THE DJOSER THERMAL SIMULATOR FOR ELECTRONIC DEVICES AND SYSTEMS: DIDACTIC FEATURES FOR IMPROVING RELIABILITY CULTURE IN ACADEMIC AND INDUSTRIAL ENVIRONMENTS. Paolo Emilio BAGNOLI (1), Fabio STEFANI (1), Maurizio BECCAI (2) (1) Dpt. of Information Engineering. University of Pisa, 56122 Pisa, Italy , p.bagnoli@iet.unipi.it, fabio.stefani@iet.unipi.it (2) EFFEI srl, Via Aldo Moro 11,13, 50019-Sesto Fiorentino, Firenze, Italy, info@effei.it Abstract The structure and the technical and didactic characteristics of a novel thermal simulation system (DJOSER) for electronic device and system mounting structures and based on an analytical resolution strategy, are shown and discussed by means of several applicative examples having large importance in the optimum reliability-oriented design of electronic power handling. Keywords Temperature, Simulation, Electronics, Power Devices, Packaging, Thermal Analysis, Didactic. 1. INTRODUCTION The thermal analysis of electronic devices and circuits and their packaging structures is becoming a crucial need in electronic design and therefore a cultural luggage for the electric and electronic engineers. This type of analysis is usually performed using the finite-element-method based (FEM) thermal simulation, which is a general purpose, widely used and reliable tool also for complex threedimensional structures. As known, it is based on the discretization of the thermal conduction equations within a close net of solid elements. Therefore the FEM analysis needs a 3-D meshing of the solid system, which is in general, not uniform, so that the model preparation for optimum results may be complex and overall requires some qualified capabilities by the operator. Since the thermal analysis must be performed not only by the device producer but also by the system assembler user which can belong to small or medium size industries, the current situation is that in these last environment there are not enough financial support to acquire the knowledge and the tools to perform it, so that the thermal problems are usually solved by adopting the strategy of over-sizing devices and systems with consequent dimension and cost penalties. Instead the importance of these reliability problems would require less expensive, less time consuming and more easily programmable thermal simulation tools in order to perform fast analyses with the same accuracy of FEM programs. Most of electronic assemblies (both devices and boards), especially for high and very high power handling, can be described by simple structures consisting of stacks of homogeneous layers of different materials and dimensions, connected directly or by means of thin soldering or attaching Fig.1. (A) Assembling structure composed by a stack of homogeneous rectangular layers. (B) The two-dimensional regular grids of cells at the interfaces of the structure. layers. The power dissipation may be assumed as two-dimensional on one or more interfaces and with the bottom side generally connected to a constant temperature heat sink, directly or through a thermalconvection coefficient. These structures are quite similar to a stepped pyramid [1], eventually with various degree of asymmetry as shown in the figure 1A. If the above simple geometrical configuration is suitable to represent the whole assembling structure, the steady-state thermal simulations may be performed using dedicated mathematical models with explicit analytical relationships [2] which can be implemented in fast and easily programmable simulation tools characterized by a shorter computing time than FEM and an almost automatic preprocessing procedure. In this communication, an analytical thermal simulation tool for the steady-state of electronic systems having the above cited benefits is presented. In the following it will be addressed as DJOSER model [3,4]. It can be used for planar electronic systems and it is able to well represent a wide range of electronic assembling configurations such as integrated circuits encapsulated in power packages, flipchip or BGA (Ball-grid-array) mounted devices, hybrid circuits, multi ‘naked’ chip assemblies, printed electronic boards assembled in surface mounting technology or on multi-layer metal interconnecting substrates and hybrid thick film technology. The theoretical and experimental validation activities performed in the past [5,6] demonstrated that the accuracy of the results is within 1% with respect to both more sophisticated FEM analyses and experimental measurements. Just because of its advantages this simulation program is particularly suitable to perform didactic activities within the university courses devoted to electronic system reliability and, in the industrial environment, to help the optimized design of electronic packaging and mounting. 