAL MICROSENSOR SYSTEM A typical use of the microsensor is measuring gas concentrations in blood. The chip is packaged such that the central part, where the sensors are aligned, is exposed to a flow channel where the blood and calibration solutions run through. This sensor alignment (centerline a–h along the long side) and the flow-channel perimeter around it (oval shape for O-ring to seal off the flow channel) can clearly be seen in Fig. 1. For a small moment during measurement, the flow is stopped such that a small blood sample is trapped at the sensor interface. After a short period needed to make sure that the sample acquires the correct temperature, measurement results are read in by a monitor. This monitor can be as simple as a laptop with a data acquisition card. Keeping the temperature of the measured samples constant is important for reproducibility of the results and for comparison with other measurements (eventually performed elsewhere). The main goal of the sensor development is twofold. First, the continuous monitoring of blood gasses must be possible with integrated sensors. Second, the system may only have a minimal amount of external connections in order to keep the wire connection to the monitor small. The minimal amount of external and ), connections is five: the power supply connections ( the clock signal, and the I/O connections. Three extra wires have been added so that a standard eight-wire connection is obtained (Fig. 2). The three extra wires are an external reference voltage for reasons of controllability, an external reference biasing current, and an extra ground connection to have a good analog reference even with currents up to 100 mA. Various versions of the system with different sensor configurations have been manufactured. Because from an electronic point of view, these ver-
0018–9200/01$10.00 © 2001 IEEE
LAUWERS et al.: CMOS MULTIPARAMETER BIOCHEMICAL MICROSENSOR
Fig. 1.
2031
Layout of the total system.
Fig. 2. Block diagram of the sensor chip.
sions do not differ much, only one is highlighted here and will be fully explained. Currently, commercial biosensors exist that provide the same range of possible measurements. A prominent commercially available biosensor is the I-STAT from the I-STAT Corporation. This biosensor works mainly according to the same principles, however, there are two important differences. The sensors are microfabricated thin-film electrodes, whereas in this work, fully integrated sensors are used. The I-STAT does not allow for continuous measurements, but uses blood samples to be placed in cartridges allowing nearly instantaneous measurements. In Fig. 2, a block diagram of the sensor chip is given. It can roughly be divided in three parts. The first part contains the sensors and their interfacing electronics. The second part contains the extra functionalities: temperature control and an EPROM.
The third part contains all the electronics necessary to control the chip and the external communication as well as the biasing. These different subsystems will be explained in Sections III–VI. III. SENSORS The chip has eight integrated sensors, indicated by letters in Fig. 1. Seven sensors are located in the centerline and one is above the centerline (e). The number of eight is determined by the available space in the flow channel. They can all be operated in parallel if required by the user. However, the idea is to make one standard chip to reduce manufacturing costs and then to program the chip according to the needs of the user. For example, if the oxygen sensor output is made available only at the chip’s output by use of an EPROM similar to the security
2032
IEEE JOURNAL OF SOLID-STATE CIRCUITS, VOL. 36, NO. 12, DECEMBER 2001
Fig. 4. Transistor schematic of the ISFET buffer: a modified symmetrical OTA with pMOS source follower.
Fig. 3. ISFET calibration setup.
