JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 12, NO. 6, DECEMBER 2003 937

A Static Turbine Flow Meter With a MicromachinedSilicon Torque Sensor

Niklas Svedin, Erik Stemme, and Göran Stemme, Member, IEEE

Abstract—A new class of flow sensors is introduced where astatic turbine converts the volume flow into a torque. The struc-ture consists of a turbine fixed to a torque sensor, which in turnis connected to the pipe wall giving a perfectly symmetric, bidi-rectional flowmeter. In contrast to conventional turbine flowme-ters, the wheel does not rotate and consequently it is insensitiveto bearing friction and wear that conventional, rotating turbinesexperience. Furthermore, the flow passing the wheel is distributedover the circumference of the wheel and levels out nonuniform flowprofiles leading to profile independent volumetric flow measure-ment. The torque-sensing element is a 300-m-thick silicon can-tilever, 2 mm wide and 16 mm long. The stiffness of the torquesensor, and thereby the sensitivity, is mainly determined by lat-eral dimensions of specially designed stiffness reduction parts de-fined by photolithography, thus giving good control of the sensi-tivity. Two polysilicon strain gauges are placed on each side of thecantilever, measuring the bending moment from the turbine wheel.The torque sensor has been evaluated for different geometries andtogether with the turbine to evaluate the flow sensing performance.A turbine with a blade length of 2.7 mm and a blade angle of 30has a sensitivity of 4.0 V V (1 min)

2 when measured usingthe silicon torque sensor. The output signal shows good symmetrybetween different flow directions. [889]

Index Terms—Flow sensor, piezoresistive, torque.

I. INTRODUCTION

WHENEVER there is a need for flow measurements,choosing an appropriate flow meter is a critical issue.

Flow velocity or flow rate is one of the physical propertiesoffering the most comprehensive choice of sensors. Apart fromconsidering what kind of fluid will be measured (gas, liquid,two-phase) and its properties (purity of the fluid, influence ofother properties), there is a wide variety of sensing principlesto choose from.

One of the boundary conditions for this work has been theflow sensor’s ability to monitor the respiratory flow of a ven-tilated patient, i.e., monitoring breathing, be it spontaneous orassisted. This specific application requires a bidirectional flowsensor with a relatively low pressure drop which must be resis-tant to humidity and temperature variations of the respiratorygas. A large variety of flow sensor principles have been realizedusing MEMS technology. These can be divided into two maincategories: thermal and nonthermal [1]. Members of the former

Manuscript received June 20, 2002; revised April 17, 2003. Subject Editor G.B. Hocker.

N. Svedin was with the Department of Signals, Sensors and Systems, RoyalInstitute of Technology S-10044 Stockholm, Sweden. He is now with Silex Mi-crosystems, Kista 16440, Sweden (e-mail: niklaus.svedin@silex.se).

E. Stemme and G. Stemme are with the Department of Signals, Sensors andSystems, Royal Institute of Technology S-10044 Stockholm, Sweden.

Digital Object Identifier 10.1109/JMEMS.2003.820271

normally have the advantages of being highly sensitive at lowflows and less intrusive than many of its nonthermal counter-parts. However, in a ventilator environment were humid air ismeasured, a mechanical flow sensor that is not depending onthe gas’ heat transfer properties, is highly attractive.

Here, a new type of mechanical gas flow meter is introducedbased on torque measurement on a static turbine wheel. In con-trast to conventional, turbine flowmeters [2], [3], the wheel doesnot rotate and consequently it is insensitive to bearing frictionand wear that a rotating wheel experiences. Furthermore, theflow passing the wheel is distributed over the circumference ofthe wheel, thus leveling out nonuniform flow profiles leading toa profile independent volumetric flow measurement [4].

II. TURBINE FLOW METERS

When analyzing the forces on the blades of a rotating tur-bine there are two approaches described in the current literature[5], [6]. The first approach describes the fluid driving torque interms of aerodynamic lift via airfoil theory, while the seconddescribes it in terms of momentum exchange. This means thatthe former approach models each blade as a wing section usingairfoil theory [see Fig. 2(a)], while the latter approach studiesthe flow deflection as it passes the space between two adjacentblades [see Fig. 2(b)].

