Miniaturised “Low Profile” Module Integrated Converter for Photovoltaic Amlications with Integrated Magnetic Components Mike Meinhardt, Terence O’Donnell, Henning Schneider, John Flannery, Cian 6 Mathuna PE1 Technologies, National Microelectronics Research Centre, Lee Maltings, Cork, Ireland, Phone: +353 21 90 42 41, Fax: +353 21 270 271 Email:
[email protected]
through the lack of a need for energy storage devices such as batteries. Grid connected Ehotovoltaic (PV) systems include in general a solar module transforming sunlight into electrical energy and a converter connecting the solar module to the grid. As shown in Figure 1 a (solar) Module Integrated Converter (MIC) for PV applications is mounted directly to the frame of a solar module
Abstract This paper deals with reliability aspects of Module Integrated Converters (MIC) for grid connected photovoltaic applications. The presented “Low Profile Design” of a MIC has almost twice the power density and a 70% longer lifetime than commercially available MICs. In order to miniaturise and increase the reliability of the MIC all magnetic components have been realised using planar cores and windings integrated into PCB. Additionally the converter’s resonant inductor has been integrated into the transformer by using the leakage inductance as the inductor. The leakage inductance of planar transformers with integrated windings can be adjusted very precisely.
The combination of MIC and solar module - often called “AC Solar Module” - allows a direct connection of the solar modules to the grid by converting the module DC voltage to a 230 V AC voltage. Recent investigations have shown that Module Integrated Converters for Photovoltaic applications have some important advantages over string orientated or central inverter concepts such as: Increased energy yield in case of systems suffering from shading effects. Reduced danger of arcs due to replacing the DC installation by an AC installation. Simplified configuration of PV systems due to standardised interface (220 V-). Nevertheless in the future central inverter concepts will still be used especially in very large-scale PV systems. Depending on the PV system’s site and properties of installation string oriented systems are likely to be used in the medium power range.
1 INTRODUCTION The use of renewable energy sources and rational use of energy are fundamental pillars of a responsible energy policy for the future. Because of their sustainable character, renewable energy sources almost exclusively based on solar energy are capable of preserving resources and providing energy services virtually without any environmental impact. Therefore, intensive research and development in solar energy technologies are justified and necessary. Grid connected Photovoltaic (PV) systems distinguish themselves
Reliability is the issue Since the MIC is part of the solar module, the electronic elements of the MIC are exposed to extreme environmental conditions (e.g. temperature, humidity and lightning). This high thermomechanical stress determines the lifetime of the MIC, which should exceed 15 years in order to approach the lifetime of the solar modules themselves. Miniaturisation and cost reduction are required. These aspects are conflicting and lead to a need for advanced power packaging solutions to find a compromise between cost, miniaturisation and reliability requirements.
Figure 1: Cross-section of a MIC mounted to the frame of a solar module
0-7803-5160-6/99/$10.00 0 1999 IEEE.
Peter Zacharias, Thomas Krieger Institut fuer Solare Energiesysteme e.V. Koenigstor 59,341 19 Kassel, Germany Phone: +49 561 7294 -0, Fax: +49 561 7294 -100 Email: pzacharias @iset.uni-kassel.de
305
2 WORLD MARKETFOR MODULE INTEGRATED CONVERTERS Figure2 shows the results of a survey of different predictions of the development of the PV market. Even considering the two most pessimistic forecasts published in 1993 the PV market in 2005 will be about 300...400 MW, world wide. If one third of PV production is used for grid connected or similar application in 2005 the potential market for AC modules would be 100.. .130MWp. Assuming that at this time two thirds of this market volume will be occupied by other converter concepts (in kW range) a more or less realistic market expectation for Module Integrated Converters (MIC) could be 55 ...65MWp per year. Considering the rated power of one MIC in a range of 100...300Wp this market volume corresponds to 160,000...500,000 pieces per year in 2005.