IV. INTERFACE ELECTRONICS A. ISFET Buffer Amplifier
EPROM, then the user only has to pay for this reduced functionality. Dimensions of the individual sensors are determined by chemical considerations from previous prototypes. The size of the micropool in polyimide around the sensor’s active area is determined by the used selective-membrane dispensing technique. More information concerning the membrane dispensing is given in [3]. It was decided to integrate six ion-sensitive field-effect transistors (ISFETs) (b, c, d, e, f, and g in Fig. 1), one oxygen sensor (h), and one conductometric sensor (a). An ISFET is basically a field-effect transistor of which the gate metalization has been omitted. The operation is based on direct contact of the electrolyte with the gate oxide. The adsorption of charged species at the solution–oxide interface is measured. The conductometric sensor, which measures the electric conductivity of an electrolyte solution, is built with two parallel sensors of which one electrode is shared (the middle one). A sinusoidal input signal is applied at the electrodes and the current through the electrodes which is dependent on the composition (number of charged elements, ions and their properties) of the solution, is measured. The ISFETs can be used to measure the voltages between the conductometric electrodes, bypassing the (interface) impedances of the current providing electrodes, hence allowing a four-point conductometric measurement. Other versions manufactured contain different sensor units such as an enzymatic or pCO sensor. The sensor chip continuously monitors ions, dissolved gases, and biomolecules. Traditionally, external reference electrodes are used for stable measurements [4]. However, stable potentials can only be obtained with a fixed chloride concentration in the sample solution. To tackle this problem, a reference electrode with an Ag–AgCl interface has been integrated in a bypass structure (Fig. 1, oval structure around a and e) next to the flow channel. The layout of this bypass structure is based on a previous design to monitor blood gasses [5]. The ISFET calibration setup is schematically shown in Fig. 3. During a calibration cycle (pump two is on, pump one is off), the bypass structure traps the calibration fluid on top of the reference ISFET. During measurement (pump two is off, pump one is on), this calibration solution then remains in place while all the other ISFETs see the fluid being measured. Different solutions for this reference problem have been published, such as, for example, a solution based on different pH-sensitive electrodes [5].
The response of an ISFET can be measured in two ways. The (Fig. 4) can be held constant and the change applied voltage in the drain current is measured as a function of the ion activity. can be changed in such Alternatively, the applied voltage a way that the drain current remains constant. Here, the current is kept constant and the gate voltage of the ISFET is measured through a buffer and then sent to the monitor for analysis. Also, for correct operation the drain-to-source voltage of the ISFET has to be kept constant. Both operations are combined in an ISFET buffer with low offset, which is a modified symmetrical OTA with pMOS source follower (Fig. 4). The ISFET and its buffer are integrated in one process and on the same device to enable a more stable operation. By introducing the bootstrap transistors T2a,b using the same technology as T1a,b, i.e., a normal CMOS FET but with an oxide–nitride gate dielectric and a Pt gate, the drain–source voltages over transistors T1a,b are kept constant. The bulk effect of transistors T1a,b is cancelled out by making this effect equal for transistors T2a,b and so the drain–source voltage of T1a,b is kept constant. The pMOS source follower T5 is added to match the input and output voltage swing and also to lower the output impedance. Measurements on previous prototypes indicated that this design is very robust toward technology variations and can handle an input (output) voltage from about 2 V up to 5 V with a supply of 5 V. The ISFETs are biased in the correct linear operating point . It was chosen to directly apply by an external connection this voltage to make sure that it is a well-controlled electrical signal. The actual measured voltage at the solution–oxide inter. In addition, this allowed having a face is indicated by 1:4 demultiplexer instead of a 1:5 demultiplexer, which in turn allowed an easier design for the controller, as will be explained in Section VI. B. Potentiostatic Setup In Fig. 5, a schematic of the potentiostatic setup is given for an amperometric measurement. The counter and working and reference electrodes are indicated by the letters , , and , respectively. The actual open sensor window is sketched by the circle around them. To prevent polarization of the reference due
LAUWERS et al.: CMOS MULTIPARAMETER BIOCHEMICAL MICROSENSOR
2033
electrode [Fig. 1(a)] is connected to an opamp and resistor structure (see left-hand side of Fig. 6) to convert the current into an output voltage. Limited by the individual blocks concerned (the ISFETs and the converter), signal frequencies up to 5 kHz can be applied. This relatively low frequency originates from low offset specifications for the different building blocks, which necessitate large transistors and therefore limit the gain bandwidth (GBW). Fig. 5. Schematic view of the potentiostatic setup.
