problems
Content
Specific Problems in Smart Power Design
ASzajfler, T.Pokniak, M.Napieralska, M.Zubert, W.Wbjciak, A.Napieralski Department of Microelectronics and Computer Science Technical University of L6&, Poland Lbcli, ul. Aleja Politechniki 11 tel.: +48 (42) 3 1 26 28 fax: +48 (42) 36 26 28; e-mail: [email protected].
Abstract. In this paper the following problems:
selection of the power device and driving circuit, sensing and protection circuits, transmission of the sensor signals, thermal problems in Smart Power module, simulation problems, of the Smart Power modules design will be presented and discussed. The deliberations will be illustrated with the example of PWM converter circuit designed in Mietec HBiMOS 2pm technology. The new electro-thermal practical model of transistors which can be use to complex electro-thermal simulation of Smart-Power structure is presented.
advantages and disadvantages and the abilities of technology have to be taken into consideration. According to the type of power device, the relevant driving circuit have to be designed. As an example of these problems the output stage of designed Smart Power module P W M for SIT control is presented. The used HBiMOS 2pm technology allows to operate with voltages up to 80-1OOV. Different kind of active and passive elements can be created using this technology: NMOS and PMOS transistors; bipolar transistors, diodes; resistors and capacitors. The following configurations of the output stage presented in (Fig.])are possible.
1. Introduction
The rapid development of semiconductor technology has an impact on the increasing interest of ,,intelligent" power modules. The area of applications and intended functions of Smart Power module determine the technology which will be used for physical realisation of the device [l]. Every technology has specific limitations, which have to be taken into consideration during the design of Smart Power module architecture [2]. Moreover, the designer of the power module should not only take into consideration the separate physical phenomena but also interactions between them. In design process the economic aspects should not also be neglected, so the adequate simulation tools have to be used to minimise the cost of design and prototyping. The following stages in Smart Power module design could be distinguished: selection of power stage configuration and behavioural specification of control part of the module, design of the control part of module, 0 mixed digital-analogue simulation of entire circuit using SPICE-like simulators, 0 design of the layout of Smart Power module, the first extraction the parameters from layout and multidomain simulations, placement of the sensors, optimisation of the design, 0 the second extraction the parameters from layout and multidomain simulations. The deliberations presented below are illustrated with the PWM converter circuit project, designed using Mietec HBiMOS 2pm technology [9].
@
'HIT
J
ov
-6OV
Fig. 1 - Possible configurations of the output power stage of the SIT driver
Taking into account that HBiMOS technology does not allow to use pnp bipolar transistor in a saturation state, the configuration presented in Fig. 1a. cannot be implemented. Several simulations have been performed in order to check the usefulness of the configuration with bipolar transistors. In order to check the behaviour of power part based on bipolar transistors, SPICE library models (created for Mietec HBiMOS technology) has been used for simulation. It has been checked, that it is possible to obtain short turn-on time of the transistor, but the results related to turn-off process are not satisfactory. Fall time is too long causing that the amplitude of the gate current of SIT is too small. The optimised version of the output stage which allows to speed-up switching time is presented in the fig 2.
2. Selection of the power device and driving circuit
The evaluation of semiconductor technology has allowed develop new power semiconductor devices such as MOSFET, IGBT, SIT, MCT. The area of Smart Power Module application affects selection of power semiconductor devices [1,2]. Their
i
f
Fig. 2 - Schematic view of the chosen power stage configuration
7803-377~Z-5/97/$10.001997 IEEE Downloaded on November 17, 2009 at 04:51 from IEEE Xplore. 0 Authorized licensed use limited to: Rajesh Panakala.
Restrictions apply.
The circuit is supplied from the external voltage source (about +18V) connected to VLOW pin and second one (about -80-9OV) connected to 1109, I1 10 transistors. Because switching on of the top transistor in push-pull power stage (see I l l 0 on Fig. 2) requires the voltage higher then VLOW, the bootstrap circuit with extemal capacitance connected to net3 and net4 is used. In spite of relatively large area of MOS transistors which have to be able to sink high gate current of SIT the chosen configuration is a satisfactory solution.
The idea of the next method is to measure the temperature and its gradient along the given distance, in a few places only on the monitored structure (method 2) (Fig.4) and evaluate obtained information in order to detect the thermal malfunction. This problem is known in the literature, especially in the field of modelling of the temperature distribution as an inverse problem. In case of detecting only overheating situations in power semiconductor devices, the complicated unit performing such computations can be simplified. Moreover, in most cases, the overheating occurs only in-one place.
