Bereich Elektrotechnik und Informationstechnik Elektrotechnik und Informationstechnik Elektrotechnik und Informationstechnik

Transcription

1 Fakultät für Mathematik und Informatik Bereich Elektrotechnik und Informationstechnik Elektrotechnik und Informationstechnik Elektrotechnik und Informationstechnik Elektrotechnik und Informationstechnik Elektrotechnik und Informationstechnik Jahresbericht 2006 Annual Report

2 Table of Contents Introduction Allgemeine und Theoretische Elektrotechnik/Fundamentals and Theory of Electromagnetics (Prof. Dr. Reinhold Pregla) Bauelemente der Elektrotechnik/Electronic Devices (Prof. Dr. Wolfgang R. Fahrner) Datenverarbeitungstechnik/Data Processing Technology (Prof. Dr. Bernd J. Krämer) Juniorprofessur für Softwaretechnik/Software Technology (Dr. Jens Krinke) Elektrische Energietechnik/Electrical Power Engineering (Prof. Dr. Detlev Hackstein) Informationstechnik/Computer Engineering, esp. Real Time Systems (Prof. Dr. Dr. Wolfgang A. Halang) Juniorprofessur für Eingebettete Systeme/Embedded Systems (Dr. Zhong Li) Kommunikationssysteme/Communication Systems (Prof. Dr. Firoz Kaderali) Optische Nachrichtentechnik/Optical Information Technology (Prof. Dr. Jürgen Jahns) Juniorprofessur für Optische Mikrosysteme/Optical Microsystems (Dr. Matthias Gruber) Prozeßsteuerung und Regelungstechnik/Control Systems Engineering (Prof. Dr. Helmut Hoyer, Dr.-Ing. Michael Gerke)

3

4 Introduction The dominant event last year was the fusion of the former Department of Electrical Engineering and Information Technology with Mathematics and Computer Science. Meanwhile this is no longer an issue but a matter of daily life. Several positive things can be reported: Prof. Unger, formerly University of Rostock, joined the faculty as a Professor of Communications Networks at the end of We wish him and his co-workers a successful start at the FernUniversität Hagen. During the year, the three Juniorprofessors Gruber (Optical Microsystems), Krinke (Software Technology) and Li (Embedded Systems) were positively evaluated and will continue their good work. Finally, the Master Program Electrical Engineering and Information Techology was accredited and has continued to attract an increasing number of students. The first degrees were given out in spring of Jürgen Jahns

5

6 formerly chair of Allgemeine und Theoretische Elektrotechnik (Fundamentals and Theory of Electromagnetics) Prof. Dr.-Ing. Reinhold Pregla Universitätsstr. 27 D Hagen Phone: FAX: Hagen.de Staff: Phone M.Sc. Agnieszka Barcz (until Oct. 2006) Wolfgang Benner 1149 Dr.-Ing. Hans Georg Bergandt (until Dec. 2006) PD Dr.-Ing. Stefan Helfert 1144 Prof. Dr.-Ing. Wilfrid Pascher 1145 Georg Schindel (until March 2006) Dipl.-Ing. Werner Schubert 1149 Doris Weyerbusch 1141

7 Left-handed materials Stefan F. Helfert Usually the electric and magnetic material parameters i.e. the permittivity and permeability are positive (at least one of them). In this case the three vectors ~ E (electric field), ~ H (magnetic field) and ~ k (direction of wave propagation) are connected via the right hand rule (right-handed media). As already predicted in 1967 by the Russian scientist Veselago [1], materials in which the mentioned parameters are negative give access to some interesting phenomena. E.g. the refractive index is negative. Therefore waves are refracted in negative direction. Fig. 1 shows the effect of the left handed material. In the upper graph we have three right-handed materials concatenated. In this case all refraction occurs in positibve angles. The lower Figure however shows a left-handed material (LHM) sandwiched between two right-handed materials leading to negative refraction angles Fig. 1: [1] V.G. Veselago, "The electrodynamics of substances with simultaneously negative values of " and μ", Soviet Physics Uspekhi, vol. 10, pp , (Originally in Russian in Uspekhi Fizicheskikh Nauk, vol. 92, no. 3, pp , July 1967). [2] A. A. Houck, J. B. Brock and I. L. Chuang, "Experimental Observations of a Left-Handed Material that Obeys Snell's Law", Physic. Rev. Lett., vol. 90, pp , [3] C. M. Soukoulis, M. Kafesaki and E. N. Economou, "Negative-Index Materials", Advanced Materials, vol. 18, pp , 2006.

8 Introducing Oblique Coordinates into various Numerical Algorithms Stefan F. Helfert Recently, oblique coordinates (see Fig. 1) were introduced into the method of lines [1] [2]. They were used particularly to compute band structures in photonic crystals with hexagonal structure. Now, these coordinates were applied to other methods as well. These were the Galerkin method (GM) and the finite difference (FD) method. In the first algorithm the fields were approximated by series of sine and cosine functions. The main difference to the Cartesian case is that both functions must be used at the same time, whereas in Cartesian coordinates the sine and the cosine series can be examined independently of each other. By determining the Floquet modes band structures are determined similar to the algorihm in the MoL. A slighly different approach is followed in the FD. All fields and derivatives are discretized with finite differences. To compute the band structures the propagation constants are enforced by introducing suitable periodic boundary conditions. Then, the corresponding frequencies are determined. This is contrast to the methods mentioned before, where the propagation constant was determined as function of frquency. Fig. 2 shows a comparison of the MoL results with those of the FDM. Slight differences can be recognized for higher modes. The parameters of the photonic crystal structure can be found in [3]. z u θ y x,v Fig. 1: Oblique coordinate system a / wavelength MoL Villeneuve FD Γ M K Γ Fig. 2: Band diagram for a hexagonal lattice [1] R. Pregla and W. Pascher, The Method of Lines", in Numerical Techniques for Microwave and Millimeter Wave Passive Structures, T. Itoh, (Ed.), pp J. Wiley Publ., New York, USA, [2] S. F. Helfert, "Applying oblique coordinates to the method of lines", Progress In Electromagnetics Research, PIER 61, pp , [3] P. R. Villeneuve, S. Fan, S. G. Johnson,, and J. D. Joannopoulos, Three dimensional photon confinement in photonic crystals of low dimensional periodicity", IEE Proc.-Optoelectron., vol. 145, no. 6, pp , 1998.

9 Impedance/Admittance Transformation and Field Propagation by Finite Differences of Second Order Accuracy Reinhold Pregla Generalized Transmission Line Equations in matrix notation for arbitrary orthogonal coordinate systems were developed some years ago for efficient analysis algorithms based on the Method of Lines [1]. These GTL equations are coupled, first order differential equations. Especially, in Cartesian coordinates they can be combined to wave equations and can be solved analytically. However, this combination and solution is not possible in cases where the coupling matrices depend on the coordinate for the analytical solution. Examples are shown in Figs. 1a,b,c. For such cases we have developed an impedance/admittance BC I II III r z φ BC d 0 s 0 r 3 ds metal r ε r0 w r1 z ε rs φ ε rl r2 s I II III IV V (a) ZKKS1020 (b) ZKKS1040 (c) OFWF116A (d) MMPL1061 Fig. 1: Cross sections of microstrip waveguide on cylindrical body (a)(b), of a bow-tie fiber (c) and a multilayered planar optical waveguide (d). transformation with Finite Differences. The rib waveguide cross section (see Fig. 1b) is suitable to check the accuracy of the algorithm because we can transform the impedances in all layers I to V analytically. The Figure 2 shows the difference of the values obtained 9 d /λ 0 = x = 0.10 n eff = n FD n analytic linear quadratic n /number of sublayers DWFW3010 eff ε re = 0.05 = 0.02 with superstrate without superstrate s/d CCZK4010 Fig. 2: Convergence of n eff for the HE 00 mode at k 0 d = 6 using linear and quadratic FD impedance transformation formulas in the rib layer of the structure in Figure 1d Fig. 3: Dispersion diagrams of the microstrip mode on a cylindrical body and finite substrate and superstrate: r 1 = 10 mm, s 0 = 0 mm, w = d S = d 0 = 1 mm, ε rs = ε r0 = 9.7, ε rl = 1.0. with the FD impedance transformation (linear and quadratic accuracy) and the analytical result as function of the number (1/number) of sub-layers in the rib [2]. The next diagram (Figure 3) shows the effective permittivity of a microstrip waveguide on a cylindrical body (Figure 1b) [3]. The substrate (and superstrate) is of finite widths. [1] R. Pregla, Modeling of Optical Waveguide Structures with General Anisotropy in Arbitrary Orthogonal Coordinate Systems, IEEE J. of Sel. Topics in Quantum Electronics, Vol. 8, No. 6. [2] R. Pregla, Modeling of Optical Waveguides and Devices by Combination of the Method of Lines and Finite Differences of Second Order Accuracy, Optical and Quantum Electronics(2006)38:3 17. [3] R. Pregla, Analysis of Microwave Structures by Combination of the Method of Lines and Finite Differences, Mikon, Int. Conf. on Microwaves, Krakow(Poland), May

10 Eigenmodes of circular waveguides with magnetized ferrite or plasma Reinhold Pregla Eigenmodes in circular hollow waveguides (see Fig. 1) of radius a with perfectly conducting walls filled in the center with a rod of ferrite material of radius b and magnetized longitudinally were determined. The case of magnetized plasma is analogous. A cylindrical coordinate system (r, φ, z) with z-axis along the geometric axis of the structure is the obvious one to use so that the perfectly conducting surface of the guide wall is a coordinate surface, described simply by r = a. ε, µ F F z a ε d, µ d r H 0 φ b The permeability or permittivity tensor is given by ν r = ν rr ν rφ 0 ν φr ν φφ ν zz ν = ε or µ Fig. 1: Circular waveguide filled with a rod of longitudinally magnetized ferrite material We use GTL-equations [1] and discretize the fields and field equations in r-direction. For the φ-direction we use analytical solutions. The material parameters (permittivities and permeabilities) are discretized at the same points as the associated field components and are combined in diagonal matrices. The values at interfaces are always arithmetic mean values. The modes must be calculated using different discretization schemes. The modes H 01 and E 01 have azimuthal independent fields. In case µ rφ = jκ = 0 the two modes are completely decoupled. The next diagrams show the computed curves for the propagation constants of the fundamental and the first higher order modes. It is interesting to see that in case of κ = 0 the curves for the H 01 and E 01 modes crosses which is not the case for κ 0 where the modes are coupled. For comparison see [2]. β / k a / λ = 0.3 κ = H E HLHC9010 b/a (a) k 0 β / a / λ = 0.3 E 01 H 01 κ =0.5 κ =0.25 κ=0 κ =0.25 κ = b/a HLHC9020 (b) Fig. 2: Normalised propagation constant for the (a) H 11 and E 11 and (b) the H 01 and E 01 modes versus b/a in ferrite filled circular waveguides: a/λ = 0.3, ɛ r = [1] R. Pregla, Modeling of Optical Waveguide Structures with General Anisotropy in Arbitrary Orthogonal Coordinate Systems, IEEE J. of Sel. Topics in Quantum Electronics, Vol. 8 (2002), No. 6. [2] R. A. Waldron, Ferrites - An Introduction for Microwave Engineers, D. Van Nostrand Company LTD, London 1961.

11

12 Chair of ELECTRONIC DEVICES (Lehrgebiet Bauelemente der Elektrotechnik) Prof. Dr. rer. nat. Wolfgang R. Fahrner Haldener Straße Hagen Tel.: , -378 Fax: Scientific Staff: Ext. Dipl.-Ing. Wolfgang Düngen 2145 Dr.-Ing. Yuelong Huang, MSc Prof. Dr. rer. nat. Reinhart Job 379 Dr.-Ing. Yue Ma, MSc Dipl.-Ing. Thomas Müller 2200 Dipl.-Ing. (FH) Maximilian L. D. Scherff 4012 Dipl.-Ing. Stefan Schwertheim 4013 Technical Staff: Dipl.-Ing. (FH) Thomas Krauß 2201 Mrs. Katrina Meusinger 4017 Dipl.-Ing. Boguslaw Wdowiak 4016 Dipl.-Ing. (FH) Holger Winkelmann 2258 Secretary: Mrs. Barbara Kleine 377 Research activities ffl Heterojunction solar cells ffl Interaction of hydrogen with silicon ffl Radiation hardness of solar cells ffl Nanoelectronic devices based on nuclear tracks ffl Silicon-on-insulator technology

13 Interface States of Heterojunction Solar Cells W. R. Fahrner, T. Mueller, M. Scherff, H. C. Neitzert * * Via Ponte Don Melillo, Fischiano (SA) - Italy The impedance of heterojunction solar cells is measured in a frequency range from 20 Hz to 1 MHz. Their equivalent network consists of the junction capacitance C 1, the junction loss R 1, and the substrate resistance R 3. For specially prepared emitters, both the series resistance and capacitance show an additional signal (fig. 1). Surface states are a good explanation of the additional signal. In the flow diagram of fig. 2 the way to check the model is described. The measured network reflects the contribution of the junction, C 1 und G 1 = 1/R 1, of the interface states, C it and G it, and of the series resistance, R 3 (fig. 2a). In a first step, R 3 is subtracted from R s, meas. Thus, only the series components Z s, meas between the upper terminal and the T-point are left over (fig. 2b). Now, this data are converted to a parallel network, Y p = 1/Z s (fig. 2c). Finally, the admittance Y it = Y p G 1 jωc 1 is obtained (fig. 2d). A D it of /Vcm-2 and a time constant τ it of 1.7 ms are obtained Series resitance (ohm] measured Fit (two time constant model) Series capacitance [F] Fit (two time constant model) measured Frequency [Hz] Frequency [Hz] Fig. 1 (a) R s vs. f, (b) C s vs. f of a solar cell Fig. 2 : Flow diagram of the interface admittance evaluation. The network at the left (outside the sample) shows the mode (parallel or serial) on which the data are based. In (a), the total network is shown. The series resistor R 3 is subtracted (b) and R s, corr, C s, corr are left. This network is transformed in the parallel representation (c). Finally, the space charge components are subtracted so that only the surface state contributions are seen (d).

14 On the Mechanisms of Hydrogen Plasma Supported Layer Exfoliation in Silicon * R. Job, W. Düngen The Smart-Cut process (Fig. 1) provides a beneficial approach for the fabrication of SOI (Silicon On Insulator) wafers. For Smart-Cut an H + -implantation is carried out into an oxidized Si wafer #1 at fairly high doses ( cm -2 ) so that the projected range R P of the implanted ions is located at an appropriate depth below the SiO 2 top layer where the layer splitting should occur. The implanted wafer #1 is directly bonded to another oxidized stiffener wafer #2 and subsequently annealed at 550 C to provide the layer splitting near R P. By this a SOI wafer structure can be obtained, since a thin Si layer is transferred to the oxidized wafer #2. The remaining wafer #1 can be refined and used as a new stiffener wafer. In frame of a project** we investigated a modification of the Smart-Cut process, called Soft-Cut (Fig. 2), where H + -implantation is done at lower doses ( cm -2 ) and the missing H-concentration (for layer splitting) is provided by H-plasma exposure at elevated temperatures up to 650 C. H + -implantation direct wafer bonding H + -implantation H-plasma T 400 C thermal oxide on Cz Si annealing at T 550 C annealing at T 550 C direct wafer bonding layer exfoliation layer exfoliation SOI structure SOI structure Fig. 1: Process schemata of the Smart-Cut (left side) and Soft-Cut approaches (right side) Within the project we presented a study on basic mechanisms of silicon layer exfoliation by ion-cut approaches. We carried out comparative analyses with regard to the Smart-Cut process for SOI wafer production and its progression, i. e. the Soft-Cut approach. Investigations were done with µ-raman spectroscopy, electron microscopy and atomic force microscopy. For the Smart-Cut scenario, H + implantations were carried out at energies of 40 or 50 kev and doses of cm -2. Post-implantation annealing was done for 10 min at temperatures between 200 C and 700 C. For the Soft-Cut scenario H + implantations were carried out at 80 kev and doses at cm -2. Post-implantation plasma hydrogenation was done at temperatures between 250 C and 650 C. Within the Smart-Cut scenario, during the H + implantation various vacancy-hydrogen (V n -H m ) and multi-vacancy (m-v) complexes are created. From the analyses of the Raman modes in dependence on the annealing temperature, we deduced that VH defects coalesce to V 2 H 6 complexes at increasing temperatures. The m-vs are stable up to 300 C, above 300 C they decompose into V 2 H 6 defects. At T > 500 C blister formation occurs. From the intensity variations of the Raman modes correlated to internal H-saturated blister surfaces and V 2 H 6 Raman modes, we concluded that V 2 H 6 defects coalesce to blisters or are absorbed by already existing blisters due to Oswald ripening. Hence, within the Smart-Cut scenario the V 2 H 6 defect complexes are identified as the precursors of blisters. For the Soft-Cut scenario the mechanism of blister formation is a little different. Within the Soft-Cut scenario (100)-oriented platelets, which are created during H-plasma exposure at T 400 C, can be identified as the prevailing precursors of blisters, but V 2 H 6 defects are also involved in blister formation. V 2 H 6 is trapped by precursors of blisters (platelets) or already existing blisters; and V 2 H 6 defects can also coalesce to form new blisters. * Published in: "High Purity Silicon IX", Editors: C. L. Claeys, P. Stallhofer, R. Falster, M. Watanabe, ECS Transactions 3 (4), 417 (2006) ** "Soft-Cut Project", funded by Deutsche Forschungsgemeinschaft (DFG), Project No. Jo787/7-1, Jo787/7-3

15 Investigation of the Impact of Hydrogen Plasma Treatments on Zinc Oxide: Evidence for Hydrogen Induced Nano-Void Formation * R. Job ZnO is a wide band gap semiconductor ( 3.3 ev at 300 K), which provides a wide range of promissing electronic, optical and mechanical applications. In particular, ZnO has a large exciton binding energy ( 60 mev), which makes the fabrication of low-threshold excitonic laser diodes possible. Furthermore, sintered ZnO is a well developed ceramic material for varistor devices, and last but not least an important material for sensor applications. The performance of ZnO based devices is highly dependent on defects and micro-structural disorder within the material. Recently, the interaction of hydrogen with ZnO became an important research focus, since hydrogen is suspected to generate shallow donor states in ZnO. H-plasma treatments are an important process tool for the hydrogen incorporation into semiconductor substrates. To investigate the impact of H-plasma on ZnO samples, H-plasma exposures at substrate temperatures between 250 C and 500 C were carried out. The H-plasma treated ZnO samples were analyzed by µ-raman spectroscopy (µrs). It turned out that this analytical tool provides a good basic insight about hydrogen related damage and the formation of nano-voids in ZnO. H-plasma: 1 h, 500 C H-plasma: 1 h, 500 C H-plasma: 1 h, 450 C 1 h, 450 C H-plasma: 1 h, 400 C I Raman (a.u.) H-plasma: 1 h, 350 C H-plasma: 1 h, 300 C I Raman (a.u.) 1 h, 400 C 1 h, 350 C H-plasma: 1 h, 250 C 1 h, 300 C as sintered 1 h, 250 C a) Raman Shift (cm -1 ) b) Raman Shift (cm -1 ) Fig. 1: a) µrs spectra of ZnO; a) spectra of as sintered and hydrogen plasma treated ZnO samples (substrate temperatures between 250 C and 500 C); b) the Raman line at about 4150 cm -1 can be attributed to H 2 molecules, which are trapped within nano-voids. The impact of the H-plasma exposure on ZnO can be noticed by the Raman modes at 540 cm -1 and at 574 cm -1 (Fig. 1a). The former one can be related to the introduction of nonspecified defects, which is quite effective at lower substrate temperatures up to 350 C. Applying higher substrate temperatures up to 500 C the intensity of the 540 cm -1 mode is reduced due to self annealing but does not vanish completely. The dependence of the Raman mode intensity at 574 cm -1 on the applied substrate temperature reveals a V O defect formation upon H-plasma exposure. The V O defects can be regarded as feasible seeds for nanovoids, which trap H 2 molecules; i.e. V O defects coalesce to nano-voids. The appearance of such nano-voids was verified be the observation of the H 2 Raman mode at 4150 cm -1 (Fig. 1b). Since H 2 molecules are trapped within the nano-voids, the appearance of the H 2 Raman signal acts as fingerprint for the nano-void formation. * Published in: "Zinc Oxide and Related Materials", Editors: J. Christen, C. Jagadish, D.C. Look, T. Yao, MRS Symposium Proceedings Series, Vol. 957, 0957-K10-40 (2006)

16 0,01 1E-3 0,46 0,48 0,50 0,52 0,54 0,56 0,58 0,60 0,50 0,52 0,54 0,56 0,58 0,60 1E-11 1E-12 DOUBLE LAYER ANTIREFLECTION COATING OF HETEROJUNCTION SOLAR CELLS M.L.D. Scherff, S. Schwertheim, W.R. Fahrner The deposition conditions and layer composition of transparent conductive oxides (TCO) can strongly influence the solar cell performance of a-si:h/c-si heterojunction solar cells. A general finding is that although the conductivity and the transparency of the sputtered ITO increase with increasing temperature, the solar cell shows an increasing series resistance, R S. The influence of an intermediate zink oxide (ZnO) layer between the a-si:h emitter and the indium tin oxide (ITO) layer is investigated and compared with solar cells which contain just a single ITO layer. For both pn-structures, a-si:h(n)/c-si(p) and a-si:h(p)/c-si(n), the open circuit voltage (V OC ), the fill factor (FF) and the efficiency (η) are improved for solar cells with an intermediate ZnO layer. Illumiation dependent J SC -V OC measurements (Fig. 1a) are performed to analyze the origin of the low FF observed in several samples. This method is used when R S and the diode ideally factor n id are voltage dependent. current density [ma/cm 2 ] 0,01 1E-3 1E-4 1E-5 ITO RT ITO 135 C ITO 200 C ITO 260 C ZnO-ITO RT ZnO-ITO 165 C ZnO-ITO 200 C ZnO-ITO 260 C 0,1 0,2 0,3 0,4 0,5 0,6 voltage [V] current density [ma/cm 2 ] 1E-5 1E-6 1E-7 1E-8 1E-9 1E-10 1E-11 1E-12 ITO RT ITO 135 C ITO 165 C ITO 200 C ITO 260 C ZnO-ITO RT ZnO-ITO 165 C ZnO-ITO 200 C ZnO-ITO 260 C 0,1 0,2 0,3 0,4 0,5 0,6 voltage [V] Fig. 1: a) Measured Suns-V OC values and b) calcu-lated J 0 (V) values of a-si:h(n)/c-si(p) solar cells. The measured illumination dependent J SC (φ) and V OC (φ) data are used to calculate the voltage dependent saturation current density J 0 (V) (Fig. 1b). With these data the resistance free IV curves (IV Rs=0 ) and the IV curves with R S = 0 and n id = 1 (IV n=1 ) are determined as described in [1] (fig. 2). As long as the measured IV curve does not coincide with both constructed ones, losses due to high R S or n id 1 occur. c-si(n) / a-si:h(p) / ZnO-ITO(RT) 0,0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0 0,016 current density [ma/cm 2 ] 1 0,1 0,01 1E-3 1E-4 1E-5 1E-6 1E-7 1E-8 ITO RT ITO 135 C ITO 200 C ITO 260 C ZnO-ITO RT ZnO-ITO 165 C ZnO-ITO 200 C ZnO ITO 260 C 0,0 0,2 0,4 0,6 0,8 1,0 voltage [V] reverse current ZnO-ITO ITO forward current RT 260 C Fig. 3: The dark IV curves of a-si:h(n)/c-si(p) solar cells present decreasing J Forw and J Rev with increasing T ITO. The influence of the front contact design on the quantum efficiency of the heterojunction cells is discussed. While single ITO samples exhibit only a small variation in IQE in the short wavelength range (350 nm 600 nm) with T ITO up to 200 C, IQE of the ZnO-ITO samples strongly depends on T ITO (Fig. 4). 0,9 current density [ma/cm 2 ] ZnO-ITO RT Rs = 0 Rs = 0, n = ,2 0,3 0,4 0,5 0,6 voltage [V] Fig. 2: Measured (black squares) and constructed IV R=0 (red circuits) and IV n=1 (green triangle) IV- and power curves for ITO/ZnO/a-Si:H(p)/c-Si(n) structure with T ITO = 260 C. The resistance free curve lies between the measured and the ideal one. The double layer structure also affects the dark IV behavior of the solar cells. With increasing deposition temperature of the ITO layer the reverse current densities (J Rev ) decrease. An increase of T ITO from RT to 260 C results in a J Rev reduced by one order of magnitude. J Forw keeps constant independent on T ITO for both ARCs up to 200 C, but for higher T ITO J Forw decreases (fig. 3). For the double layer structures overall higher forward currents are observed. 0,014 0,012 0,010 0,008 0,006 power [W] IQE 0,8 0,7 25 C 260 C ZnO-ITO RT ZnO-ITO 165 C ZnO-ITO 200 C ZnO-ITO 260 C best ITO (165 C) wavelength [nm] Fig. 4: IQE values between 350 nm and 600 nm for ZnO-ITO samples with ITO deposited at 165 C and 200 C show an improvement of the front contact. 6 CONCLUSION The improved solar cell characteristics in the case of an intermediate ZnO layer result from improved emitter properties especially at elevated T ITO. It appears that the ZnO acts as a protection layer for the emitter during the ITO process, possibly caused by reduced hydrogen outdiffusion. Low doped ZnO gives a better contact to a-si:h(n), ITO to a-si:h(p). The optimum temperature is 200 C in both cases. At 260 C both contacts are degraded. [1] P.J. Rostan, U. Rau_, V.X. Nguyen, T. Kirchartz, M.B. Schubert and J.H. Werner, Sol. Energy Mater. Sol. Cells, 90, 1345 (2006). 1,0 0,9 0,8 0,7 0,6 0,5 0,4 0,3 0,2 0,1 0,0

17 INCREASED EFFICIENCIES IN A-SI:H(N) / CZ-SI(P) HETEROJUNCTION SOLAR CELLS DUE TO GRADIENT DOPING BY THERMAL DONORS M.L.D. Scherff, Y. Ma, W. Düngen. T. Mueller, W.R. Fahrner To enhance the internal quantum efficiency (IQE) at long wavelengths a bulk drift field (BDF) is used. This BDF is produced by forming a doping gradient inside the silicon substrate. The graded doping profile is generated by hydrogen enhanced formation of thermal donors (TD) at 400 C in the front subsurface prior to the a-si:h emitter deposition of amorphous silicon / crystalline silicon (a-si:h/c-si) heterojunction solar cells (fig. 1). The effect on the characteristics of a-si:h(n)/cz-si(p) heterojunction solar cells is investigated. this curve. The BDF almost doubles the effective diffusion length (L eff ) from 179 µm of the control sample to 327 µm in the BDF case. IQE 1,0 0,8 0,6 0,4 1 / IQE 2,0 1,5 179 µm 327 µm TD graded doping standard a-si:h (n) c-si (p) a-si:h (n) c-si (n) c-si (p) 0,2 1, absorption length L a [µm] 0, wavelength [nm] Fig. 1: Standard a-si:h(n)/c-si(p) heterojunction solar cell (right) and similar structure with graded doped substrate (left). Figure 2 exhibits the SRP profiles for the control sample and the BDF wafer. The control sample shows a constant resistivity reflecting the constant bulk doping level of N a = 1.5*10 16 cm -3 of the 1 Ωcm material. The doping gradient of the BDF extends to a depth of at least 150 µm with a maximum TD concentration of N TD ~1.7*10 16 cm -3 at the top wafer surface. Due to the N TD level, which is higher than the bulk acceptor doping level, counter doping of the wafer front side appears and a pn-junction is formed at a depth of 23 µm. resistance (Ω) pn-junction H 2 plasma no Plasma depth (µm) Fig. 2: Spreading resistance profiles of the control sample and the wafer with BDF. The black line reflects the nominal resistance of a 1 Ωcm silicon material. The doping gradient extends to a depth of at least 150 µm. A pn-junction is formed at a depth of 23 µm. From spectral response and reflection values the IQE is calculated (fig. 3). Distinct improvements of the IQE in the long wavelength range are observed. The IQE is converted to the effective diffusion length (L eff ) using a plot of the inverse quantum efficiency IQE -1 (L a ) vs. the absorption length L a (λ). L eff is derived from the slope of Fig. 3: Enhancement of the spectral response of silicon solar cells with a BDF compared to the control sample. The insert shows the doubling of the effective diffusion length in solar cells with graded doping compared to the standard devices. As a consequence the short circuit current density (J SC ) increases by ~ 1.4 ma/cm 2. An efficiency improvement from 13.3 % to 13.9 % is observed (tab. 1). The resulting doping gradient causes an electric field in the silicon bulk, which extends from the front surface to the end of the TD gradient. This gradient creates an electric force on the electric carriers, which is directed perpendicular to the wafer front. This force accelerates the minority carrier electrons in direction to the front contact and thus increases the carrier collection efficiency for minority carriers, which are generated in the depth of the bulk or near the back contact. standard drift field J SC [ma/cm 2 ] VOC [mv] FF [%] η [%] R S [Ohm cm 2 ] R p [kohm] Tab. 1: IV-characteristics of solar cells with and without BDF. CONCLUSION Applying a BDF by using hydrogen enhanced TD formation is a reasonable method to create deep drift fields in oxygen rich silicon substrates in time ranges comparable to other diffusion processes but at low temperatures. Remarkably enhanced J SC and η are observed.