to current drawn through it, an inert counter electrode is introduced to provide the current for the reduction (or oxidation) at the working electrode. The voltage at the counter electrode is regulated such that the reference electrode is kept in the same as for the ISFETs. By changing the input operating point , the potential at the working electrode can be set to voltage induce an electrochemical, faradic current. This current is then converted and amplified to a voltage through a feedback resistor . This voltage is then sent to the output. The currents are very small (typically nanoamperes) and must be amplified for further signal handling. For a given input potential and chemical consets the maximal output range and centration, the size of conthus the sensitivity of the current-to-voltage converter ( verter). If
V. HEAT REGULATION For reproducible blood-gas measurements, the on-chip temperature needs to be fixed (for example, at 37 C) throughout different measurements. Also, different fluids can only be compared validly if the temperature during different measurements is the same. To keep the on-chip temperature constant, a heatregulating loop including a temperature-sensing device and a heating device have been integrated. The temperature-sensing device is a parasitic vertical p-n-p bipolar transistor. If the current through the bipolar device is kept constant, then the voltage drop over the bipolar transistor, which is connected as a diode (inset in Fig. 7), is proportional to the absolute temperature, as given by (3)
is the leakage current at zero biasing, is the biasing (1) current of the transistor, and the absolute temperature. From Fig. 7, it can be seen that for every increase in temperature of drops about 2.2 mV. This voltage (also referred to 1 , , is proportional to . For fixed ) can now be used to control the chip temperature, as as and to allow a large range To allow large variations in indicated in the inset in Fig. 7. The heating device is a normal of different chemical solutions with varying gas concentrations, nMOS transistor and is designed to be able to generate 0.25 W two possible output ranges were foreseen. This was done by or 50 mA with a 5-V power supply. Through the feedback loop, with a -resistor network, as shown in Fig. 6. the heating device is turned on to generate enough heat to keep replacing connection is left open and , then the on-chip temperature constant. The temperature is set through If the . If, however, the connection the feedback resistor is , which is an input control voltage to the system (through the is applied, then a new feedback resistance is obtained which is “Sensor in” pad in Fig. 2). . The sensitivity is then increased: larger than The thermal time constant of the regulating loop was measured by externally applying a 5-V pulse to the gate of the heater (Fig. 8). This means going transistor and then reading out from no heating power to full heating power and back. varies according to a combination of two exponential functions: or one exponential that characterizes the heat resistance and heat (2) capacitance of the silicon, and one that characterizes the heat loss to the environment [7]. or
In this design, a difference in sensitivity factor of 4.33 was taken. With the highest sensitivity selected, an input current of 100 nA of 1 V. This corresponds to an effective resistance of gives a 10 M . The bondpad in the lower right corner of Fig. 1 is connected to the extra resistor branch and can easily be connected to the bondpad on its left during packaging, which is the bondpad. The complete setup of Fig. 5 is integrated on chip. C. Conductometric Sensor As explained above, it is also possible to perform a complete four-point conductometric measurement on chip. The driving
(4) It is assumed that the thermal resistance of the silicon is small compared to the thermal resistance to the environment [7], so that the heating feedback system has to be designed with the fastest of both time constants in mind. The measurements in Fig. 8 yield time constants of 0.0095 s and 0.15 s supporting this statement. This means that wherever both are put on the same chip, no measurable effect is caused on the feedback system. The second exponential is very dependent on the chip package used. The feedback loop is designed with the first pole being the
2034
Fig. 6.
IEEE JOURNAL OF SOLID-STATE CIRCUITS, VOL. 36, NO. 12, DECEMBER 2001
Modified feedback impedance for two possible ranges of operation.
Fig. 7. Temperature regulation: measured base–emitter voltage dependence of a bipolar transistor on the environment temperature and schematic of how this is used in the heat-regulating loop.
fastest thermal pole. Also, the attenuation of the thermal path can be estimated by extrapolating the curves of Fig. 8 and is in . This information is sufficient for designing the range of the temperature-control loop. During measurements of the temperature-control loop, an electronically controllable thermochuck was used to set the ambient or environment temperature of the chip. In Fig. 9, a measurement result of the temperature-control loop is given. is the voltage drop over the bipolar transistor that is in is the voltage over a second integrated the feedback loop. is the current for the total system bipolar transistor. through the power supply. A fully automated measurement was performed by automatically changing the thermochuck temperature. The result given is an average of ten measured values. The temperature value was derived with an accuracy of values is due about 1 mV. The difference between both to the resistance of the routing. With the given heating power, between 33 C and 47 C ambient temperature, the voltage over the bipolar transistors changes only with 3.4 mV. This is equal to 0.24 mV/ C, which is almost a factor of ten smaller than without regulation. In a fully mounted and packaged chip, this value will be even better, because the heat loss to the environment will then be smaller. The chemical measurements presented in Section VII are performed on a fully packaged device.