"overheat"signal sensor cell1
3. Problems related with sensing circuits
The particular problem in the design of Smart Power modules is the monitoring and verification of actual state of the module. The power devices are working in switching mode usually with high level currents and voltages. There an: four most important physical quantities (PQ): the temperature, the electric and magnetic fields and the mechanical stresses which influence the behaviour of Smart Power Module, but in design process the designer has to take into consideration not only PQ as a separate factors but the combined influence of them on the working characteristics of Smart Power Module. 14dequate sensors for monitoring these quantities should be used. The specialised algorithms including the correlations betwleen the specified PQ for control circuits need to be established. 3.1 Thermal monitoring problems One of the most significant problems occurred in Smart Power modules are thermal problems [5,6].'The power dissipated by semiconductor power devices is the main source of heating in Smart Power module. It changes the temperature of the module and the working conditions of the devices anld sensors, so thermal monitoring of the structure is one of the most important problems in Smart Power design. The thermal monitoring of integrated semiconductor structures should be based both on temperature measurement from the long distance and on implementation of the gradient direction sensors which gives information about the position of the heat sources over the monitored structure [4]. The thermal malfunction can be detected after processing obtained information by the control unit integrated with the power module. One of the methods of monitoring the [temperatureis using many sensors placed everywhere on the monitored structure (method 1). Their output can be read simultaneously and compared with the reference voltage recognised as the overheating level (Fig.3).
I 1
Li 1
unit
control
y
monitored layout
Fig.4 - Thermal monitoring of silicon structure-method 2
The hdamental sensor cell is presented below
A
I
\
'$
single heat source
renror A
/
s
isoth-A
L
isoth-B
'%isoth-C
Fig5 - Construction of the sensor cell: a E (O",30")
For the presented method (Fig.4) four strong assumptions have been made: Al-there is only one punctual heat source on the monitored wafer. AZthe temperature is linearly distributed over the surface of silicon wafer. In this paper we will consider this assumption, however the function describing the real temperature distribution is not linear. For the given substrate one can find very precisely this function by application of the PYRTHERM software. During calculations of the heat source temperature the found function can be transformed to the linear one. A3-temperature sensors used in this method give output voltage linearly proportional to temperature. A4-the sensors in the sensor cell are placed sufficiently close one to another, that isotherms crossing them can be represented by straight lines. For two sensors, A and C placed in the distance a (Fig.5) the difference between their output voltages is proportional to the changes of the temperature value AT along the distance a. This is true only when the heat source is directly on line AC. For any
Fig.3 - Thermal monitoring of silicon structure-method 1
Authorized licensed use limited to: Rajesh Panakala. Downloaded on November 17, 2009 at 04:51 from IEEE Xplore. Restrictions apply.
other cases the value of the angle a has to be taken into account for the calculation of AT.
Ar
a.cosa
.
where: r - distance from the heat source. Equation (1) shows that actual value of the temperature gradient can be calculated if we know the output voltages from two sensors and the value of cos(@. In order to obtain the information about angle a we should introduce the third sensor. The final equation is
a - cosa
tana <
Now, using the sensor cell we can obtain the information about the temperature distribution and partly about the position of the In heat source (in this moment only the value of tan a). order to obtain the temperature of a single, punctual heat source we have to calculate the distance between the sensor and the heat source. Two sensor cells are required for this purpose. The cells are placed in a given distance (H) and each of them gives the information about the angle a (a1 and a2) to the heat source. The heat source and cells form the triangle in which the length of one side and values of the angles adjacent to this side are known. This means that we can calculate the distances between the heat source and sensors. Now we can calculate the temperature gradient along the known distance. By adding it to the temperature of the sensor we obtain the temperature of the heat source. The final placement of thermal sensors should be done after electro-thermal simulation. 3.2 Current and voltage monitoring The requirements of current and voltage sensors are following: 0 the precise measurement of current and voltage signals, 0 the detailed measurement of magnitude shape-image, 0 the fast measurement process, 0 the low power dissipation of sensor. The fulfilment of these all demands requires implementation of high-performance analogue circuits. The solution with highperformance analogue circuits is expensive especially in economical aspects (an example Fig. 6 - the third part of chip core area is occupied by analogue unit).
One of the most difficult problems is the cross talk and perturbation of the sensor signals during their transmission to the decision module. The parasitic signals due to the commutation of power devices are usually much stronger as the signals coming fiom the sensor. In the case of standard analogue transmission the signals obtained by the control block will be completely different from the signals sent by the sensors. The only way to avoid this problem is immediate conversion of analogue sensor signal to digital one. The transmission of signal in the digital form is not as sensitive to the distortion as the analogue one. Instead of D application of A converter placed close to the sensor it is possible to build the sensors giving directly the output signal in the digital form. Such approach render possible the minimisation of the circuit surface and the problems of AID placement close to the power switching devices. The development of technology and long term research activities are needed to solve the problems mentioned above.
