- Email: [email protected]

New tactile sensor chip with silicone rubber cover Michael Leineweber a,) , Georg Pelz a , Michael Schmidt b, Holger Kappert c , Gunter Zimmer a ¨ a

Department of Electrical Engineering, UniÕersity of Duisburg, FB9, FG EBS, Finkenstraße 61, D-47057 Duisburg, Germany b EPOS GmbH and Co. KG, Bismarckstraße 120, D-47057 Duisburg, Germany c Fraunhofer-Institute FhG-IMS, Finkenstraße 61, D-47057 Duisburg, Germany Received 16 September 1999; received in revised form 8 December 1999; accepted 21 December 1999

Abstract We report on a new tactile sensor chip developed for measuring distribution of forces on its surface. The chip has eight force-sensitive areas, called ‘‘taxel’’ with a pitch of 240 mm. Surface micromachining techniques are used to produce small cavities that work as pressure-sensitive capacitors. The process is CMOS compatible, therefore, on-chip switched capacitor circuits can be used for signal amplification. To enable transduction of normal forces to the sensitive areas, we cover the sensor chip surface with silicone rubber. First, measurements show that the sensor’s output can be explained by results from contact mechanics. We demonstrate this by the simple case of a hard sphere pressed in the silicone rubber cover. The center of contact can be measured within 2 mm precision. The radius of the sphere and the load working on it can be estimated with high precision from the tactile sensor output data. q 2000 Elsevier Science S.A. All rights reserved. Keywords: Tactile sensor; Silicone micromachining; Silicone rubber layer; Half-space model

1. Introduction We developed a new tactile sensor chip in order to measure the distribution of forces that act on its surface. The sensor is suitable for delicate manipulation tasks, especially in the field of micromanipulation. It provides information not only about the amount of force exerted during manipulation, but also about the position and orientation of the manipulated object. High-resolution tactile sensors based on silicon micromachining have been realised by previous authors. Investigations started maybe in 1985 with an 8 = 8 tactile array that was produced in a dissolved wafer process at the University of Michigan. The taxels consisted of force sensitive silicon membranes, which were anodic bonded on a glass substrate. The whole array was 16 = 16 mm2 and an external switched-capacitor ŽSC. circuit was used to read-out the capacitors between the membranes and metal electrodes located on the glass substrate w1x. Similar techniques have been used by Suzuki et al. w2x. Sugiyama et al. w3x realised a 32 = 32 taxel array on a single chip, where

each taxel was 250 = 250 mm2 . The pressure sensitive membranes consisted of silicone nitride and their elongation was measured by the piezoresistive effect. A readout circuit was also integrated on the tactile sensor chip. Some researchers investigated bulk micromachined sensor structures, which enable the measurement of both normal and shear stresses, which might be useful in future robotic manipulation w4,5x. Others report on surface micromachined capacitive cells for medical w6x and fingerprint applications w7x. However, these designs lack on chip signal amplification circuitry, which leads to noisy signals and extensive wiring. No attempt has been made to analyse the sensors output in terms of contact mechanics. Tactile sensors have been proposed for micromanipulation and medical purposes w8x, but none of the designs presented so far seems to be well suited for such applications.

2. System design and process 2.1. Pressure sensor process

)

Corresponding author.

The tactile sensor chip presented here is based on the FhG-IMS pressure sensor process, which is compatible

0924-4247r00r$ - see front matter q 2000 Elsevier Science S.A. All rights reserved. PII: S 0 9 2 4 - 4 2 4 7 Ž 0 0 . 0 0 3 1 0 - 1

M. Leineweber et al.r Sensors and Actuators 84 (2000) 236–245

with the standard 1.2 mm CMOS process. Therefore, integration of the sensing elements and electronics for signal conditioning and data transfer on the sensor chip is possible. The pressure-sensing elements consist of a capacitance where the top plate is represented by a polysilicon membrane. The bottom electrode is implanted in the silicon substrate. The cavity, which is 80 mm in diameter, is realized by employing a sacrificial layer Žsilicon dioxide., which is removed later by lateral etching with hydrofluoric acid. The height of the cavity is approximately 800 nm. The etching channels are subsequently sealed by a CVD process by deposition of silicon dioxide. The force sensitive areas Žwhich are sometimes called ‘‘taxel’’. consist each of the two sensor capacitors and two reference capacitors. The latter includes a much thicker membrane and, therefore, their pressure sensitivity is much lower, but parasitic capacitances and temperature effects are much the same. Fig. 1 shows a photo of the realised taxel with 240 = 240 mm2 . The process is described in details in Refs. w9,10x.