Fig. 8. T as a function of time when the heater gate voltage switches from 0 to 5 V and back.
VI. CONTROL ELECTRONICS The control electronics consist of an EPROM ( a controller, and a line driver ( in Fig. 2).
in Fig. 2),
A. EPROM The sensor chip was primarily developed for (but not limited to) medical use. For health care reasons, the chip may only be used on one patient and this for a limited time. After use, the chip must be discarded. To make sure that two patients never use the same chip, a single-bit EPROM is integrated on the microsensor chip. The first time the sensor is used, the EPROM
LAUWERS et al.: CMOS MULTIPARAMETER BIOCHEMICAL MICROSENSOR
2035
Fig. 10.
Fig. 9.
Fully automated chip temperature measurement. TABLE I OVERVIEW OF THE EPROM FUNCTION
Schematic view of the complete EPROM.
large drain–source on-resistance, the output is high (low) when the FGT has (not) been programmed. The W/L of M1 has been . If the R/W signal is in write mode, then M2 taken as is on and has to conduct a lot of current without lowering the drain–source voltage of the FGT too much. Hence, M2 needs a has been taken. larger W/L, and a value of B. Controller
is read. If it has not been used, then the chip can be used and the EPROM is programmed to avoid future use. This is a nonreversible operation and therefore provides a valid security check. Use of an EPROM can even be further exploited in future versions. If more than one bit is integrated and some logic is added, the EPROM could be programmed to indicate the kind of sensor configuration that is available on this particular chip. Then the monitor can read this and automatically set the correct parameters. The core of the EPROM is a floating-gate nMOS transistor (FGT) to which a control gate is capacitively coupled. It is not a standard EPROM circuit solution. This control gate can have two different voltages: 1.5 and 8 V. In order to generate this 8 V, a charge pump was also integrated. Before programming, the threshold voltage of the FGT is small, so that with a control-gate voltage (CGV) of 1.5 V, a current of the order of 10 A flows through it. If the CGV is now increased to 8 V during write mode, then the current increases drastically, creating hot electrons that are trapped in the gate oxide of the FGT, which induce an upward threshold voltage shift. When the CGV is now brought back to 1.5 V, only a small current of the order of 10 pA flows through the FGT. This current is measured and an output bit is sent to the monitor. The functioning of the EPROM is summarized in Table I. A schematic of the total EPROM is given in Fig. 10. The charge pump is a three-stage bootstrap structure. The output voltage (8 V in this design) can be varied easily by resizing the output capacitor. In order not to load this output capacitance too much, transistors M3 and M4 need to have a high drain–source resistance. Hence, M3 has a channel length of 20 m and M4 of 40 m, and both have a width of 2 m. The sizing of transistors M1 and M2 can be explained as follows. Assuming that a low R/W signal means “read,” then in read mode, transistor M1 is on and M2 is off. In case the FGT has not been programmed, a large current flows through M1, while only a small current flows if the FGT has been programmed. This means that if M1 has a
The minimal amount of external connections is five: the and , the clock signal, and power supply connections the I/O connections. Three wires have been added so that a standard eight-wire connection is obtained: an external for reasons of controllability, an external reference voltage reference biasing current, and an extra ground connection to have a good analog reference even with currents up to 100 mA. To this end, the controller (part 3 of Fig. 2) has to redirect 16 outputs to one output line and redirect one input line to four different inputs. This necessitates the use of a 16:1 multiplexer and a 1:4 demultiplexer. The schematics of the multiplexer and demultiplexer are identical. Only one decoder is needed if, every fourth clock cycle, the same demultiplexer input is selected, and every 16th clock cycle, the same multiplexer input is selected. This can easily be achieved by taking a 1:16 demultiplexer instead of a 1:4 and connecting every fourth output together. This increases the used area a little, but much less than a second decoder would. The synchronization is done by including three analog biasing voltages in the chain of 16 multiplexer inputs. Analog voltages eliminate the possibility of still having hard faults in the hardware without noticing, and hence provide an extra check to see if the electronics work properly. At each demultiplexer output, a hold capacitance of 40 pF has been placed to temporarily store the input voltages while the other ports are being refreshed. This value was calculated from the chosen specifications for minimal clock frequency of the system, maximal voltage droop, and a security margin of 25%. If a minimal frequency of 5 kHz and a maximal droop of 1 mV are taken, then leakage current F
(5)
The decoder and gray counter are well-known digital building blocks and will not be further explained. The controller can intrinsically be clocked at 1 MHz according to simulations, but the chip’s speed is limited by the analog electronics which have
2036
IEEE JOURNAL OF SOLID-STATE CIRCUITS, VOL. 36, NO. 12, DECEMBER 2001
Fig. 12. Offset (or total nonlinearity) of the rail-to-rail line driver.
