a 136-/spl mu/w/channel autonomous strain-gauge...

2280 IEEE JOURNAL OF SOLID-STATE CIRCUITS, VOL. 38, NO. 12, DECEMBER 2003

A 136-�W/Channel AutonomousStrain-Gauge Datalogger

Wim Claes, Student Member, IEEE, Michel De Cooman, Willy Sansen, Fellow, IEEE, andRobert Puers, Senior Member, IEEE

Abstract—A single-chip 18-channel strain-gauge datalogger ICis integrated in a 0.7- m CMOS technology. It combines a 10- mstrain-accuracy sensor interface with digital offset compensa-tion, a wireless 132-kHz/66-kHz transceiver and a 23.4-kgatesdigital unit with adjustable data processing. The datalogger’smaximum power consumption, including an external 2-Mb RAM,is 136 W/channel at 3.1 V.

Index Terms—Application-specific integrated circuit (ASIC),biomedical electronics, biomedical measurements, CMOS in-tegrated circuits, compensation, data loggers, data processing,dentistry, inductive link, low-power electronics, mixed-signal,patient monitoring, prosthetics, sensing devices and transducers,strain gauges, strain measurements, stress measurement, trans-ducers and sensing devices, wireless transceiver.

I. INTRODUCTION

I N THIS PAPER, a miniaturized battery-operated wirelessdatalogger that is part of a dental prosthesis is presented.

This datalogger is used to gain more insight in the process ofbone remodeling caused by loading oral implants that supporta dental prosthesis. Underloading as well as overloading of theimplants can cause implant failure. Because of the lack of quan-titative load data to validate the existing bone-remodeling/im-plant-failure theories,in vivo load measurements are necessary.The aim of the presented datalogger is to monitor continuouslythe in vivo loads in a prosthesis over a two-day period, inde-pendent of the hospital environment. It is capable of measuringthe unconscious nocturnal dental activities, like gnashing andclenching, which are seen as a missing link in the validation ofexisting bone remodeling models, who have to rely on restrictedand isolated measurements in an hospital environment.

Fig. 1(a) shows a schematic drawing of an oral-implantsystem with a single-tooth dental prosthesis. For a full pros-thesis, the number of implants varies between two and six.Fig. 1(b) and (c) illustrates a clinical example with five implantsbefore and after placement of a full prosthesis. To measurethe loads imposed on the different implants, every abutment( mm, mm) is equipped with three metal-filmstrain gauges ( k G.F. ) [1], positioned 120from each other with their measuring grids parallel to thecylindrical abutment’s axis, as shown in Fig. 2(a). The imposed

Manuscript received April 2, 2003; revised June 30, 2003. This work wassupported by the Fund for Scientific Research-Flanders (Belgium) under ProjectNumber G0359.99.

The authors are with the Department of Electrical Engineering, DivisionESAT-MICAS, Katholieke Universiteit Leuven, B-3001 Heverlee, Belgium(e-mail: [email protected]).

Digital Object Identifier 10.1109/JSSC.2003.819173

(a)

(b)

(c)

Fig. 1. (a) Schematic drawing of a Brånemark oral implant with a single-toothdental prosthesis: 1) implant (in the bone), 2) abutment screw, 3) abutment(in the gums), 4) set screw, 5) gold cylinder, 6) (single-tooth) prostheticsuperstructure. (b) Clinical example with five implants before and (c) afterplacement of a full prosthesis.

axial forces and bending moments on the abutments can bederived from the strains measured by the individual straingauges [2]. In Fig. 2(b), a photograph of an abutment equipped

0018-9200/03$17.00 © 2003 IEEE

CLAES et al.: 136- W/CHANNEL AUTONOMOUS STRAIN-GAUGE DATALOGGER 2281

(a)

(b)

Fig. 2. (a) Schematic overview of the strain gauges’ placement and themeasured load components. (b) Abutment equipped with strain gaugesprotected by a shrink sleeve.

with strain gauges protected by a shrink sleeve is given. Therequired bandwidth of each strain-gauge channel is 50 Hz andthe required measuring accuracy (i.e., standard deviation ofthe error) is 10 strain. This is equivalent to a strain-gauge-re-sistance measurement accuracy of 100.5 m, correspondingwith an axial-force measurement accuracy of 5.6and abending-moment measurement accuracy of 0.64cm. Sincethe strain gauges’ resistance tolerance (30) is higher thanthe resistance change (19.6) due to the maximum/minimumoccurring strain (i.e., strain), compensation for theresistance variation is required for each channel. Moreover,due to mechanical misalignments in the assembly an exces-sive pre-strain of up to strain can occur when theprosthesis is placed. This pre-strain is again higher than themaximum/minimum occurring strain. Therefore, the measure-ment system must be capable to compensate for this excessivepre-strainafterplacement of the prosthesis. This compensationin situ is carried out wirelessly, as explained further.