237

2.2. Signal amplification circuit We use SC technique to create an analog voltage signal from the pressure sensitive capacitors. Fig. 2 shows the schematic of the two-stage SC circuit. At a time, only one of the eight parallel switches is on. Therefore, the taxels are connected sequentially with the SC circuit. The output of the first stage is given by:

f1 Vout1 s Vmin y

CR

ž / CS

Ž Vmax y Vmin .

Ž 1.

f2 Vout1 s Vmin

where f 1 and f 2 are two non-overlapping clock signals f1 Ž f 2 . Ž10–50 kHz. and Vout1 Vout1 is the output voltage of the first stage if f 1Ž f 2 . is high. C R and CS are the taxel capacitances, respectively. The output of the second stage

Fig. 1. Pressure-sensitive area Ž‘‘taxel’’..

M. Leineweber et al.r Sensors and Actuators 84 (2000) 236–245

238

Fig. 2. Schematic of SC circuit.

and, hence, the output signal of the chip can be calculated as: f1 f2 Vout2 s Vout2 f2 Vout2 s Vmin q

C2

ž /Ž C1

f1 Vout1 y Vcal y Voff .

Ž 2.

Vmax and Vmin are voltage references of 4.5 and 0.5 V, respectively. The ratio of the capacitors C 1 and C 2 can be used to change signal amplification. A sample and hold capacitor C H in the second stage makes the output signal of the chip time-continuous. Vcal is an extra analog input voltage used to cancel taxel offset voltages. The realized chip layout is approximately 6000 = 400 mm2 ŽFig. 3.. 2.3. Sensor packaging After processing, the silicon wafer is carefully thinned to 160 mm, followed by sawing, die-bonding and bonding in a standard DIL-Package or on a printed circuit board. A silicone rubber sheet is glued to the chip surface. We examined several types of commercially available materials. Silicone rubber has many advantages in the field of

microelectronic packaging. It exhibits good stability with regard to most chemicals, including weak acids and polar solvents. The elastic properties are stable between y508C and 1808C. Electrical resistivity and breakdown field strength are high. Last, but not least, since the material is self-healing, we can also use the liquid material to glue the sheets onto the chip. We decided to use ELASTOSILw RT601, a two-component silicone elastomer produced by Wacker Chemical because of its durability and high resistance to tearing. The maximum tensile stress is 7.0 Nrmm2 and the maximum tensile strain reaches 100%, therefore, the modulus of elasticity E is approximately 7 = 10 6 Nrm2 . The hardness of the material is 45 Shore A w11x. We use Wacker’s primer G795 to enhance the sticking of the rubber on the silicon surface.

3. Calibration procedure In order to calibrate the sensor chip and to check whether the sensor elements work correctly, we implement

Fig. 3. Chip photo of tactile sensorship.

M. Leineweber et al.r Sensors and Actuators 84 (2000) 236–245

our first test run within an air pressure chamber. The results are plotted in Fig. 4 for air pressure between normal atmosphere Ž; 1 bar. and 3 bar Ž1 bar s 10 5 Pa, 1 Pa s 1 Nrm2 .. Note that the offset values of the sensor chip were set to 0.5 V by the external voltage Vcal . There are some differences in the slope for each taxel. The sensor’s sensitivity between 1 and 2.4 bar is almost linear Žwithin 2%. and its value for different taxels is between 1.2 and 1.5 Vrbar. The change in slope for pressure ) 2.4 bar is due to the non-linear behaviour of the plate capacitance when the upper electrode approaches the lower electrode. Software calibration was carried out with the following fit-function:

Si Ž P . s f 0,i q

1 e 0,i q e1 ,i P

Ž 3.

where Si Ž P . is the sensor’s output for the ith taxel and f 0, i , e 0,i and e1,i are fit parameters. This function is often used to describe the behaviour of a plate capacitor. If P approaches ye0, ire1,i Žwhich is equivalent to the pressure, where the membrane touches the substrate., Si Ž P . approaches infinity. We tried some more complex fit function with up to five parameters, but the results did not justify the effort. The calibration was checked again after fixing the silicone rubber sheet. No change of the output signal was observed.