Fig. 11.
Rail-to-rail line driver (biasing circuit not drawn).
to drive large loads with a reasonable power consumption. The total system speed is limited by the DAC card used in the monitor and the software. This limits the speed for the prototype system to maximally about 20 kHz. C. Line Driver To bring the analog signals off chip, a line driver is needed. This line driver must be able to drive loads up to 300 pF. The buffer must have a unity gain and must be able to drive low voltages from about 0.5 V as well as high voltages up to 5 V. Therefore, a rail-to-rail opamp was designed. Normally, when designing a rail-to-rail opamp, it is important to keep the total transconductance variation (when the input varies over the full scale) small. In this application, however, the opamp has unity feedback, so that the open-loop variation is divided by the loop gain. This means that if the gain is made large enough, it is varinot necessary to add a special structure to keep the ation small. For this purpose, the amplification was chosen to be higher than 60 dB. Large transistors are used where appropriate to provide low offsets. This limits the frequency range of the opamp but, as mentioned earlier, this is not a problem here, because the read-out software and not the hardware is the main speed-limiting factor. The transistor schematic of the line driver design is given in Fig. 11. The current of the two complementary input-differential pairs is summed and amplified. The output stage is a modified Miller-compensated opamp stage [8]. The two diodes, located next to the compensation capacitance , shift the positive zero, that is common in two-stage Miller amplifiers, to higher frequencies. In simulation, the driver has a gain of 61.6 dB and a GBW of 300 kHz for a load of 300 pF and a power consumption of 5.46 mW. The driver was also integrated separately as a test structure, and in Fig. 12 the result of an input voltage sweep is given. The total dynamic linearity error or offset variation is less than 15 mV. Only at the two extremities is a little increase seen. In these regions of operation, the open-loop gain reduces because the output transistors work in their linear region, reducing the gain. On a 5-V scale, this means that the absolute accuracy equals dB
bits
(6)
Fig. 13.
Sensor chip die microphotograph.
However, locally for one signal, the absolute accuracy can be higher. For example, if the absolute temperature information is considered, the voltage range is always between 0.5 and 0.6 V (throughout different batches and runs). In Fig. 12, it can be seen that the difference between input and output stays smaller than 1 mV for an input from 0.4 to 0.7 V. Thus, also for the temperature information range, the difference stays well within this 1-mV accuracy (or 12 bits) and not within a 15-mV accuracy that is relevant for full-swing signals only. Once above 0.7-V input voltage, there is a shift in offset due to the nMOS differential pair that starts to operate in its active region. Signals that need a higher accuracy than 15 mV and with an expected output voltage range that contains this transition region must be compensated. VII. CHEMICAL MEASUREMENTS The chip microphotograph is shown in Fig. 13. First, an amperometric oxygen measurement performed on two different packaged sensors is presented. For the measurement, the flow channel is connected to an artificial patient, enabling a solution with regulated oxygen concentration to be pumped over the sensor array. The amperometric measurement is done using the potentiostatic setup (IV-OTA and Pot-OTA). Through and O inputs, the working electrode is set at a the voltage of 0.6 V versus the reference electrode. Then the oxygen concentration in the fluid is varied. Measurement results are plotted in Fig. 14. Clearly, a linear relation of about
LAUWERS et al.: CMOS MULTIPARAMETER BIOCHEMICAL MICROSENSOR
Fig. 14.
