CAE-Tool for Optimizing Development of Switched Mode Power ...

CAE-Tool for Optimizing Development of Switched Mode Power Supplies N. Froehleke, H. Mundinger, H. Njiende, P. Wallmeier

H. Puder

Institute for Power Electronics and Electrical Drives University of Paderborn, FB-14 LEA Pohlweg 47-49, 33098 Paderborn, Germany Phone & Fax: +49 5251 60 38 81 e-mail: [email protected]

SIMEC GmbH & Co KG Blankenauer Strasse 74 09113 Chemnitz, Germany Phone: +49 371 4503450 e-mail: [email protected]

Abstract-A novel methodology and its implementation into a computer based development tool for switched mode power supplies is described. The tool consists of a number of modules for topology and component selection, circuit, controller, magnetic component and thermal design, embedded into the multi domain simulator SIMPLORER. Core of the tool is the expert system shell CLIPS, which in conjunction with a knowledge base generates choice tables for comparisons and parameterized simulation worksheets. In addition to the modules integrated into the shell, off-line tools for analysis of power electronic topologies and controller design based on MATHEMATICA supplement the tool CAE-WPS 1. A model based design optimizer for magnetic components with solenoidal or planar windings is accessible through SIMPLORER simulation sheets offering considerable improvements in respect to selected cost functions.

I. INTRODUCTION Available simulators used in the power electronic field neither offer a support for selection of switched mode power supply (SMPS) topology and controller assigned to customer specifications nor a means to perform the physical design and automated optimization of magnetic components. Thus, the design duration and the results are heavily relying on experience, qualification and motivation of the designer, which is a non satisfying, intolerable situation. Since the capability to adapt development schemes to new requirements and regulations decides about the future competitiveness of SMPS-producing industries, the objective of the underlying project CAE-WPS 1 for this contribution intends to reduce costs, time to market, required design expertise and to increase the quality of virtual prototyping. This contribution describes a new design methodology for an open and flexible computer aided design tool for SMPS supporting the following functions: • A computer aided selection (CAS) of the best matching power supply topologies using a knowledge base. 1

Funded by the European Community, Contract N° BRST-CT98-5310 (DG 12 - HIAS), Integrative CAE-tools for optimized development of welding power supplies with high power density“, For information search for „CAEWPS“ in "dbs.cordis.lu/EN_GLOBALsearch.html"

•

A computer aided DC and AC analysis (CAA) of power circuitry developed with a computer algebra program generating input for improved model libraries [1]. • Design of power electronic, control circuitry and magnetic components, taking care of high-frequency effects, using a computer aided design program (CAD) supported by the CAA-model libraries, libraries for the selection of components, materials, and circuit and system simulation using the commercial multi domain simulator SIMPLORER [2]. • Thermal design using approximate thermal modeling of power semiconductors and magnetic components, while total assemblies are characterized by thermal simulations and measurements [3] to generate models under SIMPLORER. • Automated Computer aided optimization (CAO) of magnetic components with respect to selected cost functions which might be efficiency, weight etc. by applying enhanced models and state-of-the-art numerical optimization algorithms [4], [5], [6]. Improved models for magnetic components with respect to electrical and thermal behavior upgrade the SIMPLORER and simulation results and lead subsequently to more efficient designs. Note, that presented tool assists the design engineer during the entire development process of SMPS, but since it was started only 2 years ago it does of course not contain already the multitude of PFC-switched mode rectifier topologies for 1 and 3 phases, of dc/dc converters, of controllers presented to literature nor custom specific corewinding arrangements deviating from most common solutions. Instead the designer will find a number of high qualified state-of-the-art circuits/control schemes for high power SMPS so far, while for low power the support will largely be enhanced in the near future. Fig. 1 illustrates the structure of the tool. The left side shows the modules of tool CAE-WPS, which are linked via the expert system to the design process shown on the right side. Note, that the expert system comprises both the knowledge base and the expert system shell, which was implemented using CLIPS [7].