2. THE DJOSER SIMULATION PROGRAM 2.1 Mathematical background Fig. 1A shows a typical assembling structure of a power electronic device which is composed by a stack of any number of rectangular-shaped homogeneous layers of different materials with different thermal conductivities, similar to a step pyramid, eventually also asymmetric. The thermal problem, which is the calculation of the temperature and heat flux distributions at all the interfaces between the layers, is based on the resolution of the single homogeneous slab. Here the known thermal functions are the bottom temperature T(x,y,z=Lz), the thermal flux q(x,y,z=0) entering from the top surface and the two-dimensional power sources P(x,y) and P*(x,y) located at the top and bottom surfaces respectively, while the unknown variables are the top temperature T(x,y,z=0) and the heat flux distribution q(x,y, z=Lz) at the bottom surface of the slab. The single slab solution can be obtained by means of the variable separations and superposition techniques, so that the unknown variables are expressed by the two following integral relationships: Lx Ly Lx Ly T ( x, y,0) = ∫ ∫ [ P( x′, y′) + q( x′, y′,0)] ⋅ G11 ( x′, y′ | x, y)dx′dy′ + ∫ ∫ [ R * ⋅P *( x′, y′) + T ( x′, y′, Lz )] ⋅ G12 ( x′, y′ | x, y)dx′dy′ 0 0 0 0 Lx Ly Lx Ly q( x, y, L ) = ′ ′ ′ ′ ′ ′ ′ ′ P ( x , y ) + q ( x , y ,0) ⋅ G ( x , y | x , y ) dx dy + [ ] z 21 ∫0 ∫0 ∫0 ∫0 [ R *⋅P *( x′, y′) + T (x′, y′, Lz)] ⋅ G22 ( x′, y′ | x, y)dx′dy′ (1) where R* is an eventual thermal contact resistance per unit area located at the bottom surface of the single slab. The spatial functions Gij(x’,y’|x,y) are Green functions consisting of double harmonic cosine Fourier series. Using the coordinates system shown in the figure 1B, the Gij functions are just cosine-like and may be defined as follows: ∞ ∞ Gij ( x′, y′ | x, y ) = ∑ ∑ Cij ( n, m ) cos ( β n x′) cos( µ m y ′) cos( β n x) cos( µ m y ) n =0 m =0 (2) where Cij(n,m) are suitable coefficients and the eigenvalues βn and µm are given by: βn = nπ / Lx , µm = mπ / Ly with n , m integer (3) in the adiabatic case, otherwise they are calculated from the zeros of a transcendent equation containing the top and lateral heat exchange convection coefficients. If we turn back to the case of step pyramid, it is quite evident that for a single slab of the stack the functions q(x,y,0) and T(x,y,Lz) are not really known but they are the thermal flux coming from the upper layer and the temperature distribution of the contact area with the lower layer respectively. The temperature T(x,y,Lz) at the bottom of the lowest layer is known and corresponds to the heat sink temperature. Therefore the set of relations (1) for all the layers belong to a system with 2(Ns-1) integral equations and where the 2(Ns-1) unknown variables are the functions q and T, Ns being the number of slabs belonging to the whole structure. This is the mathematical core of the DJOSER model. However the real implementation of the method and the knowledge of the q and T functions require the use of suitable numerical calculations and approximations. Firstly, the Green functions, given by equation (2), must be evaluated for a finite number of eigenvalues, being Nnm the maximum number of the indices n and m used for the calculations of the Fourier series. Secondly, the double integrals in equations (1) require the use of suitable formulas for 2-D numerical integration and therefore the heat flux and temperature functions can be known in the form of a grid of sample values. This implies that all the interfaces of the structure must be divided into a regular twodimensional grid of cells as shown in the figure 1B: each flux or temperature numerical value ( q(xi,yj) and T(xi,yj) ) in each layer corresponds or is relative to the centre of a single cell on a given interface defined by the relative coordinates (xi,yj). Therefore the system of 2(Ns-1) integral equations can be transformed into a simple algebraic linear system. In comparison with the FEM calculation strategy which uses irregular meshes of 3-D elements, the accuracy depending on both the shape and density of the elements, the DJOSER program uses only regular two-dimensional grids of rectangular cells at the interfaces, dramatically decreasing the number of variables and therefore the calculation time and memory occupation. Note that, beside the mathematical background of the DJOSER may be complex, the human operator is not compelled to have knowledge of it, but its pre-processing action is only devoted to define the geometrical characteristics of the model. Fig.2. Details and boundary conditions of the DJOSER models: (A) reference temperatures; (B) two types of power loads; (C) contact thermal resistances; (D) convection coefficients for each face; (E) passivating coatings; (F) bottom convection for the external heat spreader design. 