Amperometric measurement of oxygen.
2037
Fig. 16.
Measured resistance versus specific resistance of buffer solution.
the buffer fluid is plotted. The input is a 1-kHz signal with a 100-mV amplitude and the feedback resistance of the IV-OTA is 100 k . The resistance is calculated using (8) In Fig. 16, the output voltage amplitude thus varies between 143 and 400 mV, which is in the expected range. For example, for cm specific resistance, the theoretical a fluid with a 100value is spec resistance distance cm area electrode cm
k (9)
Fig. 15.
Measured output voltage of a fully packaged
K
sensor.
25 mV/10%[O ] full scale is observed between the oxygen concentration and the output voltage of the IV-OTA. The difference in slope between both measurements is explained by slight differences in the selective membranes deposited on top of the sensor (Clarck-type O sensor). Those membranes are deposited manually on the prototype devices, which makes it a difficult to reproduce process. A difference in thickness of the membrane can easily induce such large differences in the output slope because of the high amplification factor of the IV-OTA. Next, the measurement of the potassium concentration using an ISFET with a potassium-selective membrane is described. In Fig. 15, the output of the ISFET buffer connected directly to a plotter (bypassing the multiplexer) is shown. Hence, the absence of an absolute voltage level in this figure. It can clearly be seen that the output voltage follows the potassium concentration according to (7) This is close to the theoretical slope of 59 mV. Also, the response time is fairly rapid, typically within one minute. Finally, a conductometric measurement result is presented in Fig. 16. The measured resistance versus specific resistance of
The difference with the measured value in Fig. 16 can be readily expected for a chemical sensor as a result of the condition of the electrodes (oxidation, nonideal etching) and the nonideally parallel layout. VIII. CONCLUSION A fully integrated multiparameter sensor chip has been presented. Strong points of the system chip include low pin count, on-chip integrated ISFETs and ISFET buffers, temperature control, and built-in security EPROM. Also, a conductometric sensor and oxygen sensor are integrated on-chip with their interfacing electronics. In total, the same microsystem allows up to seven different measurements. Experimental measurement results have been presented that prove the functionality of the system and the feasibility of the integration of multiple chemical sensors on one chip, including the interfacing electronics. REFERENCES [1] F. Van Steenkiste et al., “A biochemical CMOS integrated multiparameter microsensor,” in Transducers’99, 1999, pp. 1188–1190. [2] E. Lauwers et al., “A CMOS multiparameter biochemical microsensor with temperature control and signal interfacing,” in Proc. IEEE Int. Solid-State Circuits Conf. (ISSCC), Feb. 2001, pp. 244–245. [3] C. Jorgensen and W. Kunnecke, “Fully automated membrane dispensing in nanoliter scale and its application in sensor manufacturing,” Proc. SPIE, vol. 3857, pp. 207–214, Sept. 1999. [4] Ph. Arquint, A. van den Berg, B. H. van der Schoot, and N. F. de Rooij, “Integrated bloodgas sensor for pO , pCO and pH,” Sensors and Actuators B, vol. 13, pp. 340–344, 1993.