1

II. SELECTION AND PARAMETERIZATION OF A POWER CIRCUIT

ded at this early stage using heuristic knowledge if galva-nic isolation is required, thus increasing the speed of the tool. Aspects of power electronic circuitry are treated within In the next step, stress quantities are calculated for each module CAE-PE (see Fig. 1), which comprises a topology implemented topology using fixed equations. In a later library, a component library, and selection rules. Using version, the designer will be enabled to add new topologies computer aided analysis of converter topologies based on on his own via an interface to the commercial computer Mathematica [1], [8] symbolic expressions for design rules algebra tool “Mathcad”. and stress quantities are calculated and fed into the knowImplemented topologies are: fullbridge, halfbridge, oneledge base. Before calculating evaluation criteria for each and two-switch forward-converter, Passive-Resonantconverter, pre-selection rules are applied, which provide Commutated-Pole (PRCP) - fullbridge and buck converter criteria for exclusion of topologies not suitable for the given with different rectifier configurations like M1, M2, B2 or specification from the further selection process. Non-isolated current doubler. topologies such as buck converters, for example, are discarUsing the stress quantities, components are selected from a ** Iteration loops are not shown ** Specification library by a sub-knowledgebase. Different technologies (e.g. for Topology selection semiconductors: MOSFET vs. (CAS) Topology Properties of PE-circuit & control IGBT) are compared and the best E PE-Circuit PE-circuit C solution is taken according to the X AE D user specified design goals volume, P cost and efficiency. Operating mode Design equations Design: E Parameters for chosen compoPE-circuit O R nents are derived from parameter T studies using data sheets from well C known manufacturers. Loss calcuPart selection lations include on-state-losses at S Topology: Properties of U controller control schemes 100°C and switching losses are baY C sed on linearized voltage-currentS Design: AE Modeling: PE circuit M waveforms for MOSFETs while Design equations magnetic T components IGBT-losses are calculated via E switching energy ratings. Sensors, observers/ E M filters For qualifying topologies, howSimulation: complete ever, part values and stress quanN electrical system tities are computed yielding a means Former projects to compare them in respect to the T design goals. The software module K Optimization: Magnetic core Properties of complete electrical TSelect generates a choice table (see n materials, shapes magnetic system Fig. 2) that serves to select a A o and loss models components topology. Results for expected w Magnetics: absolute and relative volume, costs Design equations l T electr. modeling Design of and losses and the total judgement e mechanical and thermal system value are shown. Also the template d Magnetics: Optimisation I C therm. modeling algorithms sheet-names for the electric- and g AE thermal simulation-sheets are e Simulation and Conducting materials O Optimization of named in this list. and shapes complete therm. and M A ready-to-use SIMPLORER B electr. system simulation model is automatically a N Type of winding s generated with part values and e initial values being parameterized, Prototyping & Verification as depicted in Fig. 3. In addition to Verification with FEanalysis assessing the topologies in respect to specifications and design goals Field Test estimated values for volume, cost, and losses are displayed.

Fig. 1: Structure of CAE-WPS. Blocks on the left side represent blocks that generate data for the knowledge base,

2

Fig. 2: Typical

UDc

C5

TR1

22n



Sap

Sbp

D1

D3

AM1 Ub

AM2

VM1

Csym 13.2u

TR4

D2

D4

LOpt3 295.93u

San

IV. THERMAL MODEL

+

V

+ TR2



A

Ua

t



A

+

ET1 UDc

TR3

based on results obtained by Mathematica packages. Package ConSeaDesT (Controller Selection & Design Tool) is used to parametrize and investigate performance of various controllers based on small and large signal models computed by package SyMap [1]. Knowledge gained by the use of this packages is fed into the knowledge base to assist the designer selecting the appropriate control scheme. To enhance the user interface, control schemes are included in the simulation templates output of the assessment of topologies by TSelect available for each topology. Like the part values, the con-trollers are automatically parameterized by TSelect when generating the electrical simulation model. So far, basic control schemes like peak-, C1 USec average-, [12], [14], and constant-on-time current mode 10u D5 LOpt4 C7 0.00055 RECT1 trOpt3W1 22n control were implemented and tested. A typical LOpt1 85.54u implementation is shown in Fig. 4, which employs both LOpt2 5.93u block diagram and state graph to model the controller and modulator.