The solar module on the left generates a maximum power of 110 W at a voltage between 26 V and 37 V. The MIC transforms the DC-power into grid conformant AC power (50/60Hz). The basic structure of a MIC is similar to conventional Switch-Mode-ACIDC-Power-Supplies.The main magnetic components are a resonant inductor, transformer and line inductors. In order to increase the efficiency and decrease the required size of the converter the resonant inductance is integrated in the transformer. This topic will be dealt in detail in section 4. Reasons for a “Low Profile” MIC Design 3.2 Although the back of the solar module gives a relatively large area for mounting a MIC and therefore there is no direct need for miniaturisation of the entire MIC, it is proposed to target a “low profile” MIC design with a high power density for the following reasons: - A much simplified thermal model [2] shows that a decreased height of a MIC leads to reduced component temperatures and consequently to a higher reliability of the MIC. - The MIC can easily be mounted to the frame of the solar module and due to the reduced weight the mechanical connection is more reliable and cheaper. - The power density of power supplies is commonly used as a general measure of technical advancement. A comparison of power densities of MICs in [3] shows that the presented lst generation MIC prototype has twice the almost power density as commercially available MIC while having a similar footprint. - The reduced amount of potting material needed for
3500
I
I
c
SU (ACC) 92 I 2000c
4
.I
2000
1995
2010
2005
encapsulation reduces the MIC costs.
Year
Figure 2: World wide PV market increase [ 11
3
3.3 Physical Realisation The Power Electronics Packaging ideas applied for the physical realisation and arrangement of the main components of the electrical circuit will be explained in the following. The side-view illustrated in Figure 4 shows that the circuitry of the MIC is spread over two layers. An Insulated Metal Substrate (IMS), with a single conduction layer, contains all power dissipating components such as Mosfets and diodes as well as all tall components such as DC capacitors and terminals. The second substrate is realised as a multi-layer FR4 Printed Circuit Board. The windings for the transformer and the inductors of the AC filter are integrated into this PCB as shown in Figure 5. The FR4 board also contains the control and measurement circuitry. This “Double Decker“ design leads to very good thermal behaviour of the MIC due to the excellent thermal conductivity of the IMS base plate combined with the opportunity to build the control circuit on both sides of a four layer PCB and to integrate the windings of the planar magnetic devices into this PCB.
MODULE INTEGRATED CONVERTER
Structure and Mode of Operation 3.1 The block diagram shown in Figure 3 illustrates the different components of the MIC.
Solar
I I
,p
Hz
kHZ 1
1
Control (5) (incl. voltage, current and temp. sensors)
Module Integrated Converter ( M E )
I
I
Grid
I
I I I MIC.dem~AI
Figure 3: Schematic of the Module Integrated Converter
306
sideview ppt
Figure 4: Side-view of the 1” generation prototype of a Module Integrated Converter for PV applications (height 1Scm)
Two rows of butt mounted pins are used for the electrical and mechanical connection of the upper FR4 board at a clearance of 7.5 mm above the lower IMS board. These pins are used as a thermal shield to screen the Electrolytic capacitors which are proven to be the most unreliable part of the converter (see section 5) from any heat running in the direction of the capacitors. The overall height of the MIC is 1.5 cm which is possible due to the use of planar magnetic cores and low profile capacitors.
The encapsulant can be used to improve the thermal performance of the MIC. Previous investigations [3] have shown, that the impact of using encapsulants with high thermal conductivities on the thermal behaviour is small when other thermal design features have been applied e.g. optimum mounting orientation of MIC while fixing MIC to the frame of the solar module. In this case the encapsulant can be optimised for other criteria such as humidity resistance or TCE match with other components or substrates. PCB tracks covered by an electrically isolating medium need a smaller minimal distance. Therefore potting can allow the size of the converter to be reduced. 3.4.2 Requirements of Encapsulants suitable for MIC Encapsulation Three types of polymers with different physical, mechanical, chemical and electrical properties can be used for encapsulation: epoxies, silicones and urethanes. The properties may vary in wide range for the same type of encapsulant. The requirements of a suitable potting material are: High thermal conductivity to ensure large lifetime of MIC. Chemical resistance to humidity and most common organic solvents are essential to provide environmental protection. The coefficient of thermal expansion (CTE) should match the CTE of IMS (aluminium) to avoid mechanical stress and ensure good adhesion. Glass transition temperature (T,) should be higher than maximum MIC operating temperature ( 100°C). Low modulus and flexural strength to eliminate stress on the MIC components. Low cure temperature and short (as much as possible) curing time to keep sensitive elements (IC, electrolytic capacitors) under low temperature stress. 0 Low level of ionic impurities (Na+, K+, Cl-) for high insulating resistance against temperature and humidity.