239

4. Contact mechanics We demonstrate the sensor’s abilities with the example of a small sphere pressed against the sensor’s elastomer cover. It is possible to calculate the stresses produced by a sphere pressed in an elastic half-space w12x. Although our elastomer is not infinite in depth, the model can be used for data analysis of the tactile sensor chip, as we will demonstrate. We take the elastic half-space to be bounded by the plane z s 0 of an rectangular coordinate system with the positive values of z located inside the half-space. The stresses produced in the elastic half-space produced by a point load P is given by w13x:

sz Ž r , z . s y

with

z3

3P 2p

Ž r2qz2 .

(

rs x2 qy2

5r2

Ž 4.

The normal stresses under an elastic strip with finite thickness z 0 produced by a point load have been calculated by the finite element software ANSYS. Fig. 5 shows that for z 0 between 100 and 1000 mm, one can reach almost perfect agreement for the stress beneath the elastic strip and the model of half-space in choosing an appropriate value for z, the depth inside the half-space. For this case, the parameter z is chosen to be 0.8 times the

Fig. 4. Output signal vs. pressure of the tactile sensor chip.

M. Leineweber et al.r Sensors and Actuators 84 (2000) 236–245

240

Fig. 5. Comparison between FEM-data for the elastic strip and half-space model Eq. Ž4..

thickness of the strip z 0 . If we would take z 0 s z Žas one might expect., the stress beneath the strip is higher than in the elastic half-space because of the reaction forces at the bottom of the strip. Therefore, it is reasonable to choose a smaller value for z. These results lead to the following tactile sensor model equation: Ss

1

ar2

br2

H H s Ž x , y, z . d xd y ab yar2 ybr2 z

Ž 5.

where S is the sensors output signal and a and b are the dimensions of the taxel in x and y directions, respectively. sz represents the stress, calculated within the half-space model. In our model, z is not a free parameter, but fixed to 0.8 z 0 throughout the paper, which leads to good correspondence with experimental results as will be demonstrated. We can always omit the integration if the variation of sz is small over the taxel area ab. In the theory of elastic contact developed by H. Hertz, the relationship between the depth of penetration d and the radius of contact between the sphere and the elastic halfspace A is given by

ds

A2

Ž 6.

R

where A might be expressed in terms of P, the load that acts on the sphere. 3

As

(

3 PR Ž 1 y n 2 . 4E

Ž 7.

where R is the radius of the sphere, E the modulus of elasticity and n Poisson’s ratio, which is nearly 0.5 for an almost incompressible material as silicone rubber w13x. From the equations above, we are able to calculate the load P. 4 E P s d 3r2 R 1r2 Ž 8. 3 Ž1yn 2 . The first point of contact between the sphere and the half-space occurs at the origin. From Ref. w12x, we get sz , the normal stress in the elastic half-space acting in vertical direction to the surface: 3P

sz Ž r , z . s

3

z

ž /

A2 u

2p A2 'u u 2 q A2 z 2 1 us r 2 q z 2 y A2 2

with

(Ž r q z y A . q 4 A z

"

2

2

2 2

2 2

(

rs x2 qy2

and

Ž 9.

For a given value of z, the stress sz has its maximum value at r s 0: 3P 1 szmax Ž z . s Ž 10 . 2 p A2 q z 2 We can use this equation to calculate the load P from szmax : 3 A2 s hP y z 2 , with h s 2pszmax

´ A s h P y 3h P z q 3hPz y z 6

3

3

2

2 2

4

6

Ž 11 .

M. Leineweber et al.r Sensors and Actuators 84 (2000) 236–245

241

hand, R can be calculated from szmax as well if P is given. From Eqs. Ž7. and Ž10.:

Rs

ž

3P

1

2p sz

max

3r2

yz

2

/

4

E

1 2

3 Ž1yn . P

Ž 14 .

If the penetration depth d is given instead of P: 2

R q

Fig. 6. Experimental set-up.

2

A s

9 Ž1 y n 2 . R2 E2

16

ž

P3y 3

z2

k q

h

h3

/

P 2 s kP 2

P2q3

Ž 12 .

z4

z6

h

h3

Py 2

s0

d

2

/

yh d Rq

z4

d2

s0

Ž 15 .

5. Experimental set-up and results

From Eq. Ž7.:

6

ž

2 z2

Ž 13 .