In the case of a nonrotating wheel, the wheel design shouldprovide maximum torque. This leads to a wheel with manyblades at a large angle or in other words, a high “solidity”. Thesolidity parameter can simply be described as how well youcan “see through” the turbine wheel. With the high solidityrequirement calling for closely spaced blades, the validity ofthe airfoil model is reduced since adjacent blades will affectthe flow around each blade in a nonnegligible way. Hence, themomentum model is best suited for analyzing the performanceof this flow sensor [6].

The number of blades, , and the blade widths, , arechosen to be large enough to assure “full guidance” [5], whichis required by the momentum model. The full guidance condi-tion means that the axially incoming flow,, leaves the turbinein the direction of the blades, , as indicated in Fig. 2(b). Forsymmetry reasons the blades are straight to produce a flowmeter with the same characteristics in both flow directions.

There are lots of similarities between the static turbine and aconventional turbine flowmeter but also some major differences.The most obvious one is that the static turbine meter does notrotate.

Since the static turbine flowmeter is not subjected to bearingfriction and other effects connected to the rotation of the tur-bine [5], the nonnegligible friction losses in the low flow regime

1057-7157/03$17.00 © 2003 IEEE

938 JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 12, NO. 6, DECEMBER 2003

Fig. 1. The static turbine consists of a turbine wheel but in contrast to conventional turbine flowmeters the wheel is fixed mounted on a torque-sensing device.

(a) (b)

Fig. 2. The turbine can be modeled in two different ways, either the airfoil approach (a) or the momentum approach (b). In this particular case, with a high solidityturbine, the latter is best suited for the analysis of the flowmeter performance.

does not exist and the sensitivity ranges down to zero flow. Incontrast to the linear flow dependence of the conventional tur-bine flowmeter the static turbine flowmeter has a quadratic flowdependence.

The viscosity of the gas can affect the turbine flowmetersin two major ways, inducing drag forces on the blades and tipclearance friction [7]. The viscous drag on the blades acts in theaxial direction and results in a counteracting torque reducingthe output signal in both the rotating and the static turbine in thesame way. For the rotating flowmeter the axial forces also resultin increased bearing friction due to the axial load.

Tip clearance viscous friction is the viscous friction betweenthe circumference of the turbine and the housing and is con-nected to rotational speed of the wheel. This effect is noticeableif the gap between the turbine and the housing is small [7]. Forthe static turbine increased tip clearance viscous friction only

increases the response time while for the rotating turbine the ro-tational speed will decrease resulting in a lower output signaland nonlinear sensitivity. Since the sensor is developed for gasflow measurement, viscous friction effects have been neglectedin the following calculations.

III. M OMENTUM MODEL

In the momentum approach the turbine wheel is seen as ”flowchannels” made up of two adjacent blades around the circum-ference of a cylinder, as illustrated in Fig. 2(b).

When the fluid passes the turbine, i.e., the blade spacing, itchanges direction providing an impulse to the guiding walls.Flow continuity gives

(1)

SVEDIN et al.: STATIC TURBINE FLOW METER WITH A MICROMACHINED SILICON TORQUE SENSOR 939

(a) (b)

Fig. 3. (a) The torque-sensing element consisting of three different parts: the mounting part, the supporting part and two stiffness reduction beams. The two sidesof the sensor are identical (b) The bridge configuration connecting the two strain gauges on each side to a full Wheatstone bridge.

where: The input and output velocities, respectively.

: The tip and hub radius, respectively. Hence,is the blade length.