Pins for connecting the lower “IMS” base plate and upper FR4 PCB
Figure 5: Top-view of the lstgeneration prototype of the Module Integrated Converter (width 13 cm, length 9.5 cm) 3.4 Encapsulatioflotting of MIC Encapsulation of power electronics is very common practice in the area of conventional power supplies. Covered by a plastic lid the electronic components of the MIC will be potted with a suitable potting compound in the final product.
3.4.1 Advantages of MIC Encapsulation The use of potting materials has different advantages. The materials used for potting provide humidity resistance. Therefore no further protection against humidity at the case of the converter is necessary which can reduce the costs of the system.
Silicone polymers seem to have most of the properties described above. However, the curing process requires rather
307
In order avoid saturation of the core caused by the 50 Hz component of line current of 0.5 A (RMS) a 50 um “air” gap has been inserted.
high temperatures and relatively long times. Meanwhile, UV curing urethanes cure at low temperature in seconds. The two most important properties - thermal conductivity and humidity resistance have a trade-off. These facts should be taken into account when choosing resins. To determine the reliability rate and make a final decision on the application the packages with different encapsulants will undergo environmental test. 3.4.3 Disadvantagesof MIC Encapsulation Potting of the MIC may have a limiting impact on the lifetime. Due to the different thermal coefficient of expansion (TCE) of encapsulant and MIC components the cycling of the MIC’s temperature caused by day-night and summer-winter variation ambient / solar frame temperature can generate an additional source of failures.
Figure 6: a.) Top view and b.) side view of realisation of AC Filter inductors 4.2.2 Measurement and comparison with wire wound inductor The measurement of the inductance of the AC-Filter inductor in Figure7.a shows a good linearity of the inductance has been achieved. A comparison with a commercially available toroidal inductor in Figure7.b shows that due to the bigger crosssection area of the toroidal core the wire wound inductor need less windings which results in smaller Copper Losses. In both cases the core losses are about 10%of the total inductor losses. Although the toroidal inductor has some advantages the planar inductor has been selected due to the lower height of the planar core and the resulting possibility to place components below the inductor. In the presented MIC design shown in Figure4 the Rectifier Mosfet is mounted on the IMS base plate below the inductors. (see also motivation in section 4.1).
4 INTEGRATION OF MAGNETIC COMPONENTS AND NOVEL TRANSFORMER DESIGN
4.1 Motivation As shown in Figure3 the MIC contains four magnetic components: two AC Filter inductors and a transformer as well as a resonant inductor. In numerous papers the advantages of using planar magnetics in combination with integration of windings into PCBs in comparison with conventional wire wound magnetic components have been pointed out: - Increased reliability, due to reduced number of connections Very good reproducibility, due to use of printed or patterned (PCB) windings instead of wire wound windings. Increased efficiency; The small variations in the manufacturing process enable the real magnetic component to be very close to the designed / optimised component. Increased power density of converter (High level of integration). Low profile magnetic components support the possibility of realising a 3 D power electronics packaging structure. As shown in Figure4 and Figure 5: the inductors are located above the Mosfets of the controlled rectifier. 4.2
10
Ucommercial toroidal core
- HO
OmA 250mA SOOmA 750mA Amplitude of the 50 Hz (60HZ) component of the line current
._l.l.l
Height
PCE Area
Losses
[mm]
[cm21
[100mW]
-._
Figure 7: a.) Measurements of planar realisation of AC Filter inductors and b.) comparison with a commercially available wire wound inductor
AC Filter inductors
4.2.1 Specifications and realisation In order to filter the harmonics generated by the switching of the DCDC-Converter (500 kHz) two AC Filter inductors with an inductance of 500 p H each are required. As shown in Figure 6.b this has been realised by using an E-PLT core combination. (Core E-PLT 22/5/16, Material 3F3, Number of turns: 20 integrated into the 4-Layer PCB).
