COMPARISON OF PLANAR AND WOUND TRANSFORMERS FOR FLYBACK FORWARD AND HALF-BRIDGE SPACE POWER CONVERTERS Thomas Björklund (1) John Andreasen (2) Finn Brøsen (3) Erik Matthiesen (4) Ole Poulsen (5) (1) Magnetics Engineer, Flux A/S, Industrivangen 5-11, DK 4550 Asnaes (Denmark), tb@flux.dk (2) Magnetics Engineer, Flux A/S, Industrivangen 5-11, DK 4550 Asnaes (Denmark), ja@flux.dk (3) Power Engineer, Flux A/S, Industrivangen 5-11, DK 4550 Asnaes (Denmark), fb@flux.dk (4) Power Manager, Flux A/S, Industrivangen 5-11, DK 4550 Asnaes (Denmark), em@flux.dk (5) Technical Manager, Flux A/S, Industrivangen 5-11, DK 4550 Asnaes (Denmark), op@flux.dk ABSTRACT Planar technology has now entered the space domain. The big advantages of planar technology are; • • • • • Low profile Excellent repeatability Economical assembly Mechanical integrity Superior thermal characteristics This is why the general power industries increasingly are using planar magnetics in more and more applications, and therefore also why we see a rising demand for the usability of the planar technology among space application developers. The differences between wound and planar transformers have been mapped with a detailed look on the various parasitic component values, such as DC- and ACresistance, Leakage Inductance and stray capacitance, and revealed the magnitude of the advantages of planar technology. This technical solution is proven in prototypes that have been built in different combination of PCB’s and copper foil, with more or less interleaving of windings. Furthermore the transformers have been designed with several outputs stacked together with a fairly high number of primary turns, in order to have planar transformers similar to the wound types that are generally used for space applications. demands regarding the thickness of copper, laminate tolerances and a strict control of etching the thick copper foils and laminating with minimal resin, in order to get a good fill factor in the final transformer. Using PCB as a part of a component, is breaking with the general thinking in the technologies of space qualified PCB - and this divagation is necessary in order to achieve functionality that can compete with traditionally wounded transformers. Where high current is occurring and thicker copper layers are needed, folded copper foil windings makes a great alternative to the PCB, and this pioneering mix of technology within planar transformers, gives a whole new flexible design frame for the space engineers. Quality assurance is of utmost importance for the planar transformer, where thermal drain, z-axis expansion and via or Plated Through Hole (PTH) reliability, during thermal cycling, are some of the most critical points. It needs to be at the same low risk as traditionally wound transformers, so tests and verification of functionality out of the normal ECSS specified frame of use is needed. This points to pioneering thermal tests that can verify the life terms of the space planar transformer. I.e. passive burn-in, power burn-in and thermal shock, but also endurance testing to show the limits of this new component technology. The different topologies are selected to represent three different usages of the ferrite shown in Fig. 1, Fig.2 and Fig. 3 INTRODUCTION Obtaining the optimal design has been impossible when engineers have developed printed circuit board (PCB) windings using the frame of The European Cooperation for Space Standardization (ECSS) standards for space qualified processes. Where the requirements towards the PCB manufacturers are driven in two main directions; fine pitch technology for microprocessors and impedance control for Radio Frequency (RF) domain circuits, which is also reflected in the Interconnecting and Packaging Electronic Circuits, now the Association Connecting Electronics Industries (IPC) standards. Here the Planar technology comes with greatly different Fig. 1 B-H Curve of a flyback transformer d) Tape & PCB vs. Coil former Using a standardised bobbin or coil former, gives a high reliability component since it is well tested with regards to creepage and clearance distances. Tape and PCB structures gives an individual design each time with tolerances that needs verification testing dependent on the tolerances. Fig. 2 B-H Curve of a forward transformer Fig. 4 Planar EE18 transformer compared with RM5 (Left side), and Planar EE22 transformer compared with RM10 (Right side) 1. Fig. 3 B-H Curve of a half-bridge transformer 2. THE TRUE COMPARISON The optimal design for transformers is normally set to be the balance of core and winding losses Fig.5, here planar transformers can be designed towards