4. Simulation problems
The design and fabrication of Smart-Power modules is very expensive fiom an economic point of view. The simulation of designed devices is a way to decrease the design time and fabrication costs, therefore the simulation of the complex circuit with the carehl consideration of the PQ is required. In real S a t mr Power modules the parameters and models related to the physical features of the system can not be considered separately. The correct design and simulation of such a system demands the multi-domain simulation in electrical, thermal, mechanical and electromagnetic domains. All physical signals can not be separated and should be treated together in one simulation environment. The particular attention should be paid to the electrothermal simulation which becomes of great importance considering still growing values of power densities dissipated in semiconductor structures. The correct interpretation of the signals coming fiom sensors, should be carehlly interpreted.
4.1 General methodology of multi-domain simulation
The electrical circuits design often requires, especially for high power devices, an electro-thermal transient simulation which would takes into account the thermal feedback in VLSI circuits where the high power density is caused by numerous power sources influencing each other. Additionally, modern VLSI systems often contain mechatronic structures like: IRSENSORs (InfraRed Sensors), ISFET (Ion Sensitive FET) and some others. In order to describe these devices, hardware description language ELDO HDL-A (ELDO Hardware Description Language - Analogue) [3] should be used. The HDL-A allows for structural programming and multi-domain including the following phenomena: mechanical, thermal (heat flow, temperature), 0 electrical (voltages and currents), 0 other phenomena not necessarily obeying the Kirchoff Laws. The basic stages of multi-domain system model creation process (proposed by the authors), are as follows. a) The design of device functional model (Fig. 7): 0 each device should be properly interconnected to its environment containing other devices,
-+
Fig. 6 - Thefinal layout of Smart Power circuit 3.3 Problems related to transmission of sensors signals
Authorized licensed use limited to: Rajesh Panakala. Downloaded on November 17, 2009 at 04:51 from IEEE Xplore. Restrictions apply.
all physical phenomena occurring in modelled device should be discerned (see P1, P2, P3, P4 phenomena in Fig. 7), all relations, signal paths and signals direction between particular phenomena should be distinguished (see Fig. 7), all functional blocks (instances) like: resistors, transistors which represent real parts of modelled devices should be separated (see Z1,22,23 in Fig. 7).
'.
*--.
' ;
I
z ,i
--.
--- -_____---------4--
/'
/'
Qbc Qbc
= aR *ZR = aR *=R
+ +
cc * cc
dV
0 O
Fig. 7 - The physical phenomena, their relations (signalpaths) and instances in a device and its environment
f
( l
v\r
q
(8
b) The implementation of a created model in hardware description language (Fig. 8): signal paths should be arranged and classified as electrical, mechanical or others, 0 inputs and outputs of a device (and its environment) should be defined, 0 mathematical description of all phenomena should be simplified (if it is possible) and implemented in a given language, device model implementation should be validated with real device measurements.
ib
+
i,
+
i, =
o
(1 1 )
Environment
TI
Using these equations, it is possible to create a new model of NPN transistor with two additional nodes (PT and MT) used for modelling of the power dissipation and the temperature influence (Is(T) and VT(T)) (see Fig. 9). Transistor power dissipation is represented by the current source in the thermal branch PT-MT. Because current ELDO HDL-A version 1.x does not allow creating structural models yet, a BJT transistor electro-thermal model was constructed using electro-thermal model of diode and other purely electrical components.
electrical thermal
...........................
Device
i
Fig. 8 1 The device implementation
4.2 Electrothermal models and macromodels All models should be written in simple and comprehensible form so as to adapt them easily to other simullators and to include new phenomena. As on example an electro-thermal model of a bipolar transistor will be presented. Let's consider a simple Ebers-Moll injection model of bipolar transistor with non-linear internal capacitors and the temperature dependence on its parameters. This model fully represents the problems, which have to be taken into consideration in electro-thermal modelling and simulation. The model can be described using the following equations.
B
Fig. 9 - Electro-thermal mod 'el of BJT transistor model transistor
E
E
Ir
I
I
thermal
!
e :
C-j,
I
Temp
!
I
!
I MT
The example of this method application is presented below. Let us consider a hybrid device which contains three bipolar transistors connected in parallel (Fig. 10). The transistor model presented in the previous section is used in this circuit description. The values of particular elements in the RC-ladder model in the thermal domain can be obtained by 3-D thermal analysis of the whole system. This analysis has to be performed "n" times considering, each time, only one from "n" temperature
Authorized licensed use limited to: Rajesh Panakala. Downloaded on November 17, 2009 at 04:51 from IEEE Xplore. Restrictions apply.
sensitive elements. On the bases of these "n" simulations the RC network containing n*(n+1)/2 resistors and "n" capacitors can be generated. The characteristic feature of this device is a very strong thermal feedback, so the occurrence of thermal instability is very likely. This phenomenon is common for bipolar devices and it is related to the positive electro-thermal feedback.