Sbn

C4

C6

C8

22n

22n

10u

D6 RECT2

S1Tr := 0.63186 S2Tr := 0.63186 PLH := 0.0310378

PRCu := 1e-6 PLS := 1u

10m RShunt

SRCu1 := 70u SLS1 := 100n

SRCu2 := 70u SLS2 := 100n

Fig. 3: Typical model of the power circuitry (excerpt) used for SIMPLORER simulations

Topologies recently presented to literature [10], [11] were investigated and will be implemented into the tool, too; but because of lack of space in this paper analytical results as well as supporting data gained on laboratory prototypes will be included in a separate publication [15]. III. CONTROLLER SELECTION AND DESIGN The controller selection and design tool (CAE-Ctrl) (see Fig. 1) used to compare different control schemes is also

i"AMSense" -1

Err

-K

KI := wz

o_GW := 0.9

no := 1

no := 1

u_GW := 0

I

P

EXT

KI := -wp

I

BEGR

Rate limiter to := 20u am := i2Set

NEG

I

V. MAGNETIC COMPONENT DESIGN AND OPTIMIZATION dt

Ramp

Pulse-Width-Modulator Start1

TRUE

On Ramp>=dt

AH:=0 to := TEND am := -i2Set/4

BH:=0

AH:=1-AH tPer##Ts/2

The tool CAE-WPS in conjunction with SIMPLORER generates temperature predictions. The use of compact thermal models with simple up to extended resistance networks became only recently a viable tool for system design. Primarily used in electronic equipment cooling the networks are derived either numerically or experimentally, see [15]. However, no other reference is known on parameterization of such models in order to generate an optimum electro-thermal system design. Using TSelect, for each power electronic part the average power dissipation is calculated, either by direct symbolic formulae or via simulations. The dissipated power is input to a thermal model of the magnetic components and the power supply assembly, which comprises power electronic parts as well as heat sinks, fans, screens, wind tunnels, etc. For verification and parameterization of latter component see [16]. Thus, it is possible to locate hot spots within magnetic devices and the assembly before the prototyping stage. An excerpt of a typical model is depicted in Fig. 5.

Off BH:=1-BH

tPer>=1

Fig. 4: Model of average current mode control (block diagram) and modulator (state-graph) (excerpt)

The design of magnetic components is performed within two steps, see [4]. Fig. 6 shows the set-up of the magnetic component design and optimization tool (CAEOMAG) including iterations loops. Starting from the circuit model a simulation is performed at steady state generating approximate stress quantities. These undergo FFT to yield starting values for pre-design/-optimization included in the expert system. It is implemented, based on the area product approach Eq. (1) using general nomenclature:

3

MAXIMUM

MAXIMUM

+

+

'LRGH

V

7UDQVLVWRU

V

MAXIMUM

Tj_max_transistor2

Tj_max_diode2

+ V

42.7259

42.7259

(( (7'

MAXIMUM

MAXIMUM

+

+

'LRGH

V

7UDQVLVWRU

V

Tj_max_diode1 Tj_max_transistor1

42.7259

42.7259

+

AMPowerdissipation_heatsink

A

2.64541 A

FingerScreenin

Fan1

+

A

V

VMPressureDropScreenin -3.06746 V

V

+

+

Heatsink1 V

+

Ambient1

VMPressureRise

VMPressureDropHeatsink

63.8004 V

-57.6654 V

AMVolFlowRate 11.644 mA

Fig. 5: Example of a thermal model used by CAE-WPS (excerpt)

G = AFe ⋅ ACu in conjunction with ∆Tmax = Rth ⋅ PV = Rth ⋅ (PCu + PFe ) with PFe

 f ⋅ kF = pn ⋅ V Fe ⋅   fn

PCu =

J Cu , RMS

κ Cu

   

ef

 Bˆ ⋅ k I ⋅  B  n

   

(1) TABLE I

(2)

DEFINITIONS, RELATIONS AND CONSTANTS eB

(3)

Variable Average winding length

2

Shape specific constant

Relation l m =k l ⋅G 0.25

k l , per unit

AW =k A ⋅G

k A , per unit

m

m

⋅ VCu

S = U1, RMS ⋅ I1, RMS =

(4)