18 Hydrogenated amorphous carbon-silicon alloys used as emitters of hetero-junction solar cells Thomas Mueller, Maximilian Scherff, Wolfgang Duengen, W. R. Fahrner More than two decades ago, the use of amorphous carbon-silicon alloys [a-si x C 1 x (n):h y ] stimulated great interest in the field of solar cells. Beside the ability of carbon to change the electronic properties in hydrogenated amorphous silicon, it enhance the optical band gap in a deposited layer and suppresses absorption in the thin film. In this investigation, hetero-junction solar cells were fabricated with specific hydrogenated amorphous carbonsilicon alloys [a-si x C 1 x (n):h y ] used as emitters. The alloys were deposited on p-type silicon (100) substrates via plasma enhanced chemical vapor deposition (PECVD) using a gas mixture of silane (SiH 4 ), phosphine (PH 3 ), methane (CH 4 ) and hydrogen (H 2 ), respectively. The influence of the additional methane and hydrogen in the feed stock during to the decomposition has been investigated by means of I-V measurements, reflection and spectral response efficiency measurements as well as spectroscopic ellipsometry. The optical band gap E G as well as the thickness of the emitter were determined by spectroscopic ellipsometry. It was confirmed that the band gap E G can be tailored by using an appropriate gas mixture during the decomposition. From IQE measurements (in the wavelength range from 350 nm to 550 nm) it could optical bandgap E G [ev] 2,3 2,2 2,1 2,0 1,9 1,8 1,7 1,6 Variation of CH 4 Variation of H 2 Variation of CH 4 with 30% H 2 Variation of H 2 with 15% CH 4 2,3 2,2 2,1 2,0 1,9 1,8 1,7 1,6 Jsc [ma/cm 2 ] 31,0 30,8 30,6 30,4 30,2 30,0 29,8 Jsc - Variation of CH 4 with 30% H 2 Jsc - Variation of H 2 with 15% CH 4 V oc - Variation of CH 4 with 30% H 2 V oc - Variation of H 2 with 15% CH Voc [mv] 1, precursor content [%] 1, precursor content [%] Fig. 1: Deduced optical band gap E G for different sample series Fig. 2: Measured IV characteristics for different sample series be deduced that the addition of hydrogen leads to a slightly better internal quantum efficiency. The addition of carbon, however, decreases the IQE slightly in the short wavelength region. As a consequence, the conversion of photon energy into electric current implicates that it is not performed efficiently by just adding carbon, due to the sp 2 /sp 3 bonding structure of C-C as well as to the decreasing photoelectronic properties, such as photo-conductivity. We conclude from this investigation that by the addition of carbon the optical band gap widens so that the suppression of light absorption in the window layer is reduced, but the density of defect states increases. By the hydrogen dilution this deterioration is minimized. From our I-V measurements, shown in Fig., and the quantum efficiency measurements it is evident that the addition of carbon and hydrogen does contribute to the photo-generated current. It is recommended that by proper valency control of the optical band gap due to the hydrogenated amorphous carbon-silicon alloys [a-si x C 1 x (n):h y ] J sc and V oc should be developed to a maximum efficiency of the solar cells. [1] T. Mueller and M. Scherff and W. Duengen and Y. Ma and W. R. Fahrner Proceedings IEEE 4th World Conference on Photovoltaic Energy Conversion, (2006)

19 µ-raman spectra analysis of hydrogenated amorphous carbon silicon alloys Thomas Mueller, Wolfgang Duengen, Y. Ma, Maximilian Scherff, W. R. Fahrner The fundamental properties of [a-si x C 1 x (n):h y ] thin films deposited using plasma enhanced chemical vacuum deposition (PECVD) with an additional methane source gas are investigated. Optical and electrical measurements have been used to characterize the deposited films regarding their properties of dark conductivity, optical refractive index, and optical band gap. µ-raman spectra analysis derives information about the incorporation of carbon and hydrogen in the deposited thin films. Intensity Raman [a.u.] N 2 10% H 2 20% H 2 30% H 2 40% H 2 10% CH 4 20% CH 4 30% CH 4 40% CH 4 30% H % CH 4 30% H % CH 4 30% H % CH 4 30% H % CH Raman Shift [cm -1 ] Fig. 1: µ-raman spectra deduced from different sample series in the 2k region (a) (b) (c) The adjustment of the resulting band gap is governed by the relative amount of trigonal, sp 2, and tetrahedral, sp 3, hybridizations. Despite of that, the population of these orbital states can be tailored by thin film deposition parameters, such as the precursor material and the method of deposition. µ-raman spectra analysis in the 2k region have shown, that additional methane as carbon source in the precursor material broadens the band at 2000 to 2100 cm 1, which can be attributed to the formation of SiH 2 species. Compared to additional hydrogen without methane, only SiH related bonds are visible, indicating the typical a- Si spectra at 2000 cm 1. Addition of hydrogen and methane shows an enhanced intensity and enlarged broad band from 2000 to 2150 cm 1, indicating trihydride SiH 3 bonds and the incorporation of C in the form of CH 3 groups. Raman spectra at 840 and 890 cm 1 indicate carbon incorporation, which can be attributed to a-sic. Spectra in the high wavenumber range at 2900 cm 1 exhibit peaks related to C-H bonds due to the high methane content in the precursor material. Hetero-junction solar cells, prepared with effective addition of hydrogen and carbon, showed a clear trend of increasing V oc and J sc. Despite of that the conductivity of the resulting thin films decreases with enhanced addition of carbon to the precursor material. The trade off between optimized electrical and optical properties of the emitter layer results in a band gap of approximately 2.0 ev and a layer thickness of 5 to 10 nm. Light in the long-wavelength region passes this layer almost without loss; the optical losses in the blue spectral range lower than 400 nm are typically around 10%. [1] T. Mueller and W. Duengen and Y. Ma and M. Scherff and W. R. Fahrner Proceedings 21st European Photovoltaic Solar Energy Conference and Exhibition, (2006)

20 ÀÝ ÖÓ Ò ØØ Ö Ò Ò ÀÝ ÖÓ Ò ÁÑÔÐ ÒØ Ò ÀÝ ÖÓ Ò ÈÐ Ñ ÌÖ Ø ÞÓ Ö Ð Ë Ð ÓÒ Ïº ĐÙÒ Ò Êº ÂÓ Ò Ïº ʺ ÖÒ Ö Ì ÒÓÖÔÓÖ Ø ÓÒ Ó Ý ÖÓ Ò ÒØÓ Ð ÓÒ Ò ØÙ Û ÐÝ Ò Ø Ð Ø ØÛÓ º ÁØ Û ÐÐ ÒÓÛÒ Ø Ø ÔÐ Ñ Ý ÖÓ Ò Ø ÓÒ Ó Ð ÓÒ Ù Ý ÖÓ Ò ÐÐ ÔÐ Ø Ð Ø Ò Ø Ù ÙÖ Ö ÓÒº ÁÒ Ø Ð Ø Ý ÖÓ Ò ÑÔÐ ÒØ Ø ÓÒ ÒØÓ Ð ÓÒ ÖÓÙ ÜØ Ò Ú ÒØ Ö Ø Ù ØÓ Ô ÒÓÑ ÒÓÒ Ó Ý ÖÓ Ò Ò Ù Ü ÓÐ Ø ÓÒ Ó Ð ÓÒ Û Ò Ù ÓÖ Ð ÓÒ¹ÓÒ¹ Ò ÙÐ ØÓÖ ËÇÁµ Ö Ø ÓÒ ÓÑÑ Ö Ð ÔÔÐ Ò Ø ËÑ Öع ÙØ Ì Å ¹ÔÖÓ º Ê ÒØÐÝ Ú Ö Ð ØÙ Ñ Ø Ø Ö ÙØ ÓÒ ÓÖ Ú Ò Ö ÔÐ Ñ ÒØ Ó Ø ÜÔ Ò Ú Ó ÓÒ ÑÔÐ ÒØ Ø ÓÒº ÔÖÓÑ Ò ÔÔÖÓ ØÓ Ö Ø Ò Ý ÖÓ Ò ØÖ Ô Ð Ý Ö Ò Ù ÕÙ ÒØ ÒØÖÓ Ù Ý ÖÓ Ò Ý ÔÐ Ñ Ý ÖÓ Ò Ø ÓÒ ØÓ Ö Ø Ð Ø Ö Ò ÓÖ Ð Ý Ö Ü ÓÐ Ø ÓÒº Ì ÔÐ Ø Ð Ø Ò Ð Ø Ö ÓÖÑ Ø ÓÒ Ý ÖÓ Ò ØÖ ÔÔ Ò Ò ÚÓÐÙØ ÓÒ Ó Ý¹ ÖÓ Ò ÓÒ Ò Ø ØØ ÑÔØ Ö Ø ÐÐ ÙÒ ØØÐ º ÁÒ ÓÙÖ ÜÔ Ö Ñ ÒØ Ø ÑÔ Ø Ó Ú Ö ÓÙ Ý ÖÓ Ò Ó ÑÔÐ ÒØ Ò Ù ÕÙ ÒØ ÔÐ Ñ Ý ÖÓ Ò Ø Ð ÓÒ Û Ö Ò ÒÚ Ø Ø Ý Ñ ÖÓ¹Ê Ñ Ò Ô ØÖÓ ÓÔÝ ¹Ê˵ ØÓ Ð Ø Ø Ý ÖÓ Ò ØÖ ÔÔ Ò Ò ØÝÔ Ó Ð ÓÒ¹ Ý ÖÓ Ò ÓÒ Ò º Ì ÑÔÐ ÒØ Ø ÓÒ Û Ô Ö ÓÖÑ Û Ø Ò ÓÒ Ñ ÙÖÖ ÒØ Ó ½¼ Ø ÖÓÓÑ Ø ÑÔ Ö ØÙÖ º Ì ÔÐ Ñ ØÙÔ ÓÒ Ø Ó ÓÐÐÓÛ Ø Ó ÔÖÓÚ Ò Ò ØÝ ÔÐ Ñ ÙÔ ØÓ ½¼ ½½ Ð ØÖÓÒ»Ñ µ Ò Ö Ó Ö ÕÙ ÒÝ Ê µ Ò Ö ØÓÖ Û Û Ø ØÓ ½ ¼ Ï Ø Ó ¹ ¼¼ Î Ò Ö ÕÙ ÒÝ ½ º ÅÀÞº Ì ÓÛ Û Ø ØÓ ¼ Ñ Ö ÓÒ Ò ¼ Ñ Ý ÖÓ Ò Ø ÔÖ ÙÖ Ó ½¼ Öº Ì Ê Ñ Ò Ô ØÖ Ó Ý ÖÓ Ò ÑÔÐ ÒØ Ò Ù ÕÙ ÒØ ÔÐ Ñ Ý ÖÓ Ò Ø Ð ÓÒ Ü Ø Ê Ñ Ò Ð Ò Ø ¾½¼¼ Ñ ½ Ò ¾½¾ Ñ ½ ØØÖ ÙØ ØÓ Ë ¹À ½ Ò Ë ¹À ¾ ÓÒ Ø ÒØ ÖÒ Ð ÙÖ º Ì ÙÖ ÐÓÒ ØÓ ÚÓ º º ÔÐ Ø Ð Ø ÓÖ Ð Ø Ö º ÐÓ ÚÓ Ö ÐÐ Ý Ý ÖÓ Ò ÑÓÐ ÙÐ Ó ÖÚ Ý Ê Ñ Ò Ø Ø ½ ½ Ñ ½ º Ê Ñ Ò Ò Ð Ø ¾½ Ñ ½ Ò ¾¾ ½ Ò Ò ØÓ Ú Òݹ Ý ÖÓ Ò Î ¾ À Ò ÎÀ Ö Ô Ø Ú Ðݺ ÅÓÐ ÙÐ Ö Ý ÖÓ Ò ØÖ ÔÔ Ò ÑÙÐØ Ú Ò ÓÑÔÐ Ü Ò ÒØ Ø Ê Ñ Ò Ø Ó ¾¾ Ñ ½ º ÓÒ Ö Ò Ø «Ö ÒØ Ð ÓÒ¹ Ý ÖÓ Ò ÓÒ Ò Ô Ò Ò Ó Ø ÑÔÐ ÒØ Ø ÓÒ Ó Ø Ø ÑÔ Ö ØÙÖ Ó ¾ ¼ Æ ÓÒ Ø Ø¹ Ö Ø ÖØ Ò Ñ Ø Ö Ð ÔÐ Ø Ð Ø Ò Ó ÖÚ º Ï Ø ÒÖ Ò ÑÔÐ ÒØ Ø ÓÒ Ó Ò ÑÓÙÒØ Ó Ø Ø Ú Òݹ Ý ÖÓ Ò ÓÒ ¹ Ò ÔÔ Ö Û Ð Ë ¹À ÓÒ Ò Ó ÒØ ÖÒ Ð ÙÖ Ö º Ø Ö ÔÐ Ñ ÜÔÓ ÙÖ Ø Ö Ù ØÖ Ø Ø ÑÔ Ö ØÙÖ Ø ÐÓÛ ÓÒ ÑÔÐ ÒØ Ø ÓÒ ÙÔ ØÓ ½ ½¼ ½ À»Ñ ¾ µ À ¾ Ò Ë ¹À Ü Ð Ò Ö ÒÓØ Ø Ø Ð Ø Ö ÑÔÐ ÒØ Ø ÓÒ Ó ÙÔ ØÓ ½¼ ½ À»Ñ ¾ µ ÔÐ Ø Ð Ø¹Ð Ø Ö Ó ÖÚ Ð Û Ö Ú Òݹ Ý ÖÓ Ò Ð Ø ÓÒÐÝ Ü Ø Ø Ø ÑÔÐ ÒØ Ø ÓÒ Ó ¾ ½¼ ½ À»Ñ ¾ µº ÁØ Ð Ó Ö Ñ Ö Ð Ø Ø Ð Ø Ö ÓÙÖ Ò Ø Ð Ø º Ë Ò Ø Ê Ñ Ò ÒØ Ò ØÝ Ó ÖØ Ò Ö ÕÙ ÒÝ ÓÖÖ Ð Ø Û Ø Ø ÓÒ ÒØÖ Ø ÓÒ Ø Ö Ð Ò ÓÖÖ Ø ÓÒµ Ø ÓÒ ÒØÖ Ø ÓÒ Ó ÑÓÐ ÙÐ Ö Ý ÖÓ Ò Ò Ö Ò Ô Ò Ò Ó Ø ÑÔÐ ÒØ Ø ÓÒ Ó Ò Ø Ù ØÖ Ø Ø ÑÔ Ö ØÙÖ Ó Ø ÔÐ Ñ ÜÔÓ ÙÖ º Ì Ê Ñ Ò Ô ÒØ Ò Ø Ó ÑÓÐ ÙÐ Ö Ý ÖÓ Ò Ø ÖÓÙÒ ½ Ñ ½ ½¼ Ñ ½ µ Ò Ë ¹À ÓÒ Ó ÒØ ÖÒ Ð ÙÖ Ø ¾½¼¼ Ñ ½ Ñ ½ µ Ö Ò ÐÝÞ º Ì Ø À ¾ ÒØ Ò Ø Ó ÖÚ Ø Ý ÖÓ Ò ÑÔÐ ÒØ Ø ÓÒ Ó Ó ½ ½¼ ½ À»Ñ ¾ ÓÖ Ø ÑÔ Ö ØÙÖ ÓÚ ¼ Æ º Ñ Ð Ö Ò Ö Ó ÔÖ ÒØ ÓÖ Ø Ë À ÒØ Ò Ø º ÖÓÑ Ø Ø Ò ÓÒÐÙ Ø Ø Ð Ý ÖÓ Ò ØÖ ÔÔ Ø ÐÓÛ Ó ÑÔÐ ÒØ Ø ÓÒ Ò Ø Ø Ð Ý Öº Ø Ó ÓÒ ÑÔÐ ÒØ Ø ÓÒ ÓÚ ½¼ ½ À»Ñ ¾ ÑÓÖ Ý ÖÓ Ò ÓÖ Ý Ø Ú ÒÝ Ø ÔÖ Ú ÒØ Ò Ø ÓÖÑ Ø ÓÒ Ó Ý ÖÓ Ò ÐÐ ÔÐ Ø Ð Ø º Ì Ö Ê Ñ Ò ÒØ Ò Ø Ø Ö Ø ÑÔ Ö ØÙÖ Ò ÜÔÐ Ò Ý Ø Ö Ý ÖÓ Ò «Ù Ú Øݺ ½ ½ Ϻ ĐÙÒ Ò Êº ÂÓ Ò Ïº ʺ ÖÒ Ö Äº Ǻ à ÐÐ Ö Âº ̺ ÀÓÖ ØÑ ÒÒ Ò Àº Ð Ö Ð ØÖÓ Ñ Ð ËÓ ØÝ ÌÖ Ò Ø ÓÒ ½ ¹½ ¾¼¼ µ

21 Anisotropic Etching of Silicon in a RIE Set up H.Winkelmann, M. Scherff, W.R. Fahrner RIE (Reactive Ion Etching) plasma etching is a commonly used process step in the semiconductor technology. Anisotropic structures in the range of nanometers to micrometers can be etched. Thus this procedure is an important process in microelectronics. The resulting etch pits can exhibit totally different features (underetching, in- and outward sloped sidewalls, surface roughnesses). These features depend on the process parameters. Etch pits with vertical sidewalls and smooth surfaces should be produced while any surface damage by ion irradiation was to be avoided. This investigation was simultaneously done for oxide covered crystalline and amorphous silicon. In a different run block cast silicon wafers were examined for which the saw damage should be removed so that a planar surface should be obtained. For this purpose the ranges of the applicable parameters (pressure, power, gas fluxes) were determined. An aluminum shadow mask was applied for the etching. Following a process matrix of ref. [1] the samples were characterized by an optical and scanning electron microscope, a profilometer, and an ellipsometer. The different material exhibited different etching behavior. The etch rates of amorphous and SiO 2 covered silicon were so small that they were without technical interest to us. The parameter variation applied on Czochralski silicon (c-si) and multicrystalline block cast silicon (mc-si) led to the intended results. A 50:15:10 gas mixture of SF 6, O 2 and CHF 3 was used, the chamber pressure was 20 mt and the HF power was 300 W. The frequency, the temperature and the etching time were MHz, 20 C and 10 min, respectively, for all the experiments. Fig. 1: c-si sample after RIE, parametes as in text Fig. 2: As in fig. 1, profilometer scan from 11 to 5 o clock The figures 1-3 were obtained from a c-si sample. They show steep sidewalls and relatively smooth surfaces. Figs. 4 and 5 exhibit a mc-si wafer prior to and after RIE treatment. Fig. 3: SEM photograph of the c-si sample of fig. 1 Fig. 5: SEM photograph of a planarized mc-si wafer The results and conclusions of the above work can be applied in various fields of the semiconductor technology. Two examples are surface defect removal after production of thermal donors and planarization of surfaces after sawing. The present work shows that the chosen variant of the dry etching Fig. 4: SEM photograph of the saw damage of a mc-si wafer procedure is more precise and straightforward than the usual wet chemical etching. [1] R. Legtenberg, H. Jansen, M. de Boer and M. Elwenspoek, Anisotropic Reactive Ion Etching of Silicon Using SF 6 /O 2 /CHF 3 Gas Mixtures, (J. Electrochem. Soc., Vol. 142, No. 6, June 1995)

22 10 10cm 2 HIT Solar Cells Contacted with Lead-Free Electrical Conductive Adhesives to Solar Cell Interconnectors Stefan Schwertheim, Maximilian Scherff, Yue Ma, Thomas Mueller, Wolfgang R. Fahrner University of Hagen, Chair of Electronic Devices, Germany Phone: , stefan.schwertheim@fernuni-hagen.de Amorphous silicon (a-si:h) / crystalline silicon (c-si) heterojunction solar cells have attracted large interest in the last decade due to their high efficiencies (21.5% in a lab scale, more than 17% in mass production) [1]. Depending on the thermal properties of the thin film layers and their interfaces, process temperatures higher than 200 C can be detrimental to HIT solar cell performance [2]. These temperatures are encountered in solar module fabrication during soldering the interconnector tabs (IT) to the front and back side of the solar cells. To prepare good solder joints temperatures higher than the melting temperature are required. July 2006 the European Union demands to use lead-free solders in electronics. These kinds of solders will exhibit melting temperatures about 30 C higher. Therefore soldering of the solar cell ITs with lead-free solders is a critical process step for a-si:h/c-si solar cells. An alternative to lead-free solders, which are already in use in homojunction solar module production, are electrically high conductive lead-free adhesives which are well known in microelectronics applications. In the present study we have investigated the electrical properties of 10 10cm 2 HIT solar cells with evaporated Cr/Ag grids. The contacts were produced with low temperature lead-free electrically conductive adhesives to standard solar cell ITs. Compared to solder joints conductive adhesives have the advantage of lower contact formation temperature (80 C 150 C) with reasonable low contact resistances (< 1mΩcm 2 ) [Fig. 1a]. Interconnector tabs glued directly on ITO [Fig. 1b] are not long term stable if the applied pressure during manufacturing is too low. Depending on the applied pressure the contact resistance increases by one to four orders of magnitude after 50 temperature cycles ( 40 C / 85 C) respectively [Fig. 1a]. (a) (b) Fig. 1: (a) Contact resistances between ITs and ITO from TLM measurements for Elecolit glue prepared at low pressure. Values are shown for samples immediately after curing, after one week in air and after 50 temperature cycles ( 40 C / 85 C). (b) Sketch of the HIT solar cell front side contacted with an IT. The IT itself works as bus bar. [1] M. Taguchi et al. (2005) An Approach for the higher Efficiency in the HIT Cells 31th IEEE PVSC, 2005, pp.866. [2] M. Scherff et. al. (2005) Dependency of Heterojunction Solar Cell Efficiency on the Indium Tin Oxide Deposition Parameters 20th Europ. PVSECE