: The blade angle.The change in direction gives changes in the velocity of each

fluid particle entering the blade spacing

(2)

This direction change leads to a momentum change pickedup by the blades giving a force on each radial sectionof each blade

(3)where is the fluid density. This results in a torque contribution

on the wheel axis from each of the blades

(4)

The total volumetric flow rate is assumed to be distributedevenly over all the flow channels around the wheel giving theflow rate in each channel. Assuming a flat flow profile in thechannels the flow velocity is given by

(5)

By using (5) and summing up the contribution from eachblade the relation between the volumetric flow and the torqueon the wheel axis is given by

(6)

Using the fact that the perimeter at the average blade spacingis

(7)

and substituting into (6) finally gives the torque as

(8)

Given the momentum model and (8) the torque is inverselyproportional to the blade length . Thus, the blade lengthshould be kept small for large torque to allow large flow sensi-tivity. The other conclusion from (8) is that the blade angleshould be as large as possible to also maximize the flow sensi-tivity.

Unfortunately, both decreasing the blade length and in-creasing the blade angle will increase the pressure drop. Asensor design optimization could be based on finding a combi-nation that results in the highest possible sensitivity in relationto the pressure loss.

IV. THE TORQUESENSOR

The torque generated when the flow passes the turbine wheelis measured with a torque sensing device. Here a speciallydesigned micromachined piezoresistive cantilever structurewas used as a torque sensor. The torque-sensing element is a300- -thick silicon cantilever, 2 mm wide and 16 mm longwhich has been DRIE-etched to form three different parts: themounting part, the supporting part and two stiffness reductionbeams, shown in Fig. 3(a). The two sides of the sensor aremechanically and electrically identical.

Two polysilicon strain gauges are placed on each stiffnessreduction beam, one on each side. The strain gauges aremade of boron doped polysilicon resistors deposited on aninsulating layer of silicon dioxide. Aluminum conductors onthe supporting part are connected to bond-pads at the otherend of the silicon beam, outside the flow. All four resistors areconnected in a full Wheatstone bridge for maximum sensitivityand suppression of common mode influences and pressure drop

940 JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 12, NO. 6, DECEMBER 2003

Fig. 4. (a) The stiffness reduction part of the torque sensor. (b) The relative dependence of each height segment on the stiffness of the torque sensor.The solidline shows the moment of inertia distribution, dI, i.e., stiffness distribution along the thickness of the stiffness reduction beam. The total moment of inertia is givenby the shaded area between the curve and they-axis. (c) Cross section of the active part of the torque sensor with a center beam reinforcement to increase thecross-sectional area and the cross-axis moment of inertia.

induced axial forces. When the static turbine wheel is subjectedto a flow, the gauges on one side are tensed and the gauges onthe other side compressed, resulting in a measurement of thetorque from the turbine wheel. The stiffness reduction beamsare aligned along the axis of the turbine wheel and the wheel isfixed to the mounting part.

The sensitivity of the torque sensor is determined by the di-mension of the stiffness reduction beams. Fig. 4(a) shows thecross section of that region where the section ’ coincideswith the wheel axis indicated in Fig. 3(a).

The bridge output signal is proportional to the stress at thesurface induced by the torque and given by

(9)

where is the flow induced torque, , and are defined inFig. 4(b) and is the moment of inertia of a stiffness reductionbeam. and are the Wheatstone bridge output andsupply voltage, respectively.

For a given thickness, , the stiffness is determined by thewidth, , of the stiffness reduction beams. This means that thesensitivity is mainly determined by lateral, photolithographi-cally defined, dimensions giving good control of the sensitivity.Narrower beams results in a less stiff structure and therebylarger strain for a given torque load.

The width is precisely defined at the surface of the waferbut may vary further down toward the center due to imperfec-tions in the etching process. The relative influence on the stiff-ness and, thereby, the sensitivity from each height segment,,of the stiffness reduction beams decreases closer to the center ofthe beam, as shown in Fig. 4(b). Hence, errors in the width canbe tolerated closer to the center without affecting the sensitivity.

Another consequence of the large height dependence is thatthe stiffness, and thereby the sensitivity, is decreased insignif-icantly by a center beam reinforcement as shown in Fig. 4(c).The mechanical resistance to influences from other directionsthan those flow induced, i.e., pulling and shear forces, dependson the cross sectional area and the moment of inertia for thecross-axis direction. The inclusion of the center beam results ina much larger cross sectional area leading to higher mechanicalstrength and lower cross-axis sensitivity.