In multi-domain simulations the most important problem is the correct physical modelling of devices. The problem is very difficult, especially where 2D or 3D thermal modelling is required. Using the above mentioned methodology the electrothermal models of several other device have been created. The electro-thermal behavioural model of MOSFET transistor in ELDO HDL-A language. This model is based on adapted electrical Shichman-Hodge model of MOSFET transistor is presented in the Fig. 12.
Fig. 10 - Electro-thermal model of hybrid device
The circuit (Fig. 10) can be divided into two separate parts. Purely electronic circuit (Yl, Y2, Y3, Rg, R1, L, etc. ) with parameters depending on temperature determined by the other part of the circuit. IRC network modelling heat flow in hybrid circuit, where: R11, R22, R33 represent transistors thermal resistance to the ambient, R12, R13, R23 represent thermal resistance between transistors, C11, C22, C33 represent the transistor's thermal capacitances. It has to be mentioned that it is possible to perform a set of simulations with different external conditions (temperature, cooling conditions, etc.), taking into account the transistors parameters dispersion. From the example of such a simulation presented below (Fig. l l ) , it can be seen that the power dissipated in each of transistors differ from each other. Without taking into account the thermal feedback this effect could not be seen. It should be underlined that this figure represents the state reached after several simulation periods. Because of the thermal model used here has been created for steady states, the transient curves obtained for the first few periods do not represent the real device behaviour.
Fig.12 - Electro-thermal MOSFET transistor
This model is used in 3D thermal transient macromodel of multilevel structures. 4.3 Simulation of the PWM module Developed Smart-Power module, which the final layout was presented in Fig.6, contains mixed: analogue, digital sections (presented in Fig.13) and power stage section (presented in Fig.2).
Fig. 13 - Analogue-digital part o designed P WM module f
aZ X
%A!?i?-.----zm
A large number of simulation have been done with taking thermal coupling into account. The results of these simulations affect on the temperature sensors placement. The main aim was verification of all Smart-Power module before send to silicon foundry. The simulations of the analogue control unit are similar to previous one and are included in paper. The current and voltage signal waveforms from digital (see Fig. 14) and power stage unit (see Fig. 15) are shown below. The results of performed simulations are in agreement with the theoretical consideration of the authors.
Fig. I I - Results of the hybrid device simulation
Authorized licensed use limited to: Rajesh Panakala. Downloaded on November 17, 2009 at 04:51 from IEEE Xplore. Restrictions apply.
STOP S FRW-WSTER TO-SLIlUE$DtoA XSoton
References
[I]. Baliga B.J.: An Overview of Smart Power Technology. IEEE Transactions on Electron Devices, Vo1.38, No7, July 1992 [2]. Napieralski A., Napieralska M.: Polowe pdprzewodnikowe przyrzqdy mocy. WNT, 1995 [3]. Orlikowski M., Zubert M., W6jciak W.: Methodology of Analog Circuit Description Using High Description Language. Fifth WORKSHOP in the frame of ESPRIT Project (European Strategic Program for Research and Development in Information Technology) CEC-Contract N O 8173-BARMINT (Basic Research for Microsystem Integration), Toulouse, France, 6-7 May 1996 [4]. W6jciak W., Napieralski A.: An analogue temperature sensor integrated in the CMO technology. International Workshop on Thermal Investigations of ICs and Microstructures- THERMINIC, Grenoble, France, 25-26" Sept., 1995, pp 15-20 [5]. Napieralski A.: La thermique des composants et les couplages thermodectriques. Congres National GRECO CNRS "Dispositifs et Systhmes Electrotechniques" Grenoble, France, 18 DCcembre 1988,30p., [6]. Napieralski A.: Computer aided design of high power semiconductor devices. Thermal and electrical aspects. Proceedings of VI Congress0 da Sociedade Brasilieira de Microeletronica, (VI Conference of the Brazilian Microelectronics Society), Belo Horizonte, Brazil, 15-19 July, 1991, 10 pages 171. HDL-A Language Reference Manual, ANACAD [8]. MAST Reference Manual, Relase 4.0, Analogy [9]. Alcatel Mietec, HBiMOS 2 ~ documentation, 1994 m
I
I I
I 1
Os
0 . 2 ~
-D
U(-ccxt)
6
U(-i-sense)
0 . 4 ~ 1 . 6 ~ 0 . h ~ l.W U(-PUl-OUT) U(-DEAD-TIlE-OUT) 0 U(-ICOIP) Time
12) .r5
Fig. 14 - The simulation o digital and analogue unit f
x$RtoD
t
i
5. Conclusions
In the Smart Power module design ,the electro-thermal phenomena have to be done. The results of the simulation will allow for correct placement of temperature sensors. In general, the multi-domain simulations have to be performed in Smart Power module design. The different kind of sensors for Smart Plower modules have to be developed. The laboratory and practical tests are ,still the best final verification of Smart Power module design.
Authorized licensed use limited to: Rajesh Panakala. Downloaded on November 17, 2009 at 04:51 from IEEE Xplore. Restrictions apply.