Eq. Ž13. can be solved at once by use of the cardanian formulas. In other words, P can be calculated out of szmax if the radius R of the indenter is known. On the other

The set-up is depicted in Fig. 6. The sensor chip is fixed on an x–y table, which allows positioning with 1 mm precision. The sphere can be moved along the z-axis by means of a position device with equally 1 mm resolution. The chip surface is identified with the X–Y plane and the direction of the taxels with the X-axis. We analyze the sensor’s output data for a fixed sphere position but increasing penetration. Fig. 7 shows the measured stresses for a sphere with 1.6 mm radius. Stresses are presented for a penetration depth ranging from 10 to 90 mm, i.e., 1% to 9% of the rubber cover. Results from the half-space model as explained in Section 4 are also included. The parameter z is chosen to be 0.8 mm for all cases. For penetration depths smaller than 80 mm, this leads to good correspon-

Fig. 7. Comparison of the tactile sensor data Žsymbols. and normal stresses sz Žlines. calculated with the elastic half-space model w8x.

242

M. Leineweber et al.r Sensors and Actuators 84 (2000) 236–245

Fig. 8. Output signal for seven taxels when changing the spheres position along the taxel line.

dence between measured data and model. Penetration larger than 80 mm results in stresses higher than the one predicted by the model at X s 0 and lower for X ) 0.4 mm. Fig. 8 shows the output signal for seven taxels when changing the sphere’s position along the X-axis. The sig-

nal of each taxel has the shape of a gaussian curve, and the maximum signal is reached when the taxel is located under the sphere. The signals of the different taxels overlap each other. This is partly because of the contact radius A Ž360 mm for 80-mm indentation., which is bigger than the

Fig. 9. Gaussian curve through three taxel data and least-square fit through all data points.

M. Leineweber et al.r Sensors and Actuators 84 (2000) 236–245

243

Table 1 Results of data fit to gaussian curve with Eqs. Ž17. – Ž19. Position of sphere wmmx

Center of gaussian curve wmmx

Center error w%x

Height of gaussian curve wbarx

Height error w%x

Width wmmx

Width error w%x

1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2.0

0.1212 0.2198 0.3184 0.4201 0.5185 0.6248 0.7263 0.8305 0.9285 1.0292 1.1313

– 1.4 1.4 0.4 0.7 0.7 0.85 1.3 0.9 0.9 1.0

1.895 1.934 1.906 1.924 1.925 1.920 1.928 1.936 1.914 1.891 1.922

1.21 0.86 0.61 0.32 0.38 0.12 0.51 0.96 0.20 1.37 0.22

0.465 0.464 0.471 0.466 0.477 0.478 0.483 0.477 0.475 0.485 0.475

1.98 2.10 0.79 1.67 0.57 0.88 1.82 0.52 0.10 2.37 0.26

width of a taxel Ž240 mm.. Second, the stress field is broadened because of the influence of the elastic layer. From Eq. Ž4., we calculate the distance r X , where the normal stress is reduced to half of its maximum value:

sz Ž r X , z . szmax

!

´ r s z'0.5 X

z5

1

s

s

5

2 X2

2

Ž r qz . y2r5

2

y 1 f z 0.565

One can see that the extension of the stress field increases proportional to the depth z inside the elastic half-space, and this is the same for the thickness z 0 of the elastic layer if we take the result of our FEM investigation. For applications that require high resolution Žfor example, object identification by tactile imaging., this so-called crosstalk between taxels has to be reduced by choosing a

very thin elastic layer w14x. For other applications, an overlap of the taxel signals is needed to avoid ‘‘death zones’’ between the taxels. One example is precise position measurement. This is carried out for the tactile data shown in Fig. 8. In this case, position measurement does not even require a time-consuming least-square algorithm fit. Instead, we make use of the data of only three taxels for an analytical calculation of a gaussian curve that describes the slope of our measured data well. We choose the data point with the highest signal and its two neighbours to determine central position x 0 , width w and height A of the gaussian curve y Ž xyx 0 . 2

G Ž x . s Ae

w2

Fig. 10. Dependence of position error from signal error.

Ž 16 .

M. Leineweber et al.r Sensors and Actuators 84 (2000) 236–245

244

Table 2 Calculated data for the sphere’s radius and load Position of sphere wmmx

Height of gaussian curve wbarx

Radius R wmmx

Radius R error w%x

Load P wNx

Load P error w%x

1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2.0

1.895 1.934 1.906 1.924 1.925 1.920 1.928 1.936 1.914 1.891 1.922

1.59241 1.69392 1.62035 1.66723 1.66988 1.65669 1.67785 1.69932 1.64101 1.58239 1.66195

0.5 5.9 1.3 4.2 4.4 3.5 4.9 6.2 2.6 1.1 3.9

0.302 0.308 0.303 0.306 0.307 0.305 0.307 0.308 0.304 0.3 0.307

1.0 0.9 0.7 0.3 0.6 0.1 0.6 0.9 0.4 1.7 0.6

by simple algebraic calculation: x 12 y x 22 y

ln Ž Sig 2 y Sig 1 .