2.2 Model builder characteristics The figure 2 schematically shows all the characteristics and boundary conditions which can be inserted into the DJOSER model using its pre-processing unit and which make the model itself able to realistically represent a lot of device mounting configurations. A) Reference temperatures. There are two different reference temperatures Ta and To (figure 2A). Ta is the temperature of the external environment surrounding the lateral faces of each layer and the top ones not covered by an upper layer. To is the temperature of a heat sink placed in contact with the bottom face of the lowest layer. Note that, while Ta has a unique uniform value, To can have a not uniform distribution on the assigned plane. B) Power loads. The assigned electrical power loads (figure 2B), i.e. the surfaces corresponding to the resistances or the active devices acting as heat sources are supposed to be localized at the interfaces of the structure (eventually also on the top surfaces not covered by an upper layer) and with any two-dimensional distributions. The power distributions may be organized in separated islands each dissipating an assigned power or power density. In each layer two different power sources can be inserted: those on the top surface P(x,y) and those on the bottom surface P*(x,y), which, of course, have a true meaning only if the given layer is thermally separated from the lower one by the presence of a contact thermal resistance. This last type of power source allows to represent also particular device mounting configurations as flip-chip and BGA, in which the silicon chip is mounted top-sidedown and whose active surface is separated from the underlying substrate by a thin polymeric insulating layer (underfilling). C) Thermal contact resistance. The presence of a thermal contact resistance R* (figure 2C) between two adjacent layers or at the bottom surface may be included in the model. This parameter represent the contribution of very thin soldering or gluing layers or the effects of an imperfect contact due to the surface roughness of the two elements of the mounting stack. All these factors, causing a drawback for the conduction heat flux and therefore an abrupt temperature drop within the structure, are generally responsible for the uncertainties of the experimental values of the thermal resistance of real assemblies. D) Convective boundary conditions. The DJOSER model builder allows the setting of convective boundary conditions on all the surfaces exposed to the environment (figure 2D) with the only limitation that the heat exchange coefficient H must be uniform on the whole single surface. Different H coefficients can be set for different layers and different surfaces (top, north, south, west and east) The possibility to take into account also the convection heat exchange is very useful from the real applications since it allows to verify the effectiveness of the internal or external vent in the removal of excess heat and in the decrease of the device maximum temperature. E) Passivating coatings. On all the surfaces exposed to the external environment the presence of an insulating or passivating layer (figure 2E) may be inserted. From the practical point of view this presence is directly translated into a modification (decrease) of the heat exchange coefficient of the corresponding surface, since the insulating layer thermally acts as a thermal resistance per unit area in series to that due to the convection mechanism. F) Bottom convection. The possibility to set the To temperature equal to that of the environment Ta and to place a heat convection coefficient Hb on the bottom surface of the structure allows us to verify the effect of an external dissipater on the device maximum temperature (figure 2F) and therefore to exactly design the proper dissipating structure needed to decrease the device temperature under the safe area limits. The details of this procedure is reported in the application section. 3 APPLICATIONS OF THE DJOSER THERMAL SIMULATOR The DJOSER program was implemented in the MATLAB 7.0 environment which is provided with very useful and powerful graphic interfaces helping to make the data pre-processing and the results visualization easy and friendly from the user point of view. In this section a not-exhaustive set of thermal problems concerning power electronic devices and packaging are shown and discussed. They are presented just to demonstrate the utility of this simulation approach in solving thermal problems involving both the conduction and convection heat exchange mechanisms. Especially thanks to the high speed of the calculation procedures with respect to other simulation strategies this software provides the possibility of performing a large number of thermal analyses and hence exploring a large set of mounting solutions. Furthermore, still thanks to all its benefits, the DJOSER program may also be retained as a powerful, reliable and low-cost didactic tool in both academic and industrial environments in which the design activities include the reliability analysis of the electronic systems. 