23 Development of a Method for the Estimation of a Automobile Velocity Based on Kalman Filtering * P. Goerg 1, 2, 3, J. Fuchs 3, R. Job 1 1 Department of Mathematics and Computer Science, University of Hagen, Germany 2 Altek GmbH, München, Germany, 3 Audi AG, Ingolstadt, Germany In frame of a master thesis a method for the estimation of a reference velocity for automobiles is developed, which bases on a Kalman filter approach. A Kalman filter is recursive filter which efficiently estimates the state of dynamic systems from a series of incomplete and noisy measurements. Input values for the filter are given by the four velocities of the wheels and the signal of a longitudinal acceleration sensor. The basic problem for the determination of the reference velocity is the different behavior of the wheels during acceleration and braking. As an example one might think about an acceleration event on partially glaciated roads (Fig. 1) or about abrupt changes of tire-road friction coefficients due to a change of the paving properties. Employing longitudinal acceleration sensors one has to take also into account that beside the desired acceleration provoked by the driver additional acceleration signals can occur due to a downhill or uphill ride, which can add some error to the calculations. Furthermore, the sensor might also hand out some slightly corrupted signals due to manufacturing/fitting tolerances. Therefore, first the sensor has to be controlled to assure a proper operation and then it has to be matched. However, the sensor signal can not be used until the actual ascent of the car ride is estimated to correct the sensor signal with this regard. Kalman Filter velocity control signal (meas.) Fig. 1: Slippery road conditions ( Audi AG) Fig. 2: Kalman filtering To allow for the above mentioned issues, using a longitudinal acceleration sensor the nominal value of the acceleration as desired by the driver, which is composed of the engine respective the brake torques, has to be compared with the effectively achieved acceleration as deduced from the velocities of the wheels. If both acceleration values are coming close, the actual sensor value can be synchronized to the average of both values. If both acceleration values are significantly different one can conclude that an ascent of the road figures up. Furthermore, for the controlling of the sensor the suitable physical gradients and boundaries have to be checked permanently. One has to underline that the whole procedure is only valid for a stable ride. For the completion of the Kalman filtering process, the velocities of the wheels have to be converted to the longitudinal direction and axis and filtered to minimize the impact of the driving dynamics, and therefore, to enhance the availability of the system. The signal of the longitudinal acceleration sensor is pre-filtered by a PT1 filter to reduce white noise. A typical result of an estimated reference velocity based on a discrete Kalman filter is shown in Fig. 2. One can state that the Kalman filter delivers much better results than other analytical methods. However, the results can still be optimized. To find the most appropriate type of Kalman filter for the described problem, the results delivered by a discrete Kalman filter have to be compared with the approaches of extended and unscented Kalman filters. * Master thesis of Patrick Goerg (to be finished in 2007)

24 Development of a Sensor System Integrated into an Active Medical Implant for the Therapy of Hydrocephalus * M. Gier 1, 2, O. Scholz 2, R. Job 1 1 Department of Mathematics and Computer Science, University of Hagen, Germany 2 Fraunhofer Institute for Biomedical Engineering (FhG-IBMT), St. Ingbert, Germany The central nervous system (CNS) consists of the brain and the spinal cord which are embedded in a liquor (i.e. liquor cerebrospinalis). Due to dysplasia, disease or accident an imbalance between the production and resorption rate of the liquor, and therefore, a liquor excess can occur, which results in an elevated brain pressure and the clinical picture of the hydrocephalus. To avoid brain damage the elevated brain pressure has to be reduced. Usually this is done by an implantation of a so called shunt. This shunt is a catheter which connects the cerebral ventricle in the brain containing the liquor with the abdominal cavity. Via the shunt excessive liquor can be abducted from the brain, and therefore, a pressure balance can be achieved. The drain of the liquor is controlled by a valve, which is integrated into the shunt system. Fig. 1: Prototype of the 'Brain-Shunt' developed at FhG-IBMT Commercially available shunt devices are passive mechanical valves with the disadvantage that after the implantation their aperture can be only slightly modified, if ever. Those shunts can not react on physiologic pressure fluctuations, which also appear in the sane body. Within the project Brain Shunt, FhG-IBMT, in collaboration with Miethke GmbH and Aesculap AG, developed an active electromechanical implant for the therapy of hydrocephalus. In contradiction to the conventional passive shunt devices this system should react on the physiological condition of the patient. Furthermore, the system should provide the possibility of a non-invasive intervention, where the physician can change the system parameters via a simple telemetric interface. The first prototype of the "Brain Shunt" system (Fig. 1) operates time controlled so that the valve aperture is adapted in dependence on the time of the day to allow for the necessities of the patient. In frame of a master thesis this system was extended by a functional concept with regard to the integration of various sensor components, which provide important physiological data of the patient as well as function control options. I. e. significant physiological parameters like the brain pressure and liquor flux are detected for surveillance and for an individually adjusted attendance of the patient as well as a function control of the valve to realize malfunction of the system in advance. In a final state the "Brain Shunt" system is expected to enhance the amenity of the hydrocephalus patients significantly. * Master thesis of Markus Gier (successfully finished in Dec. 2006)

25 Systemic Analysis of Electronic Flight Instrument Systems in Small Aircrafts * S. Höchtl, R. Job In Frame of a master thesis devoted to advanced aircraft guiding systems, especially to Electronic Flight Instrument Systems (EFIS), a systemic investigation was carried out with regard to evaluate the applicability especially for small aircrafts. EFIS is also denominated as glass cockpit system due to the monitors displaying the flight information data. EFIS instrumentation is used in airliners since quite a lot of years. However, for small aircrafts the utilization of EFIS just starts. Small aircrafts, which are equipped with EFIS are called Technically Advanced Aircrafts (TAA) or more lapidary aircrafts with a glass cockpit. EFIS is more efficient in use of space and it is cheaper to maintain than the traditional clocklike instrumentation displaying flight data in older cockpits. EFIS is also supposed to be easier understandable for the pilot. Nevertheless, the launch of EFIS in a small aircraft comes along with some challenges especially for the amateur pilot, and in particular for those who got their basic pilot training not on a TAA, since the appearance of the cockpit in a TAA is significantly changed by EFIS as compared to the traditional cockpit equipped with traditional clocklike instrumentation (see Fig. 1). Fig. 1: Technical Advanced Aircraft (TAA) with EFIS instrumentation The onboard instrumentation, which is imperative for a safe flight, is analysed with regard to the basic technical performance and system properties. In general EFIS consists of two display sets, delivering essential information about the aircraft/flight status: (i) EADI (Electronic Attitude Display Indicator) or PFD (Primary Flight Display) concerning mainly information about the 'vertical' status, i. e. attitude, yaw, altitude, flight direction, etc., and (ii) EHSI (Electronic Horizontal Status Indicator) or ND (Navigation Display) concerning the 'horizontal' status, i. e. heading, ground track, distance measuring equipment (DME), etc. With regard to a systemic approach for the evaluation of new technologies for guiding systems, which are recently introduced into small aircrafts, one of the authors (S. Höchtl), who is an ambitious amateur pilot, carried out a transition training for pilots, who received their pilot license with classical clocklike instrumentation. During the transition training it was proofed how the change from clocklike instruments to EFIS affects the burden of the pilot during a flight. The description of the various phases of the transition training implying the changeover towards correct EFIS handling is one of the key points of the investigation and underlines its systemic pretension, which does not only consider the technical aspects of modern guiding systems in aircrafts but also the problems arising at the human-machine-interface. * Master thesis of Stefan Höchtl (to be finished in 2007)

26 Tactile Sensors for Robot Guided Intraoperative Neurosurgery * R. Stroop 1, 2, R. Westphal 3, D. Oliva Uribe 2, T. Hemsel 2, F. Wahl 3, R. Job 1 1 Department of Mathematics and Computer Science, University of Hagen, Germany 2 Heinz Nixdorf Institute, Mechatronics and Dynamics, University of Paderborn, Germany 3 Institute for Robotics and Process Control, Technical University of Braunschweig, Germany Introduction: Intraoperative robotic applications are in the majority of cases manipulator based systems. These are in terms of level of automation passively acting systems providing the surgeon a precise trajectory for tissue manipulation. Indications for actively intervening robots are usually limited to hard tissue robotic applications like bone processing as in craniotomy or orthopaedic surgery or for interventions in geometrically fixed situs as stereotactic procedures. Introduction of robotics for more complex soft tissue surgery with the necessity of real time reaction on tissue deformation and dislocation are considerably limited not only to intelligent decision making algorithms or to heavy-handed dexterity as an efferent motor problem but also by lacking an efficient sensory system. Sensory perception has to capture the visual aspect of the surgical situs as well as further physicochemical properties of the tissue under investigation. The development of a tool for intraoperative tactile perception will be a highly challenging task to perform and improve the safety during surgery. Fig. 1: Tactile sensor mounted on a robot Fig. 2: Sensor response on elasticity variations of test specimen Material and Method: A resonant vibrating piezoelectric bimorph sensor was implemented for a high precision tissue elasticity determination. The bimorph sensor was attached via a force torque sensor to a 6 axis, articulated industrial robot (TX40, Stäubli). The force torque sensor was interconnected to ensure a defined, non-harming bimorph sensor load. Differences in brain tissue elasticity were simulated using gelatine-gel resp. PVC-gel preparations with minute concentration gradients. Results: Using the robot guided bimorph sensor system (Fig. 1), subtle differences in tissue phantoms elasticity could clearly be differentiated by evaluating frequency shift and amplitude changes against human ability to discriminate elasticity changes (Fig. 2). Discussion: Providing tactile sensory perception systems is not yet a sufficient but a crucial precondition for soft tissue surgical robotic applications. More sophisticated sensor systems have to be established to even determine dignity of tumorous alterated tissue. * Master thesis of Dr. med. Ralf Stroop (to be finished in 2007)

27 Thermal donors formed in N-type HR magnetic Czochralski silicon exposed to hydrogen plasma Y. L. Huang N-type high-resistivity (HR) magnetic Czochralski (MCz) silicon wafers are hydrogenated via a direct hydrogen plasma exposure at 270 C. N-type HR (>500 Ω cm) MCz silicon wafers with an interstitial oxygen concentration of cm -3 are used. After the hydrogenation, isochronal annealing (20 min) is carried out up to 450 o C. Capacitancevoltage (C-V) measurement shows an introduction of high dose excess electrons into the silicon wafers due to the hydrogen plasma exposure, inferring the formation of shallow donors. The diffusion-like concentration profiles of free electrons suggest that these shallow donors are hydrogen related shallow donors (STDH). Deep level transient spectroscopy (DLTS) measurement reveals that oxygen thermal donors (OTD) were already after an annealing (20min) at C. The STDH centers are the dominant donors in as hydrogenated samples. The concentration of the OTD centers becomes comparable to that of the STDH centers as the annealing temperature is 400 C or higher. Fig. 1 plots the concentrations of the electrons from the different shallow-level centers normalized to the total concentrations of excess electron as a function of the annealing temperature. Here, we assume that each donor level contributes one electron to the excess electrons, i.e., STDH, OTD 0/+ and OTD +/++. It can be seen more than 90% excess electrons are contributed by the STDH centers at the annealing temperatures of 350 and 380 o C. In comparison, the contributions from the STDH centers to excess electrons are reduced to 50% and 31% for the cases of 400 and 450 o C, respectively. These results are in line with the information given by the C-T measurement [1]. The evident changes in the doping level due to the formation of the STDH and OTD centers indicate that serious care should be taken for the use of HR MCz substrates for application in high energy physics experiments. Relative Concentration OTD 0/+ OTD +/++ STDH Annealing Temperature ( o C) Fig. 1 Evolution of the concentrations of various shallow donor species in hydrogenated and annealed high resistivity magnetic Cz-grown silicon. The values are deduced from the results of C-V and DLTS measurements. [1] Y. L. Huang, E. Simoen, C. Claeys, J.M. Rafí, P. Clauws, R. Job, and W. R. Fahrner, Appl. Phys. Lett. 89, (2006). 1

28 Chair of Data Processing Technology (Datenverarbeitungstechnik) Prof. Dr.-Ing. Bernd J. Krämer Universitätsstr Hagen Phone: Fax: WWW: Staff Michael Averstegge Thorsten Blum Dr. Timo Borst Dr.-Ing. Klaus Gotthardt Susanne Heyer Marc Jelitto Dr. Han Peng Dr. Georg Ströhlein -371 Alexander Stuckenholz Xia Wang Volker Winkler Dr.-Ing. Fan Yang Renate Zielinski -371 Research Areas Distributed Systems and Software Engineering service-oriented computing component-based software engineering, in particular architectural design and compositional reasoning, software evolution and software maintenance. E-Learning Tools and Environments: component-based content authoring learning with mobile devices reusable learning objects group formation and recommendation software

29 Cross-Production of Learning Content for PDAs and Mobile Phones Bernd J. Krämer The EU funded project m-learn2 1 set out to investigate the potential and limitations of mobile devices for different types of mobile learning applications. From the students perspective, m-learning services have the potential to increase their options in time and location independent learning and allow them to access content and services on demand. In principle, all educational services, including provision of learning content, support and management of learning activities, maintenance of personal data, tutoring and consulting, communication and collaboration, performance of exercises and tests, access to virtual and physical laboratories (wherever reasonable) and access to the university library, can be provided for students on the move. The idea underlying this research is not to migrate all learning activities completely on mobile devices but expand the learning space to the mobile world, focusing particularly on the support of spontaneous short learning phases. Figure 1: Flash animation on a PDA In the first year of the project we had developed Histobrick, a game-like application on descriptive statistics for use on mobile phones. This application won the innovation price of the MKWI 06 conference [1]. The project as a whole has been nominated for a European Quality Award. A particular idea underlying the design of our second year application was to use a novel learning content management system, FuXML, and experimentally extend its cross-media publishing features from paper and laptop to mobile devices. A decent study of different m-learning projects revealed a range of alternatives to make learning content fit the constraints of mobile devices, which can be divided into serverside, intermediate and client-side adaptation techniques. For the sake of cost-effectiveness we chose a server-side solution relying on a pre-processed presentation version of our sample course for use one PDAs and mobile phone running an HTMLcapable browser (see Fig. 1). As we also wanted to experiment with special viewer software and explore the possibility of offline reading as well, we chose Palm OS as the target platform and used Plucker to produce a Palm pdb file. Plucker is a free offline Web and e-book viewer for PalmOS. As newer PDAs are equipped with flash cards, downloading a whole course on a Palm or other type of PDA and then accessing the materials offline may be a cheaper and more convenient way of m-learning. [1] Georg Ströhlein: Mobile learning using mobiles: hype or tripe? Multikonferenz Wirtschaftsinformatik 2006 (MKWI 06) 1 This research was financially supported by the EU under contract number 2003-IRL/03/B/F/PP

30 Service Component Architecture - Old Wine in New Bottles? Bernd J. Krämer Service-oriented computing (SOC) [1] has gained a lot of attention as a novel paradigm for coping with interoperability problems in heterogeneous enterprise application integration. In many ways SOC resembles component-based software engineering (CBSE) in its, which has focused on design methods and implementation tools supporting the construction of large software systems as assemblies of reusable collections of functionality. CBSE is exemplified in popular technologies including J2EE,.Net, Corba and others. But SOC differs from these predecessors not only by putting the concept of (software) service in center but also by proving vendor, platform and technology independent standards for inter-service infrastructure necessary to enable service interaction and communication. Utilizing services as lightweight constructs supports the development of rapid, low-cost distributed applications. The visionary promise of SOC is a world of cooperating services where application components are assembled with little effort into a network of services that can be loosely coupled to create dynamic business processes and agile applications that span organizations and computing platforms. Other than traditional component architectures, service-based applications are inherently dynamic assemblies relying on the fact that SOA includes facilities for service publication, discovery and dynamic binding. However, SOA still lacks the engineering comfort that component technologies provide including component containers providing a range of predefined basic services addressing component lifecycle, concurrency, or transaction issues. Standard techniques to compose services into larger applications include orchestration and choreography. Through orchestration individual services are organized around a central business process that coordinates the invocation of local and remote services dynamically. Choreography does not rely on central coordination but each service involved knows exactly when and with what other service it has to interaction. Late 2005 a first attempt was made to combine the advantages of SOA with those of CBSE through as new model for the construction of SOA-based business applications, called service component architecture (SCA). The SCA assembly model allows both tight coupling of functionally into service components running under the control of a particular organization and loose coupling of cross-organizational service-oriented business solutions. Service implementations are the elementary building blocks of the assembly model. SCA encompasses a growing range of technologies to provide service component implementations including Java, C++, BPEL, SCA can use (Web) services but also other software solutions including legacy systems and it is not tied to a specific runtime environment or communication infrastructure. Rather a range of mature technologies is integrated in the upcoming standard. Like in CBSE, component implementations can be defined recursively as assemblies of other service components. At first glance, SCA looks like a promising marriage of two successful software engineering approaches. As the standardization process is far from complete and only rudimentary support tools exist so far, it is too early, however, for a concluding evaluation. A more detailed examination of the pros and cons of SCA is presented in [2]. [1] F. Cubera, B. J. Krämer, M. P. Papazoglou (Eds.): Service Oriented Computing, Dagstuhl Seminar Proceedings 05462, 2005, [2] B.J. Krämer: Component meets Service - What does the bastard look like? In 2nd International Symposium on Leveraging Applications of Formal Methods, Verification and Validation, IEEE Computer Society Press, 2007

31 Contract Based Testing of Web Services Michael Averstegge As Web services become more prevalent the inherent paradigm of loose coupling causes extra effort in testing. Context and enclosing process are not predictable and therefor not solely testable at design time, e.g. a Web service can be a substitute of any unavailable Web service. Since web services act as black boxes, contracts are the only formal functional specification visible to the client beyond the mere syntactic level offered by WSDL specification. The lack of semantic specification of web services is often stated in literature, Testing distributed web services means testing external services with no realistic chance on introspection or modifying source (e.g. instrumenting) for test purposes. Solely focusing on detailed pre-production testing of web services tends to get an anachronistic reputation. In order to offer reliable web services, one needs formal semantic specification to derive test criterions. To plan testing of web services, focussing on traditional development steps is not sufficient, because discovery, matching and dynamic composition can cause functional and non-functional behavior unforeseen at design time or development time. Therefore the introduced approach additionally uses web service runtime information to increase the quality of test data by approximating them cost-effectivly to the expected use of the web service, owing to the fact that pre-production tests are not sufficiently executed in practice. To overcome the drawbacks of creating a huge set of testing data out of user input data and which has to be reduced afterwards by applying heuristic methods we try to reduce testing data by deriving contracts from input data first and deriving test data from them afterwards. We use the approach of in combination with learning strategies for black-box systems. Only in the field of contract specification, a manual intervention is necessary. Postconditions are used as a test oracle and preconditions are used to automatically derive test cases. Invariants are used in either case. Figure 1: Test layer architecture for contract based black-box testing. Further functionality can be added by third parties to the test system using the visitor design pattern. Visitors implemented from third parties can be hooked into the test system, since the test system dynamically loads visitors from the file system at runtime. The loaded visitor is called back depending on the visitor s parent class.

32 CampusContent - a research project on its way to a community building repository for digital learning resources Timo Borst In its second year of existence, the research work in CampusContent has lead to significant results in developing a competence centre for e-learning and research information. First, we manifested our conceptual work regarding elementary concepts like learning object, learning objective or didactic scenario by means of formal modelling in UML. This provided a basis for the layout of the database and the actions to be taken on different object types. Moreover, our understanding of the whole conceptual model was sharpened. To prepare the prototype implementation of a content repository, we specified the user interface of the application by means of screen designs and HTML demonstrators. At that time it was already clear that we develop a repository in the sense of a database application with content management, user management and editors for educational purposes. The user interface also provides functionalities to create learning objectives and learning objects, the latter consisting of digital learning resources (or, in our words, information objects) embedded in learning activities which relate to certain types of resources. Besides searching, browsing, uploading, composing and editing of learning resources we collected digital learning material both from the FernUniversität and other universities and institutions in order to build up an initial library. The material was deliberately taken from different disciplines just to prove the overall concept of reusability and to attract users from different subject areas. These efforts finally resulted in the completion of a first software prototype of CampusContent as a platform for sharing and reusing digital learning resources. The prototype has been evaluated and tested by different users, so it could be significantly enhanced. Also progress on the isolation and metadata indexing of so-called information objects has been made, with a more strict distinction between the context-independent parts of a learning object and its context-related information, which is not part of the object itself but of its metadata. With regard to the latter, the effort of metadata input has been minimized by using contextual information and predefined values. Finally, first community functions have been implemented just to convey a slight feeling of membership to other users learning objects. Figure 2: Sreenshot of CampusContent prototype application

33 Learning Objective Editor in CampusContent Marc Jelitto Basore [1] developed an objective builder for formulating educational objectives based on the taxonomy of Bloom [2]. In CampusContent we decided to use the revised version of Bloom s taxonomy by Anderson and Krathwohl [3]. It reflects the operating experience of 50 years of using the taxonomy. More interesting for us was, besides to the cognitive process dimension, the usage of a knowledge dimension. We conjecture that this will make the development and use of a learning objective editor more difficult, but we hope also more helpful for users and for our research about using learning objects. The result of the first developments can be seen in [4] and in figure 1. Figure 1: Screenshots of the Learning Objective Editor in the Prototyp of CampusContent. Figure 2 shows the table of the Anderson and Krathwohl taxonomy which helps users to analyse a group of learning objectives. This will be implemented in the next prototype version. Figure 2: The table of the dimensions of Anderson and Krathwohl. [1] Basore, James (2006). CyberCampus Objectives Builder. Online: (Last Visit: ). [2] Bloom, Benjamin S., Engelhart, M.D., Furst, E.J., et al. (1956). Taxonomy of educational objectives: Handbook I: Cognitive domain. New York, David McKay. [3] Anderson, Lorin W. und Krathwohl, David R., Hg. (2001). A Taxonomy for Learning, Teaching, and Assessing. A Revision of Bloom s Taxonomy of Educational Objectives. New York, Addison- Wesley. [4] Jelitto, Marc (2006): Learning Objective Editor - A First View into the Prototype. Online: (Last Visit: ).

34 Exploit Community Oriented Social Collaboration in Information Retrieval Peng Han With the continuous emergence and increasing popularity of social software like BBS, Blog, Wiki and Podcasting, the concept of community has played a more important role in the design and implementation of the next generation of web applications. In our recent work [1], we investigated the significance of community information to the searching of academic resources by performing an empirical study on the community structure of the LON-CAPA system, one of the leading online learning content management and assessment systems. We used a subset of the learning resources usage information kept in LON-CAPA as our experimental data. It refers to 253,972 learning resources developed by 539 authors and their usage in 2275 courses composed by 2120 course instructors. From the statistic analysis on the data set we found that the popularity of authors does not have direct relations with how many resources they contributed. By defining the Normalized Contribution Popularity (NCP) as denoted by equation 1, which represents the average usage frequencies of each resource the author contributed, we found that of the top 10 most contributing authors, only two appear in this top 10 most popular author list. Furthermore, resources contributed by the top 10 highest NCP authors account for more than 40% of the total resource usage in the system, while they add up to only approximately 6% of the whole number of available resources. NCP = Num Used Resources Num Authored Resources (1) Another interesting observation from the LON-CAPA usage data is that the top 10 authors with the highest NCP also have a high Co-Contribution Association with each other, which means the resources contributed by them are frequently used together in the same course. It somehow proved the existence of community within the information retrieval processes. In order to exploit social collaboration within learning resource reuse communities, we conducted an study on how user added descriptions associated with web resources can be used to enhance the existing resource search paradigms [2]. Based on a study on the data collected from the del.icio.us website, we found that user added keywords on the resources can be categorized into search keywords and exploration keywords. The search keywords can help the search engine to locate corresponding resources more accurately and the exploration keywords are helpful for the exploration process as they reflects the personalized search context and information need of an individual user. Based on that, we then designed and implemented a collaborative tagging enhanced exploratory search interface. When a user starts a search scenario, we first use Google API to get the initial result set. Then the URLs in the initial result set are passed to the Del.ici.ous website s URL check function to extract the corresponding tags which are displayed as clickable hints behind the associated URL. When the user clicks on the search keyword, we expand his original query by adding the clicked keyword into it. For each click on an exploration keyword we then directly return the tagged URL in the del.icio.us system to the user. [1] P. Han, B. Krämer and G. Kortemeyer, Latent Social Networks and Communities: An Empirical Analysis of LON-CAPA User Data, submitted to the 3rd international conference on communities and technologies, Michigan, USA, [2] P. Han, Z. M. Wang, Z. Y. Li, B. Krämer and F. Yang, Substitution or Complement: An Empirical Analysis on the Impact of Collaborative Tagging on Web Search. In Proceeding of the 2006 IEEE/WIC/ACM International Conference on Web Intelligence, IEEE Computer Society, 2006, pages , Hong Kong, China, December 18-21, 2006

35 Synthesis of component-based software configurations Alexander Stuckenholz Component-based software systems still evolve. Because of bug-fixes, performance enhancements or changes in requirements, components frequently have to be replaced by newer or other versions. In order to avoid system breakdowns, system compatibility has to be ensured. To accomplish system compatibility, it is usually checked, if the new components are subtypes of the currently used ones and can therefore be introduced into the system configuration without risk. In [1] we proposed a different solution for the problem in which system configurations are automatically generated with the intention to integrate otherwise incompatible components. The problem of finding a well-configured and against different objective functions optimal system configuration is comparable to the backpacking-problem. From a huge repository a small subset of components has to be selected in order to form an optimal configuration. We proposed two search strategies with different qualities: On the one hand the boolean search which is guaranteed to find the global optimal solution but with inadequate performance. On the other hand a heuristic method which combines greedy- with local-search. The heuristic method is not able to generate the same quality of results but is much faster. Figure 1: Behavior of the analyzed synthesis methods during search Figure 1 shows the behavior of the different strategies during search. Where the objective function reaches the optimal value during heuristic search without indirection, the boolean search has to evaluate the complete state space. Both strategies have been implemented in our prototype-tool Componentor which is now able to support system-administrators in order to ensure safe component updates. During evaluation in an open-source project, we additionally collected experiences for the practical usage of the tool and the integration into the development process. [1] A. Stuckenholz, A. Osterloh: Safe component updates, In: Proceedings of the Fifth International Conference on Generative Programming and Component Engineering (GPCE 06), ACM Press, 2006, To appear.