Ž x 2 y x 32 . ln Ž Sig 3 y Sig 2 . 2 x0 s ln Ž Sig 2 y Sig 1 . 2Ž x1 y x 2 . y 2Ž x 2 y x 3 . ln Ž Sig 3 y Sig 2 . w2 s

sensitivity, which is 1.35 Vrbar Žfrom Fig. 4.. Therefore, the maximum position error is:

x 22 y x 32 y 2 x 0 Ž x 2 y x 3 . ln Ž Sig 3 y Sig 2 .

Ž 17 .

w2

Ž 18 .

Ž 19 .

where x 1, . . . ,3 are taxel positions along the x-axis and Sig 1, . . . ,3 taxel output data. Fig. 9 compares this analytical procedure with the result of the least-square fit. Concerning the center position x 0 and the height A, there is no difference for the two gaussian curves. Table 1 shows the result of the analysis for the central part of the tactile sensor with the data of Fig. 8. Errors are given in percent deviation of the mean value, except for the center of the gaussian curve, where the error is taken as the deviation of the x-step value, which was 0.1 mm. The error in the sphere’s position is smaller than 2 mm. This result can be supported theoretically by the calculation of the maximum error in the measurement of the center x 0 resulting from the noise of the sensor chip output signal. The derivation of the gaussian curve ŽEq. Ž16.. is: X

g Ž x . s y2 A

x w

2

y Ž xyx 0 . 2

e

w2

g X Ž 240 mm .

s

3.125 = 10 8 Nrm3

s 2.4 mm

Ž 21 .

Ž x 1 yx 0 . 2

A s Sig 1e

750 Nrm2

DY D Xs

Ž 20 .

Fig. 10 illustrates that the maximum position error occurs if the first and last of the three taxels exhibit maximum signal error with different signs. The maximum signal error observed was 10 mV, which is equivalent to 750 Nrm2 derived from the mean value of the sensor’s

which agrees well with our experimental results. To calculate either the radius R of the sphere or the load P from the height A in Table 1, we make use of the formulas Ž13. and Ž15. where we take zX as 0.8 mm again. Results are listed in Table 2. The radius R can be measured with "6% accuracy. We could not determine the exact value of the load P to compare it with our results, but we notice that with the data of Table 1, the result for P is stable within 2% for the central part of the tactile sensor chip. It is possible to calculate radius R and load P together from height A and width w of the gaussian curve by using numerical calculation methods.

6. Conclusion A new tactile sensor chip with micromachined force sensitive areas is presented. These taxels are covered with elastic rubber. The stresses under the cover can be well understood by means of the elastic half-space model, which has been demonstrated by the example of a sphere pressed into the rubber. The center of the sphere’s contact can be determined with 2 mm precision. The algorithm used for this task consists of simple algebraic calculation, no time-consuming least-square algorithm fit is needed, which is important for control purposes. The load acting on the sphere and the sphere’s radius can be calculated from the sensors output data as well within 6% and 2% accuracy, respectively. Other contact geometries might re-

M. Leineweber et al.r Sensors and Actuators 84 (2000) 236–245

sult in more difficult expressions for the load within the elastic half-space model. If exact solutions are not available, one has to calibrate the sensor for each contact geometry again, what seems to be no major problem. Therefore, the sensor is well suited for force as well as position control in micromanipulation applications.