a higher core loss in order to gain less winding loss, because of the better thermal drain possibility. When comparing a standard transformer with a planar transformer there are actual four different points of comparison: a) THERMAL PERSPECTIVE EELP vs. RM Here the core shapes are different in standard sizes. Designing the winding structure sets the criteria for the needed window, which results in selecting a bigger core than initially needed. RM cores enclose the windings better than the EE cores, which gives a better EMI performance. b) PCB vs. Wire Here the PCB is limited in winding-techniques and has a thermal expansion factor, where the wire can be wound in wrong number of turns from time to time, and has pulling and bending issues. c) Fig. 5 Power loss balance (1) Flat vs. Round A thin flat conductor doesn’t suffer from skin effects in high frequency alternating current (AC), like a round conductor. But Flat conductors couples better to each other hence generates a larger parasitic capacitance. (2) There is also the possibility of better thermal drain of the winding structure, when using PCB for windings. 3. MECHANICAL INTEGRATION One of the most obvious advantages of planar magnetics is the different possibilities of integrating the transformer with the converter PCB. Stand-alone is the common use of wound transformers, but sunken wound transformers are also seen. The winding construction of a PCB transformer can also never reach as high a mass density as the wound transformer. This is due to the maximum possible fill factor Fig.8 and the material density Fig.9, where polyimide and plastics are lighter than copper. Fig. 8 Typical Max Fill factors (Ku) for standard and planar magnetics Fig. 9 Density of transformer materials Fig. 6 PCB mounting options Planar transformers are often integrated, but here the tradeoffs between PCB layer build-up and transformer performance becomes an issue. Even a small modification in the transformer, demands for a whole new layout, where a stand-alone model easily can be prototyped to the right fit, by removing or adding a single turn. A big advantage with stand-alone transformers is the possibility of using several multilayer PCBs stacked together, and even in combination with foil or wire windings. Here the Hybrid model also has advantages, when high secondary current can be placed in thick copper foil combined with primary PCB windings, or when multiple primary windings can be placed in thin wire combined with the lower number of turns secondary PCB windings in the converter board. This points on a more mechanical robust transformer, however only if they are constructed with the same size of core and same window area. It will then give a higher current density in the planar transformer, and this is also often the case in the design as the thermal transfer is more efficient. 4. PCB TOLERANCES When designing transformers, it is very important to take thermal expansion into consideration due to the fragile core that easily breaks in stressed conditions. Introducing PCB is actually rather troublesome, as the glass woven fibres are preventing x-y expansion and instead transfer all the thermal expansion to the zdirection. Which is exactly what is needed for a stable Surface Mount Device (SMD) on a PCB, but absolutely a bad idea for fitting into a transformer core. 3.1. Fill Factor and Robustness When looking at the RM wound transformer and the EELP PCB in Fig.7 from a vibration point of view, it is clear that the Centre of Mass Point is lower on the planar core, as the core is lower. This gives a more robust mechanical construction, despite of the windings. Fig. 7 Centre of Mass Point The core dimensions comes with big tolerances of window height shown in Fig.10 for EELP18 which is ±5%, but it is worse for the final PCB where the tolerances is ±10%. Fig. 10 EELP18 Core tolerances Examining the manufacturing process reveals that the laminate and prepreg process is dependent on the resin content, which depends on the heat and flow when filling the bare woven fibre sheets, where the viscosity also depends on the stock-age of the resin. For a 100µm laminate the tolerance is as large as ±25% according to IPC as shown in Fig.11 clearance. In Fig.14 is shown the microsection of the EELP18 transformer windings. The ECSS standards states max ±100µm. Fig. 11 Raw laminate tolerances Additional the copper plating also has tolerances and since copper is expensive, the manufacturer aims very precise on the minimum tolerances Fig.12. Fig. 14 Flyback transformer microsection 5. DIELECTRIC CONSTANT When it is mentioned that the Planar transformer repeatability is