In the past, an external nonportable measuring unit [2], con-nected to the strain gauges by wires, has been employed to quan-tify the loads in dental prostheses. Compared with the existingsystem, the new autonomous datalogger, being embedded inthe prosthesis, has several advantages. First of all, the measure-ments are no longer restricted to the hospital environment, sothat the patient can be monitored in his normal living condi-tions. In this way, unconscious nocturnal dental activities alsocan be monitored. Moreover, the patient is no longer hindered

Fig. 3. Total system overview.

Fig. 4. Block diagram of the complete datalogger. The external RAM isindicated with a dashed box. A single-chip datalogger IC combines the otherbuilding blocks.

by strain-gauge wires coming out of his mouth and the measure-ments are no longer carried out on command, so that artificialchewing behavior is avoided. Instead, the natural load behaviorcan now be traced.

II. TOTAL SYSTEM OVERVIEW

To perform the compensation wirelessly, a bidirectionaltransceiver is included in the datalogger. This allows wirelessactivation of the compensationafter placement of the pros-thesis. Moreover, it allows to (re-)program the datalogger’soperation-mode settings (e.g., the number of strain gaugechannels, the data processing algorithm, etc.). By the incorpo-ration of flexibility into the datalogger and the possibility ofreprogramming the device settings wirelessly, a highly flexibledatalogger is obtained which can be adapted toward eachindividual patient with the implant placedin situ.

Fig. 3 shows an overview of the complete bidirectionaltelemetry system. At the measurements startup, the data-logger’s internal transceiver communicates with an external RFunit connected to a PC. Dedicated software runs on the PC toprogram/read out the datalogger and to store and visualize thecollected data. The datalogger is a transponder-type device; it isable to pick up a nearby low-frequency programming/readingfield (132 kHz) and respond to it. The datalogger can be(re-)programmed by this field or can be instructed to send thecollected data to the external RF unit. The receiver (Rx) andtransmitter (Tx) antennas of the external RF unit areLC circuitstuned, respectively, to 66 and 132 kHz. In the monitoring mode,this field is neither present nor needed, and the device switchesto an autonomous mode, powered by the incorporated battery.It continues to collect and process data, until the field is presentagain. This is done, for example, at consultation.


Fig. 5. Block diagram of the sensor interface.

III. D ATALOGGER

A. Sensor Interface

An overview of the complete datalogger is shown in Fig. 4.It consists of four major parts: a sensor interface, a digital part,and a wireless transceiver, integrated on a single chip, and anexternal 2-Mb SRAM. In [3] and [4], the sensor interface chip,the selection/design of the sensor interface building blocks, andthe system concept have been discussed. This work presents forthe first time the single-chip integration and experimental ver-ification of the complete datalogger IC with on-chip digital in-telligence.

A circuit block diagram of the sensor interface is given inFig. 5. (22 A) is a self-biased thermal-voltage-ref-erenced current reference (biasing and startup circuits arenot shown) from which the currents (176 A),(308 A) and the digitally controllable current of abinary weighted current-steering digital-to-analog converter(DAC) ( nA) are deduced. To measure the straingauges , a current-driven Wheatstone configuration, con-sisting of a reference branch - and a measurementbranch - , is applied. This has a doubled sen-sitivity-per-total-current ratio compared to the voltage-drivenone [3] and thus allows a lower power consumption. The

measurement of the 18 different channels is carried out consec-utively by means of two 18-channel multiplexers (MUX) and a5-bit channel-select decoder, included in the PROG/SEL-block.The overall sampling frequency of the sensor interface is2 kHz, resulting in a sampling frequency of the 18 individualmultiplexed strain-gauge channels of 111 Hz. To cope withthe offsets introduced by the resistance tolerance and the highpre-strains, the total current through a strain gauge

under measurement is adjustable. By applying the correctdigital word at the input of the 8-bit DAC, becomesequal to (limited by the resolution level of the DAC) andcompensation is obtained. Compensation for every strain-gaugechannel is carried out after placement of the prosthesis and thedigital words needed for compensation are programmed intoan on-chip nulling memory REG using the PROG/SEL-block.In the measurement mode, when a particular strain-gaugechannel is selected, the digital word belonging to that particularchannel is fetched from REG and offered to the DAC, so thatoffset-compensated measurements are performed.