References w1x K.J. Chun, K.D. Wise, A capacitive silicon tactile imaging array, 3rd Int. Conf. on Solid-State Sensors and Actuators, 1985, pp. S22–S25. w2x K. Suzuki, K. Najafi, K.D. Wise, Process alternatives and scaling limits for high-density silicon tactile imagers, Sensors and Actuators A 21–23 Ž1990. S915–S918. w3x S. Sugiyama, K. Kawatha, M. Yoneda, I. Igarashi, Tactile image detection using a 1k-element silicon pressure sensor array, Sensors and Actuators A 21 Ž1990. S397–S400. w4x Z. Chu, P.M. Sarro, S.M. Middelhoek, Silicon three-axial tactile sensor, Sensors and Actuators A 54 Ž1996. S505–S510. w5x B. Kane, R. Cutkosky, CMOS-compatible traction stress sensor for use in high-resolution tactile imaging, Sensors and Actuators A 54 Ž1996. 511–516. w6x B.L. Gray, R.S. Fearing, A surface micromachined microtactile sensor array, Proceedings of the IEEE International Conference on Robotics and Automation Ž1996. 1–6. w7x P. Rey, A high density capacitive pressure sensor array for fingerprint sensor application, Transducers 97 Ž1997.. w8x M. Schuenemann, Anthropomorphic tactile sensors for tactile feedback systems, Proceedings of the SPIE 3206 Ž1997. 82–97. w9 x M. Kandler, CMOS kompatibler Siliziumdrucksensor in Oberflachenmikromechanik, Doctors Thesis, University of Duisburg, ¨ 1992. w10x H. Dudaicevs, M. Kandler, Y. Manoli, W. Mokwa, Surface micromachined pressure sensor with integrated cmos read-out electronics, Sensors and Actuators A 43 Ž1994. 157–163. w11x Wacker Chemie, Munich, Product information silicon rubber ‘‘Elastosil’’. w12x M.T. Huber, Comments on the theory of the contact between solid elastic bodies,wZur theorie der beruhrung fester elastischer korper ¨ ¨ x Annalen der Physik 14 Ž1904. 153. w13x K.L. Johnson, Contact Mechanics, Cambridge Univ. Press, 1992. w14x M. Shimojo, Mechanical filtering effect of elastic cover for tactile sensor, IEEE Transactions on Robotics and Automation 13 Ž1. Ž1997. 128–132, Feb.

245

Biographies Michael Leineweber was born in 1969. He studied solid-state physics at the University of Duisburg, receiving his diploma in 1995. He is now a PhD student at the Department of Electrical Engineering at the University of Duisburg. His research interests include fabrication and simulation of micromechanical systems, especially tactile sensors. Georg Pelz received his diploma degree in computer science from the University of Dortmund, Germany in 1988 and doctor’s degree in 1993 from the University-GH Duisburg, Germany. From 1989 to 1993, he was with the Fraunhofer Institute of Microelectronic Circuits and Systems, Duisburg. Presently, he is with the Department of Electronic Devices and Circuits, Gerhard Mercator University-GH Duisburg, Germany. His interests of research include modeling and simulation of electromechanical systems, computational geometry, and VLSI circuit design and verification. Dr. Pelz is member of the German Society of Computer Science ŽGesellschaft fur ¨ Informatik.. He received a distinguished paper citation at the International Conference on Computer-Aided Design ŽICCAD. in 1991 and a nomination for the best paper award at the Design Automation Conference ŽDAC. in 1992. Michael Schmidt was born in Bottrop, Germany, in 1966. He received his Dipl.Ing. degree in electrical engineering and PhD from the University of Duisburg, Germany, in 1993 and 1997, respectively. He joined the Fraunhofer Institute of Microelectronic Circuits for Systems ŽIMS. in 1993, where he worked on analog circuits for sensor applications in CMOS and SOI techniques. Since 1998, he has been with EPOS Žembedded core and power systems. in Duisburg working on automotive integrated circuit design. Holger Kappert was born in 1968. He has received his Dipl.Ing. degree in electrical engineering from the University of Bochum in 1993. Since then, he has been with the Fraunhofer-Institute of Microelectronic Circuits and Systems where he is working on embedded systems. His special research interest is the design for testability of those systems. Gunter Zimmer received his diploma in physics from the Technische ¨ Hochschule Darmstadt in 1966 and PhD from the Technische Hochschule Munchen in 1968. Until 1970, he was an assistant in the Physics ¨ Department of the Technische Hochschule Munchen. In 1970, he joined ¨ the Semiconductor Division of Siemens, where he worked on bipolar and MOS technology and on the application of ion implantation in various device technologies. From 1973 to 1984, he was a chief engineer and lecturer at the University of Dortmund with research activities in integrated circuit technology, particularly using MOS devices. In 1984, he was appointed professor at the University of Duisburg and director of the Fraunhofer Institute of Microelectronic Circuits and Systems, Duisburg. Since 1991, he has also been director of the Fraunhofer Institute of Microelectronic Circuits and Systems in Dresden.