3.1 Power device assembling thermal characterization The figure 3 shows the cross-section (fig. 3A) of a typical mounting structure for a power electronic device in surface mounting technology having a bottom copper frame. The whole structure is almost asymmetrical, as can be seen in the figure 3B depicting the corresponding DJOSER model. Neglecting as a first approximation the thermal effects due to the presence of the device pins, the whole assembly corresponds in practice to a stack of homogeneous layers, some conducting and some insulating, attached each other by means of solders or gluing pastes. FIG. 3. Cross-section (A) of the sample indicating the layers and their thicknesses, the contact resistances and the silicon power source map. (B) View of the corresponding DJOSER model. FIG. 4. Colour-scaled temperature maps on the top surfaces of plastic, silicon, copper and alumina. Each map has its own color scale. FIG. 5. (A) Temperature distributions of the five layers along the horizontal axis crossing the centre of the silicon chip. (B) Temperature distributions of the top plastic surface under four convection 2 conditions ( H = 8, 50, 200,800 W/m °C). The thicknesses of all the layers, the localization and values of the contact thermal resistances at the interfaces, the two reference temperatures and the power load map on the top of the silicon layer are directly reported in the same figure. All the surfaces of all the layers, except those of the silicon chip and the top surface of the copper frame, are supposed to exchange heat with the environment with a 2 coefficient H = 8 W/m °C. The temperature maps are shown in the figure 4, while the figure 5A shows the temperature distributions along the horizontal axis crossing the silicon centre. These types of figures are some of those available in the DJOSER post-processor for the visualization of the results. The first step of the calculation procedure, devoted to the algebraic matrix construction and solution, required 158 seconds, while that for obtaining the five more resolved temperature maps required 518 seconds on a lap-top computer having a clock frequency of 2.1 GHz. The effects of the different convection coefficients (some of which are unrealistic) were found to be almost week on the silicon top temperature distribution but more evident on those of the top plastic. Instead these effects can be clearly seen in the 2-D temperature distribution plots calculated along the main vertical axis of the plastic layer. The convective heat exchange coefficients were 8, 50, 200 and 2 800 W/m °C. 3.2 Characterization of the insulator thermal conductivity in IMS substrates The use of metal substrates (IMS) [7] for circuit mountings is becoming irreplaceable for all the FIG. 6. (A) Cross-section of the sample used for the IMS insulator layer characterization. (B) Top view of the sample. (C) Results of DJOSER simulations: total thermal resistance vs insulator layer thermal resistivity. The labels indicate the corresponding thermal conductivity values. electronic systems handling high and medium powers or which are located in thermally unfavourable positions such as for automotive electronics [8]. This type of substrate with high heat dissipating capabilities is composed by a thick aluminium slab covered by a thin insulating layer whose thickness and composition are crucial for its thermal, thermo-mechanical and dielectric properties. Several methods are currently used for a reliable deposition of the insulating layer and for keeping its thermal conductivities as high as possible. These techniques include the use of thin silica tissues in a polymeric matrix or the use of fine alumina powder mixed to the polymer material. However the accurate experimental knowledge of the insulating layer specific thermal conductivity is very important not only for the customer, in order to choose the most suitable substrate for his application, but also for the producer which must provide these data and must take the deposition technological process under control. The joined use of an experimental thermal resistance measurement method and of the DJOSER thermal simulations provides an easy and highly accurate technique for the characterization of the specific thermal properties of the thin insulating layer of an IMS substrate, which is independent from the sample geometry. The figure 6A shows the cross-section of a power device in a SO10P package soldered to an IMS substrate and placed on a thermostatic plate. The same system is shown in the top view photograph of figure 6B where the encapsulating plastic of the device was eroded with the aim to highlight the geometry of the inner silicon chip. The sample was kept in vacuum in order to eliminate any heat exchange by convection. The power device provided by ST-Microelectronics is dedicated to package thermal resistance measurements and contains only a power resistor acting as electrical heather and a temperature sensing diode whose sensitivity is about 1.85 mV/°C. The thermal resistance of the