36 A QoS-aware Selection Model for Semantic Web Services Xia Wang Service-oriented computing (SoC) is an emerging paradigm for distributed computing that is changing the way software applications are designed, architected, delivered and consumed. Automating the selection of service is, however, a crucial challenge, in particular, as more services are becoming available and match a user s functional requirements. In such situation the selection of the best service among all candidates depends on a combined evaluation of the services qualities. To address this problem, we propose a quality of service- (QoS-) based selection technique. A QoS ontology and its vocabulary are specified in terms of the Web Services Modeling Ontology (WSMO). This ontology is then used for annotating service descriptions with QoS data. We then define quality attributes and their respective measurements along with a QoS selection model, which provides the basis for a fair and dynamic selection mechanism, using an optimum normalization algorithm. The scenario of QoS-based service selection assumes that a user specifies his requirements (including non-functional, functional, and quality properties) for the expected service. These requirements are mapped into a requirements profile, denoted as s R = (NF R, F R, Q R, C R ). Figure 1: QoS-based selection of services Similarly, our scenario assumes that the properties of published services can be represented in an advertisement structure of the form s A = (NF A, F A, Q A, C A ). Our selection method employs a series of filters that first tries to match non-functional (basically the service name and service category) and functional features (including inputs, outputs, preconditions and effects) of s R with corresponding properties of any s A available in selected repositories or peer-to-peer environments. We assume that this filter yields the result set S = {s 1, s 2,..., s m }, m N. The second filter then considers all quality features to select those services in S satisfying the user s requirements best. We assume that Q R = {r 1, r 2,..., r k } expresses the profile of a user s quality requirements, which includes k quality metrics. Similarly, the quality profile of m candidate services in set S is denoted by Q S = {Q A1, Q A2,..., Q Am }, where Q Ai = {q i1, q i2,..., q ij }, i, j N, i.e., the advertisement of service s i provides j quality metrics. Our matchmaking algorithm defined in [1] compares the QoS requirements Q R pairwise with the quality profiles Q A of candidates services s A S, as illustrated in Fig. 1. Our evaluation with a hypothetical phone service proved that the algorithm for evaluating multiple quality metrics in combination is fair and simple. [1] Xia Wang, Tomas Vitvar, Mick Kerriga, Ioan Toma: Synthetical Evaluation of Multiple Qualities for Service Selection. In Proc. of the 4th International Conference on Service Oriented Computing, , Springer-Verlag, 2006.

37 Enhancing Collaborative Learning based on Personalized Recommendation Among Communities Fan Yang Collaborative learning provides a good solution to help the e-learners to communicate with each other and share learning resources or experiences. However most of current approaches are still missing one important issue an e-learning system should address. That is, collaborative learning may enhance learning effect only when proper learners are grouped together. Since learners always have different status and background, it makes difficulty to decide which e- learners are similar and suitable to learn collaboratively [1]. Based on the former research on learning communities self-organization [2], we propose an innovative collaborative learning platform connecting similar e-learners. The main purpose of this system is to help e-learners to collaboratively recommend useful and interesting materials with other similar e-learners based on their learning backgrounds, preferences, purposes and other meaningful attributes [3]. This platform is implemented based on the intelligent agent platform JADE, which generates a collaborative learning agent (LA) for each e-learner [4-5]. The system consists of three functional modules: user information collection and modeling module, user community structure exploitation module and community based recommendation module. We chose 160 volunteers and divided them into two groups, A and B, where each group had 80 learners.we chose the course Graduate English as the test subject and designed a sequence of tests. Learners of group A learned in a normal way. While each learner of group B was supported by a LA to monitor the performance of test which is considered as the learning status of relative concept. Thus the system ran the community-organization algorithm and grouped learners with similar learning status into the same community. At the same time, the learners of B were encouraged to recommend the learning resource through the recommendation platform along with the tests. Each learner also could receive, read and evaluate the recommendation materials from other learners. At the end of the semester we compared the learning curves of each group and found that the learners in group B have an ascending trend with higher acceleration than for the learners in group A. Experiments results have shown that this system really helps e-learners to enhance their learning effect [3]. [1] F. Yang. Analysis, Design and Implementation of Personalized Recommendation Algorithms Supporting Self-organized Communities. VDI-Verlag, Düsseldorf, [2] F. Yang, B. Krämer, Z.M. Wang. An Novel E-Learner Community Construction Algorithm in Point View of Learning Interests. WSEAS Transactions on Computers, v 5, n 12, pages , December (Ei Core Journal; Ei No ) [3] F. Yang, B. Krämer, P. Han. Enhancing Collaborative Learning based on Personalized Recommendation among Communities. Advanced Technology for Learning (ATL) International Journal, Special Issue on Technology for Collaborative Learning, v 3, n 4, pages , [4] F. Yang, B. Krämer, P. Han. Book Chapter: Common-Interest Based Self-organizing E-Learner Communities. B.J. Krämer and W.A. Halang (Eds.), Contributions to Ubiquitous Computing, Springer, Studies in Computational Intelligence, [5] F. Yang, B. Krämer, Z.M. Wang, P. Han. An Improved E-learner Community Construction Algorithm Based on Learning Interest Feature Vectors. In Proceedings of the 6th WSEAS International Conference on APPLIED INFORMATICS AND COMMUNICATIONS (AIC 06), pages , Crete Island, Greece, August 18-20, 2006.

38 Juniorprofessur für Softwaretechnik Prof. Dr. Jens Krinke Universitätsstr. 27 (PRG) Hagen Tel.: Fax: Staff: Ext. Philipp Bouillon 4314 Stefan Winkler 4314

39 A Dynamic Program Analysis Framework in Eclipse 1 Philipp Bouillon Software systems become increasingly complex. As programmers need to understand them in order to extend them further, we developed a program analysis framework in Eclipse. This framework is able to trace programs dynamically at runtime and record the method calls. These execution traces can be analyzed with any algorithm according to the users needs. We have implemented a first analysis algorithm which can find classes which are used in different ways, indicating a need for refactoring. On top of our framework, we created a plug-in for Eclipse, which dynamically analyzes existing plug-ins and records their usage of the Eclipse API. After the usage has been recorded, our tool interprets the recordings, or execution traces, and presents its findings to the user. Thus, a programmer can easily see how a specific part of the Eclipse API is used by already existing plug-ins and it is therefore possible to determine how method calls are executed while the plug-in is running. The advantage of this technique is that a programmer can see the implicit relationships between method calls: Rules like close has to be called after open can be discovered. Those rules can be determined by other tools as well, but what makes our tool different is that it analyzes the run-time behavior of the program in question and does not, like other tools, rely on statical inspection of the source code. More specifically, our tool first starts the program in question and logs every method call into a file. Only after the program has ended, the trace file is analyzed and the results are presented to the user. After an execution trace for a specific program run has been recorded, the user can either choose to execute the program a few more times to produce more trace data for several program runs, or she can choose to let our plug-in analyze the trace data. As a first useful analysis, we have implemented an algorithm to check if a class is used in more than one way. Often, two or more disjoint ways of using a class are an indicator for a possible refactoring. For example, consider a badly designed class, which is either used to read from a file or used to write to a file. It can be argued that splitting this class into a reader class and a writer class is sensible. To determine if a class is used in two or more ways, we have to collect all execution events of a single object. These execution events can be grouped to form an execution trace of this specific object. By collecting all execution events of all objects belonging to the same class, we get an idea of the usage of this class. In order to determine if the execution traces of two objects are similar, we use the Jaro-Winkler String similarity function. This metric compares strings by considering fixed elements in both strings and comparing the similarity of the rest. Applied to execution traces, every execution event in a trace is considered as a character and the so formed strings are then compared using the Jaro-Winkler metric. If the score is high enough (we experimented with values 0.85), the classes represented by those strings are considered to be used in the same way. Accordingly, all pairs of objects with a Jaro-Winkler score less than our threshold are considered to use the class in different ways. After the Jaro-Winkler score has been determined, we split the object execution traces into several clusters. Each cluster contains the traces, which are similar to each other. We have applied this analysis scheme to JHotDraw and Eclipse and our results indicate that both applications contain badly designed classes. We are in the process of incorporating more existing tools into the analysis framework and can then provide a tool, which can run overnight and provide the developer with detailed statistics on what is going on in a piece of software. 1

40 How information flows between requirement artifacts Stefan Winkler Requirements engineering is still an area of software engineering in which theory and practice greatly differ. An important aspect when dealing with requirements is documentation. According to our own perception of industrial software projects we see that requirements are often scattered between different artifacts. This leads to inconsistencies and consequently to higher costs and lower software quality. To address these issues we have set up a research project to improve the information flow between the requirement artifacts. We have conducted an online questionnaire with the goal to support our perceptions and conclusions, to justify our research intention, and to get input for the research. The key questions and the outcome can be summarized as follows: 1. Which requirement artifacts are created and used during the requirements phase in software development? Meeting notes and structured textual documents (like the requirements specification) play a central role as requirements artifacts. Additionally, requirements lists are a form of documentation which is used as often as meeting notes. 2. How many different types of artifacts are created and used? Most projects use several different types of artifacts to document requirements. We found an average and a median of seven artifact types. 3. How big is the problem of inconsistencies between different artifacts? Inconsistencies between artifacts are often introduced without being noticed when an artifact is updated. But as it turned out, most participants were aware of the introduction of inconsistencies and even consciously accepted it. It seems that the effort to keep different artifacts consistent is regarded as too high compared with the impact of inconsistencies. 4. How does information flow between the artifacts? This topic is illustrated by the figure below. The nodes represent the artifacts. They are sized according to the number of participants who named the artifact. The directed edges represent the flow of information. If an edge is drawn from one node to another, this means that information from the artifact represented by the first node was used to create the artifact represented by the second node. The stronger the edge is drawn, the more projects use this path of information. To make the graph more readable, only edges named by seven or more participants are shown. Consequently isolated nodes are omitted. Scenarios Requirements Lists Rough Concept Dictionary Use Cases Meeting Notes GUI Layouts SRS Technical Spec. Functional Spec. These results will now be used to drive the further research and exploration in the field of an integration of the requirements artifacts on the basis of models.

41

42 Chair of Electrical Power Engineering Prof. Dr.-Ing. D. Hackstein Elektrische Energietechnik FernUniversität Universitätsstr. 27/PRG D Hagen Phone: Fax: homepage: Staff members Phone SECRETARY Marianne Lehmann SCIENTIFIC STAFF Dipl.-Ing. Henning Bohn TECHNICAL STAFF Wolfgang Köhler

43 Closed loop control of newly developed six phase permanent magnet synchronous machines for a mobile application T. Treichl Today the power trains of tractors in agricultural applications consist of a combustion engine followed by a hybrid realised hydraulic mechanic gearbox. The most disadvantage of this realisation is that the generated hydraulic power cannot be used to supply vehicle internal components or vehicle external implements. A few years ago the idea for a new concept was born: beside the mechanical connection from the combustion engine to the gearbox it should also be possible to transport the maximum power that can be produced by the combustion engine via an open electric 540V dc circuit to further inputs of a new gearbox as well as to vehicle internal components and external implements. All the hydraulic paths at the power train completely had to disappear (cf. figure 1: concept of the new electric mechanic power train). To succeed with this research and development project new electrical drive systems with a nominal power of 130kW were needed because electrical drive systems that can be found in industry applications are too large to fit in a tractor. Each component of such electrical Figure 1: electric-mechanic power train drive systems could be considered and optimized by itself to obtain minimum overall frame sizes. This procedure led to a new fluid stator cooling concept at the electrical machines and also led to a fluid cooling concept for the power semiconductors of the converters. This procedure failed when considering the large dc capacitors for themselves that are needed to reduce dc voltage ripple caused by switching power semiconductors. The overall inspection of electrical drive systems illustrated that the size of the dc capacitors under previous conditions can only usefully be reduced if the electrical machines are build six phase (same frame size as three phase machines at same power) and the converters also are expanded to six phases (twice as much semiconductors at half size again result in same frame size and same power as three phase converters). Further an optimized six phase pulse width modulation pattern resp. an optimized six phase space vector modulation has to be used to achieve a size and cost reduction of C dc,6~ = 0.42 C dc,3~ (cf. figure 2: three phase pulse width modulation pattern and resulting dc current in contrast to a six phase pulse width modulation pattern). Figure 2: pwm patterns A new mathematical framework for six phase components had to be developed for the representation of their physical context. The realisation and the verification of three six phase electrical drive systems illustrated that these systems are well adapted for the usage in the introduced mobile application (cf. figure 3: six phase electrical machine, six phase converter and reduced size dc capacitor). Further information is given by the EET institute of the FernUniversität Hagen and can also be found in [1]. Figure 3: electrical drive system [1] T. Treichl, German Phd. thesis submission: Regelung von sechssträngigen permanenterregten Synchronmaschinen für den mobilen Anwendungsfall, FernUniversität Hagen, 2005

44 Requirements of hydraulic and thermal power plants under the influence of regenerative energies R. Kühne, D. Hackstein In the last years a rapid increase of the installed electrical power gained by wind energy power stations is to be registered particularly in Europe and in Germany. This development is an outcome of the politically promoted utilization of wind energy. Its goal is to reduce the fraction of fossil primary energy carriers leading to a reduction of CO 2 - emissions. This will essentially attribute to the underlying similar objective of the European Union and will presumably lead to an increase of the installed electrical power out of wind energy plants from about 30 GW nowadays to at least 40 GW in The rapid development of the wind power stations leads however to changed, not yet quantified requirements for power station reserve and dynamics. This is due to their strong fluctuation, badly predictable output and very limited controllability. Those new requirements are technical and economic boundary conditions according to access times as second or minute reserve and provisions of payment and/or continuous reserve (Figure 1). Change by conventional power plants Change by wind power stations Load prognosis error Variations in load Power station losses Wind prognosis error ordered dependently production Fluctuations of wind power Second reserve Minute reserve Continuous reserve Figure 1 Reasons for increased power plants reserve and dynamic of power plants There within the next years a further development of wind power will take place requiring more power station reserve. Since this is connected with substantial costs, it is important to know as exactly as possible the requirements of power station dynamics and reserve which are necessary for the adherence to a desired reliability of supply. For this present practice of power station employment using system services was analyzed regarding possible changes for high wind power supply. In addition the technical characteristics of the existing conventional power station pool and its interaction with wind energy plants and the load over the transmission network was examined and gave approaches for the operation within the defined system range.

45

46 Chair of Computer Engineering, esp. Real Time Systems Prof. Dr. Dr. Wolfgang A. Halang Faculty of Electrical and Computer Engineering FernUniversität in Hagen Hagen, Germany Tel: Fax: Internet: World Wide Web: Staff members Name Extension Internet Dipl.-Ing. Jutta Düring 4529 Dr. Harald Fock 4003 Shourong Lu, MSc 1731 Dipl.-Ing. Martin Skambraks 4527 Min Xiong, MSc 1732 Wei Zhang, MSc 1192 Dr. Yi Zhao 4526 Research interests Dependable platforms for real time computing Safety related embedded control systems emphasising licensable software Engineering of distributed and embedded control systems based on UML Malware protection by means of hardware Semantic web

47 Ubiquitous Computing for Control W.A. Halang, B.J. Krämer, R. Sanz, R. Babuška and H. Roth A new approach in control engineering, viz., Information Processing for Action, is presented in [1], in which control, computers, communication and cognition play equal rôles in addressing real-life problems from very small-scale devices to very large-scale industrial processes and non-technical applications. Thus, the paradigm Computers for Control is shifting towards the paradigm Computers, Communication and Cognition for Control providing an integrated perspective on the rôle computers play in control systems and control plays in computer systems. This change is mainly due to new developments in computers and knowledge management, and the rapidly emerging field of telecommunications providing a number of possible applications in control. Control engineers need to master computer and software technologies to be able to build the systems of the future, and software engineers need to use control concepts to master the ever-increasing complexity of computing systems. Essential design issues for computer control systems Software-intensive control systems are becoming progressively more complex, prevalent in numerous industrial sectors, and decisive for human safety and for reliability. As there is an accelerated growth of demands for functionality and dependability of such systems, our intellectual and engineering abilities are being challenged to come up with practical solutions to the problems faced in the design and development of complex realtime and safety-related control systems. Even though software testing is an essential part of manufacturing embedded systems, and significant research efforts have been devoted to it, the current state of the practice in system and software verification and validation (V&V) leaves much to be desired. In many industries, V&V is not receiving the attention deserved, and other important activities such as testing functional and non-functional requirements, modeling large software systems, conformance, acceptance and qualification testing, or measuring the effectiveness of different V&V approaches have not yet been incorporated into the V&V processes employed. Therefore, in [2], a holistic view is taken on mobile and distributed computing systems. All system layers ranging from hardware over vertical and horizontal infrastructure services and novel middleware solutions to various types of application software are touched to the end of mastering their complexity. References: [1] W.A. Halang, R. Sanz, R. Babuška and H. Roth: Information and Communication Technology Embraces Control, Annual Reviews in Control 30, 31 40, 2006 [2] B.J. Krämer and W.A. Halang (Eds.): Contributions to Ubiquitous Computing, Series Studies in Computational Intelligence, Vol. 42, Berlin-Heidelberg-New York: Springer- Verlag 2006

48 Clock Synchronisation in Nodes on a Ring Bus T. Erdner, W.A. Halang, K.C. Chan and J.K. Ng As wired interconnections of stationary distributed (sub-) systems often dual ring buses are employed, as they provide for fault tolerance by redundancy and for predictable data transmission delays. On such a dual ring bus every telegram is transmitted in parallel on both data channels (channel 1 and channel 2), but in opposite directions, as shown in the figure. The transmission delay of a telegram sent from the bus master to a slave depends on only two configuration constants, viz., one holding for the entire bus (T CCT ), and the other one being a function of the slave s position on it ( T SiS ). master GPS-receiver clock-module master-control channel 1 channel 2 master i/o-register slave slave clock-module slave-control i/o-register T CCT time period clock-module slave-control i/o-register slave clock-module slave-control i/o-register T SiS T C1Mi T C2Mi slave clock-module slave-control i/o-register Clock synchronisation in nodes on a dual ring bus Given these constants, each slave can synchronise [1] its local clock with the clock of the master node every time it receives a telegram, provided the latter contains the information when it was sent. Temperature effects can easily be compensated for by measuring the difference between the arrival times of a telegram s versions travelling on both rings. Hence, for a distributed system interconnected by a fieldbus, only one receiver for official time signals as broadcast by, for instance, the Global Positioning System (GPS) is required in its master unit to provide for accurate timing synchronous with the official and legal Universal Time Co-ordinated (UTC). Reference: [1] Th. Erdner, W.A. Halang, K.-Y. Ng and K.-C. Chan: Synchronisation der lokalen Uhren an ringförmigen Übertragungsmedien angeschlossener Einheiten. German patent No , 9 February 2006

49 Splines and Fractional Differentiation W.A. Halang Polynomial spline functions have become a widely used tool in numerical analysis. They are also very well suited for interpolation purposes. Owing to the number of interpolation and continuity conditions on one hand and the number of polynomial parameters on the other, odd degree splines usually interpolate at the partition points, and even degree splines half-way in between. For interpolating polynomial splines of odd degree and their generalisations remarkable properties have been shown: the orthogonality relation in the space L 2, the minimum norm properties, and the two integral relations. These properties have also been applied to derive error estimations for the approximation of functions by interpolating odd degree polynomial with respect to arbitrary partitions. In case these splines fulfill appropriate boundary conditions, the validity of the second integral relation follows from the fact, that the splines are identical with a polynomial in each partition subinterval. By the theory of spline systems, the proof of the other properties was reduced to the one of the orthogonality relation. Let now k 1, I R a compact interval, X a subspace of C k 1 (I) and P : X P (X) the linear and idempotent spline operator of degree 2k with respect to the partition π of I, i.e., the spline P f interpolates a given function f at the centre points of the partition s subintervals. In [1], its is considered whether there is a linear differentiation operator U of order m N, such that the functions in P (X) reduce in the partition s partial intervals to the elements of Ker U U and that Im UP Im U(Id P ) holds with respect to the inner product of L 2 (I). The answer obtained to this question is negative, which means that properties analogous to the ones holding for odd degree splines cannot be shown. This, however, raises the further question whether there may be other polynomial splines of even degree defined by different interpolation conditions, which have the sought characteristics. This is indeed the case for those splines interpolating the local integrals of given functions with respect to the partition subintervals and, thus, at some interior, a priori not determined, points of these subintervals. Applying the theory of spline systems, minimum norm properties are derived, analogous to the ones which are known for odd degree splines already for a long time. To circumvent insurmountable problems in applying calculus of fractional degree, this is not done directly, but a corresponding generalised problem within the framework of the functional analytic theory of spline systems is treated. To this end, the notion of a derived spline system is introduced and its properties are investigated. As corollaries, all properties sought for even degree polynomial splines with respect to differentiation operators of fractional degrees are obtained. Reference: [1] W.A. Halang: Minimum Norm Properties of Even Degree Polynomial Splines with Respect to Fractional Differentiation Operators, Proc. 2nd IFAC Workshop on Fractional Differentiation and its Application, S. Samko and B. Vinagre (Eds.), pp , Porto, 2006

50 Disjunctive Dataflow Computing in Automation D. Verber, W.A. Halang and B. Lent When distributed control applications are executed in normal operation mode, the only varying factor is the data. Thus, the dataflow principle seems to be a natural choice to organise such executions. By monitoring the data, one can achieve shorter response times and better throughput. However, in a traditional dataflow computer system, a function block is only executed when new data are available on all inputs. According to the disjunctive dataflow paradigm, howver, a function block is executed when there is a change in any of its input data values. Otherwise, the block s inputs and outputs are preserved. Thus, the overall processing load is greatly reduced, and shorter response times are attained. This idea was brought further by modifying the triggering condition: a change of data is detected only if there is a significant change in data values. Conceptual diagram of the firmware logic to support disjunctive dataflow execution In [1], execution control was migrated into firmware to reduce response times even further. For this, an experimental hardware platform was employed which consists of several processing and I/O nodes interconnected by a dual TTCAN bus. Control applications are divided into several function blocks that may be allocated to different processing units. In order to support distributed architectures, a distributed shared memory model was introduced that allows for communication between different function blocks. A function block seeking to communicate with other ones writes its message into a shared memory cell. This value is then distributed throughout the system and read by the recipient block(s). A middleware layer is responsible to perform this communication transparently and in a fault-tolerant fashion. To utilise the disjunctive dataflow model, the shared memory model was modified to perform the triggering of function blocks based on this principle. Using the original design, each node in the system is associated with an FPGA chip that allows different parts of the operating system to be executed in firmware. Reference: [1] D. Verber, W.A. Halang and B. Lent: Firmware Support for Disjunctive Dataflow Driven Distributed Control Applications, Proc. IFAC Workshop on Programmable Devices and Embedded Systems, Z. Bradác et al. (Eds.), pp , Brno, 2006

51 Quality of Service of Real-Time Systems R. Gumzej, M. Colnarič and W.A. Halang Computers work in the real-time operating mode when the programs for the processing of data arriving from the outside are permanently ready, so that their results will be available within predetermined periods of time. This gives rise to four fundamental qualitative requirements real time-computing systems have to meet: timely and simultaneous operation even under extreme load conditions, and predictable and dependable behaviour. Quantitative performance measures for real-time systems, on the other hand, did not receive much attention, yet. The few existing methods to evaluate the Quality of Service were reviewed and assessed. It turned out that they are highly inappropriate, and that a more thorough consideration of the very nature of real-time systems is necessary, which leads to meaningful performance criteria. Compute speed (Whetstones/second) Workstation performance Typical real time application requirements I/O Bandwidth (MB/second) Requirements space of real-time applications Deterministic response (Interrupts/second) Surprising to those who used to confound real-time with fast computing, much more important than quantitative criteria (worst case response and error handling times, MTBF, MTDF, MTTF, MTTR, capacity reserves, costs) are qualitative binary criteria (timeliness, bounded execution times, safety licensability, functional correctness, deterministic behaviour, permanent readiness, meeting physical constraints, static design, deadlock prevention), i.e., a system may either fulfil them or not, as well as qualitative gradual criteria (safety, dependability, behavioural predictabiliy, simplicity, reliability, robustness, fault tolerance, graceful degradation, portability, flexibility). Based on these criteria, in [1] the critical points in the design and development process of real-time systems were determined, it was investigated where the criteria should be applied and/or checked and how, software development and certification standards where assessed to which extent they support the criteria, and it was proposed how to integrate the criteria into them. Reference: [1] R. Gumzej, M. Colnarič and W.A. Halang: Introducing QoS Parameters into Design and Development of Real-time Systems. Proc. 29th IEEE Intl. Convention MIPRO 2006, S. Golubic, B. Mikac und V. Hudek (Eds.), pp , Rijeka: MIPRO HU 2006

52 Designing Control Applications with the Language PEARL E. Steinle and W.A. Halang For frequent power outages caused by extreme weather and ailing power lines, an emergency power generator was set up on a farm to secure milking and feeding. As other agricultural equipment, the generator is driven by a tractor s motor via a cardan shaft. The control of this generator has to prevent or, at least, strongly reduce frequency variations, and to automatically switch on and off sensitive electronics when the normal frequency is reached again. In this respect problems are mainly caused by bigger load changes, which lead to intolerable frequency deviations, and by the deadtime as result of the motor s inertia. If the desired frequency is exceeded, fast regulation is possible as the motor s inertia can be exploited; if the desired frequency is under-run, however, it must be regulated slowly, because the motor s intertia must be overcome as well. A very robust control algorithm was designed [1], since the driving machine employed is generally not known, and since there are many unpredictable loads and, hence, current and frequency variations for which only very few dependecies are known. After a strong rise or decline, the frequency can be led to its set-point in an asymptotic way, only, as otherwise the regulation and, thus, the frequency may be in danger of oscillations. Tractor with stationarily operated generator Digital control of the tractor s motor was implemented in the real-time programming language PEARL running under the operating system RTOS-UH on a single-board computer. In addition, a PLC was installed to switch all motors, heatings and the light, and to prevent machines to be switched on at the same time. The PEARL computer notifies the PLC via 4 static control lines on the frequency status, and orders it to take measures supporting the regulation as well as to carry out switching-off sequences if required for safety reasons in case of frequency variations. If the regulation cannot reach the set-point for whatever reason, the PLC switches off all machines after a specified time. Reference: [1] E. Steinle and W.A. Halang: Ein PEARL-Einplatinenrechner hält Einzug auf dem Bauernhof, in Echtzeitsysteme im Alltag, P. Holleczek and B. Vogel-Heuser (Eds.), pp , Reihe Informatik aktuell, Berlin-Heidelberg-New York: Springer-Verlag 2006