excellent, it is needed to take a deeper look into predicting the parasitic capacitance according to production tolerances. The variations of dielectric constants for polyimide PCB Fig.15 appears when the resin has a lower value than the fibre glass, and the resin content varies. Fig. 12 Innerlayer copper thickness When plating the drilled through holes, extra copper is added to the entire surface. This gives an additional thickness Fig.13 Fig. 15 Dielectric constant and resin content of prepreg Calculations due to laminate thickness tolerances and dielectric constant variation results in huge differences. Fig. 13 Outerlayer copper thickness after plating (class 3 is for space and aerospace) A final thickness of a 70µm winding is therefore in the area of 55,7µm to 83,7µm depending on the PTH structure. Fig. 16 Capacitance variations of laminate thickness Fig. 17 Capacitance variations of moist content 4.1. Milling and registration Other PCB tolerances, such as milling of the edge and registration of the layers in-between are also the cause for poor fill-factor, and this is a critical point regarding Polyimide is hygroscopic, and water has a dielectric constant of 80,40, so within the max accepted value of 0,28% the variation is a few extra pF as in Fig.17 6. EMI AND EMC OF CORE SHAPES Electromagnetic Interference (EMI) and Electromagnetic Compatibility (EMC) are the unintentional generation, propagation and reception of electro-magnetic energy of the transformer interacting with the surroundings. This is the almost impossible part to calculate when designing the prototype, and it very much depends on the core shape and usage. 6.1. Induction formulas When looking at the standard textbook formulas for transformers and how they are constructed it appears that the core-shape differences are not a part of the calculations. The induction formulas are derived from two different systems, where the primary system is based on a magnetic field from an endless long straight lin shown in Eq.3-5 And Fig.20 and the self-inductance is based on a toroidal solenoid Eq.9-10 And Fig.22 The energy propagates in a combination of magnetic (Bfield) and electric (E-field) radiation as shown in Fig. 18 Fig. 18 Electromagnetic(EM) wave The waves are only polarised in the Far-field area, which is more than 2 wavelengths from the source, and at 125kHz the wavelength is 2.4km long. In the Nearfield the EM wave is much more complex and can therefor not be measured exactly on the two kinds of transformers. Fig. 20 Endless long line The E-core shapes in general are open and the windings are wound outside the core structure. The Pot-core is very closed and has the best EMI performance, however the closed core makes it difficult to enter and exit the windings. The RM core is therefore a fairly good solution for EMI shielding of the windings. Fig. 19 shows the differences drawn with a Mean Length Turn (MLT). (3) (4) (5) Figure 21 reveals that the distance 2 x, where x is the radius, is the same as the Magnetic Path Length (MPL), when calculating the transformer. Fig. 19 Pot-, RM- and EELP cores with windings and B-field lines indicated (Blue arrow) Then of course there is the factor of relative ferrite permeability µ r vs. the vacuum permeability, which is also the permeability of air µ 0. In N87 and 3F3 the factor is around 2000. Fig. 21 relationship of 2 r and MPL In the Ferroxcube handbook, Eq.6 Is given for the magnetic field strength based on the effective MPL denoted le. (Square root 2 is used only for a sinus) 7. TEST RESULTS The prototypes in Fig. 4. are defined by the winding turns (T) shown in Fig.24 RM5 (6) * EELP18 * 31T (7) (8) 3T 16T * * 7T Where Eq.7 and Eq.8, From the Fundamentals of Power Electronics theory book, shows that this is based on a straight endless line. The principle of a toroidal solenoid Fig.22 gives the base for equations where the magnetic path length is a circle. 2T 5T Fig. 24 Flyback Transformer turns ratio The mass budget is: RM5: 5,12g of which the core and coil former is 3,61g EELP18: 8,01g of which the core is 4,81g Dimensions are: RM5: H=10,52mm W=15,81mm EELP18: H= 8,09mm W=19,23mm L=15,95mm L=22,52mm Measurements have been performed on the prototypes revealing the magnitude of the design. Fig. 22 Toroidal solenoid (9) (10) When 2 r equals the MPL, it is only true for toroidal cores. Not for RM and EELP core structures. Hence the differences in shape is ignored in the transformer calculations Fig.23. Fig. 25 Flyback Transformer measurements Fig. 23 Transformer equation (MPL is denoted lm) When the shape of the magnetic path length is ignored, the contribution, due to one or another shape, cannot be defined. Comparison of the measurements Fig.25 shows that they are overall the same, but the planar transformer has lower leakage, which is around half of the standard transformer, but also around four times as large capacitance’s. These values are also the commonly expected. The prototypes shown in Fig. 4. are defined by the windings in Fig.26 RM10 * EELP22 * 5T 3T 4T * 3T * 7T 7T 5T Comparison of the measurements Fig.27 shows that the two forward transformers over all are the same. However here the wound transformer has lower leakage and higher capacity than the planar transformer Fig.28. This was anticipated, but not the most intuitive result. 