The multigauge interface is followed by an offset-cancelledswitched-capacitor (SC) resettable-gain amplifier (AMP) andsample-and-hold (SH), proceeded by an offset-cancelled 9-bitsuccessive approximation analog-to-digital converter (ADC).The total integrated input-referred noise of these buildingblocks is 9 V. The sensor interface also contains a 5-bit


Fig. 6. Overview of the digital part of the datalogger.

TABLE IPROPERTIES OF THEIMPLEMENTED DATA PROCESSINGALGORITHMS

programmable 128-kHz relaxation oscillator with two clockdividers (/2 and /64) and two nonoverlapping clock generators( and ).

B. Digital Part

Fig. 6 shows a block diagram of the digital part, which isclocked by the sensor interface’s 128-kHz clock. It contains aprogrammable data processing unit including selectable algo-rithms with adjustable parameters. This unit can be programmedto store only clinical relevant data, reducing the necessary data-storage-capacity drastically, and to employ an optimized dataprocessing algorithm for each individual patient. Without dataprocessing, a memory capacity of 388.8 MB would be requiredto store all the raw data of the 18 channels over a two-day pe-riod. Because the onboard memory is limited to 2 Mb due to theavailable space for the datalogger embedded in the dental pros-thesis, it is clear that data processing is a requisite. Table I givesan overview of the properties of the 8 selectable algorithms im-plemented in the data processing unit. The first algorithm storesthe raw data without further processing and can be used to derivethe optimal parameters for the other algorithms. This algorithmdoes not allow memory savings, which results in a 116.5-s datacollection time for 18 channels. It is used in the learning cycleto retrieve patient-specific load behavior. The second algorithmstores the average value of the data in each channel using a pro-grammable number of data points for the average calculation.The last algorithm stores the duration during which a thresholdis trespassed, the time of occurrence, and the strains of the dif-ferent channels when the maximum strain occurs in one of thechannels. The other algorithms store a programmable numberof raw data or the average values with/without time informa-tion and with/without storing the maximum strain set after tres-passing a threshold. To achieve a very flexible data processingunit, the thresholds as well as the number of data points used(RAW/MEAN) are adjustable.

Fig. 7. Integrated wireless transceiver.

The digital unit also includes a programming unit to pro-gram the compensation words directly into the on-chip nullingmemory REG via the PROG/SEL block (Fig. 5). The dataloggeris also capable of compensating itself for the offsets introducedin the strain-gauge channels due to the strain gauges’ resis-tance tolerance and due to the placement of the prosthesis. Thiscompensation is carried out automatically by commanding thedatalogger IC wirelessly to compensate toward a user-defin-able output value for a selectable strain-gauge channel. Thisis performed by the nulling block, which employs successiveapproximation to determine the required compensation wordto be stored. The sampling unit controls the 5-bit channel-se-lect decoder, included in the PROG/SEL block (Fig. 5) of thesensor interface in the measurement mode and also controls thestorage of processed data in the RAM, which is dependent onthe number of selected strain-gauge channels. The transmissionunit transmits the stored data in selectable data packages of 256bytes. To achieve a correct communication, the data bytes areManchester-encoded and an extra 3-bit header per byte is addedas well as an extra parity byte per data page. A commerciallyavailable external transceiver [5] has been modified to commu-nicate with the transponder-type datalogger. This wireless trans-ceiver, controlled by a PC, allows to program the dataloggerwith Manchester-encoded 15-bit commands with an extra 4-bitheader and two extra parity bits. The receiving unit of the data-logger takes care of the reception and validation of these com-mands and appropriate actions are taken by the controller if acorrect command is received. After programming, the actualstatus of the datalogger can be verified by calling the devicestatus bytes. The controller also ensures that normal operationis reassumed after a possible lock during communication bymeans of programmable watch-dog timers (WDT). The digitalpart has been implemented together with the sensor interface onthe same chip. It contains approximately 23 400 gates and occu-pies 22.5 mm. A single power supply of 3.1 V is used for logiccore and I/O.