Eff. core cross section

1 ⋅ kT ⋅ G ⋅ Bˆ Fe ⋅ J Cu , RMS ⋅ f 2

[ ]

1 1 − +   2 e B

 

  

  S  ⋅ [VA]   

(5)

eG

[ ]

W

3 4

 G  l ⋅ L + k Fe ⋅  4  m  [m]  

[ ]

er +

3 − eB 4

l  ⋅  L   [m] 

AFe = k A ⋅G

k AFe = 1 k , ACu

0.5

Fe

Core volume

per unit k VFe , per unit

VFe = k V ⋅G 0.5 Fe

(6)

e /e    1 1  2 ⋅ k Ff B  e  3  kG = +  er +    1 + B  , eG = 1 −    kT kW k K  2  4       e B 2  er −

0.5

W

ef  −1   f  eB  ∆T G    = kG ⋅   ⋅  max 4  m  [K ]  fn    for transformers with

 G  ∆T (G, ll ) = kCu ⋅  4  m  [K ]  

Winding window

−

eB 2

−1

(7)

Thermal resistance

 G  Rth = R0 ⋅  4  m 

er

er , per unit R0 in K/W

R0 K  2 W  L  I eff      and ⋅ ⋅ kCu = k l m ⋅ µ0 κ [H ]  [A]  ⋅ ⋅kf  Vs   1   Am   Ωm     

for inductors with

k Fe =

eB 2 kVFe k Cu

 B   2µ 0  n   p n R0   [T ]   ⋅    W   K    Vs    3   W    Am     m     

eB 2

2   1 L  k I ⋅ iˆ   ⋅ ⋅  2 [H ]  [ A]       

eB 2

 kF ⋅ f     f   n 

ef

4

This pre-optimization algorithm provides data about the expected volume, losses and temperature rise of a fixed core and winding set-up of inductors and transformers, inputted via a graphical user interface, see Fig.7 a). Thus, initial values for a subsequent optional parameter optimization of these components are derived. The core and winding dimensions, like e.g. the air gap width, the layer thickness in gapped inductors and in transformers are optimized with respect to design objectives such as minimum size, minimum temperature rise, costs, etc., summarized in objective function F, see [4], and left loop of Fig. 6 and exemplary shown in Fig. 7 b). The underlying electrical winding model, briefed in [4] and

simulation until f(t)=f(t+T)

minimisation of F(X)

F(X)

parametervector x

cost function

boundary conditions g(x)

starting values

stress quantities

temperature rise losses (coupling)capacitance (short circuit inductance) mass / volume

circuit model

FFT

Z(ω)

calculation 1. geometric dimensions 2. mass 3. volume 4. flux density 5. core losses 6. winding losses 7. magnetic permeance 8. capacitances 9. impedances 10. temperature rise

model core model winding thermal models

Fig. 6: Design and Optimization tool CAOMAG – setup and iteration loops

a)

b)

Tmax

Pv

Orig.

163 °C

112 W

Optim.

136 °C

85 W

c)

Fig. 7: : a) A part of a SIMPLORER-worksheet of the 5 kW-welder (10 kW peak) and the Wizards for a transformer b) Result of a parameter study and c) comparison of optimum and current design.

elaborated in [5], supports planar and solenoidal wound gapped inductors and transformers using foil, litz or round wires and considers all relevant parasitic effects like eddy current losses, leakage, gap effects and core losses including rate dependencies and capacitive effects. Skin and

proximity effects, caused respectively by inner current density displacement and by eddy currents, are considered in a way that considerable accuracy improvements result [6], [13]. Edge effects are considered by manipulating the skin effect factor after FEA analysis of foil windings in solenoidal

5

10

1000 Rk01=38mΩ

Rk02=41mΩ

Core EE 65

Lk µH

FAC

2

Simulated

2

10

1

1 Simulated

1

1 1

10

100

1000 f/[kHz]

Connection Data Cs: 1000000000000000000000 0000000000000000000001 0111111111100000000000 0000000000011111111110

10000

1

10

100

1000

10000

f/[kHz] Cp: 1100 0010 0001

Layers: Primary: 1.4 mm round wire, Secondary: 0.2 x 36 mm Cu-foil

Ni: 20,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,20 leg_info: 0000000000000000000000