7T 4T 7T Fig. 26 Forward Transformer turns ratio Fig. 28 Forward Planar Transformer The mass budget is: RM10: 33,75g of which core and coil former is 27,97g EELP22: 21,74g of which the core is 12,96g Dimensions are: RM10: H=18,61mm W=29,61mm EELP18: H=11,76mm W=24,15mm The reason for having a better performance on the wound transformer lies within the possibility of more complex interleaving structure Fig.29 and Fig.30. L=39,50mm L=33,96mm Measurements have been performed on the prototypes to reveal the magnitude of the design. Fig. 29 Interleaving structure of Planar Transformer Fig. 30 Interleaving structure of RM Transformer (T is short for Turn) 7.1. Measuring tolerances When measuring the leakage, it is ideal to short the secondaries, so that only the none-coupled inductance is measured Fig.31. Shorting with a wire adds inductance to the circuit, hence the measurement has a large tolerance relating from the inductance of the short. Fig. 27 Forward Transformer measurements Fig. 31 Simple and complex attempt to achieve a perfect ac short 8. DISCUSSION For both traditional and planar transformers, it is possible to have different structures of interleaving, which is a choice of leakage vs. stray capacitance. Prototyping the planar transformer is very costly, but here simulation SW really comes in handy. • • • Low building height Mechanical integrity Superior thermal characteristics - The drawbacks are; EMC and winding area When looking on introducing PCB windings, the thin layers and photo-etching process has its advantages: • • • • • Low building height Excellent repeatability Economical assembly Mechanical integrity High frequency performance (to some extend) - The drawbacks are; prototype costs, reduces possible number of turns. PTH or VIA reliability. Moist sensitivity, thermal expansion and production tolerances. 10. REFERENCES Fig. 32 Test converter mock-up Test converters Fig.32 has been built so that the RM transformer and the EELP transformer can be swapped. This gives the possibility of testing load and efficiency capabilities, and the general picture is as expected; the one with lower leakage performs better. But as they were designed on the same criteria they are overall the same on performance. Simulating the transformer makes it possible to predict the leakage and stray capacitance by Finite Element Analysis (FEA) but here tolerances are not included, and Monte-Carlo plots is not possible due to standard Central Processing Unit (CPU) performance. 1. Robert W. Erickson & Dragan Maksimovic. Fundamentals of Power Electronics, Springer; 2nd edition (January 2001). 2. Colonel Wm. T. McLyman. Transformer and Inductor design Handbook, Third Edition (2004) 3. Soft Ferrites and Accessories Data Handbook. Ferroxcube (2008) 4. Hugh D. Young, Roger A. Freedman. University Physics with Modern Physics with Mastering Physics. Addison Wesley; 11 edition (August 8, 2003) 5. Generic Standard on Printed Board Design. IPC2221A, IPC (May 2003) 6. Chet Guiles, Rancho Cucamonga. Everything You Ever Wanted to Know About Laminates, but Were Afraid to Ask. Arlon-Materials for Electronics, 9th Edition (Nov 2008) Fig. 33 Capacitance model of a transformer The EMC performance of the actual core can neither be calculated nor measured precisely. The only way of verifying the differences, due to core shapes, is to design a PCB for each converter, with individually optimised snubber circuits due to the different pin positions on the RM and EELP transformer and the layout. This part must be an issue that needs to wait for a future study. 9. 7. Rafael Asensi, Roberto Prieto, José A. Cobos, and Javier Uceda. Modeling High-Frequency Multiwinding Magnetic Components Using FiniteElement Analysis. UPM, División de Ingeniería Electrónica, Madrid 2008 Thomas Björklund Magnetics Engineer, Flux A/S www.flux.dk CONCLUSION When looking on the core structure, the lower and plane core has its mechanical and thermal advantages: Web discussion group :