The stress at the strain gauges of the reinforced torque sensoris given by

(10)

where is the thickness of the center beam reinforcement.The sensitivity effect from the center beam can be neglected if

SVEDIN et al.: STATIC TURBINE FLOW METER WITH A MICROMACHINED SILICON TORQUE SENSOR 941

, i.e., the second term of the denominator of (10) canbe negected

(11)

For example, a 20- -thick center beam gives a sensi-tivity reduction of only 1.5%. The cross-axis moment of inertiais given by

(12)

and with the given dimensions this results in about 200 timeshigher stiffness in the cross-axis direction.

The torque sensor described here uses “lateral weakening”to control the torque sensitivity, i.e., the stiffness is reduced bydecreasing the lateral dimensions. This can be compared to themore commonly used “vertical weakening” where the stiffnessis controlled by reducing the vertical dimensions, i.e., makingthe structure thinner.

V. TORQUESENSORFABRICATION

The fabrication of the silicon torque sensor includes the for-mation of the three major elements: the mounting part, the stiff-ness reduction beams and the support part. Furthermore, straingauges and their metal conductors are formed. The process con-tains four mask steps where every mask is used twice, one timeon each side.

The starting point is a 300- , double-sided polished waferwith a 1 thermally grown oxide as electrical insulation. A5000 layer of polysilicon is deposited on both sides and dopedusing ion implantation, and 80 keV [8], [9]. Straingauges are formed on both sides using double sided alignmentfollowed by a polysilicon dry etch. The precision of the polysil-icon alignment determines the symmetry of the structure andwas performed in a Karl Suss MA/BA 6 double-sided aligner.A low temperature oxide (LTO) layer is deposited and the resis-tors are annealed to activate the dopants. After opening contactholes to the resistors, a 1.5- -thick aluminum layer is sputterdeposited and patterned photolithographically to form conduc-tors and bond pads. Every step is performed simultaneously onboth sides forming a structure symmetric with respect to the twosides of the wafer.

Finally, the stiffness reduction beams are patterned andformed using deep-reactive ion etching (DRIE) and 5-photoresist mask. The DRIE is performed first from one side,halfway through the wafer and then the other half from theother side. The versions with a center beam are created by notetching all the way through the wafer leaving a core of desiredthickness, as shown in Figs. 5 and 6.

VI. TORQUESENSORMEASUREMENTS

The torque sensor is glued and wire-bonded to a printed cir-cuit board with separate connections for each strain gauge re-sistor. The four strain gauges are then connected externally toform a full Wheatstone bridge and supplied with 2 V from aPowerbox 3000 stabilized power supply. The resistors have re-sistances between 900and 5 depending on the dimensions

Fig. 5. Side-view SEM of the torque sensor showing the center reinforcementbeam and the good alignment of the double sided etching.

of the stiffness reduction beams. The bridge output was mea-sured using a HP34001A digital multimeter.

Before assembling the flow meter, the torque sensitivitywas measured for the torque sensors by simply applying fixedweights to the outer tip of the mounting part of the siliconpiece, shown in Fig. 3, resulting in the calibration curvesshown in Fig. 7. The measurements depicted in Figs. 7 and 8were done on torque sensors with different lengths and widthsof the stiffness reduction beams and without center beamreinforcement.

The output signal from the torque tests shows good linearitywith the highest sensitivity for the versions with the narroweststiffness reduction beams. As predicted by (9) and (10) the sen-sitivity is independent of the length and inversely propor-tional to the width of the stiffness reduction beams. Themeasurements plotted in Fig. 8 represent three different lengths

: 50 , 100 and 200 , all at four widths showingthe expected inverse proportional relation to the beam width.

Even if the length does not influence the sensitivity it canplay an important role for the sensor characteristics. The lengthdetermines the deflection for a given torque load and thus, thespring constant. Hence, the deflection and the spring constantcan to some extent be chosen independently from the sensi-tivity. The spring constant will affect the dynamical properties,i.e., the response time of the sensor. A shorter length gives afaster response, but making the beams too short will decreasethe validity of the model on which the sensitivity calculationsare based.