ASzajfler, T.Pokniak, M.Napieralska, M.Zubert, W.Wbjciak, A.Napieralski Department of Microelectronics and Computer Science Technical University of L6&, Poland Lbcli, ul. Aleja Politechniki 11 tel.: +48 (42) 3 1 26 28 fax: +48 (42) 36 26 28; e-mail: [email protected].
Abstract. In this paper the following problems:
selection of the power device and driving circuit, sensing and protection circuits, transmission of the sensor signals, thermal problems in Smart Power module, simulation problems, of the Smart Power modules design will be presented and discussed. The deliberations will be illustrated with the example of PWM converter circuit designed in Mietec HBiMOS 2pm technology. The new electro-thermal practical model of transistors which can be use to complex electro-thermal simulation of Smart-Power structure is presented.
advantages and disadvantages and the abilities of technology have to be taken into consideration. According to the type of power device, the relevant driving circuit have to be designed. As an example of these problems the output stage of designed Smart Power module P W M for SIT control is presented. The used HBiMOS 2pm technology allows to operate with voltages up to 80-1OOV. Different kind of active and passive elements can be created using this technology: NMOS and PMOS transistors; bipolar transistors, diodes; resistors and capacitors. The following configurations of the output stage presented in (Fig.])are possible.
1. Introduction
The rapid development of semiconductor technology has an impact on the increasing interest of ,,intelligent" power modules. The area of applications and intended functions of Smart Power module determine the technology which will be used for physical realisation of the device [l]. Every technology has specific limitations, which have to be taken into consideration during the design of Smart Power module architecture [2]. Moreover, the designer of the power module should not only take into consideration the separate physical phenomena but also interactions between them. In design process the economic aspects should not also be neglected, so the adequate simulation tools have to be used to minimise the cost of design and prototyping. The following stages in Smart Power module design could be distinguished: selection of power stage configuration and behavioural specification of control part of the module, design of the control part of module, 0 mixed digital-analogue simulation of entire circuit using SPICE-like simulators, 0 design of the layout of Smart Power module, the first extraction the parameters from layout and multidomain simulations, placement of the sensors, optimisation of the design, 0 the second extraction the parameters from layout and multidomain simulations. The deliberations presented below are illustrated with the PWM converter circuit project, designed using Mietec HBiMOS 2pm technology [9].
@
'HIT
J
ov
-6OV
Fig. 1 - Possible configurations of the output power stage of the SIT driver
Taking into account that HBiMOS technology does not allow to use pnp bipolar transistor in a saturation state, the configuration presented in Fig. 1a. cannot be implemented. Several simulations have been performed in order to check the usefulness of the configuration with bipolar transistors. In order to check the behaviour of power part based on bipolar transistors, SPICE library models (created for Mietec HBiMOS technology) has been used for simulation. It has been checked, that it is possible to obtain short turn-on time of the transistor, but the results related to turn-off process are not satisfactory. Fall time is too long causing that the amplitude of the gate current of SIT is too small. The optimised version of the output stage which allows to speed-up switching time is presented in the fig 2.
2. Selection of the power device and driving circuit
The evaluation of semiconductor technology has allowed develop new power semiconductor devices such as MOSFET, IGBT, SIT, MCT. The area of Smart Power Module application affects selection of power semiconductor devices [1,2]. Their
i
f
Fig. 2 - Schematic view of the chosen power stage configuration
7803-377~Z-5/97/$10.001997 IEEE Downloaded on November 17, 2009 at 04:51 from IEEE Xplore. 0 Authorized licensed use limited to: Rajesh Panakala.
Restrictions apply.
The circuit is supplied from the external voltage source (about +18V) connected to VLOW pin and second one (about -80-9OV) connected to 1109, I1 10 transistors. Because switching on of the top transistor in push-pull power stage (see I l l 0 on Fig. 2) requires the voltage higher then VLOW, the bootstrap circuit with extemal capacitance connected to net3 and net4 is used. In spite of relatively large area of MOS transistors which have to be able to sink high gate current of SIT the chosen configuration is a satisfactory solution.
The idea of the next method is to measure the temperature and its gradient along the given distance, in a few places only on the monitored structure (method 2) (Fig.4) and evaluate obtained information in order to detect the thermal malfunction. This problem is known in the literature, especially in the field of modelling of the temperature distribution as an inverse problem. In case of detecting only overheating situations in power semiconductor devices, the complicated unit performing such computations can be simplified. Moreover, in most cases, the overheating occurs only in-one place.