53 Hardware-Interpretation of Graphically Formulated Safety-related Real-time Programs J. Baron von Stackelberg A dataflow-oriented graphical programming language to express safety-related automation applications in terms of standardised function blocks was devised in [1]. Selecting blocks and setting dataflow lines between them ares governed by a small number of strict rules. The two-dimensionally displayed programs directly serve as machine code for a specially designed interpreter implemented completely in hardware. Thus, there is no need to translate the human-readable programs into a machine-readable form eliminating a possible source of errors. Further, as a precondition for safety licensing, the very code processed is easy to understand. X 101 Wert A 101 X 101 X 101 A Ab1 For16 A 101 B Ab1 B 101 C 101 X X Ab1 Ab1 WertDatei Wert0050,0050 A graphically formulated program The hardware interpreter uses a conventional microprocessor to convert data into different formats and operate on them. The algorithms implementing these functions were written in the microprocessor s assembley language. They were rigorously verified to be employed for safety-related applications, which also included their worst-case execution time behaviour. The interpreter provides functions to read, debug, and process programs. Before switching into the execution mode, the interpreter checks syntactic correctness of the application programs. Hence, it is not possible to run incomplete code, or to access the operating system or other application programs in a hazardous way. On the other hand, the interpreter enables for a wide range of memory accesses, supports several kinds of standardised communication ports, allows to enhance the function block set, and can be run in a multiprocessor mode. Reference: [1] J. Baron von Stackelberg: Hardware-Interpretation graphisch formulierter sicherheitsgerichteter Echtzeitprogramme, Fortschr.-Ber. VDI Reihe 10 Nr Düsseldorf: VDI- Verlag 2006

54 Designing Safety-Related Real-Time Systems with UML S. Lu, W.A. Halang, R. Gumzej and M. Colnarič Embedded real-time control systems are often safety-critical. In order to assure safety, it must be considered right from the conceptual phase of system design. To enable the description of safety-related software, in [1] and [2] it was explored to extend UML by defining a framework of safety mechanisms based on the safety standard IEC to be collected in a UML profile apt for developing safety-critical systems. In [1], the safety mechanisms were constructed by a fault-tolerant structure which is based on well-proven fault-tolerance techniques and FT-CORBA. In [2], a safe structure associated with well proven constructs, e.g., subsets of the real-time programming language PEARL and of Function Block Diagrams according to IEC and IEC 61499, resp., and an executive safety architecture based on concepts taken from the PEARL were constructed. These constructs are ordered in nested sets to fulfill the respective requirements of the four Safety Integrity Levels of IEC Use of the framework enables reasoning about system dependability already at an early stage of system development, and to customise fault-tolerance strategies according to application characteristies. An application architecture A reconfiguration management pattern for UML real-time projects inspired by the Specification PEARL methodology was defined in [3]. It its parameterised by defining the properties of its components as well as by defining how software is mapped to hardware architectures. The pattern itself is a UML representation of a Specification PEARL architectural specification, together with the definition of the initial and alternative configurations as well as of methods of switching between operation scenarios. The goal followed in designing the pattern was to obtain clearly, specified operation scenarios with well-defined transitions among them. In order to achieve safe and timely operation, the pattern provides a safety shell for all scenarios, i.e., enables their deterministic, temporally predictable operation, and transitions between them. References: [1] S. Lu and W.A. Halang: A UML Framework for Safety Mechanisms Based on IEC 61508, in Sicherheit 2006, J. Dittmann (Ed.), pp , Lecture Notes in Informatics, Vol. P-77, Bonn: Gesellschaft für Informatik 2006 [2] S. Lu and W.A. Halang: A UML Profile to Model Safety-Critical Embedded Real-Time Control Systems, in Contributions to Ubiquitous Computing, B.J. Krämer and W.A. Halang (Eds.), pp , Series Studies in Computational Intelligence, Berlin- Heidelberg-New York: Springer-Verlag 2006 [3] R. Gumzej, M. Colnarič und W.A. Halang: Safe and Timely Scenario Switching in UML Real-time Projects, Proc. 9th IEEE Intl. Symp. on Object and Component-oriented Realtime Distributed Computing, S. Lee, U. Brinkschulte and B. Thuraisingham (Eds.), pp , Los Alamitos: IEEE Computer Society Press 2006

55 An Architecture for Hard Real-Time Multitasking Conforming with IEC M. Skambraks and W.A. Halang Most contemporary real-time systems use interrupts to co-ordinate process-driven software execution. This approach of asynchronous programming clearly conflicts with the international safety standard IEC 61508, which strives for design simplicity by restricting the use of interrupts and other complexity-increasing processing methods. The approach of synchronous programming, i.e., a cyclic operating style that always processes program code completely within any cycle, meets the software design guidelines of IEC best, but limits the field of application to simple control tasks. Task Administration Unit (TAU) Begin Output Synchronous begin of Execution Cycles IDs of Task and Execution Block to process Begin Input Application Processing Unit (APU) Input Either task completion notification or ID of task's next Execution Block Task-Execution: - Processing the Execution Block identified by the IDs submitted - In case this block does not complete this task s execution, the block that needs to be executed next is determined Output Execution Phase Task-Administration: - For the task just processed: storing the new ID of task's next Execution Block in Task List - Control of task-states (reading event inputs, checking activation characteristics, inducing state transitions) - Computation of deadlines - Determination of the activated task that needs to be executed in the next cycle, i.e. that has the closest deadline Administration Phase End End The novel operating principle Motivated by the objective to combine the advantages of synchronous and asynchronous programming, a concept of a programmable electronic system was devised [1,2,3], which supports task-oriented real-time execution, but avoids the use of asynchronous interrupts. The application software is organised in tasks but, contrary to conventional task-oriented systems, the tasks are subdivided into a number of blocks which are executed in discrete intervals of constant duration. This execution concept renders special synchronisation mechanisms such as semaphores superfluous, since any task has non-interrupted exclusive access to the processor during an execution interval. As a result, the temporal behaviour as well as the hardware structure are simiplified, and a higher conformity with IEC for systems of highest safety criticality (SIL3/SIL4) is achieved. References: [1] M. Skambraks: A Safety-Licensable PES Architecture for Task-Oriented Real-Time Execution without Asynchronous Interrupts, Düsseldorf: VDI-Verlag 2006 [2] M. Skambraks and W.A. Halang: Task-Oriented Real-Time Execution without Asynchronous Interrupts combined with Runtime State Restoration, Proc. Intl. Multiconf. on Computer Science and Information Technology, pp , Warsaw: Polish Information Processing Society 2006 [3] M. Skambraks and W.A. Halang: Einheit zur Verwaltung von Echtzeitprozessen ohne asynchrone Unterbrechungen. German patent no , 28 Dec. 2006

56 State Restoration at Runtime in a Safety-Related Hard Real-Time Environment M. Skambraks and W.A. Halang In addition to minimise the failure probability of a Programmable Electronic System (PES) by internal redundancy, it is also desirable to provide a redundant configuration of PESs with the capability to re-start a unit at runtime. This requires copying a PES s internal state at runtime, since a re-started unit must equalise its internal state with that of its redundant counterparts before redundant processing can be rejoined. As a result, redundancy attrition due to transient faults is prevented, since failed channels can be brought back on line. The PES architecture proposed in [1,2,3] features this capability of state restoration at runtime. Its operating principle complies perfectly well with the strategy of cyclic processing, which is usually applied to minimise the transfer bandwidth required for this special form of forward recovery. Faulty State Detected Waiting for next RSI to start state restoration process Fragmented transfer of Task List and data memory PES rejoins processing Waiting for cycle that ends with all Modification Bits being reset TPU TAU SDS t Cycle m t Cycle I m II tcycle m t I Cycle m t II Cycle t m t Cycle m t Cycle RSI RSI The state restoration process RSI The PES transfers at any cycle s end a fixed amount of data words modified in one of the recent cycles. Thus, the frequency of permissible state changes does not need to be restricted to a number that allows their exchange within the on-going cycle. This transfer policy bases on modification bits, which indicate state changes until they have been copied to the redundant units. Strict separation of task-administration and -execution, along with dedicated digital circuitry, enables a particularly efficient integration of modification bit handling. After a cycle s program code execution, the state data needing to be transferred are accessible without any delay. Hence, the transfer of recent state changes can already start while task administration is still in progress. References: [1] M. Skambraks: A Real-Time PES Supporting Runtime State Restoration after Transient Hardware-Faults, Proc. 20th IEEE/ACM Intl. Parallel and Distributed Processing Symposium, p. 173, Los Alamitos: IEEE Computer Society Press 2006 [2] M. Skambraks: An Architecture for Runtime State Restoration after Transient Hardware- Faults in Redundant Real-Time Systems, Proc. 11th IEEE Intl. Conf. on Emerging Technologies and Factory Automation, Los Alamitos: IEEE Computer Society Press 2006 [3] M. Skambraks and W.A. Halang: State Restoration at Runtime after Transient Hardware- Faults in Redundant Real-Time Systems, in Discrete-Event System Design 2006, M. Adamski, L. Gomes, M. Wȩgrzyn and G. Labiak (Eds.), pp , Zielona Góra: University of Zielona Góra Press 2006

57 A Communication Scheme for State Restoration at Runtime in a Redundant PES Configuration M. Skambraks and P. Neumann A novel fieldbus concept was devised for a new safety-related real-time architecture [1]. What distinguishes this fieldbus from others is that it inherently supports a forward recovery technique called state restoration at runtime, which enables a redundant configuration of Programmable Electronic Systems (PESs) to re-start single units at runtime, and involves special communication between the redundant PESs. Hence, the fieldbus was not only designed for data exchange with process peripherals, but also for the communication necessary to restore a PES s state without interrupting the processing going on. Similar to the proprietary fieldbus Interbus, all system nodes, i.e., the redundant PES instances as well as sensors and actuators, are connected to a ring, and data are transferred from node to node as in a shift-register. The transfer is organised in cycles that match the execution cycles of the dedicated PES architecture. A special multi-ring technique guarantees uninterrupted operation in case of physical failures like wire breaks or node outages. Internal read access Alternating Storage of Process Image Communication Node n Memory Block Bus Interface FIFO- Structure Internal write access Control for alternating access Predecessor node Memory Block FIFO- Structure I/O Channel Successor node Input Pointer Output Pointer The nodes I/O interfaces consisting of two memory blocks and two FIFO-structures The fieldbus fulfills real-time demands and supports state restoration of error-affected or replaced processing nodes at runtime. Its tight integration into the safety-related PES architecture features this form of forward recovery without the drawback of inadequately long system response times. The multi-ring connection scheme combines high, scalable reliability and low wiring expenses with an especially simple structure. Communication is performed in discrete intervals, and in synchrony with the cyclically organised application processing, leading to a simple and easy-to-model temporal behaviour of the entire system. This eases verification and reduces the cost for safety licensing. Reference: [1] M. Skambraks and P.Neumann: A Fieldbus for Safety-Related Real-Time Operation Supporting Forward Recovery from Redundant Nodes, in Fieldbus Systems and their Applications, M.L. Chavez (Ed.), Oxford: Elsevier-Science 2006

58 Function Block Paradigm for Safety-related Applications W. Zhang Currently, safety-related systems become more and more essential in industrial automation. With software increasingly appearing in those applications, the current understanding of safety as a probability of failures is not sufficient any longer, as software systems inherently have no failures but designed-in errors. Therefore, in the development of safety-related systems, it is vital to make sure that applications are safe. To this end, the Function Block (FB) paradigm originating in automation engineering was selected as a suitable approach [1], both due to its long tradition in automation practices, and the intrinsic simplicity provided, which is understood to be the very principle to cope with safety problems caused by system complexity. Compared with the object-oriented (OO) approach and component-based development (CBD), the current FB paradigm has many features similar or equal to those of OO and CBD. Based on that, the FB paradigm was enhanced to be test-oriented, integration-friendly, and library-based. Strict development processes and verification based on formal methods are generally understood as the main ways towards safety. Supported by the Unified Modelling Language, specification of safety-related applications and corresponding integration testing was considered, and design guidelines for safe FBs and safe applications were given. Safety-related activities in the V-model Knowledge management is a method to improve the global competitiveness of an enterprise. It is made difficult by problems such as knowledge/information exchange and system complexity. In [2], it was shown that knowledge management can be facilitated by information management. An FB-based framework of multi-agent systems was introduced to make information/knowledge exchange available. All information sources in industrial automation systems were analysed and modeled by agents. The relationships between these sources were emphasised in the context of especially proactive maintenance. References: [1] W. Zhang: Specification and Verification of Safety-related Automation Applications based on Function Blocks, Fortschr.-Ber. VDI Reihe 10 Nr Düsseldorf: VDI-Verlag 2006 [2] W. Zhang: Knowledge Management for E-Maintenance of Industrial Automation Systems, in Contributions to Ubiquitous Computing, B.J. Krämer and W.A. Halang (Eds.), pp , Series Studies in Computational Intelligence, Berlin-Heidelberg-New York: Springer-Verlag 2006

59 Measuring Similarities of Ontologies Y. Zhao, X. Wang, Y. Ding and W.A. Halang Measurement of semantic similarity between web services is an important factor for web service discovery, composition, and even execution. Semantic web services are usually specified based on ontologies. In [2], a measure for the similarity of single and multiple ontologies was proposed, which is based on ontology concept distance, the fuzzy-weighted associative network and the information-theoretic method: dis = ω 1 Dis s + ω 2 Dis i + ω 3 Dis c, with ω i = 1 where Dis s is the distance basing on the concept structure in the service ontology, Dis i is basing on the common contents shared by concepts, and Dis c is only used to measure the compound concept terms by clustering concepts, basing on co-occurrences of concept elements co-occurrence. The measure does not only consider the different concept naming features, but also makes up for the losses of any single approach, because the context of service description is just a structure and a short piece of text, and neither a corpus nor thesaurus. Number of found mappings Similarity threshold (%) Mapping results with different similarity thresholds With the rapid development of the semantic web, it is likely that the number of ontologies will greatly increase during the next few years, which leads to a rising demand for rapid ontology mapping. A novel similarity measurement method based on rough sets and concept lattices was proposed in [1] to realise ontology mapping tasks. A reference concept lattice is first constructed with the combination of two normalised contexts. Rough set theory is then employed to calculate the similarity measure of the two ontology nodes. With a specified threshold, the final ontology mapping can be obtained. Compared with other mapping algorithms, the proposed ontology mapping method is featural and structural, and experiments show the performance of the mapping method. References: [1] Y. Zhao and W.A. Halang: Rough Concept Lattice-based Ontology Similarity Measure, Proc. First ACM International Conference on Scalable Information Systems, A. Chowdhury, F. Lau and Z. Wang (Eds.), ACM International Conference Proceedings Series Vol. 152, New York: ACM Press 2006 [2] X. Wang, Y. Ding and Y. Zhao: Similarity Measurement about Ontology-based Semantic Web Services, Proc. Workshop on Semantics for Web Services, O. Shafiq (Ed.), pp , Zürich, 2006

60 Ontology Mapping Y. Zhao, X. Wang and W.A. Halang The concept of the Semantic Web suggests to annotate web resources with machineprocessable metadata. As means to conceptualise and structure knowledge, ontologies are considered as the key to realise the vision of the Semantic Web. Ontologies themselves, however, do not provide semantic interoperability, since a single ontology cannot be used to represent all kinds of domains and applications. Therefore, ontology mapping is introduced to achieve knowledge sharing and semantic integration in an environment with different underlying ontologies. Similarity measure Ontology A Mapping discovery Mapping result Ontology B Principle of ontology mapping An overview of the ontology mapping approaches to information integration developed by researchers in the community of ontology mapping was provided in [1]. After introducing the backgrounds and concepts of ontology mapping, a comprehensive and detailed treatment of different ontology mapping schemes was presented. Aspects closely related to ontology mapping such as describing mapping results, similarity measures, and performance evaluation of algorithms were also addressed. In [2], a novel similarity measuring method was presented apt to realise ontology mapping tasks: Sim(a, b) = (a b) LA /( (a b) LA + α a LA b LA + (1 α) b LA a LA ) where a LA means the lower approximation of set a and. represents the cardinality of a set. With a specified threshold, the final result of ontology mapping can be obtained. Compared with other mapping algorithms, the proposed ontology mapping method provides more accurate results of higher confidence. References: [1] Y. Zhao, X. Wang and W.A. Halang: Ontology Mapping Techniques in Information Integration, in Adaptive Technologies and Business Integration: Social, Managerial and Organizational Dimension, M.M. Cunha, B.C. Cortes and G.D. Putnik (Eds.), Hershey: Idea Group Reference 2006 [2] Y. Zhao, X. Wang and W.A. Halang: Ontology Mapping based on Rough Formal Concept Analysis, Proc. International Conference on Internet and Web Applications and Services, P. Dini, P. Lorenz, D. Roman and M. Freire (Eds.), Los Alamitos: IEEE Computer Society Press 2006

61 Knowledge Discovery and Semantic Web Service Y. Zhao, X. Wang, and W.A. Halang Widely supported by industry and recent developments in research, semantic web services appear to constitute the next generation of the service-oriented paradigm. Although much work has been done on selecting semantic services, an exact selection model, which could describe services and their selection in a powerful way, is still an open issue. Moreover, little consideration was given so far to quantify the selection of services basing on both capabilities and quality of service (QoS). Employing formal concept analysis [1], therefore in [2] a QoS-based selection model for semantic services was proposed, and a service matching algorithm working under this model presented. Hasse diagram of a context built with Galois Lattice Interactive Constructor (Galicia) In [1], the effect the theory of formal concept analysis (FCA) has on semantic web mining, and the contribution of semantic web mining techniques to building concept lattices were described. In more detail, semantic web mining techniques such as ontology building and visualisation based on FCA, and ontology interoperability techniques such as ontology merging and mapping based on FCA were reviewed. The challenge of information retrieval is to extract information from the large amount of data available. Semantic web techniques together with FCA can help to improve the quality of the results retrieved. Furthermore, semantic web techniques can be used for FCA applications. References: [1] Y. Zhao: Using Formal Concept Analysis for Semantic Web Applications, in Contributions to Ubiquitous Computing, B.J. Krämer and W.A. Halang (Eds.), pp , Series Studies in Computational Intelligence, Berlin-Heidelberg-New York: Springer-Verlag 2006 [2] X. Wang, Y. Zhao and W.A. Halang: Selection Model of Semantic Web Services, in Applied Artificial Intelligence, D. Ruan, P. D hondt, P.F. Fantoni, M. De Cock, M. Nachtegael and E.E. Kerre (Eds.), pp , Singapore: World Scientific 2006

62 Junior professorship of Embedded Systems Jun.-Prof. Dr. Zhong Li Faculty of Electrical and Computer Engineering FernUniversität in Hagen Hagen, Germany Tel: Fax: Internet: World Wide Web: Staff members Name Extension Internet Dr. Juan Gonzalo Barajas Ramírez 4530 Hong Li, MSc 1193 Ping Li, MSc 1732 Research interests Embedded systems Applications of chaos control and synchronisation to information and communication technology Chaos based cryptography Applications of chaos theory in DC-DC converters Group co-ordination and co-operative control

63 Integration of Fuzzy Logic and Chaos Theory Z. Li, W.A. Halang and G. Chen In the 1960s scientists faced difficulties in dealing with imprecise information and complex dynamics. A set theory and, then, an infinite-valued logic of L.A. Zadeh were so confusing that they were called fuzzy set theory and fuzzy logic; a deterministic system found by E.N. Lorenz to have random behaviour was so unusual that it was later named a chaotic system. Just like irrational and imaginary numbers, negative energy, anti-matter,..., fuzzy logic and chaos were gradually and eventually accepted by many, if not all, scientists and engineers as fundamental concepts, theories, as well as technologies. With widespread industrial, commercial and technical applications, ranging from control, automation and artificial intelligence to image/signal processing, pattern recognition, and electronic commerce fuzzy systems technology has reached maturity. Chaos, on the other hand, was considered one of the three monumental discoveries of the 20th century together with the theory of relativity and quantum mechanics. As a very special nonlinear dynamical phenomenon, chaos has reached its current outstanding status from being merely a scientific curiosity in the mid-1960s to an applicable technology in the late 1990s x(k) k A chaotic time series generated by a fuzzy system Finding the intrinsic relation between fuzzy logic and chaos theory is certainly of significant interest and of potential importance. Current research progress and some results on the interplay of fuzzy logic and chaos theory are presented in [1]. The coverage includes the fuzzy definition of chaos, fuzzy modeling and control of chaotic systems using both Mamdani and Takagi-Sugeno models, fuzzy model identification using genetic algorithms and neural network schemes, bifurcation phenomena and self-referencing in fuzzy systems, complex fuzzy systems and their collective behaviours, as well as some applications of combining fuzzy logic and chaotic dynamics, such as fuzzy-chaos hybrid controllers for nonlinear dynamic systems, and fuzzy-model-based chaotic cryptosystems. Reference: [1] Z. Li, W.A. Halang and G. Chen: Integration of Fuzzy Logic and Chaos Theory. Series Studies in Fuzziness and Soft Computing, Vol. 187, Berlin-Heidelberg-New York: Springer-Verlag 2006

64 Fuzzy Chaotic Systems Modeling, Control and Applications Z. Li Over the past few decades, fuzzy systems technology and chaos theory have received ever increasing research interests from, respectively, systems and control engineers, theoretical and experimental physicists, applied mathematicians, physiologists, and other communities of researchers. Especially, as one of the emerging information processing technologies, fuzzy systems technology has achieved widespread applications in many industries and technical fields, ranging from control, automation, and artificial intelligence to image/signal processing and pattern recognition. On the other hand, in engineering systems, chaos theory has evolved from being simply a curious phenomenon to one with real, practical significance and utilisation. We are now standing at the threshold of major advances in the control and synchronisation of chaos with new applications across a broad range of engineering disciplines, where chaos control promises to have a major impact on novel time- and energy-critical engineering applications. A controller for Chua s circuit In [1], it was tried to provide some heuristic research achievements and insightful ideas to attract more attention on the topic. To this end, the fundamental concepts of fuzzy logic and fuzzy control were reviewd, and chaos theory and chaos control as well as the definition of chaos on the metric space of fuzzy sets treated. It was emphasised that the original definition of Li and Yorke and its fine-tuning by Devaney seem to be the most popular definitions of chaos among many others. Various generalisations of these definitions to difference equations in R n were reviewed as well as higher-dimensional conditions ensuring the existence of chaos, specifically the snap-back repeller condition of Marotto and its counterpart for saddle-points. Then, further generalisations to mappings in Banach spaces and complete metric spaces were considered. These were illustrated with various simple examples including an application to chaotic dynamics on the metric space (E n, D) of fuzzy sets on the base space R n. Then, fuzzy modeling and (adaptive) fuzzy control of chaotic systems, all based on both Mamdani fuzzy models and Takagi-Sugeno (TS) fuzzy models, and topics such as synchronisation, anti-control of chaos, intelligent digital redesign, all for TS fuzzy systems, and spatiotemporal chaos and synchronisation in complex fuzzy systems were discussed. Reference: [1] Z. Li: Fuzzy Logic Chaotic Systems. Series Studies in Fuzziness and Soft Computing, Vol. 199, Berlin-Heidelberg-New York: Springer-Verlag 2006

65 Li-Yorke Chaos in Higher Dimensions P.E. Kloeden and Z. Li Chaos, specifically chaotic dynamics, may well be considered together with relativity and quantum mechanics as one of the three monumental scientific discoveries of the 20th century. In the past four decades since Li and Yorke s seminal paper and the equally influential work of Loren which partly motivated it, chaos has matured as a science (indeed, is still evolving) and has provided many deep insights into previously intractable and inherently nonlinear natural phenomena. The term chaos was first formally introduced into mathematics in connection with an interval map by Li and Yorke in 1975, where they established a simple criterion for the existence of chaos in one-dimensional difference equations, their now famous period three implies chaos result. H. Poincare was well aware of the possibility of very complicated, chaotic dynamical behaviour already in the 1890s. Indeed, his writings stimulated major advances in mathematics during the first half of the last century, including the work of Birkhoff, the KAM theory of Kolmogorov, Arnold and Moser, and the sensational horseshoe diffeomorphism of Smale in the 1960s. The novelty of the results of Lorenz and of Li and Yorke was to show that such behaviour could arise in quite simple systems and could be generated by quite simple mechanisms. In particular, the period three implies chaos result in scalar difference equations involved noninvertible, continuous mappings rather than more demanding diffeomorphisms Lyapunov exponent μ Lyapunov exponent for the logistic map as a function of µ Surveys on the definitions of chaos were provided in [1,2], and it was pointed out that in practice, however, establishing the presence of chaos in a particular dynamical system often still depends mainly on numerical calculations to estimate quantities such as the maximum Lyapunov exponent and topological entropy. A unified, well accepted, easy-totest, and rigorous mathematical definition of chaos is still in the process of being revealed. References: [1] P.E. Kloeden and Z. Li: Beyond the Li-Yorke Definition of Chaos, in Integration of Fuzzy Logic and Chaos Theory, Z. Li, W.A. Halang and G. Chen (Eds.), Series Studies in Fuzziness and Soft Computing, Vol. 187, pp. 1 23, Berlin-Heidelberg-New York: Springer-Verlag 2006 [2] P.E. Kloeden and Z. Li: Li-Yorke Chaos in Higher Dimensions: A Review, Journal Difference Equations and Applications 12, 3/4, , 2006

66 Anticontrol of Chaos for Takagi-Sugeno Fuzzy Systems Z. Li, G. Chen and W.A. Halang The process of chaos control is now understood as a transition between chaos and order, depending on the application of interest. The task of purposely creating chaos, sometimes called chaotification or anticontrol of chaos, has attracted increasing attention in recent years due to its great potential in non-traditional applications such as those found within the context of physical, chemical, mechanical, electrical, optical and, particularly, biological and medical systems x2(t) x2(t) x1(t) Projection on x 1 x 2 plane and 3-D phase portrait of anticontrolled Chen s TS fuzzy system In [1], one simple yet mathematically rigorous control method was developed to chaotify discrete Takagi-Sugeno (TS) fuzzy systems. Moreover, such a controller can be designed for an arbitrarily given, n-dimensional dynamical system that could originally be nonchaotic or even asymptotically stable. The goal of chaotification is finally achieved with a simple modulus operation or a sawtooth (or even a sine) function. The design criterion is to use the definition of chaos given by Devaney or Li-Yorke, where for the n-dimensional case the Marotto theorem was used for a proof. For the continuous-time case, a general approach to make an arbitrarily given autonomous TS fuzzy system chaotic has also been proposed. Here, the main tool to use is time-delay feedback perturbation on a system parameter or as an exogenous input. It can make an arbitrarily given continuoustime system chaotic via fuzzy feedback linearisation, differential geometric control theory, and time-delay feedback perturbations. In this approach, asymptotic analysis is used to establish an approximate relationship between a time-delay differential equation and a discrete chaotic map, so that a time-delay feedback control term can be constructed to make the controlled system chaotic, where the generated chaos is in the sense of Li and Yorke. Systems with time-delay are inherently infinite-dimensional; therefore, it is possible to produce complicated dynamics such as bifurcation and chaos, even in a very simple first-order system. Reference: [1] Z. Li, W.A. Halang and G. Chen: Anticontrol of Chaos for Takagi-Sugeno Fuzzy Systems, in Integration of Fuzzy Logic and Chaos Theory, Z. Li, W.A. Halang and G. Chen (Eds.), Series Studies in Fuzziness and Soft Computing, Vol. 187, pp , Berlin- Heidelberg-New York: Springer-Verlag x1(t) x3(t) 4 6 8