The torque measurements plotted in Fig. 9 show the influ-ence of different center beam reinforcement thicknesses whichincreases the cross-sectional area and the cross-axis moment ofinertia. The measurements were done with 50long and20 wide stiffness reduction beams and a center beamthicknesses between 18 and 47 . The theoretical sensitivitycurves show good agreement with the measurements with onlya minor loss in sensitivity relative to nonreinforced versions,e.g., with a center beam 27 thick the loss in sensitivity isonly 3% while the moment of inertia is 1.7 times higher and the

942 JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 12, NO. 6, DECEMBER 2003

Fig. 6. SEMs of the weakened part of the silicon structure with its stiffness reduction beams and the polysilicon strain gauges.

Fig. 7. Sensitivity of the torque sensor for different lengths (L) and widths(b ) of the stiffness reduction beams. The torque sensitivity shows good linearity.

cross-sectional area 2.5 times higher. The moment of inertia iscalculated from (10).

VII. FLOW METER MEASUREMENTS

The torque sensor is assembled with the turbine wheel andthe pipe section housing to form the flow meter, as shown in

Fig. 10. The turbine wheel is made of polysulfone and fabri-cated by milling the blades out of a 6-mm-thick cylinder usingconventional precision machining. A 0.5-mm and 2-mm-deepgroove is milled in the wheel for the torque sensor to fit in. Thestiffness reduction beams with the strain gauges are aligned tothe axis of the turbine and the wheel and the mounting part ofthe torque sensor is glued together with epoxy right above the

SVEDIN et al.: STATIC TURBINE FLOW METER WITH A MICROMACHINED SILICON TORQUE SENSOR 943

Fig. 8. Three curves show the sensitivity as a function of the stiffness beam width(b ) for three different lengths (L). As predicted from the theory the sensitivityis virtually independent of the length and linear to the inverse of the width.

Fig. 9. The influence of the center beam thickness(h ) on sensitivity and cross-axis moment of inertia normalized to a torque sensor without the center beamreinforcement.

stiffness reduction beams, as illustrated in Fig. 11. The align-ment of the mass center of the turbine, i.e., the turbine axis, tothe stiffness reduction part is important to minimize unwantedinertial effects. The other end of the torque sensor, the mountingpart, is glued to the pipe fixing the wheel inside the pipe sectionand leaving the bond pads outside the flow.

An experimental setup for testing the flow sensor was builtand is schematically outlined in Fig. 12. Pressurized air is fedthrough an electrically controlled valve to the two flow meters: athermal reference flow meter (TSI 800635) and the static turbineflow meter. The valve and the reference meter are connectedto the control computer by means of an I/O-card constituting

944 JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 12, NO. 6, DECEMBER 2003

Fig. 10. (a) The torque sensor and turbine wheel assembled to a flow meter module. (b) The flow meter module mounted in a 16-mm-flow channel. Pressure tapsare placed upstream and downstream of the sensor for pressure loss measurements.

Fig. 11. The torque sensor is inserted in the grove and the mounting part is glued to the turbine with the stiffness reduction beams aligned to the turbine’s centeraxis.

a feedback loop for accurate flow control. The static turbinewas connected using the same electrical configuration as for thetorque calibration measurements with a bridge supply voltage of2 V. The bridge output was measured with a HP34001A digitalmultimeter, connected via GPIB to a computer for automatedsignal acquisition. The nonrecoverable pressure loss over theturbine is measured through a pressure tap mounted at a distanceof two pipe diameters upstream from the flow meter. A differen-tial pressure sensor (Motorola MPX2010DP) was connected tothe pressure tap measuring the upstream pressure in relation tothe atmospheric pressure downstream where the static pressurerecovery is complete.