device-packagesubstrate system is easily statically measured from the ratio between the temperature drop between the silicon chip and the heat sink and the electrical power dissipated in the heating resistor. Also the thermal resistance of the device itself was experimentally measured. The thermal resistance of the structure depends not only on the specific thermal conductivities of all the layers but also on geometrical factors. It means that this value depends on the spreading of the heat flux within the structure. Note that in the structure of fig. 6A the only unknown parameter is just the thermal conductivity of the thin insulating layer since all the geometrical (thicknesses, lateral sizes) and thermal factors of the standard materials are known. Therefore using the DJOSER program it is possible to build the model (the effects of the device pins are supposed to be negligible) and perform several simulations by changing the thermal conductivity of the insulator layer. The figure 6C shows the result of the DJOSER simulations which is the plot of the maximum device temperature (i.e. the device thermal resistance for 1 W of power dissipation) as a function of the specific thermal resistivity (the corresponding thermal conductivities are reported in the plot as point labels) used in the DJOSER models. The pursued conductivity value can be easily interpolated within this plot at the experimental thermal resistance. However for a best accuracy, the device model was previously modified by adding a proper thermal contact resistance between the silicon chip and the lower copper frame so that the calculated thermal resistance of the model is exactly the same of that obtained from the experiment. Also note that, for this characterization only the maximum temperature in the silicon chip centre is needed so that the simulation procedure is very fast since in this case the time consuming temperature maps calculations are disabled. 3.3 Evaluation of the external heat spreader One of the most important technical need concerning the power electronic device assembling in very large range of applications is the design of a proper heat spreader for cooling the device and keeping the maximum temperature inside the safe range. In particular this need is now becoming crucial for the lighting systems in which the usual traditional lamps are replaced by groups of high efficiency LED devices, thus gaining advantages in terms of compactness and lower energy consumption. However the external cooling apparatus is always needed in applications for public lighting requiring high luminance (especially for tropical countries with higher air temperatures) since the heat dissipation in quiet air is very poor and the lamp cannot be provided with an active electrical refrigerating system; therefore the use of an external metal heat spreader is a mandatory choose. To perform the design of this component some practical formulas are generally provided by the manufactures with which the maximum device temperature can be roughly foreseen. However the use of these formulas does not allow to directly see the effects of the heat spreader presence on the temperature distributions within the inner parts of the packaging structure and the effects of the quality of the thermal connection between the heat spreader and the substrate. The DJOSER software provides the possibility to simulate the presence of a heat spreader on the bottom or top surfaces of the whole assembling structure, taking into account the quality of the contact with the heat spreader (by setting a proper thermal contact resistance at the interface) and in the same time to still obtain the corresponding temperature maps at all the structure interfaces. The figure 7A shows an assembled power device with a metal heat spreader at the bottom surface 2 exchanging heat with quiet air whose coefficient Ho is typically in the range 4-8 W/m °C. Although the geometrical configuration of the heat spreader cannot be represented in the DJOSER model, however its presence can be modelled using the below described strategy. The heat exchange by convection at a surface in contact with a fluid is ruled by the following basic relationship: Q = Ho ⋅η ⋅ As ⋅ (Tj − Ta ) , (4) where Q is the heat flux exchanged, Tj and Ta are the temperature of the surface and the fluid respectively, As is the total exposed surface of the spreader and η is a number lower than unit representing the efficiency of the metal fins. In other words, the dissipating structure is able to decrease Tj by increasing the area of the heat exchange process. However the equation (4) can be rewritten in the following way: Q = H * ⋅ Ao ⋅ (Tj − Ta ) being H * ⋅ Ao = Ho ⋅η ⋅ As , (5) where Ao is the area of the flat interface and H* is an increased value for the convection heat exchange coefficient, eventually also very unrealistic. Therefore the effect of an external heat spreader having an effective area (η As) exchanging