4.2.3 Costs of planar inductors using windings integrated into a PCB Besides technical issues, cost is always a very important criteria for assessment of a technical solution. Based on the manufacturing process suggested in Table 1 the costs for two inductors can be calculated assuming a production of 100,000pieces per year and labour costs of 18 ECU / h.
308
The overall costs of the toroidal inductor are very similar with only the division between material and labour costs being different (labour costs 7 40% of total costs).
Variation of distance between primary and secondary winding in order to adiust the leakage
4-Layer PCB (70 pm copper) for secondary windina 117 turns) and
Table 1: Costs related with manufacturing of two planar inductors designed in section 4.2.1 (1 ECU $ 1.2)
-
Costs / ECU Materia[ Labour 2*0.2+0.03 45ec
0.02
0.02
0.03
5 sec
zsec
2*0.24+0.28
4 sec
zsec
0.02 5sec zsec zsec
1.23
+
E-core E32/6/20
Step Manufacturing process 1 Place 2 I cores + spacer Glue spacer to 1 cores
2 0.01 3 0.02 4 0.01 5 0.03 6 0.01 7 0.01 8
Llakage inductance !-U = f(x)
Cure glue Place I cores + E cores + PCB Close Jig Glue Cure glue Remove Jig and Parts
0.14 = 1.37 ECU
drimary winding: only one because of
high current (ZOA)
\
Secondary winding
1
IMS base Plate Tr.".lomrr
WI
Figure 8: Cross-section through transformer with adjustable leakage inductance 3.3.3 Realisation of primary and secondary winding of the planar transformer As shown in Figure8 the lower PCB contains the primary winding with a single turn In order to reduce losses a "thick" copper layer (140 pm) is necessary because of the high current of 20A. Therefore the upper PCB contains the secondary winding which is divided on 4 layers of 70pm copper. The upper PCB is also used for attaching the control and measurement circuitry and the windings of the AC filter inductors.
YI
4.3 Integration of Transformer and Resonant Inductor In order achieve a reduction in costs, losses and size the required resonant inductor is realised by adjusting the leakage inductance of the transformer instead of having an additional discrete inductor. According to [4] the integration of resonant inductor and transformer save nearly 50% of the space compared to a discrete realisation. 3.3.1 Specifications The use of planar magnetics is necessary to meet the specifications of a maximum permissible height of the converter of 15 mm. Since the DC filter is located at the primary (low voltage) side of the transformer the transformer has to transmit an average power of 110 W fluctuating with 100 Hz. The switching frequency is 500 kHz and the amplitude of the input current is 20 A. A transformation ration of 1:17 is required. The resonant converter is designed with a resonant inductance of 150 nH. Since the leakage inductance of the remaining circuit including the resonant capacitors, MOSFETs and copper tracks amounts to approx. 50 nH an additional inductance of 100 nH is necessary.
3.3.4 Realising the resonant inductor by adjusting the leakage inductance of the planar transformer Due to the fact that the leakage inductance of a transformer is mainly determined by the distance x between primary and secondary tracks the windings are divided on two PCBs to gain an adjustable leakage inductance. The transformer core is assembled as an E-E core instead of an E-PLT combination in order to get a larger range of adjustable inductance. In order to determine the leakage inductance of the E-E-core transformer it has been simulated with the Maxwell 2D Field Simulator. A plot of the leakage flux for transformers with maximum and minimum separation of primary and secondary winding in rated mode of operation is shown in Figure 9a/b.. The leakage inductance L, can be derived based on the magnetic energy density shown in Figure 9cld. The simulation results match the measurements with an accuracy of 10%. The range of leakage inductance which can be realised by separating the primary and secondary winding is between 20 nH to 60 nH. Therefore other methods have to be investigated in order to increase the leakage inductance of the transformer to meet the specification of this particular application (100 nH).