The flow rate/output signal relation in Fig. 13 was acquiredwith the static turbine connected to the test set-up. The measure-ments were done using a torque sensor with 100long

and 20 wide stiffness reduction beams giving the rela-tion between the bridge output signal and the torque generatedby the flow presented in the two scales on the left hand side ofthe diagram. The measurements show the expected quadraticflow dependence (8). Negative flow rates were measured byconnecting the flow meter in the reverse direction showing goodsymmetry between positive and negative flows with a sensitivityof in both directions. The graph of Fig. 13also presents the pressure loss or the nonrecoverable pressuredrop over the turbine being the measure of flow resistance ofthe static turbine flow meter.

Piezoresistive strain gauges are known for their temperaturesensitivity. In theory, the full Wheatstone bridge would be in-sensitive to temperature variations. The polysilicon resistorsitself has a temperature coefficient of about 600 ppm/. A

SVEDIN et al.: STATIC TURBINE FLOW METER WITH A MICROMACHINED SILICON TORQUE SENSOR 945

Fig. 12. The flow measurement setup with a valve and reference flow meter connected to the computer constituting a feedback loop controlling the flow rate.

Fig. 13. Measured flow sensor output and pressure loss for a static turbine wheel with 30blade angle and 2.7 mm blade length mounted together with a torquesensor with 20�m wide and 100�m long stiffness reduction beam.

single resistor in a quarter bridge configuration would henceyield a temperature sensitivity of 150 . The staticturbine flow sensors have shown a temperature dependence of0.2–2.5 , depending on how well matched the resis-tors are. Alternatively this can be expressed as 0.2-0.8 L/\min/ zero flow offset. The accuracy of the flow measurementslies below 1% for flows rates above 10 L/min, for lower flowrates the uncertainty is higher, up to about 10% at the lowerend of the range. The reason for this is the relatively low signallevels measured which is an issue for the measurement systemas a whole.

VIII. D ISCUSSION

The static turbine flowmeter is mainly targeted toward med-ical gas flow measurements in applications such as ventilatorsand spirometry. The desired flow range for this application is0.6-120 L/min. The static turbine flowmeter has been tested forflows between 0.5-80 L/min and works well in this range, givingit a dynamic range above 1:100. The sensitivity and range caneasily be controlled by varying the dimension of the stiffnessreduction beams on the silicon torque sensor.

Like most other mechanical flow sensor, the static turbinesensor extracts energy from the flow in order to measure the

946 JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 12, NO. 6, DECEMBER 2003

flow rate. This interaction with the flow results in a sensor thatis more or less intrusive. The degree of intrusiveness is in turnconnected to the pressure losses. For ventilator applications, thepressure losses should be in the order of 10Pa at 60 L/min.This value is 190 Pa for the turbine dimensions of the sensorpresented here, The previously reported lift force sensor targetedtoward the same application is less intrusive, i.e., it has lowerpressure losses. On the other hand, it also has considerably lowersensitivity [10].

The measurements on the new flow meter have also shownthe required bidirectionality, which is a result of the fully sym-metrical, linear torque sensor and the symmetric turbine.

These features indicate that this new static turbine sensor iswell suited for the intended applications.

IX. CONCLUSION

We have presented a new type of flow sensor where a staticturbine converts the volume flow into a torque resulting in anonthermal, bidirectional, volumetric flow meter. The flow-in-duced torque is measured using a symmetric, double-sided sil-icon torque sensor developed for this purpose.

The sensor in this study has a turbine wheel with a bladelength of 2.7 mm and a blade angle of 30giving a sensitivityof 4.0 when measured with a torque sensorhaving 20 wide and 100 long stiffness reduction beams.

The sensitivity along with a relatively low pressure drop in-dicate that this new sensor is well applicable for ventilator mea-surements.

REFERENCES

[1] N. T. Nguyen, “Micromachined flow sensors—a review,”Flow Mea-sure. Instrum., vol. 8, pp. 7–16, 1997.

[2] R. C. Baker, “Turbine and related flowmeters. I. industrial practice,”Flow Measure. Instrum., vol. 2, pp. 147–61, 1991.