"overheat"signal sensor cell1
3. Problems related with sensing circuits
The particular problem in the design of Smart Power modules is the monitoring and verification of actual state of the module. The power devices are working in switching mode usually with high level currents and voltages. There an: four most important physical quantities (PQ): the temperature, the electric and magnetic fields and the mechanical stresses which influence the behaviour of Smart Power Module, but in design process the designer has to take into consideration not only PQ as a separate factors but the combined influence of them on the working characteristics of Smart Power Module. 14dequate sensors for monitoring these quantities should be used. The specialised algorithms including the correlations betwleen the specified PQ for control circuits need to be established. 3.1 Thermal monitoring problems One of the most significant problems occurred in Smart Power modules are thermal problems [5,6].'The power dissipated by semiconductor power devices is the main source of heating in Smart Power module. It changes the temperature of the module and the working conditions of the devices anld sensors, so thermal monitoring of the structure is one of the most important problems in Smart Power design. The thermal monitoring of integrated semiconductor structures should be based both on temperature measurement from the long distance and on implementation of the gradient direction sensors which gives information about the position of the heat sources over the monitored structure [4]. The thermal malfunction can be detected after processing obtained information by the control unit integrated with the power module. One of the methods of monitoring the [temperatureis using many sensors placed everywhere on the monitored structure (method 1). Their output can be read simultaneously and compared with the reference voltage recognised as the overheating level (Fig.3).
I 1
Li 1
unit
control
y
monitored layout
Fig.4 - Thermal monitoring of silicon structure-method 2
The hdamental sensor cell is presented below
A
I
\
'$
single heat source
renror A
/
s
isoth-A
L
isoth-B
'%isoth-C
Fig5 - Construction of the sensor cell: a E (O",30")
For the presented method (Fig.4) four strong assumptions have been made: Al-there is only one punctual heat source on the monitored wafer. AZthe temperature is linearly distributed over the surface of silicon wafer. In this paper we will consider this assumption, however the function describing the real temperature distribution is not linear. For the given substrate one can find very precisely this function by application of the PYRTHERM software. During calculations of the heat source temperature the found function can be transformed to the linear one. A3-temperature sensors used in this method give output voltage linearly proportional to temperature. A4-the sensors in the sensor cell are placed sufficiently close one to another, that isotherms crossing them can be represented by straight lines. For two sensors, A and C placed in the distance a (Fig.5) the difference between their output voltages is proportional to the changes of the temperature value AT along the distance a. This is true only when the heat source is directly on line AC. For any
Fig.3 - Thermal monitoring of silicon structure-method 1
Authorized licensed use limited to: Rajesh Panakala. Downloaded on November 17, 2009 at 04:51 from IEEE Xplore. Restrictions apply.
other cases the value of the angle a has to be taken into account for the calculation of AT.
Ar
a.cosa
.
where: r - distance from the heat source. Equation (1) shows that actual value of the temperature gradient can be calculated if we know the output voltages from two sensors and the value of cos(@. In order to obtain the information about angle a we should introduce the third sensor. The final equation is
a - cosa
tana <
Now, using the sensor cell we can obtain the information about the temperature distribution and partly about the position of the In heat source (in this moment only the value of tan a). order to obtain the temperature of a single, punctual heat source we have to calculate the distance between the sensor and the heat source. Two sensor cells are required for this purpose. The cells are placed in a given distance (H) and each of them gives the information about the angle a (a1 and a2) to the heat source. The heat source and cells form the triangle in which the length of one side and values of the angles adjacent to this side are known. This means that we can calculate the distances between the heat source and sensors. Now we can calculate the temperature gradient along the known distance. By adding it to the temperature of the sensor we obtain the temperature of the heat source. The final placement of thermal sensors should be done after electro-thermal simulation. 3.2 Current and voltage monitoring The requirements of current and voltage sensors are following: 0 the precise measurement of current and voltage signals, 0 the detailed measurement of magnitude shape-image, 0 the fast measurement process, 0 the low power dissipation of sensor. The fulfilment of these all demands requires implementation of high-performance analogue circuits. The solution with highperformance analogue circuits is expensive especially in economical aspects (an example Fig. 6 - the third part of chip core area is occupied by analogue unit).
One of the most difficult problems is the cross talk and perturbation of the sensor signals during their transmission to the decision module. The parasitic signals due to the commutation of power devices are usually much stronger as the signals coming fiom the sensor. In the case of standard analogue transmission the signals obtained by the control block will be completely different from the signals sent by the sensors. The only way to avoid this problem is immediate conversion of analogue sensor signal to digital one. The transmission of signal in the digital form is not as sensitive to the distortion as the analogue one. Instead of D application of A converter placed close to the sensor it is possible to build the sensors giving directly the output signal in the digital form. Such approach render possible the minimisation of the circuit surface and the problems of AID placement close to the power switching devices. The development of technology and long term research activities are needed to solve the problems mentioned above.