3.3.2 Transformer Design A cross-section of the physical realisation of the transformer is shown in Figure 8. The transformer core chosen is an E32 core by Philips. The material is 3F3, which is suitable for power conversion in the specified frequency range. One reason for choosing an E-E-core combination is to provide the space necessary to place the MOSFETs on the lower IMS base plate and at the same time the control circuitry on the bottom of the upper FR4 board as shown in Figure 4 and Figure 5. 309
The main reason for selecting the planar transformer realisation is the lower overall height of the MIC based on the reasons mentioned in section 4.1.
5 RELIABILITY ASSESSMENT OF MIC 5.1 Lifetime limiting factors In general the lifetime of MIC is limited by a number of factors. The failures can be categorised into design related failures and failures which are triggered by operational conditions. In most cases failures are caused by a combination of both factors. The main design related factors are selection of components and materials, number, rating and physical placement of components. Failures related to operational conditions are either of deterministic or unpredictable nature. For a MIC the maximum absolute temperature and the number and amplitude of temperature cycles and the humidity during operation is very much predictable. Whereas the electrical disturbances coming from the mains or in the case of lightning from the Solar Module are less predictable since they are caused by the weather or other stochastic events in the electric grid such as failures in other appliances connected to the same grid.
Lo= 67 nH m-".
Figure 9: FEA Simulation results for determination of leakage inductance of the transformer for minimum (a) and maximum (b) distance between primary and secondary winding
3.3.5 Comparison of different transformer realisations Figure 10 shows a comparison of 3 different transformer realisations. The leakage inductance of the ETD29 realisation is not exactly reproducible whereas the leakage inductance of the planar realisation can be defined very precisely.
5.2 Thermal model and simulation As a first step, only design related issues are discussed since the impact of the power electronics designer on the operational factors such as temperature (of the ambient air or the frame of solar module) or humidity is very small.
70
60 50
10
40 30
5
20 10
0
0 Leakage
Inductance [nH]
0K%ound)
Losses Iwl
-
EE32l6i22 single PCB 146nm(4 01) Cu Tracks
0
Frame of the
PCB Area [cm2]
Height [mm]
e P = 22ow
a
EE32/6/P. 2 PCBs oz) Cu Tracks for sec. w h a
70 "(2
Figure 10: Comparison of different transformer realisations
The copper losses have been calculated based on ANSOFT simulations. The number of layers of the multi-layer PCB should be limited to 4 in order to minimise the costs. By separating primary and secondary onto 2 PCBs the copper losses can reduced by approx. 100 mW because in that case the secondary winding can be spread over 4 layers even when 70 pm Cu tracks instead of 140 pm Cu tracks are used. If a secondary winding with 35 pm Cu tracks is used (not shown in Figure 10) the Copper losses would be increased by a further 0.4 W compared to the 70 pm (2 0 2 ) realisation. The core losses are approx. 60% of the total losses. They are calculated based on core data sheets. For the high frequencies used in the presented application (500 kHz) a planar transformer shows a better efficiency than wire wound ETD version.
Figure 11: Temperature distribution at the surface of the MIC
According to [ 5 ] , which is often used for reliability estimations of electronic equipment, the component's temperature has a major impact on it's failure rate. The model for a 3D thermal simulation (representing the physical structure of the MIC presented in Figure4 and Figure 5) and the results of this simulation are presented in Figure 11.The MIC is assumed to be running in rated mode
310
The influence of using encapsulants on the performance of the MIC has been discussed. The failure rate of the MIC has been calculated only taking into account the ageing effects which depend on the absolute component temperature. The real lifetime of the MIC can be determined better by carrying out additional investigations on the impact of temperature cycling and unpredictable electrical failures on the failure rate.