[3] D. W. Spitzer,Flow Measurement: Practical Guides for Measurementand Control: Research Triangle Park: Instrument Society of America,1991.

[4] E. Stemme,Static Turbine Flow Meter Swedish Patent Application.[5] R. E. Thompson and J. Grey, “Turbine flowmeter performance model,”

J. Basic Eng., Trans. ASME, vol. 92, pp. 712–723, 1970.[6] M. Rubin, R. W. Miller, and W. G. Fox, “Driving torques in a theoret-

ical model of a turbine meter,”J. Basic Eng., Trans. ASME, vol. 87, pp.413–420, 1965.

[7] Y. Xu, “A model for the prediction of turbine flowmeter performance,”Flow Measure. Instrum., vol. 3, pp. 37–43, 1992.

[8] E. Obermaier and P. Kopystynski, “Polysilicon as a material for mi-crosensor applications,”Sens. Actuators, Phys. A, vol. 30, pp. 149–155,1992.

[9] V. Mosser, J. Suski, J. Goss, and E. Obermeier, “Piezoresistive pressuresensor based on polycrystalline silicon,”Sens. Actuators, Phys. A, vol.28, pp. 113–132, 1991.

[10] N. Svedin, E. Kälvesten, E. Stemme, and G. Stemme, “A new Silicongas-flow sensor based on lift force,”J. Microelectromech. Syst., vol. 7,pp. 303–308, 1998.

Niklas Svedin was born in Stockholm, Sweden, on October 7, 1971. He re-ceived the M.Sc. degree in electrical engineering from the Royal Institute ofTechnology, Sweden, in 1995. Presently, he is a graduate student in the SiliconSensor Research Group at the Department of Signals, Sensors and Systems atthe Royal Institute of Technology, Stockholm, Sweden.

His research is in the field of silicon sensors and actuators, especially for flowmeasurements.

Erik Stemmewas born in Halmstad, Sweden, on October 21, 1921. He receivedthe M.Sc. degree in electrical engineering from Chalmers University of Tech-nology, Gothenburg, Sweden, in 1946 and the honorary Dr.-Ing degree from theTechnische Hochschule, Darmstadt, West Germany, in 1959.

From 1946 to 1950, he was employed by the Research Institute of NationalDefence (FOA), Stockholm. During 1947 and 1948, he received a fellowshipand took a leave of absence from FOA to join the Institute for Advanced Study’sElectronic Computer Group and the RCA Laboratories, Princeton, NJ. In 1950,he joined the Swedish State Board for Computing Machinery where he wasresponsible for the development and design of the computer BESK. In 1956, hecame to Facit, Stockholm, where he was the head of the electronics division. In1963, he was appointed a Professor at Chalmers University of Technology.

Dr. Stemme is Member of the Royal Swedish Academy of Engineering Sci-ences. In 1954, he received the Gold Medal of the Royal Swedish Academy ofEngineering Sciences and in 1965 the Gustaf Dalén medal.

Göran Stemme(M’98) was born in Stockholm, Sweden, on February 4, 1958.He received the M.Sc. degree in electrical engineering and the Ph.D. degreein solid-state electronics both from the Chalmers University of Technology,Gothenburg, Sweden, in 1981 and 1987, respectively.

In 1981, he joined the Department of Solid State Electronics, Chalmers Uni-versity of Technology. In 1990, he became an Associate Professor heading theSilicon Sensor Research Group. In 1991, he was appointed Professor at theRoyal Institute of Technology, Stockholm, Sweden. He heads research on sen-sors and actuators based on micromachining of silicon in the Department ofSignals, Sensors and Systems.

本文献由“学霸图书馆-文献云下载”收集自网络，仅供学习交流使用。

学霸图书馆（www.xuebalib.com）是一个“整合众多图书馆数据库资源，

提供一站式文献检索和下载服务”的24 小时在线不限IP

图书馆。

图书馆致力于便利、促进学习与科研，提供最强文献下载服务。

图书馆导航：

图书馆首页 文献云下载 图书馆入口 外文数据库大全 疑难文献辅助工具