C. Transceiver

The datalogger’s integrated transceiver is shown in Fig. 7.The external transceiver, controlled by a PC, employs amplitudemodulation of a 132-kHz carrier to program the datalogger. Theincoming amplitude-modulated 132-kHz field is received by areceptionLC-tank (RX), demodulated by a buffer (BUF) and anexternal rectifier (RECT), and further processed by the receivingunit. During data retrieval of stored data or status bytes fromthe datalogger, a 66-kHz carrier is derived from an incoming


TABLE IIOVERVIEW OF THE COMMANDS

nonmodulated 132-kHz field. The data are sent to the externaltransceiver by phase modulation of the 66-kHz carrier, carriedout by EXORing the data with the carrier, and by buffering themodulated signal into a transmissionLC-tank (TX). The internaltransceiver is able to transmit data to the external transceiverover a distance of 30 cm at a data rate of 4 kB/s with a meanpower consumption of 2.3 mW.

D. Commands

An overview of all the commands that can be issued to the dat-alogger is given in Table II. Each command has a 5-bit commandcode and 10 databits, used to program the parameter values. Thecommands in Table II are divided into five subgroups, as fol-lows:

1) commands related to the general operation mode of thedevice;

2) commands employed to store the digital compensationwords into REG or to activate the automatic nulling of aselectable strain-gauge channel toward a selectable outputvalue;

3) commands used to set up the data processing algorithmand to select the number of strain-gauge channels;

4) commands to set up the communication mode and to acti-vate the transmission of a selectable data page or the statusbytes;

5) command for the programmation of the WDT intervals.

IV. EXPERIMENTAL RESULTS

Fig. 8 shows a micrograph of the realized datalogger IC in-cluding the sensor interface, the digital part, and the transceiver.The chip has been fabricated in a 0.7-m CMOS technology, of-fering a lower cost compared with modern state-of-the-art tech-nologies. The measured maximum mean power consumptionof the complete datalogger including the SRAM [6] is a mere

Fig. 8. Chip micrograph.

136 W per strain-gauge channel, which is to our knowledgethe lowest ever presented for comparable systems [7]–[10].

To perform static and dynamic measurements, a voltage-con-trolled piezoelectric actuator [11], composed of PZT ceramicstacks, has been employed. This actuator is able to impose dis-placements with a maximum of 90m and an accuracy better


Fig. 9. Static measurement result. Dots: measurement data. Solid line:least-squares fit.�(error) < 6:5 �strain.

(a)

(b)

Fig. 10. Measurement setup with (a) a beam supported at both sides and loadedby two point loads and (b) a real abutment.

than 0.2 m. The maximum load of the actuator is equal to1000 N. To measure the forces imposed by the actuator, a mon-itoring cell is employed. Fig. 9 gives the output data for a staticmeasurement where an incrementally-increasing strain has beenimposed to a strain gauge. The dots represent the measurementdata and the solid line the least-squares fit. This measurementshows a standard deviation of the error smaller than 6.5strain.

Fig. 10 depicts the two measurement setups employed for thedynamic measurements. In the measurement setup in Fig. 10(a),the strain gauge is installed on a rigid PVC beam supported atboth sides. The beam is loaded by two point loads, symmetri-cally distributed around the center of the beam. The benefit ofthis approach is that the strain between the two point loads isconstant [12], so that the alignment of the point loads and thestrain gauge is not critical. The applied strain is proportional tothe displacement of the piezo stack.

(a)

(b)

Fig. 11. (a) Output data and (b) PSD for a sinusoidal strain (f = 4 Hz,peak-to-peakamplitude = 1005 �strain,f = 118 Hz, SNDR =35 dB). �(error) = 6:1 �strain.

Fig. 11(a) shows a window of the measured output datafor a sinusoidal strain with a peak-to-peak amplitude of1005 strain and a 4-Hz frequency performed with thismeasurement setup. The amplitude corresponds with themaximum actuator displacement. Fig. 11(b) illustrates thepower spectral density (PSD) of the measured output data,yielding a signal-to-noise-and-distortion ratio (SNDR) of35 dB, corresponding with a standard deviation of the errorequal to 6.1 strain.