of operation i.e. temperature of frame of solar module 75OC, DC-power 110 W, MIC efficiency 87%. A detailed description of the modelling of MIC and solar module can be found in [3]. MIC Lifetime for worst case and annual average temperature Table 2 summarises resulting hotspot temperature of the critical components and the failure rates of the operational groups including these components. The calculated MIC failure rate according the Military Handbook 217 [5] is 35.7 failure / 106hoperation time which leads to an estimated lifetime of the MIC at a frame temperature of 75°C of only 28,000 h (< 4 years). Investigations of the annual temperature frequency distribution of solar modules show that the annual average temperature of the solar module frame [2,6] and therefore the MIC components is more than 30 K smaller than the worst case (rated frame temperature 75°C). Consequently the failure rate due to decelerated ageing of the components (e.g. electrolyte of electrolytic capacitors) is smaller. The reduced average annual temperature would reduce the failure rate by more than 80% [2]. Hence the lifetime could reach the targeted 20 years if there were no other failure mechanisms, which is certainly not the case. Therefore detailed investigations in particular on the impact of temperature cycling on the failure rate of MIC in real mode of operation have to be initiated.
5.3
7
LITERATURE [l] Development of PV Power Supplies using modular systems design, Report Institute for solar Energieversorgungstechnik e.V., httD:l/www.iset.unikassel.de/, Kassel, Germany, 1995
Table 2: Temperatures and failure rates of MIC components I (Temperature IFailure rate based on [SI I (@ rated mode of operation) 60 % (4 electrolyte capacitors) DC Capacitors 77.0°C 14 % (incl. rest of components) Control 83.5OC 8% (incl. Diodes) Rectifier MOSFET 85.8”C 83.9OC 10% (incl. transformer) Converter MOSFET I 86.2”C I 8% (incl. filter capacitors) 1100 % (= 35.7 failures/106h)
I
[2] Meinhardt, M., et. al.: Reliability of Module Integrated Converters for Photovoltaic Applications, PCIM98, Nuremberg, Germany, May 26 - 28, 1998 [3] Meinhardt, M., et. al.: Power Electronics Packaging of “Low Profile” Module Integrated Converters for Photovoltaic Applications with Improved Reliability, MAPS 98, 31’‘ Symposium on Microelectronics, San Diego, Nov 1-4, 1998 [4] Kats, A.: Application of integrated magnetics in resonant converters, Proceedings of the 1997 IEEE 12th Applied Power Electronics Conference, Atlanta, GA, USA, Feb 23-27 1997 [SIDepartment of defence, United States of America Military Handbook 217 E, Reliability Prediction of Electronic Equipment, October 27, 1986 [6] W. Knaupp: Operational behaviour of roof installed photovoltaic modules, 25* PVSC, May 13 - 17 1996, Washington, DC 8 ACKNOWNLEDGMENTS This work was partly funded by the European Commision in the framework of the JOULE 111 Programme under the contract number EC-DG12- JOR3-CT97-0148 (HICAAP). The authors would like to thank all members of the HZCAAP working group very much for their contribution to this paper -especially: L. Frisson, H. Hotlens, Soltec, Lapeldreef 60, 3001 Leuven, Belgium B. de Meys, G. Vanvijnsberge, IMEC, Kapeldreef 75, 3001 Leuven, Belgium A. De Broe, C.F.A Frumenau, M. Jantsch, ECN, Westerduinweg 3, 1755 ZG Petten, The Netherlands Frank Phlippen, ISET, Koenigstor 59, 34 119 Kassel, Germany
6 CONCLUSIONS A “Low Profile Design” for a Module Integrated Converter (MIC) for Photovoltaic Applications is presented. For the first generation prototype a power density of 0.6 W/cm3 (3.7 W/in3), which is comparable to commercially available AC-DC power supplies, has been achieved by using integrated planar magnetic components. The different design possibilities of the magnetic components are compared according to losses, size and costs. The design has been proven by measurements. The integration of transformer and inductor promises a reduction of size and losses compared to a discrete realisation of transformer and resonant inductor. 31 1