2038
IEEE JOURNAL OF SOLID-STATE CIRCUITS, VOL. 36, NO. 12, DECEMBER 2001
[5] W. Gumbrecht, D. Peters, and W. Schelter, “Integrated pO , pCO , pH sensor system for online blood monitoring,” Sensors and Actuators B, vol. 8–19, pp. 704–708, 1994. [6] H. Wong and M. H. White, “A CMOS-integrated “ISFET-operational amplifier” chemical sensor employing differential sensing,” IEEE Trans. Electron Devices, vol. 36, pp. 479–487, Mar. 1989. [7] W. Van Petegem, B. Geeraerts, W. Sansen, and B. Graindourze, “Electrothermal simulation and design of integrated circuits,” IEEE J. SolidState Circuits, vol. 29, pp. 143–146, Feb. 1994. [8] M. Steyaert and W. Sansen, “A high-dynamic-range CMOS opamp with low-distortion output structure,” IEEE J. Solid-State Circuits, vol. 22, pp. 1204–1207, Dec. 1987.
Erik Lauwers was born in Leuven, Belgium, in 1973. He received the M.Sc. degree in electrical engineering in 1997 from the Katholieke Universiteit Leuven, Belgium. Since 1997, he has been working toward the Ph.D. degree as a Research Assistant at the ESAT-MICAS Laboratories of the Katholieke Universiteit Leuven. His research interests are mainly in mixed-signal and analog base-band integrated circuits.
Jan Suls received the Master’s degree in chemistry in 1979 from the Katholieke Universiteit Leuven (KULeuven), Belgium. The work involved the modeling of the ground state of Co(II)N4 Schiffs Base complexes. Until 1983, he studied the photochemical reaction mechanism of transition metal complexes after pulsed laser excitation under an I.W.O.N.L. fellowship. Since 1984, he has been a Delegate Scientist from IMEC at the KULeuven, Electrotechnical Department, ESAT. His interests are in the integration of biomembranes and biorecognition systems in planar chemical sensor applications.
Walter Gumbrecht received the Ph.D. degree in physical chemistry in 1983 from the University of Erlangen, Nurnberg, Germany. He joined the Corporate Technology Department of the Siemens AG, Erlangen, Germany, in 1984. He is working on the development of semiconductorbased chemical and biochemical sensors and microsystems for medical applications.
David Maes received the degree in electrical engineering from the University of Leuven, Belgium, in 1996. In 1996, he joined IMEC as a Process Engineer to work on the CMOS integration of chemical sensors and of ferroelectric memories.
Georges Gielen was born in Heist-op-den-Berg, Belgium. He received the M.Sc. and Ph.D. degrees in electrical engineering from the Katholieke Universiteit Leuven, Belgium, in 1986 and 1990, respectively. After being a Postdoctoral Research Assistant and Visiting Lecturer at the University of California, Berkeley, he returned to the Department of Electrical Engineering of the Katholieke Universiteit Leuven, where he is currently a Professor. His research interests include design and computer-aided design (simulation, modeling, synthesis, layout, and test) of analog and mixed-signal integrated circuits. He serves regularly on the program committees of international conferences and is an associate editor for several journals.
Willy Sansen (M’72–SM’86–F’95) received the M.Sc. degree in electrical engineering from the Katholieke Universiteit Leuven (K. U. Leuven) in 1967 and the Ph.D. degree in electronics from the University of California, Berkeley, in 1972. In 1969, he received a BAEF fellowship. In 1972 he was appointed by the National Fund of Scientific Research (Belgium) at the ESAT laboratory of the K. U. Leuven, where he has been a full Professor since 1980. During 1984–1990, he was the head of the Electrical Engineering Department. Since 1984, he has headed the ESAT-MICAS Laboratory on analog design, which includes about fifty members and which is active in research projects with industry. He is a member of several boards of directors. In 1978, he was a Visiting Professor at Stanford University, Stanford, CA, in 1981 at the EPFL Lausanne, in 1985 at the University of Pennsylvania, Philadelphia, and in 1994 at the T. H. Ulm. He is cofounder and organizer of the workshops on Advances in Analog Circuit Design in Europe. He has been involved in design automation and in numerous analogue integrated circuit designs for telecom, consumer electronics, medical applications, and sensors, and has supervised over forty Ph.D. theses in these fields. He is a member of several editorial and program committees of journals and conferences. He has authored and coauthored eleven books and more than 550 papers in international journals and conference proceedings. He is a member of the executive and program committees of the IEEE ISSCC conference, and is the program chair of the ISSCC 2002 conference.