4. Simulation problems
The design and fabrication of Smart-Power modules is very expensive fiom an economic point of view. The simulation of designed devices is a way to decrease the design time and fabrication costs, therefore the simulation of the complex circuit with the carehl consideration of the PQ is required. In real S a t mr Power modules the parameters and models related to the physical features of the system can not be considered separately. The correct design and simulation of such a system demands the multi-domain simulation in electrical, thermal, mechanical and electromagnetic domains. All physical signals can not be separated and should be treated together in one simulation environment. The particular attention should be paid to the electrothermal simulation which becomes of great importance considering still growing values of power densities dissipated in semiconductor structures. The correct interpretation of the signals coming fiom sensors, should be carehlly interpreted.
4.1 General methodology of multi-domain simulation
The electrical circuits design often requires, especially for high power devices, an electro-thermal transient simulation which would takes into account the thermal feedback in VLSI circuits where the high power density is caused by numerous power sources influencing each other. Additionally, modern VLSI systems often contain mechatronic structures like: IRSENSORs (InfraRed Sensors), ISFET (Ion Sensitive FET) and some others. In order to describe these devices, hardware description language ELDO HDL-A (ELDO Hardware Description Language - Analogue) [3] should be used. The HDL-A allows for structural programming and multi-domain including the following phenomena: mechanical, thermal (heat flow, temperature), 0 electrical (voltages and currents), 0 other phenomena not necessarily obeying the Kirchoff Laws. The basic stages of multi-domain system model creation process (proposed by the authors), are as follows. a) The design of device functional model (Fig. 7): 0 each device should be properly interconnected to its environment containing other devices,
-+
Fig. 6 - Thefinal layout of Smart Power circuit 3.3 Problems related to transmission of sensors signals
Authorized licensed use limited to: Rajesh Panakala. Downloaded on November 17, 2009 at 04:51 from IEEE Xplore. Restrictions apply.
all physical phenomena occurring in modelled device should be discerned (see P1, P2, P3, P4 phenomena in Fig. 7), all relations, signal paths and signals direction between particular phenomena should be distinguished (see Fig. 7), all functional blocks (instances) like: resistors, transistors which represent real parts of modelled devices should be separated (see Z1,22,23 in Fig. 7).
'.
*--.
' ;
I
z ,i
--.
--- -_____---------4--
/'
/'
Qbc Qbc
= aR *ZR = aR *=R
+ +
cc * cc
dV
0 O
Fig. 7 - The physical phenomena, their relations (signalpaths) and instances in a device and its environment
f
( l
v\r
q
(8
b) The implementation of a created model in hardware description language (Fig. 8): signal paths should be arranged and classified as electrical, mechanical or others, 0 inputs and outputs of a device (and its environment) should be defined, 0 mathematical description of all phenomena should be simplified (if it is possible) and implemented in a given language, device model implementation should be validated with real device measurements.
ib
+
i,
+
i, =
o
(1 1 )
Environment
TI
Using these equations, it is possible to create a new model of NPN transistor with two additional nodes (PT and MT) used for modelling of the power dissipation and the temperature influence (Is(T) and VT(T)) (see Fig. 9). Transistor power dissipation is represented by the current source in the thermal branch PT-MT. Because current ELDO HDL-A version 1.x does not allow creating structural models yet, a BJT transistor electro-thermal model was constructed using electro-thermal model of diode and other purely electrical components.
electrical thermal
...........................
Device
i
Fig. 8 1 The device implementation
4.2 Electrothermal models and macromodels All models should be written in simple and comprehensible form so as to adapt them easily to other simullators and to include new phenomena. As on example an electro-thermal model of a bipolar transistor will be presented. Let's consider a simple Ebers-Moll injection model of bipolar transistor with non-linear internal capacitors and the temperature dependence on its parameters. This model fully represents the problems, which have to be taken into consideration in electro-thermal modelling and simulation. The model can be described using the following equations.
B
Fig. 9 - Electro-thermal mod 'el of BJT transistor model transistor
E
E
Ir
I
I
thermal
!
e :
C-j,
I
Temp
!
I
!
I MT
The example of this method application is presented below. Let us consider a hybrid device which contains three bipolar transistors connected in parallel (Fig. 10). The transistor model presented in the previous section is used in this circuit description. The values of particular elements in the RC-ladder model in the thermal domain can be obtained by 3-D thermal analysis of the whole system. This analysis has to be performed "n" times considering, each time, only one from "n" temperature
Authorized licensed use limited to: Rajesh Panakala. Downloaded on November 17, 2009 at 04:51 from IEEE Xplore. Restrictions apply.
sensitive elements. On the bases of these "n" simulations the RC network containing n*(n+1)/2 resistors and "n" capacitors can be generated. The characteristic feature of this device is a very strong thermal feedback, so the occurrence of thermal instability is very likely. This phenomenon is common for bipolar devices and it is related to the positive electro-thermal feedback.
In multi-domain simulations the most important problem is the correct physical modelling of devices. The problem is very difficult, especially where 2D or 3D thermal modelling is required. Using the above mentioned methodology the electrothermal models of several other device have been created. The electro-thermal behavioural model of MOSFET transistor in ELDO HDL-A language. This model is based on adapted electrical Shichman-Hodge model of MOSFET transistor is presented in the Fig. 12.