67 Fighting Internet Congestion by Means of Chaos Theory Z. Li Congestion control mechanisms in today s Internet already represent one of the largest artificial feedback systems deployed. As the Internet continues to expand in size, diversity and reach, playing an ever-increasing ro le in the integration of other networks (transportation, finance, etc.), it becomes ever more crucial to have a solid understanding of how this fundamental resource is controlled. Given the scale and complexity of the network, however, and the heuristic, intricate nature of many deployed control mechanisms, until recently this problem appeared to be well beyond the reach of analytical modeling and feedback control theory. Hence, a reliable and accurate traffic model is essential which involves time-delays inevitably being introduced in communication networks, both due to limited bandwidth, but also due to overhead in the communication nodes and in the network. Such a traffic model is needed in order to evaluate the performance of the Internet s networks, and for use in network planning and designs b Avg. Queue Length Norm Avg. Queue Length b w Time Bifurcation diagram and congestion control Recently, it has been shown that TCP-RED could approximately be modeled as a firstorder nonlinear map, and exhibit a rich variety of irregular behaviours such as bifurcation and chaos. In [1], the TCP-RED mechanism was introduced in a control system framework its discrete dynamical system model is derived. Then, chaotic properties of the TCPRED map were analysed and the chaotic behaviour was shown under some parameter conditions. Finally, a time-delayed feedback control method was proposed to stabilise the average queue length on a desired value, whose effectiveness was shown by a simulation example. Taking the exponential averaging weight or the RED parameter as bifurcation parameter, the bifurcation behaviour of the TCP-RED system is illustrated above on the left. On the right it is shown that, by the proposed control law, the system is stabilised on the fixed point q e and, thereafter, the control input becomes zero. Reference: [1] Z. Li: Fighting Internet Congestion by Means of Chaos Control, in Contributions to Ubiquitous Computing, B.J. Kra mer and W.A. Halang (Eds.), pp , Series Studies in Computational Intelligence, Berlin-Heidelberg-New York: Springer-Verlag, 2006

68 Li-Yorke Chaos in a Spatiotemporal System P. Li, Z. Li, W.A. Halang and G. Chen To be useful in cryptography, coupled map lattices need to exhibit spatiotemporal chaos. But, does spatiotemporal chaos in a coupled map lattice really exist? Or is it just an artefact due to finite precision in computer simulations? To lay a theoretical foundation for further study and for the exploitation of spatiotemporal chaotic systems, a one-way coupled logistic-map lattice was considered, which has already been intensively investigated and which is described by x j k+1 = (1 ɛ)f(x j k ) + ɛf(xj 1 k ), where f(x) = ax(1 x), x j k represents the state variable for the jth site (j = 1, 2,..., L, with L being the number of the sites in the coupled map lattice) at time k (k = 0, 1, 2,...), ɛ (0, 1) is the coupling strength, and a (0, 4] is the parameter of the logistic map. It is assumed that the coupled map lattice satisfies the periodic boundary condition, x 0 k = xl k for all k x i λ i i Lyapunov exponent and state space i For this coupled map lattice it was shown in [1], that it is chaotic according to the definition of Li and Yorke, provided that (2 a)(1 2ɛ) > 1, a (a 1, 4] and c (0, 4 2 ( 3), where a 1 = 1 3 2(c + 4) + c2 + 8c + 4 ) and c = 1/ L(1 2ɛ + 2ɛ 2 ). Reference: [1] P. Li, Z. Li, W.A. Halang and G. Chen: Li-Yorke Chaos in a Spatiotemporal Chaotic System, in Applied Artificial Intelligence, D. Ruan, P. D hondt, P.F. Fantoni, M. De Cock, M. Nachtegael and E.E. Kerre (Eds.), pp , Singapore: World Scientific 2006

69 A Multiple Pseudo-Random-Bit Generator Based on a Spatiotemporal Chaotic Maps P. Li, Z. Li, W.A. Halang and G. Chen Chaotic systems are extensively applied in cryptography, such as in designing pseudorandom-bit generators (PRBGs). So far, many chaos-based PRBGs were proposed, among which spatiotemporal chaotic systems have prominent merits when high encryption speed is required. This is due to some inherent features of spatiotemporal chaos, viz., long period orbits, high complexity of system dynamics yielding random orbits, and multiple sites in spatiotemporal chaotic systems, which can be used to generate unrelated pseudorandom-bit sequences (PRBSs) simultaneously. In [1], a multiple-output PRBG based on a spatiotemporal chaotic system is proposed. A nearest-neighbour coupled map lattice (CML) consisting of logistic maps is used as a spatiotemporal chaotic system: x n+1,i = (1 ɛ)f(x n,i ) + ɛ 2 [f(x n,i+1) + f(x n,i 1 )], where n = 1, 2,... is the time index, i = 1, 2,..., L is the lattice site index with a periodic boundary condition, f is a local chaotic map in the interval I, and ɛ [0, 1] is a coupling constant. Here, the logistic map, i.e., f(x) = rx(1 x)(r (0, 4]) is taken as the local map. The construction of the CML-based PRBG is shown on the left, where the output of the CML is digitised to generate multiple PRBSs in the way as shown on the right: CML Site 1 Site 2 Site L LM 1 {x 1 n } LM 2 {x 2 n } LM L {x L n } PRNS 1 PRNS 2 PRNS L Digitizing Digitizing Digitizing {b n1 } 1 {b n1 } 2 {b n1 } P {b n2 } 1 {b n2 } 2 {b n2 } P {b nl } 1 {b nl } 2 {b nl } P 1 PRBS 1 1 PRBS 2 PRBS 1 P 2 PRBS 1 2 PRBS 2 2 PRBS P L PRBS 1 L PRBS 2 L PRBS P x 1 x 2 0. = = x n = b 1 1 b 2. 1 b n,,, 2 b 1 2 b b n,,, P b1 P b 2.. P b n Construction of CML-based PRBG Digitisation method Thus, P L PRBSs can be generated simultaneously from the CML with L sites. By analysing [2] the largest Lyapunov exponent and the ergodicity of this and 5 similar CMLs, their parameter ranges were determined for which corresponding PRBGs render good cryptographic properties. In-depth investigations indicate that the PRBSs are almost balanced, that their linear complexity is about N/2, that their auto-correlation is δ- like, that their cross-correlation is close to zero, and that their probability distribution function is uniform, which indicates that the PRBGs have very satisfactory cryptographic properties and high efficiency. This analysis resulted in criteria to design PRBGs based on these six CMLs. References: [1] P. Li, Z. Li, W.A. Halang and G. Chen: A Multiple Pseudo-Random Bit Generator based on a Spatiotemporal Chaotic Map, Physics Letters A 346, , 2006 [2] P. Li, Z. Li, W.A. Halang and G. Chen: Analysis of a Multiple-Output Pseudo-Random- Bit Generator Based on a Spatiotemporal Chaotic System, International Journal of Bifurcation and Chaos 16, 10, , 2006

70 A Stream Cipher Based on a Spatiotemporal Chaotic System P. Li, Z. Li, W.A. Halang and G.R. Chen A stream cipher based on a one-way Coupled Map Lattice (CML) consisting of logistic maps used as a spatiotemporal chaotic system was proposed [1]. The configuration and parameters used for decryption are the same ones as those for encryption. The most favourable parameters of the cipher were determined by analysing its overall performance. The size of the keyspace, that is, the number of all possible keys, was evaluated as 2 48L by analysing the error function, where L is the size of the CML. Already for L > 3, the keyspace is cryptographically large enough. Cryptographic properties of the keystreams generated from the CML, such as order-1 and order-2 probability distribution functions, run and auto-correlation, were investigated numerically and are shown here: ρ ρ k 0 0 k n n k n P 0/ C ii (τ) n τ Cryptographic properties of keystreams These results are satisfactory. The security of the cipher was evaluated by analysing its confusion and diffusion properties, and by applying the error function, known-plaintext, differential, brute-force and the chosen-plaintext/ciphertext attacks, respectively. It turned out that the cost to break the cipher is about 2 40L, so the security of the cipher with L > 3 is high enough from the cryptographic point of view. In addition, efficient parallel operation of the cipher is guaranteed by close-to-zero cross-correlation among multiple keystreams. Hence, the encryption speed is sufficiently high even if the size of the CML in the cipher is small. As compared to ciphers based on spatiotemporal chaotic systems with much more complicated configurations, the proposed cipher has higher security, higher encryption speed and lower computation cost as well as other similar good cryptographic properties. Reference: [1] P. Li, Z. Li, W.A. Halang and G. Chen: A Stream Cipher based on a Spatiotemporal Chaotic System, Proc. 1st IFAC Conference on Analysis and Control of Chaotic Systems, pp , Université de Reims, 2006

71 Using Chaos-based Encryption in Electronic Banking P. Li, Z. Li, W.A. Halang and G. Chen To explore feasibility and efficiency of a stream cipher based on spatiotemporal chaos as proposed in [1], it is applied in Internet-based online payments, where the pseudorandom-bit generator is used to generate TransAction Numbers (TAN). The TAN generator was implemented [3] in a commercial system, which issues electronic cash codes for Internet-based online payments, embedded in a project to automate the logistics process to distribute these codes via Internet communication. These cash codes are delivered by the issuer to distributors who sell them to customers. The latter use the cash codes for Internet-based online transactions. Changing from physical to electronic delivery of the cash codes opens up more possibilities of distribution channels and models (e.g., PIN on-demand). The interface for the distributors to connect to the distribution system is web-based. The accessibility of these processes via public Internet necessitates a high level of security to protect transactions representing very high economic values. Key-Control En/Decryption p 0 (c 0 ) x 0 (0) E 0 b 0 Data Input p 1 (c 1 ) f 0 b 0 * x 1 (0) E 1 b 1 Data Output f 1 b 1 * p n (c n ) x n (0) E n b n c f n b n * Vcc Reset Clock GND Block diagram of a chip for chaotic stream encryption The security concept provides for the following three-step authorisation and authentication process. At the connection layer, the link to the system is controlled by digital certificates with client authentication via the https protocol. At the application layer, the use of the functions is controlled by access data, i.e., username and password. For sending an order, which is an actual high-risk operation, single-use cryptocodes (TANs) are utilised to prevent any misuse by un-authorised persons, to authorise the process by means of a digital signature and to prevent malfunctions, such as unintended multiple orders. Once connected to the interface, the distributor can order sets of TANs, download and activate them. To encrypt and decrypt the data transferred in course of the above process, a chip implementing the stream cipher of [1] in hardware [2] is employed. References: [1] P. Li, Z. Li, W.A. Halang and G. Chen: Securing Communication by Chaos-based Encryption, in Contributions to Ubiquitous Computing, B.J. Krämer and W.A. Halang (Eds.), pp , Berlin-Heidelberg-New York: Springer-Verlag 2006 [2] Y. Mao, W. Liu, Z. Li, P. Li and W.A. Halang: A Chip Performing Chaotic Stream Encryption, in Contributions to Ubiquitous Computing, B.J. Krämer and W.A. Halang (Eds.), pp , Berlin-Heidelberg-New York: Springer-Verlag 2006 [3] P. Li, Z. Li, S. Fettinger, Y. Mao and W.A. Halang: A Pseudo-Random Bit Generator based on a Spatiotemporal Chaotic System and its Application in Internet-based Online Payments, in E-Service Intelligence: Methodologies, Technologies and Applications, J. Lu, D. Ruan, G. Zhang (Eds.), pp , Heidelberg: Springer-Verlag 2006

72 Analysing Chaotic Spectra of DC-DC Converters with the Prony Method H. Li, Z. Li, W.A. Halang, B. Zhang and G. Chen Direct current (DC-DC) converters produce electromagnetic interference and, thus, also electromagnetic pollution which is considered to be harmful. On the other hand, when DC-DC converters operate in chaotic modes, they can generate spread spectra, which are useful for reducing electromagnetic interference. Conventionally, the Fast Fourier Transform (FFT) is used to analyse spectra. This approach is computationally efficient and, in most cases, can provide reasonable results for signal processing. However, it is not applicable to inner-harmonics, i.e., the non-integral multiples of the fundamental frequency, which is a prominent feature of chaotic signals. To alleviate the limitations of FFT, several new spectral estimation methods have been proposed. Power spectral density (PSD) obtained by using the Prony and FFT methods at I ref = 4A In [1], the Prony method is suggested for chaotic spectral estimation of DC-DC converters. Numerical simulations showed its advantages over the traditional FFT. In addition, the frequencies, phases, amplitudes, and damping factors of the harmonics of currents or voltages in DC-DC converters could be obtained with the Prony method. Further, the Prony method can distinguish a signal s DC component and its AC component. Therefore, it is recommended that the Prony method should be employed instead of the popular FFT in such applications as the spectral analysis of converters involving chaotic signals. Simulation results of chaotic spectral estimation using the Prony method are illustrated on the left above. For comparison, a similar simulation using the FFT method was also carried out and the results are shown on the right. It is obvious that the Prony method can much more accurately locate the frequencies of the harmonics corresponding to the peaks. Reference: [1] H. Li, Z. Li, W.A. Halang and G. Chen: Analyzing Chaotic Spectra of DC-DC Converters using Prony Method, Proc. Intl. Symposium on Power Electronics, Electrical Drives, Automation and Motion, pp , Piscataway: IEEE Press 2006

73 Calculating the Invariant Density of the Chaotic Mapping in a DC-DC Converter and its Applications H. Li, Z. Li and W.A. Halang In direct current (DC-DC) converters some singular or abnormal phenomena occur often, such as unexpected collapse of the operating mode, strange electromagnetic interference, unstable operation, or failure after a long time running, which can result in system failures and external random disturbance. These effects greatly limit the applications of DC-DC converters. Studies indicate that these phenomena are related to chaotic behaviour. It is known that chaotic motion is an unstable, aperiodic behaviour in a bounded area, and its long-term behaviour shows random-like characteristics. Hence, it is possible to describe and study chaotic behaviour with probability theory. Chaotic mapping, bifurcation diagram, and the invariant density at α = 1.52 Thus, in [1], a Boost converter operating in a chaotic mode was represented as a onedimensional mapping, based on which the invariant density of the chaotic mapping is calculated using the eigenvector method. Comparing this invariant density with its phase portrait and its bifurcation diagram showed that the method is appropriate to calculate the invariant density of the DC-DC converter s chaotic mapping. The calculation results can further be used to estimate the power spectral density of the input, to calculate the average switching frequency of DC-DC converters, and to accurately design the system parameters. On the left of the figure above it can be seen that there are no orbit points in the intervals [0.13, 0.91] and [1.10, 1.15], corresponding to the zero invariant density in these intervals. Simulation results illustrate the accuracy of the eigenvector method in calculating the invariant density. Moreover, according to the latter, the average switching frequency of the boost converter can be expressed as s = 2, where α = V O 1+α VI 1. Reference: [1] H. Li, Z. Li and W.A. Halang: Calculation of Invariant Probability Density Distribution for Chaotic Maps in DC-DC Converter and its Applications, Proc. 51. Internationales Wissenschaftliches Kolloquium, pp , Ilmenau: Technische Universität 2006

74 Chair of Communication Systems (Kommunikationssysteme) Prof. Dr.-Ing. Firoz Kaderali Universitätsstr. 11 (TGZ) Hagen Tel.: Fax: Consultant: Ext. Prof. Dr. rer. nat. Werner Poguntke 2781 Academic staff: Ext. Dipl.-Ing. Alex Didier Essoh 4172 Dipl.-Inform. Jan Gellweiler 4115 Dipl.-Ing. Emir Güleç 4132 Dipl.-Ing. Thorsten Kisner 4163 Jessica Pater, M.A Dr.-Ing. Michael Stepping 4104 Dipl.-Ing. Suphithat Songsiri 4155 Technical staff: Ext. Andrea Frank 4116 Stefan Gebhard 4104 Julian Kuhn 2782 Christian Ney 2782 Christoph Pater 4124 Dipl.-Ing. Gerd Tübben 4109 Ulrike Schmidt 4303 Secretary: Ext. Jessica Gabski 4119

75 Statistical Texture Analysis Methods for Network Traffic Classification Alex Essoh The issue of IT-Awareness has been of great interest in the last few years. Not surprisingly when the dramatic increase of incidents and vulnerabilities over the past ten years is taken into consideration. Therefore, many organisations try to protect themselves by enforcing security guidelines or even a security policy for their network. In most cases there is a security officer (network administrator) responsible for the implementation of this security policy. As an example the firewall policy of an organisation could be to block diverse ports: the telnet port (23) and common P2P ports or only allow incoming traffic on Port 80. This solution only provides limited protection with regard to enforcing security policies for several reasons. Firstly, malicious users are able to bypass the firewall by configuring their server to listen to port 80 although the service offered is not HTTP and forbidden. Secondly using HTTP Tunneling, encapsulated in HTTP packets, data packets of applications which are normally not permitted can reach their destination by bypassing the firewall and violating the security policy of the organisation. Since many security officers responsible for network administration mainly use simple port analysis to detect applications which are normally not permitted, they are not able to detect the security violations described in the first two scenarios as well as some others (see [1] for more). When it comes to the detection of traffic which is not allowed, a simple port analysis cannot really help as has been confirmed in several studies carried out in the past [2]. A signature based Intrusion Detection System such as Snort ( can be used to access the content of the packet and highlight the differences between port indication and the content as such. But such systems only work well on links where the speed is not too high. Furthermore, through the increase in use of cryptographic algorithms to protect applications, the payload of the packets and even the header cannot be accessed. In those cases the security officer is blind and has no chance to detect applications carried by the packets. To address this problem, diverse methods have been proposed with the aim of identifying the service related to traffic flow (traffic classification). The idea behind traffic classification is, that each application has a profile, an extensive analysis of this profile can be used to identify applications running on top of the transport protocols. It has been shown that port based, but also payload based analysis have their limitations with regard to application identification. Therefore, transport layer statistics have been used to build traffic profiles and diverse methods (Bayesian networks, densitybased clustering, Hidden Markov Models, etc.) for the classification of traffic. Hereby, a differentiation has to be made with regard to which information is used to perform the classification: header, payload or just metainformation based on header and or payload. In our approach we introduce a method to classify encrypted network traffic using statistical texture analysis methods. We analyse time-aggregated network traffic and related statistical metrics to determine both HTTP and SMTP traffic. Since they considerably differ from each other, we were able to classify traffic as HTTP or SMTP. [1] J. P. Early, C. E. Brodley and C. Roseberg, Behavioral Authentication of Server Flows, Proceedings of the 19th Annual Computer Security Applications Conference, Las Vegas, USA, December , [2] A.W. Moore, D. Zuev, Internet Traffic Classification Using Bayesian Analysis Techniques, Proceedings of the 2005 ACM SIGMETRICS, Banff, Canada, June ,

76 Network Traffic Visualization Thorsten Kisner Text based tools are widely used to estimate the network status. Most of the system administrators interviewed in [1] answered they use these tools, but most of them realize the necessity of better methods. The following text output of a network traffic capturing tool (see below) demonstrates this immediately: It is not possible for a human to gather all relevant information to directly understand the security state of the network, especially for heavily loaded networks with thousands of packets per minute. 14:11: > ack[0x63a11130]psh l=20,337] 14:11: > ack[0x63a11130]fin l=20,0] 14:11: > ack[0x2ca3f9c1]l=20,0] 14:11: > ack[0x2ca3f9c1] fin l=20,0] Although humans cannot understand huge amounts of text data at once, they have magnificent capacities in image processing, thus a suitable visualization of this data is needed. There are three traditional ways pointed out in [2] to reduce the amount of network data to be visualized: aggregation is used for a large number of nodes or links, averaging is used for large numbers of time periods and thresholds and/or exceptions are used to detect changes and differences from normal behaivior. Classic approaches to visualize network data are based on grid representations, where different features like IP addresses, arrival times or port numbers are mapped to the axes of the coordinate system in 2D or 3D. Then a color-coding for each position is normally used to represent several kinds of network activity. For example NVisionIP [4] uses network flow information as input data and both axes correspond to internet addresses, the communication relationship between both hosts with the dependency of the number of packets are plotted in the grid. PortVis [5] uses ip address ranges and port numbers and the frequency of occurrence is coded with different colors. Both approaches are able to detect large scale network security events and provide functions for detailed analysis. Our approach is different from many other visualization methods. We use the same basic data, network traffic, but transfer them to the domain of digital image processing and are able to analyse this resulting dataset with different statistical texture analysis methods. We can process detailed network packet data, e. g. TCP flows, or can use aggregated data where only the packet arrival times and packet sizes are available. [1] Robert Ball, Glenn A. Fink and Chris North: Home-Centric Visualization of Network Traffic for Security Administration, Proceedings of the 2004 ACM Workshop on Visualization and Data Mining for Computer Security (VizSec/DMSec 2004) [2] R. Becker, S. Eick and A. Wilks: Visualizing network data. IEEE Transactions on Visualization and Computer Graphics, 1(1):16?28, [3] S. T. Teoh, K.-L. Ma, S. F. Wu and X. Zhao: Case study: Interactive visualization for internet security. In Proc. IEEE Visualization, [4] Kiran Lakkaraju, William Yurcik, Ratna Bearvolu, and Adam J. Lee, NVisionIP: An Interactive Network Flow Visualization Tool for Security, IEEE International Conference on Systems, Man and Cybernetics (SMC), [5] J. McPherson, K.-L. Ma, P. Krystosk, T. Bartoletti, M. Christensen: PortVis: A Tool for Port-Based Detection of Security Events, VizSEC/DMSEC 04, October 29, 2004, Washington, DC, USA

77 The Open Source Initiative CampusSource Dr. Michael Stepping During the last decade a number of Virtual University systems have been developed at different stages. Today s state-of-the-art systems usually contains elements for content management (i.e. multimedia content) and technical infrastructure (i.e. distribution and access to content) to support computer and web-based learning and teaching time and location independent. These different systems were bundled under the head of CampusSource Initiative. CampusSource supports and welcomes the set-up of cooperative networks in universities and with industrial partners to a broad use. Introducing the Open Source strategy, CampusSource takes up the academic tradition of presenting results of latest research in a suitable form. Within the academic tradition, scientific results are described as precisely as possible to allow for the verification by third parties. The aim of the Open Source initiative is to unite the efforts of several projects and make the results freely available for use and further development to those who are interested. Each system developed is freely available under the CampusSource license. This license enforces the publication of all source code and allows every licensee to modify the code as long as every modification is returned into the public software pool. There is no license fee for the use of the software, which is called free software. Another advantage of the availability of source code is the adaptability of the program to the user s individual needs including functional extensions and adjustments to new technologies. Users of Open Source software can modify the source code at any time, and therefore be independent from manufacturers. Success of CampusSource Criteria for success of CampusSource are a wide distribution of the software and acquiring dedicated and competent developers to improve the software. Therefore CampusSource publishes the software products nation wide as well as internationally. Directions of software development have to be put forward through discussion with all engaged users and developers. This includes development and use of a decent reference architecture for infrastructure-software in educational institutions. In addition to that a quality standard has been defined and achieved. CampusSource has prooved itself as a high quality level organization which supports and leads the development of technical platforms for operating virtual universities. To achieve these aims the initiative created an internet portal at Integrated into the portal is the software exchange market. This is the area where software is available for download including all documents, demos, installation scripts, documentation and test cases. Current Developments in CampusSource The emerging need of access to external data sources like university databases (students, professors), HIS or SAP lead to the set-up of the CampusSource Engine working group CSE. This working group identified common functionalities of platforms and systems. The architecture for the collaboration and message interchange of the systems is released. Adaptation works and further investigations are still ongoing. Besides the infrastructure for distribution of e-learning content a new task for the content itself is started. The requirements for content exchange is still an issue under research and development.