heat with a Ho coefficient is equivalent to that of a flat surface having Ao area exchanging heat with the fluid with an increased H* coefficient. The figure 7B FIG. 7. (A) Cross-section of an assembling structure with a bottom thermal dissipator. (B) Results of the DJOSER simulations: plot of the device maximum temperature vs the normalized dissipator area. shows the plot of the maximum temperature calculated by DJOSER within the silicon surface of thedevice, having a small heat source area but high power density, as a function of the ratio (H*/Ho) which is equal to the ratio (η As/Ao) and can be called “normalized dissipator area”. This type of plot allows choosing the proper heat spreader characteristics in order to decrease the device maximum temperature below a given safe value. The DJOSER pre-processor is just provided with a routine for the automatic generation and running of the input files for this analysis and the production of the plot of the figure 7B. 3.4 Multi-pyramidal applications: coupling LEDs with IMS substrates Although the DJOSER program can be applied to simulate single pyramidal structures (a single stack of rectangular layers), many real application cases may be addressed as “multi-pyramidal structures”. It means that, as shown in the figure 8A, many devices are placed on the same multi-layered substrate, all of which dissipate a given power, inject an heat flux on the common substrate and therefore contribute to warm up the substrate and the other surrounding devices. This is the case, for instance, of a lighting system using a matrix of power LEDs (see the figure 8B showing the view of a device manufactured by LUXEON), all mounted on the same substrate. However the thermal analysis of the whole system cannot be performed by taking into account only one device. There are two ways to still use DJOSER overcoming this impediment. The first one is provided by the linear nature of the thermal system and consists of many simulation as the number of the devices on the same substrate, each device located on the substrate on its own position. The temperature map of the common substrate can be obtained by summing all the partial maps due to the various devices. Furthermore, the temperature maps internal to a single device can be obtained by simulating only the device, without the substrate, but imposing on its bottom surface the total temperature distribution on that substrate area covered by the device itself. The solution provided by this method is exact also in presence of convection heat exchange on any surface of the solid system. The second method is more practical and fast but almost approximated. The thermal analysis of the multi-pyramidal system can be carried out as shown in the figure 8C. The structure is composed by only one device, chosen as that located in the worst position from the thermal point of view (in the centre for the present case). The other devices are represented by 2-D uniformly distributed heat source islands located on the substrate at the positions of the other devices and equal to the heat fluxes injected by them. In this way the temperature maps on all the layers of the structure, including those internal to the central device, can be directly obtained with only one simulation run. However the solution is exact if the whole structure or at least the device are adiabatic (does not exchange heat with the ambient). This hypothesis can always be retained valid in the cases of thermally not-resistive contact with the heat sink and in presence of poor values of the convection coefficient as those corresponding to the quiet air. As an application example of this latter strategy, the simulations corresponding to the coupling of seven LEDs to an IMS substrate are reported. The IMS substrates have 1.6 thick aluminium (AL6082) with thermal conductivity 170 W/m°C and 125 microns thick insulating layer with three different values for the thermal conductivity ki : 0.6 , 1.4 and 2.2 W/m°C. The LED device, dissipating a net electrical power of 1 W and considered as adiabatic, was represented by superposition of the semiconductor chip and of three different copper layers just for faithfully reproducing the shape of the internal heat spreader. The substrate was supposed to be in perfect contact with the heat sink (To = 0 °C). The figure 9 shows the results and the comparison among the three simulation runs. The figures 9A and 9B show the temperature colour-scaled maps on the top semiconductor surface (where the maximum temperature is located, about 16.53 °C) and on the top of the insulating layer (Tmax = 2.68 °C) respectively. The two maps are referred to the sample with ki = 1.4 W/m°C. The figure 9C shows the two-dimensional temperature distributions for the three cases and along the medium horizontal axis of the semiconductor chip in the central LED. The comparison among the curves clearly describes the dependence of the device self heating