Fig. 10 - Electro-thermal model of hybrid device
The circuit (Fig. 10) can be divided into two separate parts. Purely electronic circuit (Yl, Y2, Y3, Rg, R1, L, etc. ) with parameters depending on temperature determined by the other part of the circuit. IRC network modelling heat flow in hybrid circuit, where: R11, R22, R33 represent transistors thermal resistance to the ambient, R12, R13, R23 represent thermal resistance between transistors, C11, C22, C33 represent the transistor's thermal capacitances. It has to be mentioned that it is possible to perform a set of simulations with different external conditions (temperature, cooling conditions, etc.), taking into account the transistors parameters dispersion. From the example of such a simulation presented below (Fig. l l ) , it can be seen that the power dissipated in each of transistors differ from each other. Without taking into account the thermal feedback this effect could not be seen. It should be underlined that this figure represents the state reached after several simulation periods. Because of the thermal model used here has been created for steady states, the transient curves obtained for the first few periods do not represent the real device behaviour.
Fig.12 - Electro-thermal MOSFET transistor
This model is used in 3D thermal transient macromodel of multilevel structures. 4.3 Simulation of the PWM module Developed Smart-Power module, which the final layout was presented in Fig.6, contains mixed: analogue, digital sections (presented in Fig.13) and power stage section (presented in Fig.2).
Fig. 13 - Analogue-digital part o designed P WM module f
aZ X
%A!?i?-.----zm
A large number of simulation have been done with taking thermal coupling into account. The results of these simulations affect on the temperature sensors placement. The main aim was verification of all Smart-Power module before send to silicon foundry. The simulations of the analogue control unit are similar to previous one and are included in paper. The current and voltage signal waveforms from digital (see Fig. 14) and power stage unit (see Fig. 15) are shown below. The results of performed simulations are in agreement with the theoretical consideration of the authors.
Fig. I I - Results of the hybrid device simulation
Authorized licensed use limited to: Rajesh Panakala. Downloaded on November 17, 2009 at 04:51 from IEEE Xplore. Restrictions apply.
STOP S FRW-WSTER TO-SLIlUE$DtoA XSoton
References
[I]. Baliga B.J.: An Overview of Smart Power Technology. IEEE Transactions on Electron Devices, Vo1.38, No7, July 1992 [2]. Napieralski A., Napieralska M.: Polowe pdprzewodnikowe przyrzqdy mocy. WNT, 1995 [3]. Orlikowski M., Zubert M., W6jciak W.: Methodology of Analog Circuit Description Using High Description Language. Fifth WORKSHOP in the frame of ESPRIT Project (European Strategic Program for Research and Development in Information Technology) CEC-Contract N O 8173-BARMINT (Basic Research for Microsystem Integration), Toulouse, France, 6-7 May 1996 [4]. W6jciak W., Napieralski A.: An analogue temperature sensor integrated in the CMO technology. International Workshop on Thermal Investigations of ICs and Microstructures- THERMINIC, Grenoble, France, 25-26" Sept., 1995, pp 15-20 [5]. Napieralski A.: La thermique des composants et les couplages thermodectriques. Congres National GRECO CNRS "Dispositifs et Systhmes Electrotechniques" Grenoble, France, 18 DCcembre 1988,30p., [6]. Napieralski A.: Computer aided design of high power semiconductor devices. Thermal and electrical aspects. Proceedings of VI Congress0 da Sociedade Brasilieira de Microeletronica, (VI Conference of the Brazilian Microelectronics Society), Belo Horizonte, Brazil, 15-19 July, 1991, 10 pages 171. HDL-A Language Reference Manual, ANACAD [8]. MAST Reference Manual, Relase 4.0, Analogy [9]. Alcatel Mietec, HBiMOS 2 ~ documentation, 1994 m
I
I I
I 1
Os
0 . 2 ~
-D
U(-ccxt)
6
U(-i-sense)
0 . 4 ~ 1 . 6 ~ 0 . h ~ l.W U(-PUl-OUT) U(-DEAD-TIlE-OUT) 0 U(-ICOIP) Time
12) .r5
Fig. 14 - The simulation o digital and analogue unit f
x$RtoD
t
i
5. Conclusions
In the Smart Power module design ,the electro-thermal phenomena have to be done. The results of the simulation will allow for correct placement of temperature sensors. In general, the multi-domain simulations have to be performed in Smart Power module design. The different kind of sensors for Smart Plower modules have to be developed. The laboratory and practical tests are ,still the best final verification of Smart Power module design.
Authorized licensed use limited to: Rajesh Panakala. Downloaded on November 17, 2009 at 04:51 from IEEE Xplore. Restrictions apply.