The second measurement setup is depicted in Fig. 10(b). Inthis case, the PZT actuator applies a displacement to a steel discwhich is fixed by an M2 screw to an abutment of a dental pros-thesis. The weak connection with the M2 screw and the bendingof the steel disc introduce nonlinearities in the measurements,resulting in a lower accuracy of this test setup compared withthe first measurement setup. Despite this drawback, the secondtest setup allows to perform measurements withreal abutments.An example of such a measurement with a real abutment isshown in Fig. 12. In this measurement, a sinusoidal strain witha peak-to-peak amplitude of 479strain and a 30-Hz frequencyhas been imposed. [Note that in Fig. 12(a) the data points areconnected with each other so that it seems that the measuredsignal is modulated. This, however, is not the case and is due tothe fact that the 30-Hz signal is close to the Nyquist frequency.]The SNDR is equal to 27.2 dB, equivalent to a standard devia-tion of the error of 7 strain.

To investigate the influence of the integration of the digitalpart and transceiver on the same chip as the sensor interface,the datalogger has been tested under similar conditions as the


(a)

(b)

Fig. 12. (a) Output data and (b) PSD for a sinusoidal strain (f = 30 Hz,peak-to-peakamplitude = 479 �strain,f = 118 Hz, SNDR =27:2 dB). �(error) = 7 �strain.

TABLE IIIPERFORMANCESUMMARY

sensor interface chip [3]. This measurement yields a standarddeviation of the error equal to 6.8strain, which shows that theaccuracy degradation due to the single-chip integration of thedatalogger IC is limited to 0.6strain. This has been achievedby extensive (in-circuit) substrate contacts, guard rings and theuse of a conductive glue, grounding the chip’s substrate. Also,the influence of using a single supply instead of separate analog

and digital supplies has been investigated. The accuracy degra-dation due the use of a single supply is limited to 0.3strain,demonstrating that also this influence is negligible.

V. CONCLUSION

The performance of the datalogger is summarized in Table III.It has been shown in this paper that the single-chip integrationin a 0.7- m CMOS technology of a 18-channel strain-gaugedatalogger, combining a 10-strain-accuracy sensor interface,an intelligent digital unit, and a wireless transceiver is feasible.The combination of these building blocks yields a flexible au-tonomous datalogger with integrated automatic offset compen-sation and programmable data processing expanding the intelli-gence of the device. Note that the application of the presented in-telligent datalogger concept is not only restricted to dental pros-theses, but it may also be applied to other medical applicationareas, such as, for example, orthopedics.

REFERENCES

[1] FSM-A6306S-500, Catalog no. 305, BLH, Canton, MA.[2] J. Duycket al., “Magnitude and distribution of occlusal forces on oral

implants supporting fixed prostheses: An in vivo study,”Clin. Oral Im-plants Res., vol. 11, pp. 465–475, 2000.

[3] W. Claes, W. Sansen, and R. Puers, “A 40�A/channel compensated18-channel strain-gauge measurement system for stress monitoring indental implants,”IEEE J. Solid-State Circuits, vol. 37, pp. 293–301,Mar. 2002.

[4] W. Claeset al., “A low power miniaturized autonomous data logger fordental implants,”Sensors Actuators A (Issue on Tranducers’01/Eurosen-sors XV), vol. 97–98, pp. 548–556, 2002.

[5] Eureka 411 Tag Industrial Decoder, Avonwood Developments Ltd.,Wimborne, U.K.

[6] BSLV2000, Brilliance Semiconductor Inc., Irvine, CA.[7] J.-B. Begueretet al., “Converters dedicated to long-term monitoring

of strain gauge transducers,”IEEE J. Solid-State Circuits, vol. 32, pp.349–356, Mar. 1997.

[8] G. Bergmannet al., “Multichannel strain gauge telemetry for or-thopaedic implants,”J. Biomechan., vol. 21, pp. 169–176, 1988.

[9] P. Cappa, Z. Del Prete, and F. Marinozzi, “Experimental analysis ofa new strain-gauge signal conditioner based on a constant-currentmethod,”Sensors Actuators A, vol. 55, pp. 173–178, 1996.