78 CampusSource Engine Convergence and Modularisation of Open Source e-learning Systems Jan Gellweiler In the last decade a multiplicity of open source e-learning platforms and tools have been developed at German universities. But a large, competitive amount of such systems means a small, decreasing community of developers per system. It also implies a repetitive development of functionalities and consequently a waste of resources. In addition most of these systems are faced with the emerging need of access to external data sources like university databases (students, professors), HIS or SAP. Therefore the CampusSource- Systems metacoon (Bauhaus-Universität Weimar, metacoon-services), MILESS (Universität Duisburg-Essen), EdoWorkSpace (Universität Dortmund), OpenUSS and Freestyle Learning (Westfälische Wilhelms-Universität Münster), STUD.IP (data-quest GmbH), Uni Open Platform and openfuxml (FernUniversität Hagen), ILIAS open source (Universität zu Köln) and HIS GmbH joined together and set-up the CampusSource Engine (CSE) working group. This working group identifies, specifies and implements common functionalities of platforms and systems and investigates the architecture for the collaboration and message interchange of the systems ([1], [2], [3] and [4]). As a first result of the working group the interconnection of the management system HIS- LSF with four platforms of the CampusSource portfolio (EdoWorkSpace, ILIAS open source, metacoon and OpenUSS) was demonstrated at Learntec At the end of the year the first installation of the CampusSource Engine went into production at the FHW Berlin. The current function range of this installation enables remote controlled creation of courses and course participants in the connected e-learning system by HIS-LSF. Further functions are under development. The following figure sketches the architecture of the CampusSource Engine as intermediate system between e-learning systems and internal as well as external data sources i.e. of universities. Individual Functionality Individual Functionality Individual Functionality Individual Functionality Individual Functionality Individual Functionality Individual Functionality Individual Functionality HIS Data Other Data Sources Local Data In addition to the development of the HIS-interface, work on an XML-based document schema, a common solution for identity management, etc. has been started. [1] J. Hüvelmeyer et al.: A Vision For Distributed Modular Platforms aktuelles/publikationen/platform-vision.pdf, 2003 [2] E.-E. Doberkat et al.: Anforderungen an eine elearning-plattform elp.pdf, 2002 [3] J. Hüvelmeyer et al.: CampusSource Engine - Überlegungen zur Entwicklung einer modularen Lernplattform [4] M. Stepping et al.: CampusSource Engine - Die Schnittstelle von e-learning Systemen zum HIS-GX System der HIS GmbH

79 MTrust: a reputation-based trust model for a mobile agent system Suphithat Songsiri This work aims to promote MTrust: a reputation-based trust model that will ensure cooperative interactions amongst mobile agents and visited hosts in a mobile agent system. MTrust is composed of a reputation system and a trust formation system. A reputation system relies on two components; a truthful feedback submission algorithm and a set of distributed feedback information storages. The first encourages participants to submit truthful transaction feedbacks and punishes them when pairs of contradictive feedbacks are reported. The second constitutes the distributed storage system, where each storage location is assigned to contain feedback information of one visited host. A trust formation system enables a truster to compute a trustee s trustworthiness. It has two components. A FAM calculates a trust value from feedback information when there is a lack of direct experiences using Beta distribution. A TCM calculates a trust value using Bayesian Network (BN). In a reputation system, a truthful feedback submission must involve following steps namely: encouraging feedback submission, identifying bias feedbacks from all received feedbacks, and punishing entities who submitted a bias feedback, which can be explained as follows. Incentive based timely feedback submission: after a transaction is completed, the visited host submits its feedback to its feedback storage server. A mobile agent submits its feedback to its owner to be used as direct experience and the visited host s feedback storage server only after migrating to the next host. This is to prevent the current visited host from snooping on a mobile agent s feedback. If any participant fails to submit its feedback, it will be banned from engaging itself in any future transaction. A fair punishment scheme: The storage server of the current visited host will then compare the feedbacks for consistency or contradiction. For contradictive feedbacks, a fair punishment scheme must be exercised by prohibiting all the mobile agents from the owner of a mobile agent who has submitted a feedback and the corresponding visited host from engaging in new transactions for a period of time. In a trust formation system, we propose five methods (i.e. a predefined trust value PT, a feedback aggregation method for an experienced truster FAEX, a feedback aggregation method for an inexperienced truster FAIN, a general trust value GT, and a combination of Bayesian network and a feedback aggregation method for an experienced truster CO) for a truster to calculate each trustee s trustworthiness according to the types of a truster and status of a trustee towards a truster should be considered. A truster can be categorized into an inexperienced truster; i.e. a truster who is new to the system or an experienced truster; i.e. a truster who has had some transactions with some trustees. Status of a trustee towards a truster can be explained as a newcomer to the system (type1), never transacts with a truster but not new to the system (type2), or has committed some transactions with a truster (type3).

80 Chair of OPTICAL INFORMATION TECHNOLOGY (Lehrgebiet Optische Nachrichtentechnik) Prof. Dr. Jürgen Jahns Universitätsstraße Hagen Tel.: Fax: Staff: Ext. Dr. Qing Cao 1133 Dr. Matthias Gruber 1131 Tina Heldt 1121 Manfred Jarczynski (until 09/2006) 1123 Hans-W. Knuppertz 1128 Thomas Seiler 1122 Graduate Assistant: Michael Bohling 1132 Henry Benjamin Ngimbis (until 06/2006) 1132 Students: Ulrich Lohmann 1132 Secretary: Susanne Oetzel (until 10/2006) 1120 Visitor: Arash Sabatyan (01 06/2006) Inst. of Advanced Studies in Basic Sciences, Zanjan, Iran

81 Perturbative method for effective refractive indic of subwavelength gratings Q. Cao Subwavelength dielectric gratings are important photonic components. They have various applications, such as spectral filtering, beam spliting, and anti-reflection. A schemetic view of a subwavelength dielectric grating is given in the left of Figure 1. As shown, the light is incident from Region 1 with an uniform refractive index n 0. Region 3 is the transmission region with a refractive index n s. Region 2 is the grating region. Its periodic refractive index function n(x) has the property n(x + Λ) = n(x), where Λ is the period. Region 1 n 0 R Region 1 n 0 R Region 2 n(x+λ)=n(x) = Region 2 n eff Region 3 n s T Region 3 n s T Fig. 1: Schemetic view of the effective medium theory for a subwavelength dielectric grating. According to effective medium theory, Region 2 is equivalent to a homogeneous medium with an uniform refractive index n eff. This property is shown in the right of Figure 1. Here I develop a simple method to calculate the effective indice n eff. This method is based on a perturbative treatment and the periodic condition of diffraction gratings. From the viewpoint of waveguide theory, the effective index n eff is directly related to the eigenvalue of the fundamental eigenmode. For a dielectric waveguide, the eigenvalue and the eigenfunction are both real. Accordingly, the eigenfunction φ(x) can be expressed as [ x ] φ(x) = exp P (x)dx, (1) 0 where P (x) is a real function. This expression is similar to that used in the WKB method. Because of the periodicity, P (x) is a periodic function of Λ. Accordingly, Λ For the TE polarization, in terms of Eq. (1), one can obtain 0 P (x)dx = 0. (2) P βk 2 εk 2 P 2, (3) where β = n 2 eff, ε = n 2 (x), k = 2π/λ and λ is the vacuum wavelength. By use of the perturbative method, one can expand β and P as β = β 0 + β 1 + β 2 and P = P 0 + P 1 + P 2, respectively. Then one can establish the following recurrence equations P 0 = β 0 k 2 εk 2, (4) P 1 = β 1 k 2 P 2 0, (5) P 2 = β 2 k 2 2P 0 P 1, (6) By use of the periodic conditions Λ 0 P 0dx = 0, Λ 0 P 1dx = 0, Λ 0 P 2dx = 0, and numerical integrals, one can accurately get the effective index n eff, without solving the complicated eigenproblem. With some modifications, this method may be extended to the TM polarization.

82 Pulse width modulated phase relief for diffractive optical elements H. Knuppertz Inspired by a paper of A.W. Lohmann [1], we designed a diffractive optical element (DOE) using pulse width modulation (PWM) techniques (Fig. 2). In photonics this is an unusual design approach, the conventional technique to encode DOEs is pulse amplitude modulation (PAM) and generates staggered relief profiles (Fig. 1). Up to now PWM is little used for phase objects, some implicit hints are found in the fundamental work about computer holography [2], published by A.W. Lohmann in We investigated the design of a DOE that generates three intensity rings in the far field (focal plane of a lens) with a spatial spectrum given as: 3 ũ(ν) = δ ( ν ) nν 1 (1) n=1 The corresponding Fourier transform is a superposition of three real sinusoidal transmission curves which is illustrated in Fig. 1 (analogue curve). This function must be digitized. One technique to convert this analogue item to a digital PWM-profile is to perform a comparison with a sawtooth reference signal, (Fig. 2). If the reference signal exceeds the source signal then the result is the output signal being high otherwise low. The frequency of the sampling sawtooth in our design was 5 times the lowest signal frequency. Because of the radial symmetry of the desired far field pattern, the design of a one dimensional profile is sufficient. The binary profile (rectangle curve in Fig. 2) then represents the cross section of the phase profile for the likewise radial symmetric DOE. We have to mention here that the sawtooth sampling in fact creates an amplitude object. However, every binary phase object can be composed by two amplitude objects with inverted opaque and transparent sections and of opposite phase. Therefore PWM is applicable for phase binary objects with two phase levels only. In Fig. 3 we see the diffraction pattern of the PWM designed DOE. The bright spot in the center represents the 0 th order. Outside of the field shown in Fig. 3 lies a ring at the carrier frequency ν C = 5ν 1. It can be removed by spatial filtering. In benign cases one may choose the sampling frequency large enough to get the carrier frequency evanescent Fig. 1: Standard phase quantization technique for DOEs. Fig. 2: Conversion of an analog signal to a PWM-signal by comparison with a sawtooth signal (only a small section of the complete signal is shown). Fig. 3: Diffraction pattern of a radial symmetric phase object designed with pulse width modulated phase profile. [1] A. W. Lohmann, Teaching How to Invent Is It Possible?, Israel Journal of Technology 18 (1980) [2] A. W. Lohmann and D. P. Paris, Binary Fraunhofer Holograms Generated by Computer, Appl. Opt. 5 (1967)

83 Optical microsystem for signal distribution made of plastic M. Bohling, J. Jahns To overcome the problems of electrical interconnection in computer systems, like the limited number of in- and outputs of a single chip and the limited temporal bandwidth, the use of optical interconnection has been suggested. A specific interconnection problem is signal distribution, for example, for clocking purposes. For the practical implementation of optical interconnections, the concept of planar integrated free-space optics (PIFSO) has been investigated [1]. In PIFSO systems, optical elements are located on the surfaces of a substrate (e.g. SiO2, Si or plastic). The light propagates inside the substrate along a folded optical axis. PIFSO-interconnects offer the advantage of planar fabrication just like the use of the third dimension for signal guiding which is in general difficult for electrical wiring. A successful demonstration for this concept is the HOLMS-Project (high-speed opto-electronic memory system) [2]. Usually lithographic techniques are applied to the fabrication of PIFSO systems especially for Si- und SiO2 -substrates. Recently, however, there has been a strong interest in polymer optics. For this class of materials other fabrication techniques are needed for mass fabrication like, for example, injection moulding. For the fabrication of the moulds, high precision diamond turning or diamond milling can be used. These techniques based on ultra precise machine techniques can be also applied directly to fabricate prototypes [3]. A simple PMMA-PIFSO system has been fabricated recently to demonstrate this fabrication approach. It consists of several microprism arrays which were fabricated with the diamond milling technique. These microprism arrays are used in reflection as beam splitters and beam deflectors, respectively (figure 1 a). By means of a suitable configuration of several 1 2 beam splitters and beam deflectors a 1 4-distribution system was realised and tested (figure 1b). Results 1:2 clock distributor subsystem calculated: P= -5.7 dbm measured: P= -5.2 dbm complete 1:4 clock distributor system calculated: P= dbm measured: P= -16 dbm Fig. 1: a) In reflection used micro prism array as 1 2 beam splitter. b) Experimental set up for a 1 4 signal distribution. [1] J. Jahns, A. Huang, Planar integration of free-space optical components, Appl. Opt. 28 (1989) 1602 [2] M. Jarczynski, J. Jahns, Integrated three-dimensional optical multilayer using free-space optics, Appl. Opt. 45 (2006) [3] E. Brinksmeier, O. Riemer, Wirkmechanismen bei der Mikrozerspanung, Mat.-wiss. u. Werkstofftech. 31 (2000)

84 Substrate coupling by alignment structures fabricated in thick photoresist Th. Seiler, M. Jarczynski, J. Jahns An important issue for the implementation of microoptics in smart packages is the assembly procedure. In general, the concept of planar-integrated free-space optics is used at ONT to perform the integration and packaging of microoptics and optoelectronics. We normally use an active visual procedure for the alignment of planar substrates [1]. This procedure requires high precision alignment equipment and is time-expensive. Therefore, it is of interest to use passive mechanical alignment in order to reduce time and processing steps. Here we demonstrate the feasibility to use micromechanical structures made of thick-film photoresist (AR-N ). The structures were designed like a plug and a socket, on opposite substrates. The accuracy of the alignment between the plug and the socket is given by the photolithographic process to less than 1µm. The air gap between plug and socket was designed to be 1, 2, 3 and 5µmm for test purposes. In Fig. 1 a corner of an aligned plug with a tolerance of 1µm to the socket is shown. Substrate coupling was done by hand without any further equipment. To fix both substrates together viscose glue is necessary. The area inside the plug (square ring) is designed to carry the glue volume. By attaching the substrates the glue wets the second substrate and the aligned substrates can be fixed by a UV-curing lamp. Fig. 1: The sketch on the left shows the principle method of substrate assembly that is accomplished by hand without any further equipment. The result is observed with a microscope at a magnification of 100. A section of the image is magnified by a factor of 5, it is shown in the right bottom and top photograph, respectively. The process parameters for the Photoresist AR-N were: 8ml photoresist; spin coating with Gyrset 400rpm 10s and 1000rpm 120s; oven 90 C 4h drying; expose UVsensitive 150mJ/cm2, oven 100 C 1h linking, developed with AR for 50min. A similar approach is used to attach two thin substrates to a metallic spacer. Precise boreholes as sockets and photoresist cylinders as plugs were used. The result is shown in [2]. The above passive alignment procedure leads to a fundamental reduction in complexity. The practical demonstrations have shown that the tolerancing is feasible and the achieved accuracy is adequate. The method is transferable to the assembly of optoelectronic substrates, too. [1] M. Gruber, D. Hagedorn, W. Eckert, Precise and simple optical alignment method for double-sided lithography, Appl. Opt. 40 (2001) [2] M. Jarczynski, M. Bohling, U. Lohmann, J. Jahns, Design and test of a PIFSO system for multichannel communication, this Annual Report.

85 Design and test of a PIFSO system for multichannel communication M. Jarczynski, M. Bohling, U. Lohmann, J. Jahns In a project sponsored by ESA, we demonstrate the use of planar-integrated free-space optics (PIFSO) for signal splitting, fan-out, interconnection and processing. In addition, it is foreseen to combine the PIFSO architecture with surface-normal multiple-quantumwell (MQW) spatial light modulators in proof-of-concept experiments, see figure 1. In this particular case, we chose a design where the light propagates in air between two substrates. This represents a deviation from the usual case, were light propagates inside a substrate. The purpose behind this approach is the fact that we want to demonstrate a system with minimal material dispersion. The specific requirement that results is the need for a mechanical coupling of the two substrates, here denoted as S1 and S2 (Fig. 1). For achieving good coupling between the two substrates S1 and S2 a CNC-fabricated metal ring was used [1]. The achieved alignment tolerance between both substrates) was approximately 18.1µm in the lateral and 5µm in the longitudinal direction. These tolerances seems to be sufficiently small for many purposes. In our case, we were able to test and achieve the desired optical functionality. The test setup is shown in Fig. 2. The optical elements were fabricated as diffractive multilevel phase elements. The optics demonstrates a fanout to six channels and free-space-to-fibre coupling at the output. Fig. 1: The integration approach for beam propagation between two glass substrates in air. The optical elements are fabricated on the inner surfaces of the substrates S1 and S2. The substrates are held in position by a spacer. Fig. 2: Integration of the PIFSO (consisting of substrate S1 (here back side is shown), metal ring and substrate S2 (front side). In the background the MT-connector, in the front the camera objective. [1] Th. Seiler, M. Jarcynski, J. Jahns, Substrate coupling by alignment structures fabricated in thick photoresist, this Annual Report.

86 Retroreflector array as tapped delay-line filter for ultra-short optical pulses A. Sabatyan and J. Jahns Earlier we reported about the Talbot band experiment [1] and its modification by using special microoptical elements [2]. Talbot bands are observed in the spectrum of a broad band light source when optical phase delays are introduced across the illuminating beam of a spectral apparatus like a grating spectrometer. In the classical Talbot band experiment a glass plate of suitable thickness is entered halfway into the beam and generates a sinusoidal modulation of the spectrum, typical for a two-beam interferometer. In [2], we showed that other, more complex, elements lead to a different modulation. For example, we can insert a staircase of glass plates to introduce multiple phase delays. We then observe sharper peaks, typical for a multi-beam interferometer. So far, the Talbot band experiment was carried out with transmissive elements. Here, we consider the case of a specific reflective device, a retroreflector array (RA). The setup is shown in Fig. 1. The beamlets reflected from the RA will be delayed by tilting the RA around the axis indicated in the figure. The tilt can be varied continuously, thus we can vary the time delays between the beamlets. This leads to a variation of the spectral modulation as it is, in fact, observed. Figure 2 shows different situations of the spectral bands for different tilt angles. The device can be useful to implement a tapped delay-line filter for ultrashort optical pulses. As described in [3], the impulse response can be varied by using a simple aperture across the RA. tilted retroreflector array 0 source grating diffraction plane Fig. 1: Talbot band experiment in reflection Fig. 2: Modulated spectra observed in the diffraction plane for different tilt angles of the retroreflector array. [1] H. F. Talbot, An experiment on the interference of light, Phil. Mag. 10 (1837) 364 [2] J. Jahns, A. W. Lohmann, M. Bohling, Talbot bands and temporal processing of optical signals, JEOS-RP 1 (2006) [3] A. Sabatyan, J. Jahns, Retroreflector array as tapped delay-line filter for ultra-short optical pulses, JEOS-RP 1 (2006) 06022

87 Dispersion compensated pulse shaper for fs-pulses H. Knuppertz, M. Bohling, J. Jahns The usage of ultrashort pulses in the femtosecond range for information processing requires appropriate hardware. Optical pulse shaping in the frequency domain was presented already in the late 1960s [1] and later demonstrated by [2]. For practical applications, small and rugged setups are of interest. In [3] the usability of planar integrated free space optics (PIFSO) with diffractive elements was analyzed for fs-pulse shaping. Here, we present the design of a PIFSO-based pulse shaper using mostly reflective surfaces for achieving minimal dispersion. In order to equal propagation delays for all optical rays, we use an 8f setup. Fig. 1 shows the unfolded setup with a tilted optical axis. Note that the outermost rays (solid and dashed lines) have equal paths. This design provides two spectral planes a the positions 2f and 6f for the modulation of the spectrum. reflective grating incident beam spectrum grating lenses modulator grating spectrum mirror mirror f f f f f f f f exit beam reflective grating Fig.1 Fig. 2 Fig. 3 Fig. 1: Unfolded 8f setup for pulse shaping; Fig. 2: Simulation model for the PIFSO-based pulse shaper; Fig. 3: Ray tracing picture for the phase distribution of the output beam. We performed a ray tracing analysis to verify that dispersion is minimal. The simulations are performed for the folded system illustrated in Fig. 2. At the left side four parabolic mirrors are placed. The right surface contains plane mirrors and two gratings. In the focal planes a modulator can be mounted. The ray tracing of Fig. 2 is calculated for the central wavelength, so the rays are focused in a single spot in the focal planes. With the complete spectrum of a short pulse we have there a spatially extended band with all the wavelengths side by side. The simulation results (Fig. 3) indicate a homogeneous phase front for the exit beam. The differences in pathlength for the center ray compared to the outermost rays for the beam with a diameter of 1 mm are less than 2nm. For a wavelength of λ = 850nm this means a phase front distortion clearly less than λ/100, a value that usually must be undercut to call a system aberration free. [1] C. Froehly, B. Colombeau and M. Vampouille, Shaping and analysis of picosecond pulses in: Progress in Optics XX (1983), ed.: E. Wolf, North-Holland, Amsterdam. [2] A. Weiner and A. Kanan, Femtosecond Pulse Shaping for Synthesis, Processing, and Time-to-Space Conversion of Ultrafast Optical Waveforms, IEEE JSTQE 4 (1998) [3] M. Testorf and U. Österberg, Planar-integrated systems for pulse shaping, OSA Techn. Digest Diffractive Optics and Microoptics 10 (1998)

88 Interference of ultra short pulses in a Talbot interferometer H. Knuppertz, J. Jahns We model the propagation of diffracted ultrashort pulses in order to evaluate the temporal impulse response of the Talbot interferometer [1]. The Talbot interferometer consists of two gratings separated by a suitable distance. A simple model for the operation of the interferometer was derived earlier [2], here we want to consider the interference of ultrashort pulses in a Talbot interferometer. For a stationary wave field one observes behind a grating diffraction orders at discrete spatial frequencies ν = 0, 1/p, 2/p,..., 1/λ (λ: wavelength, p: period of the grating). The diffraction order is determined by a discrete wavelength shift induced by every period. Thus the limit for the highest order is given by ν = 1/λ. But is this true for ultrashort pulses with only a few cycles? With the 0 th order the pulse remains unchanged (Fig. 1a), only its intensity is weakened. Direction of pulse front (envelope, k) and wave front (carrier, k) are straight in incident direction. For higher orders this is no more valid, pulse fronts directs furthermore straight on, but wave fronts are tilted with an angle sin(α) = mλ/p where m gives the m-fold shift by one wavelength supplied by one period. Like Fig. 1c illustrates for the 1 st order this makes a pulse interfere with a shifted copy of itself. The shift corresponds to the diffraction order mλ or delay of m/f 0 (f 0 temporal carrier frequency). There is no interference any more if the shift exceeds the puls length. To calculate this fact we may consider the interference equation for intensities for two by τ time shifted pulses E(t): I(t) = E(t) + E(t + τ) 2 (1) In Fig. 2 the graphic solution for an example is provided with normalized spatial frequencies ν norm = ν/ν 1 = νp ( ν 1 : spatial frequency of the first order, p: period of the grating). The normalized spatial frequencies are directly associated with diffraction orders. Significant orders in this example thus exists only up to the 10 th order and as a rule of thumb for the last significant order is valid N tf 0 ( t pulse duration, f 0 : carrier frequency). Δk k G 1 G 2 Δk k a) b) c) [a.u.] [νp] wave front pulse front Fig. 1: Diffraction of a short pulse in a Talbot interferometer, a) 0 th order, b) 1 st order, c) pulse interference Fig. 2: Interference curve (ordinate relative intensieties, abscissa with normalized diffraction orders ν/ν 1 = νp t) for the used pulse to indicate orders with sufficient interference. [1] H. Knuppertz, J. Jahns, Temporal impulse response of the Talbot interferometer, Annual report of 2005 [2] J. Jahns, E. ElJoudi, D. Hagedorn, S. Kinne, Talbot interferometer as a time filter, Optik 112 (2001)

89

90 OPTICAL MICROSYSTEMS (OMS) (Juniorprofessur für Optische Mikrosysteme) Prof. Dr. Matthias Gruber Universitätsstraße 27 / PRG Hagen Tel.: Fax: matthias.gruber@fernuni-hagen.de Secretary (via Chair ONT): Ext. Susanne Oetzel (until 12/2006) 1120 Susanne.Oetzel@fernuni-hagen.de Anna Sobczyk (since 01/2007) 1120 Anna.Sobczyk@fernuni-hagen.de Staff / Students: Wolfgang Hanke (until 9/2006) 1125 Wolfgang.Hanke@fernuni-hagen.de Richard Heming 2354 Richard.Heming@fernuni-hagen.de Udo Vieth (since 11/2006) 1134

91 Bifurcating Bessel-type optical vortex beams generated by a single DOE Matthias Gruber The design of a diffractive optical element (DOE) is tremendously facilitated by the possibility to synthesize a highly complex transmission function from various simple basis functions with known optical effect. This approach was chosen to design a DOE that generates two dark optical needles with well-defined mutual inclination which can be used for 3-D optical alignment of two wafers with finite spacing as indicated in Fig. 1c [1]. "axicon" a) Bessel beam "needle" b) grating 1 2 splitting alignment mark bifurcating "needles" c) observation from top DOE wafer 2 wafer 1 thick solder bump Fig. 1: Principle of the generation of dark optical needles with an axicon and a grating (a,b) and the possible use of such DOEs to perform a 3-D optical alignment of two wafers with finite spacing (c). The DOE is synthesized from three basic optical functions. As indicated in Figs. 1a,b an axicon is combined with a Ronchi-type phase grating to perform (in good approximation) a 1 2 fanout of a finite Bessel beam. In order to generate an optical vortex the axicon contains a polar phase rotor term e iφ in addition to the optical functionalty of a convex glas cone resulting in the binary phase pattern of Fig. 2a. The mask pattern for the DOE (Fig. 2c) is obtaind through a XOR combination with the Ronchi pattern. DOE = Al a) b) c) d) Fig. 2: Binary mask patterns of an axicon with a 2π optical vortex, a Ronchi ruling, and their XOR combination for the bifurcating Bessel beams DOE (a-c) as well as two fabricated components (d). Two such DOEs were fabricated on a glass substrate (Fig. 2d) and illuminated with a HeNe laser. The resulting far-field diffraction pattern is depicted in Fig. 3a. By far the most intensity is accumulated in the two inner rings corresponding to the first grating and axicon diffraction orders. In the near-field behind the DOE these two orders have the expected appearance of Fig. 3b with a central dark spot indicating the optical vortex. axicon orders 3 2 back side reflex 1 a) grating orders b) Fig. 3: Fraunhofer diffraction pattern of the DOEs of Fig. 2d showing various grating and axicon orders (a), characteristics of a higher-order Bessel beam in the near-field of the +1 and 1 grating orders (b). [1] M. Gruber, Optical vortex beams for metrology and optical trapping applications, Proc. ICO Top. Meeting on Optoinformatics/Information Photonics 2006, St. Petersburg, Russia (2006)

92 Radio-frequency plasma generator for atomic emission spectroscopy R. Heming, S. Tombrink, J. Franzke, M. Miclea, M. Gruber Atomic emission spectrometry (AES) is a tool to measure analyte concentrations in an aerosol in the ppb range via the characteristic optical emission spectra of the atomic and molecular constituents. In gas mixtures the fragmentation of the constituents and the atomic excitation can be conveniently accomplished by means of a plasma discharge, usually generated with a DC current source. Experimental setups used for this purpose can be miniaturized, a prominent example is the microhollow cathode (MHC) discharge [1]. In MHC modules the durability of the electrodes is a critical issue; it can be improved by replacing the DC power source with an AC driver. For the MHC module shown in Fig. 1 a specially adapted RF generator that exploits the lumped components of the electrical circuit to act as an intrinsic amplifier was designed and demonstrated. DC power RF generator PDMS chamber gas inlet MHC module He microplasma Fig. 1: Microhollow cathode (MHC) module powered by a RF generator that is capable of sweeping, tracking, and maintaining the maximum value of the delivered RF power. The key concept of this RF amplifier is to work in a tuning mode in order to form a high-q series resonance. The LC tank is formed by the MHC geometry. The basic theory of a tuned RF amplifier is well known [2] and with a modified robust switching scheme it is possible to deliver more than enough RF power to ignite and to maintain a steady-state microplasma. The RF amplifier utilizes several RF matching techniques to transform the low-impedance amplifier output into a relatively high-impedance plasma load [3]. In series resonance a power amplifier with an output in the range of volts is Q-amplified up to 900 Vpp and the amplifier sees a purely resistive load. In theory, the inherent resonant circuit allows one to apply nonlinear power amplifier methodology to construct a 100% efficient device. A tuned series resonant circuit is a low-impedance path for the fundamental frequency while it is suppressing other parts of the spectrum. Using this property it is possible to built a power amplifier from many push-pull switches in parallel. This outlines a class-d switching amplifier configuration depending on CMOS-FET technology. These switches incorporate low on-resistance in parallel while keeping the parasitic capacitance low, because of their small size. A careful balance of these aspects guarantees low power dissipation in the output stage in practice. [1] J. Franzke et al., Sample analysis with miniaturized plasmas, Appl. Spectroscopy 60 (2006) [2] T. H. Lee, Planar Mircowave Engineering, Stanford University Press (2004) [3] A. Niknejad, Electromagnetics for high-speed communication circuits, UCB, EECS 142 suppl, Department of Miniaturisation, ISAS Institute for Analytical Sciences, Dortmund Fachbereich Physik, Martin-Luther-Universität Halle-Wittenberg