on the thermal conductivity of the IMS insulating layer which is the most crucial factor of the mounting system. However the low maximum temperature values showed in the present example are mainly due to the not-resistive contact with the heat sink. Much higher values could be found if the bottom aluminium surface exchanges heat by convection. FIG. 8. (A) View of the multi-pyramidal structure composed by power LEDs. (B) View of the power LED by LUXEON. (C) Equivalent DJOSER model used for the thermal simulations. FIG. 9. Results of the DJOSER simulations. Colour-scaled temperature maps for the top semiconductor (A) and the insulator (B) (case of ki = 1.4 W/m°C). (C) Medium temperature horizontal distributions on the semiconductor for three different values of the insulator layer thermal conductivity. 4 CONCLUSIONS In the present paper the structure and characteristics of thermal simulation system for electronic assembling structures was briefly presented and discussed. This software tools, which in the past was presented to the scientific community, is mainly based on the analytical resolution of the basic equations of heat transfer in a geometrical system which well represents most of the mounting configurations of power electronic devices and their usual boundary conditions. It was designed just with the aim to replace or to grant an alternative tool to the programs using the traditional finite-element calculation strategy, while maintaining the same accuracy degree and hence to spread the thermal analysis culture also in the academic and industrial environments not traditionally involved in FEM analyses. The successful realization of this aim was testified by the increased number of applications in which the DJOSER program is currently called to operate, ranging from hybrid technology sensor prototypes analysis [9] and flat thick film heathers for consumer electronics to the design of mounting structures using highly conducive substrates such as IMS, both for automotive and lighting applications. This activity is currently carried out not only by our academic laboratories but also within a close collaboration with an industrial subject (EFFEI s.r.l. of Sesto Fiorentino, Florence, Italy) involved in the electronic board production, assembling and design. This industry uses the software to offer thermal analysis services to their customers, helping them in the thermal design and optimization of their products. As far as the didactic characteristics are concerned, it is very significant that the DJOSER program is currently used as a learning tool within the Laurea degree courses not only in Electronics (Industrial Electronic course) but also in Engineering Management (Applied Electronics course) which is devoted to the preparation of engineers specialized in quality control of industrial processes. Many applicative theses about the thermal analysis performed with the DJOSER program were recently assigned to students belonging just to this second Laurea degree course. References [1] “Djoser” was the Egyptian king of the third dynasty who built the stepped pyramid in Sakkara. [2] P.E. Bagnoli, M. Montesi, C. Casarosa, G. Pasquinelli, Fast analytical thermal modelling of electronic devices and circuits with multi-layer stack mountings. 5th EPTC Conference, Singapore, December 2003. [3] M. Montesi, P.E. Bagnoli, C. Casarosa, G. Pasquinelli, Steady-State thermal mapping of electronic devices with multi-layer stack mountings by analytical relationships, ITSS II ASME–ZSIS Conference, Bled, Slovenia, June 13-16, 2004. [4] P.E. Bagnoli, C.Casarosa, F.Stefani, 2007, “DJOSER: Analytical Thermal Simulator for Multilayer Electronic Structures. Theory and Numerical Implementation, Proceedings of Thermal Issues in Emerging Technologies Conference, ThETA 1, Cairo, Egypt, Jan 3-6th 2007; also published on IEEE Xplore, Digital Object Identifier: 10.1109/THETA.2007.363413, May 2007. [5] P.E.Bagnoli, C.Bartoli, F.Stefani, 2007, Validation of the DJOSER Analytical thermal simulator for electronic power devices and assembling structures, Microelectronics Journal, vol. 38, pp. 185196. [6] P.E.Bagnoli, C.Bartoli, G.Pasquinelli, F.Stefani, DJOSER: Verifica teorica e sperimentale dell’accuratezza del simulatore termico stazionario per l’elettronica di potenza , XXIII Congresso Nazionale U.I.T., vol. 1, pp. 33-38, Parma 2005 (in Italian). [7] M. März, Thermal Management in High-Density Power Converters , International Conference on Industrial Technology ICIT'03 Maribor, Slovenia, December 10 - 12, 2003. [8] S. Di Pascoli, P.E.Bagnoli, C.Casarosa, Thermal Analysis of Insulated Metal Substrates for Automotive Electronics, Microelectronic Journal (30), 1999, 1129-1135. [9] F.Stefani, P.E.Bagnoli, S.Luschi, Study of Water Speed Sensitivity in a Multifunctional Thick-film Sensor by Analytical Thermal Simulations and Experiments, Proceedings of THERMINIC-07 Conference, Budapest, Hungary, September 17-19, 2007.