[10] D. H. Follett, “An externally powered six channel strain gauge transcu-taneous telemetry system,”Implantable Telemetry in Orthopaedics, pp.87–92, 1990.

[11] P-841.60 Preloaded PZT Translators. Physikinstrumente (PI), Karls-ruhe/Palmbach, Germany. [Online]. Available: http://www.physikin-strumente.de/products/

[12] R. J. Roark and W. C. Young,Formulas for Stress and Strain. Tokyo,Japan: MacGraw-Hill, 1975.

Wim Claes (S’96) was born in 1973 in Geel,Belgium. He received the M.S. degree in micro-electronics in 1996 from the Katholieke UniversiteitLeuven, Belgium. He has been a Research Assistantwith the ESAT-MICAS Laboratories since 1996,where he is currently working toward a Ph.D. degree.

His main research interests are low-power sensorinterfaces and data acquisition systems for biomed-ical applications.


Michel De Cooman was born in 1966 in Leuven,Belgium. He received the B.S. degree in electrical en-gineering from the Polytechnical Institute I.H.B. ofBrussels, Belgium, in 1988 and the M.S. degree inbiomedical engineering from the Katholieke Univer-siteit Leuven, Belgium, in 1989.

In May 1990, he joined the ESAT-MICAS group,where he was mainly involved in thick-film andinterconnection research. His main interests arebiomedical electronics and active implantabledevices. At present he is in charge of the intercon-

nection and packaging laboratory of the sensor group at MICAS. He is amember of Imaps Benelux.

Willy Sansen (S’66–M’72–SM’86–F’95) receivedthe M.Sc. degree from the Katholieke UniversiteitLeuven (K.U. Leuven), Belgium, in 1967 and thePh.D. degree in electronics from the University ofCalifornia, Berkeley, in 1972.

In 1969, he received a BAEF fellowship. In 1972,he was appointed by the National Fund of ScientificResearch (Belgium) at the ESAT Laboratory, K.U.Leuven, where he has been a full Professor since1980. During 1984–1990, he was the Head of theElectrical Engineering Department. Since 1984, he

has headed the ESAT-MICAS Laboratory on analog design. He is a member ofseveral boards of directors. In 1978, he was a Visiting Professor at StanfordUniversity, Stanford, CA, in 1981 at the EPFL Lausanne, in 1985 at theUniversity of Pennsylvania, Philadelphia, and in 1994 at the T.H. Ulm. He hasbeen involved in design automation and in numerous analog integrated circuitdesigns for telecommunications, consumer electronics, medical applications,and sensors. He has been supervisor of over 50 Ph.D. dissertations in thesefields. He has authored and coauthored twelve books and more than 550papers in international journals and conference proceedings. He is a memberof several editorial and program committees of journals and conferences. Heis cofounder and organizer of the workshops on Advances in Analog CircuitDesign in Europe.

Dr. Sansen is a member of the executive and program committees of the IEEEInternational Solid-State Circuits Conference (ISSCC) conference, and was Pro-gram Chair of the ISSCC 2002 conference.

Robert Puers (M’88–SM’96) was born inAntwerpen, Belgium in 1953. He received theB.S. degree in electrical engineering from thePolytechnical Institute HRITHO, Ghent, Belgium,in 1974, and the M.S. and Ph.D. degrees from theKatholieke Universiteit Leuven (K.U. Leuven),Belgium in 1977 and 1986, respectively.

He is currently a Professor with the K.U. Leuven,and Director of the clean room facilities for siliconand hybrid circuit technology at the ESAT-MICASlaboratories of the same University. His main re-

search interests are biomedical implant systems (monitoring and stimulation),transducer and actuator principles, mechanical sensors (pressure, accelerome-ters, strain gauges, temperature, etc.), silicon micromachining, packaging andinterconnection techniques (hybrids, hermetic and biocompatible sealing, etc.),and biotelemetry (inductive power delivery, bidirectional data communication).He is involved in the organization, reviewing, and publishing activities of manyconferences, journals and workshops in the field of biotelemetry, sensors,actuators, micromachining, and microsystems. He is the author of more then220 papers on biotelemetry, sensors, or packaging in journals and internationalconferences.

a 136-/spl mu/w/channel autonomous strain-gauge...

Documents