93 Design of a micro-plasma generator with a planar-integrated optical collector M. Gruber, R. Heming, J. Franzke Plasma discharges have many different technical applications, ranging from display technology and lighting to material processing and chemical analysis. In the latter field, atomic emission spectrometry (AES) is one of the standard methods to measure analyte concentrations in aerosols with high accuracy. The role of the plasma in AES is two-fold: It is used for the molecular and atomic fragmentation of the constituents in a sample and for the energetic excitation of these frangments. The resulting characteristic opanode hollow cathode classical insulator electrical field plasma discharge hole-type cavity-type MHC Fig. 1:Evolutionfrom the classical hollow cathode plasma generator to the microhollow cathode (MHC) setup in planar technology. microhole analyte gas flow planar substrate thin film electrodes plasma discharge optical imaging fiber pigtail of spectrometer Fig. 2: Excitation of a gaseous analyte with a MHC plasma generator and standard optical coupling to a fiber spectrometer. tical emission spectra are then detected quantitatively with a spectrometer. For an AES analysis of gaseous samples, hollow cathode plasma generators the name of which is derived from their classical shape are particularly well suited because they can be conveniently miniaturized in a way that makes them compatible with planar microfabrication technology (Fig. 1). A MHC plasma generator essentially consists of a planar ceramic insulator with a tiny ( 50 µm) hole and metal thin-film electrodes on both sides. The miniaturization has the beneficial secondary effect that stable plasma discharges can be obtained with comparatively low voltages ( 300 V) at atmospheric pressure [1]. For AES measurements an arrangement where the analyte gas is blown through the hole is used (Fig. 2). An imaging-type optical setup is usually employed to couple the radiation emitted from the plasma into a (fiber-based) spectrometer. microplasma ellipticallyshaped substrate a) cathode focal points ray paths out-coupling y x reflective coating b) plasma electrodes reflective coating ray path... spectrometer deflector elements fiber to... Fig. 3: Top view (a) and cross section (b) of a MHC plasma generator with integrated optical collector. To improve the coupling efficiency and to obtain a more compact and rugged setup we propose [2] the modified systems approach shown in Fig. 3 where the MHC plasma generator is combined with a planar-integrated optical collector on the same optically transparent substrate. The elliptically-shaped collector adopts the whispering gallery principle to concentrate (almost) all light emitted from the plasma in the second focus where it can be out-coupled into the attached fiber pigtail of the spectrometer by (refractive or diffractive) deflector elements. [1] See also the article Radio-frequency plasma generator... in this Annual Report. [2] J. Franzke, M. Gruber, R. Heming, German patent application pending. Department of Miniaturisation, ISAS Institute for Analytical Sciences, Dortmund z x

94 Halftone-based grey-scale lithography with optimized mask patterns M. Gruber, W. Hanke, J. Jahns Analog lithography followed by reactive ion etching is one of the approaches to realize microoptical elements with continuous surface profiles. The smooth continuous photoresist preform of the desired optical components or systems can thereby be obtained in two different ways: By exposing a (typically several tens of µm) thick photoresist layer through a grey-scale mask, based, for example, on the well-known HEBS glass technique [1], or by use of a halftone mask [2] in combination with an optical method to implement a low-pass filtering operation that suppresses the transfer of the high raster frequency of the binary mask into the photoresist during exposure. A simple way to achieve this effect is an exposure with a large distance of the order of tens or hundreds of µm between mask and photoresist (Fig. 1a). The interdependence and optimization of the particular raster patterns and the lithographic process parameters of this approach was investigated. mask x L R Fresnel propagation photoresist a) I(x) b) Fig. 1: a) Generation of a grey-scale intensity distribution by means of spatial low-pass filtering. The filtering step was implemented here simply by using the Fresnel propagation behind the binary mask; b) two tiling methods (3 3 cells shown) used for experiments to generate a linear grey wedge. We applied a halftone scheme that divides the mask into a cartesian grid of cells (size 4.8 µm 4.8 µm) each of which is further subdivided into binary pixels. The desired grey-level distribution is coded in the average transmittance of each cell and controled by the ratio of opaque and transparent pixels. The design freedom lies in their spatial arrangement; various tiling methods as illustrated in Fig. 1b were investigated in lithographic exposure experiments with linear grey-wedges. Comparison criteria were the achievable linearity of the process, the smoothness of photoresist profiles, the quality of edges, and the exposure distance. The best results (see Fig. 2) were obtained with the tiling schemes of Fig. 1b at an exposure distance of 1.3 mm [3]. Interestingly, the two photoresist wedges are laterally displaced with respect to each other despite identical transmittance values of cells at the same lateral position. Further investigations are necessary to explain this phenomenon and to linearize the analog lithography process. L L R R Fig. 2: Surface profiles in photoresist obtaind from two linear grey-scale wedges of equal transmittance along the vertical direction but composed with the two tiling methods L and R of Fig. 1b, respectively. [1] W. Däschner et al., Appl. Opt. 36 (1997) [2] K. Reimer et al., Proc. SPIE 3226 (1997) 2. [3] W. Hanke, J. Jahns, M. Gruber, Proc. DGaO 2006, p19.pdf.

95 Planar-integrated microfluidic pump based on electro-acoustic actuation M. Gruber, T. Seiler, R. Heming, E.-M. Schnäker A microfluidic pump can be conveniently realized in planar technology by means of the piezoelectric effect of materials like LiNbO 3 [1]. There are certain crystal orientations (e.g. 128 y-cut LiNbO 3 ) for which a HF voltage applied to interdigitally-shaped thin-film electrodes on the planar wafer will excite surface acoustic waves (SAW) of the Rayleightype with both a transverse and a longitudinal elongation similar to that of water waves. U O (10 8 Hz) interdigital thin-film electrodes O (10 µm) λ LiNbO 3 crystal Rayleigh-type SAW z y x O (nm) induced BAW propagating SAW momentum transfer H 2 O LiNbO 3 attenuation Fig. 1: Physical principle of microfluidic actuation by means of surface acoustic waves (SAW) that are excited on a piezoelectric crystal with an electro-acoustic interdigital transducer (IDT). In contact with a liquid the SAW is strongly attenuated and induces a bulk acoustic wave (BAW). The BAW propagation through the liquid is accompanied by a momentum transfer which can be used for actuation (acoustic streaming effect). We have designed and fabricated such a microfluidic pump for a test chamber to be used for microbiological investigations of living cells in an aqueous solution under laminar flow. Fig. 2 shows several tested interdigital transducers (IDT) and the test chamber. The aqueous solution circulates in a ring-like channel patterned in a polymer material (PDMS). IDT LiNbO 3 wafer PDMS flow channel HF input LiNbO 3 Fig. 2: Several fabricated IDTs and the experimental test chamber. The operability of the setup was verified with spherical silicone beads ( 10 µm) immersed in water (Fig. 3). Streaming velocities of the order of mm/s could be demonstrated. IDT flow direction beads Fig. 3: Experimental test of the IDT-based pump. The frames were recorded with the microscope of Fig. 2 and show that the silicone beads (note the encircled pair) move downward with the actuated liquid. [1] A. Wixforth, Acoustically driven planar microfluidics, Superlattices and Microstructures 33 (2004) Klinik und Poliklinik für Hautkrankheiten, Universitätsklinikum Münster

96 Control Systems Engineering Prof. Dr.-Ing. Helmut Hoyer Mechatronics Prof. Dr.-Ing. Michael Gerke Universitätsstraße Hagen Phone: +49-(0) Fax: +49-(0) Staff: Phone: Dr.-Ing. Andreas Bischoff Dr.-Ing. Ulrich Borgolte Wolfgang Gülland Gabriele Kappenstein Dipl.-Ing. Dimitrios Lepentsiotis Dipl.-Ing. Ivan Masár Josef Tschuden Dipl.-Inform. Bernd Winnige -1108

97 Actuation and Sensorics for a Robotic Airship C. Föcking, I. Masár and M. Gerke The blimp-type robotic airship of our research group has undergone some major constructive changes and technical improvements, leading to a new vectored thrust system, a symmetrical stern propulsion, and a more efficient and light-weighted technical equipment. These changes increase the airship s manoeuvrability in restricted in-door environments. Furthermore, a sensor cube has been developed by our research group, which includes 3Daccelerometers and 3D-gyroscopes. These sensorics act as an internal low-cost Inertial Navigation System (INS) in order to estimate the airship s position and orientation with the on-board computer system. In-flight tests and performance analysis show sufficient 3D data resolution to provide the necessary information for the blimp controllers to handle disturbances caused by airflow and even to build a rudimentary dead-reckoning for small positional changes. Figure 1: The airship s main thrust system Figure 2: Sensor cube for inertial navigation A Fuzzy flight controller has been derived in order to establish a basic control loop for stationary hovering above the ground. The Fuzzy controller works fine for certain disturbances, but it is not sufficient for other types of deviations from the hovering position. This is mainly due to the fact, that airships are under-actuated systems, which need more complex menoeuvres for lateral changes of their position. [ 1 ] Christoph Föcking: Aufbau eines Roboter-Luftschiffes auf der Basis eines digitalen Flugreglers und unter Benutzung bordeigener Beschleunigungssensorik Diploma Thesis 2006, FernUniversität in Hagen (in German) [ 2 ] M. Gerke, I. Masár and C. Föcking: Control of an Unmanned Blimp with Acceleration Sensors International Symposium on Robotics / Robotik 2006, Munich, Germany, May 2006

98 Dynamic Modeling of a Robotic Airship S. Schnabel and M. Gerke A basic flight control for a robotic airship can be set up with a Fuzzy rulebase for simple types of motion (e.g. stationary hovering with disturbances). However, the derivation of a complete mathematical model of the blimp dynamics is mandatory for the control of complex flight missions and for path planning. In general, the non-linear dynamic model of an airship considers its geometric shape, its aerostatical and aerodynamical behavior, and its actuation system: F dx dt = C(x)+A(x)+G(λ, W, B)+P Here, vector x(t) represents the 6 1 velocity vector of the flight system with respect to any invariant frame system (e.g. fixed on earth), which is connected to the local airship frame via transformation λ. C(x) is a force vector considering Coriolis effects and centrifugal forces, while A(x) is related to all aerodynamical forces effective on the airship. Vector G(λ, W, B) includes all external physical forces, which are caused by gravity and buoyance. Finally, vector P represents the influence of thrust forces and momentum, which are produced by the actuation system of the blimp itself. Figure 1: Airship s longitudinal section Figure 2: Airship s transverse section While we are still working to derive the complete dynamical model of our small in-door airship, some basic principles can already be observed and understood from this first model. This will facilitate the development of appropriate controllers in the near future. [ 1 ] S.B. Varella Gomes and J. Jr. G. Ramos: Airship Dynamic Modeling for Autonomous Operation Proc IEEE Intern. Conf. on Robotics and Automation Leuven, Belgium, 1998 [ 2 ] Stefan Schnabel: Herleitung eines dynamischen Modells für ein Luftschiff unter Verwendung von MATLAB/Simulink Diploma Thesis 2006, FernUniversität in Hagen (in German)

99 NAMUR Award 2006 Michael Gerke Our student Dipl.-Ing. Marcus Nohr has received the NAMUR Award 2006 for his outstanding Diploma thesis in the field of process control. The topic of his thesis is on Modelling and Observation of Semi- Batch Reactors. NAMUR is an international user association of automation technology in process industries, which represents approximately 15,000 experts in process control from more than 100 member companies in the chemical and pharmaceutical industry, including global players such as BASF AG, Bayer-Schering Pharma AG, BP, Degussa, Merck, and Pfizer. In his laudatio, Dr. Kuschnerus (president) stressed the highly demanding scientific character of his Diploma thesis with respect to actual and practically relevant problems in the field of intelligent process control. Our research group congratulates Mr. Nohr for receiving this honorable prize. [ 1 ] M. Nohr: Zur Modellierung des Kühlsystems für die Beobachtung von Semi-Batch-Reaktoren im Produktionsmaßstab Diploma Thesis 2005, FernUniversität in Hagen (in German)

100 Stepping-Motor Control for Mobile Infusion Pumps U. Fechner and M. Gerke This work describes the optimization of an existing stepping-motor control for mobile infusion pumps, which are used for patients in pain therapy. Based on the original hardware of a German manufacturer in the field of medical equipment, an advanced concept was made and then realized as a technical prototype. This prototype concept considers processor redundancy using AVR microcontrollers. These processors are especially suitable for controller applications because of their PWM interface, on-chip debugging features and their low quiescent current consumption. Furthermore they can be operated with different (e.g. open-source) compilers and RTOS. Beneath the intrinsic controller tasks carried out on each processor, the software architecture includes mutual process supervision. Pump motor actuation is optimized via variable chopping efficiency, adjustable speed, and steppingincrements. Persistent increment counters and measurement of motor current enhance the reliability and operational safety of the medical device. Hardware implementation proves a decrease of energy consumption by the factor 2-3 and a 50 % reduction of peak-currents. At the same time, the motor torques could be increased by a factor of 2-3. Figure 1: Comparison of peak currents in the process of time (left=old, right=new) Furthermore, the prototype promises a cost reduction for the device, which makes it very likely that our concept is included into the next-generation devices. [ 1 ] Uwe Fechner: Optimierung einer sicherheitskritischen Schrittmotorsteuerung für mobile Infusionspumpen Bachelor Thesis 2006, FernUniversität in Hagen (in German)

101 Self-Localization of Mobile Robots Using Visual Landmarks and Extended Kalman Filter Torsten Bucco, Ivan Masár One of the most important problems to be solved in mobile robotics is the self-localization of the robots in their environment. In the past years, several methods were developed for this purpose. The most famous system for absolute global localization is GPS. However, this system is not usable for indoor mobile robots, because of the absence of satellite signals inside buildings. Therefore, the self-localization methods for indoor navigation are mostly based on dead reckoning (DR) algorithms. The DR uses information from internal sensors of the robot, like incremental encoders or tachometer generators. These sensors measure the actual speeds and steering angles of the robot wheels, which are used for computation of the relative robot position with respect to its initial location. However, the measurement error leads to accumulation of position errors. Therefore, the estimated robot position must be corrected by use of some additional sensor information. One possibility for improvement of the position estimation is using landmarks, that can be active or passive. Moreover, the measurement errors can be estimated using Extended Kalman Filter (EKF) and thereby the overall performance of the localization method can be improved considerably [2]. To verify the proposed methods with a real robot, we use our mobile robot F.A.A.K. Because this robot is equipped with an on-board vision system [1], it can detect the landmarks as those suggested by ARToolkit [3]. Exhausting simulations approved the developed concept of the robot localization, as the resulting position errors with EKF and landmarks were times smaller then the errors of DR. x, y, Φ x, y, Φ Dead reckoning algorithm x 10 3 EKF with landmarks Traveled distance x [m] y [m] Φ [rad] x [m] y [m] Φ [rad] Figure 1: Comparision of the estimated position error by using dead reckoning and EKF with landmarks In the future work, the self-localization will be improved by tracking of several landmarks at the same time. Moreover, the robot should be able to define and recognize the easy-to-detect landmarks in its surroundings by itself. [1] I. Masár: An Image Processing System for Visual Servoing of Mobile Robots and Object Tracking, The European DSP Education and Research Symposium EDERS-2006, Munich, Germany, April 2006 [2] T. Bucco: Navigation und Kartenbildung für einen mobilen Roboter Diploma thesis, Hagen, September 2006 (in German) [3] ARToolKit Plus hompage: ar/artoolkitplus.php

102 On the Efficiency of Online Seminars U. Borgolte and A. Bischoff On the background of a number of online seminars run by the Control Systems Engineering Group, the efficiency of approaches for synchronous, online, and interactive collaboration in open and distance learning (ODL) environments was evaluated. Based on interviews with participants of online seminars, the assets and drawbacks of online seminars in relation to conventional face-to-face seminars were discussed. Traditionally, higher education is based on synchronous communication, on real-time discussions of a certain topic. On the other hand, conventional elearning is based on written courses. Communication in ODL is mainly asynchronous ( , newsgroups), or bilateral synchronous (telephone). Several attempts have been made to integrate multilateral synchronous communication in ODL. Most of them have the disadvantage of expensive equipment and high bandwidth requirements. In certain cases, there are some good reasons for using online seminars instead of face-to-face seminars. Sometimes, participants are not able to travel to a common meeting place. This may be due to personal reasons (e.g. disabilities, care for family members, imprisonment), or to vocational reasons (no time for travel). In the long run, online seminars can be successful only if communication costs can be kept lower than expenses for travel. The availability of high bandwidth connections is a problem in some rural environments. A second dimension to keep in mind is that of stability and reliability of the technical seminar environment. If these can not be guaranteed to a high degree, the acceptance of online seminars is low. The last crucial factor is efficiency of the whole solution. This does not only cover technical aspects, but also the learning success with respect to the subject matter. An online seminar system was implemented as a prototype in It is now run on a regular base. A server at the university is acting as main hub. It hosts an audio reflector and a virtual reality server. In addition, this server hosts the shared applications (Acrobat and PowerPoint) which are remotely controlled by the participants. Though most users in Germany are now connected via DSL, some of them are still using modem-line connections only. The system works very reliable and offers high quality communication to all participants. This results are based on voluntary feedback of participants of online seminars. Though not all of them filled in the questionnaire, all were asked in an informal way immediately after the event about their view on it. The result of both the questionnaires and the interviews was that in general, the participants do like this method of communication. Hardware is not the main issue in online seminars. Broadband internet connections are desirable, but are no indispensable condition for a successful event. A thorough technical introduction is much more important. Even if users with modem connections are involved, it is possible to get a sufficient throughput if presentations are restricted to the essence. Nevertheless, it is necessary to put further effort in data compression and parallel distribution of content to a number of participants with heterogeneous internet connections. Especially if students need a great effort to convene at a certain time and location, online seminars are efficient ways of learning. [1] On the Efficiency of Online Seminars, 20th Asian Association of Open Universities Annual Conference (AAOU 06), Kunming, China 2006, CD-ROM.

103 Pediaphon - a Speech Interface to the free Wikipedia Encyclopedia Andreas Bischoff, bischoff/ The ubiquitous availability of mobile communication devices which are connected to the Internet, makes it possible to use small amounts of spare time for mobile learning (m-learning). Travel and latency times can be used for the so called microlearning. The term microlearning describes a new e-learning paradigm with small or very small and short learning units. Learning material can be based on web logs (blogs) and social bookmarking. The main reasons which limit the usage of m-learning services for the end-user are usability problems, mainly the limited screen size and input facilities of highly mobile devices like smartphones and PDAs. Figure 1: Pediaphon audio interface As an alternative to display large text documents on very small displays, audio based learning material can be a solution for hand held devices. The usage of audio based learning material in distance education is state of the art since the seventies. Nevertheless the production of audio learning material is expensive and time-consuming. As an alternative approach, automatically generated audio material can replace time-consuming audio reproduction. Despite the fact that the quality of text-to-speech generation is not perfect for m- and e-learning purposes, it is still usable for rapid prototyping of learning material. Especially to generate an audio representation of a text, dynamically text-to-speech conversion is the only solution. The growing amount of high quality articles available via the online encyclopedia Wikipedia [1] is very suitable as dynamic content for microlearning purposes. Pediaphon is a web based service which generates audio representations of Wikipedia articles dynamically. The tool is usable on- and off-line, as web based service to listen the articles directly in the web browser as well as to download MP3 files for later use in mobile devices like MP3-players and mobile phones. The realization of Pediaphon combines different techniques like text to speech audio generation [3][4], on the fly MP3 compression, Meta file generation for Winamp and Windows media player, Podcast generation, and Pseudo streaming (progressive download). [1] [2] bischoff/radiopedia/index_en. html [3] [4] HADIFIXforMBROLA.html [5] A.Bischoff: Podcast based m-learning with Pediaphon - A Web based Text-to-Speech Interface for the free Wikipedia Encyclopedia, Virtual University 2006, Bratislava 2006

104 A Mobile Phone Interface to the Pediaphon Service Andreas Bischoff, bischoff/ For mobile phone access we have established an WAP interface to the Pediaphon [1] service, but the communication costs are still a limiting factor for mobile users. The WAP interface to the Pediaphon service provides a WML 1.1 input mask as well as a WML server response. The response includes a link to the corresponding article at the Wapedia WAP service (a WAP representation of all Wikipedia articles) as well as a link to the generated MP3 audio file. Downloading the usually huge MP3 files can be very expensive for users without a GPRS or UMTS flat fee contract. To avoid those communication costs and the limitation to users of the mobile internet we have created a pure SMS and GSM Figure 1: WAP interface of the Pediaphon speech based service, too. No (probably expensive) GPRS or UMTS mobile Internet ac- output of Wikipedia articles and 2D barcode representation of the corresponding URL cess is required. The service works with all mobile phones today without the need of additional configuration or expert knowledge. The costs are transparent for the users, only communications fees for an SMS request and a land line call have to be paid. Since the users are paying directly to the communications service provider, the Pediaphon service can be established without any costs for the university. The very successful Pediaphon GSM interface [2][3][4][5][6] was realized by usage of voice over IP technology (VoIP) and the Asterisk[7] open source PBX (private branch exchange). [1] bischoff/radiopedia/sms.html [2] [3] Handy-Nutzern_Lexikon-Artikel_vor.html [4] [5] [6] [7]

105 Mobile Learning with PDA s and Mobile Phones Andreas Bischoff, bischoff/ Mobile devices such as notebooks and PDA s are very interesting tools for webbased teaching and distant teaching today. By the development of wireless communication networks like portable mobile phone networks (GSM, GPRS, UMTS) and local wireless networks (WLAN), electronic teaching material can be accessed from any location. Web based remote laboratory environments have been adapted to mobile devices like PDAs and smartphones [1]. Since not all implementations of PocketPC are supporting an ActiveX control integration to a web page and ActiveX is not available at all on other platforms than Windows, an alternative solution was chosen. On typical PDAs the Java virtual machine (JVM) is not fast enough to decode MPEG4 in real time. But with the help of the open source kuttpech applet [2], it is possible to decode up to 3 frames per second of a MPEG1 video stream on a PDA. For Linux Figure 1: Pure Java PDA interface to online experiments based PDA s, the VideoLan player is an alternative to display MPEG4 content. Personaljava is a Java runtime environment for mobile devices with limited resources. The newer and better known J2ME standard is divided in different profiles and configurations. The Mobile Information Device Profile (MIDP) in combination with the Connected Limited Device Configuration (CLDC) is implemented on most of todays mobile phones. The J2ME (CLDC) application environment (MIDlets) does not allow a seamless integration of streaming clients with an MIDlet user interface (GUI). Different to J2ME (CLDC) the J2ME Connected Device Configuration (CDC) and its predecessor Personaljava are providing the full AWT (Abstract Window Toolkit) API. Like on standard PCs Personaljava allows an integration of a Java virtual machine into the web browser of a mobile device. Implementations of Personaljava are available for Windows CE, Linux, Palm and Symbian OS based PDA and mobile phone platforms. With the help of the described techniques an existing online experiment (remote control of a mobile robot) has been adapted to a pure Java PDA and smartphone environment (Fig. 1). The modified client Java applets are running on a Personaljava virtual machine. [1] [2] [3] A.Bischoff: M-Learning with Remotely Operated Laboratories, Virtual University 2006, Bratislava 2006

106 Podcasts and RSS Feeds in Teaching Andreas Bischoff, bischoff/ To provide students instantly with information about updates in asynchronous changeable media like wikis and blogs RRS feeds are a suitable solution. We are using a Wiki based environment to prepare our online seminars. For announcements of events like online tests or meetings Wiki based tools are inconvenient for learners and teachers. To avoid extra effort, an automatically announcement of all user was needed. We have integrated a RRS (Really Simple Syndication) feed to our Wiki [1]. RSS feeds for notification were also added to the news page on the PRT web site [2] and the news page of the university [3]. A RSS file is a computer readable XML based description of any changeable web content. Especially for highly dynamic changeable news sites, a RSS feed becomes an interface to poll the changes automatically with a so called feedreader. Figure 1: Podcast based teaching material, Reactive robot programming Audio based teaching material has a long tradition in distant teaching. The upcoming Podcasting euphoria reestablish the focus on audio based material. The podcast technology, which is basically an extension of existing RSS techniques for deployment of audio (and video) data makes the distribution of audio based teaching material very convenient for the users. The word Podcast was derived from ipod and broadcast. In a technical sense a Podcast is nothing more than a MP3 file and a XML based RSS-Newsfeed meta description. But for users, with no or little knowledge about file formats and meta data, a Podcast is a very convenient way to synchronize their audio data to their mobile devices or to create an own Podcast. Nevertheless the production of audio based teaching material is time consuming and expensive. Since the teaching material at University of Hagen will be generated out of a XML based source format (FUXML) in the future, an automatically conversion to audio would be helpful. To evaluate these techniques in teaching, text based course material was automatic converted to speech by the use of text to speech (TTS) technology. Since math containing material is not very suitable to this task a more information technology based course (a manual for a robotics practice) was chosen as an example course to establish a Podcast service in PRT s teaching [4]. [1] bischoff/feed/ onlineseminarws xml [2] bischoff/feed/prt-aktuell.rss [3] bischoff/feed/fu-aktuell.rss [4] Reactive robot programming, M.Sc. courseware