Fig. 8: Measured and simulated resistance ratio of 5 kW - transformer design example and its winding data

structures and multi-turn windings in planar structures were evaluated. To account for gap effects caused by the fringing field in air gaps, a method utilizing the current sheet technique and a separate field solution for induced fields was implemented for linear gapped solenoidal inductors in [6] and [13]. The core is divided into straight and curved segments whose respective losses are summarized to the total core losses. Rate dependencies of losses are taken in account, too. The reluctances of the segments are calculated using mean cross section and path length formulations. The influence of air gap fringing is considered by applying conformal mapping techniques. The capacitance model uses a plate capacitor approximation assuming a radial electric field for solenoidal and an axial field for planar structures. VI. EXPERIMENTAL RESULTS Fig. 7 a) shows an excerpt of a schematic of a power stage of a 5 kW (10 kW peak) welding power supply and the user interface of the transformer design and optimization together with the objective function (see Fig. 7 b)) using a constant core but variable primary and secondary conductor thickness. Note, the two major subminima caused by the discontinuous change in parallel connected wires to fill the window as good as possible. Losses of the original design are reduced by at least 24 %, which leads to a temperature decrease of 27 °C. The model accuracy compared to measurements of the original transformer is depicted in Fig. 8 using the short circuit impedance as model quality indicator. The model reveals good agreement with the measured values as well for the resistance as for the inductance. Please note, that the winding consists of two different winding types, complicating the modeling tremendously. The matrices Cs and Cp code the series and parallel connection scheme of the layers. Entries of 1 in the

i-th and j-th column of a row in Cs indicate a series connection of these layers. Analogous is true for the parallel connection. For a planar transformer based on 4 EE80 cores with 1.16 kg on cores and 0.34 kg on winding designed by a non automated optimization tool the novel tool facilitated an optimized design based on a single E64/10/54 core with only 0.21 kg on core and 0.16 kg on winding. The thermal flow increases of course for latter design as compared to the conservative first design but still the peak temperature rises only to 57° C at an ambient temperature of 40° C. VII. SUMMARY A survey about a CAE-tool for the optimized development of SMPS is presented incl. its underlying methodology, supporting the selection of power and control topologies by an expert system, the magnetic component and thermal design by an enhanced modeling base for transformers and inductors, embedded into the multi domain simulator SIMPLORER. Additionally, off-line tools for analysis of the ever increasing multitude of new power electronic topologies and controllers based on MATHEMATICA supplement the tool to facilitate an efficient evaluation and comparison. A design optimizer for magnetic components with solenoidal or planar windings is accessible through the SIMPLORER simulation sheet, which proved its capability on various applications by considerable power loss or volume reductions. The modularity of the tool allows to adapt it for future requirements. ACKNOWLEDGEMENT The authors acknowledge gratefully the financial support granted by the European Commission under project BRST-

6

CT98 5310, the co-operation with University of Leuven and the support by the industrial partners of the consortium.

[14]

J. SUN, R. M. BASS: “Modeling and Practical Design Issues for Average Current Control,” APEC 1999, Dallas TX, USA, March 1999,

REFERENCES

[15]

F. CHRISTIAENS: “Thermal Modelling and Characterization of Electronic Components: Steady-State and Transient Analysis,” Ph-D Thesis, Kath. Univ. of Leuven, Belgium, 1998.

[1]

J. SUN, H. GROTSTOLLEN: “A Symbolic Computation Package for Averaged Analysis of Power Electronic Systems,” IEEE Applied Power Electronics Conference (APEC) 1996, pp. 96-102.

[2]

SIMEC GMBH: “Reference Manual to the simulator system SIMPLORER,” Chemnitz, Feb. 1999.

[3]

J. HENNISSEN, W. TEMMERMAN, J. BERGHMANS, K. ALLAERT: "Modelling of axial fans for electronic equipment", Electronics cooling, Vol. 3, 1997.

[4]

P. WALLMEIER; N. FRÖHLEKE; D. HAHM; H. MUNDINGER: “Integrating magnetic component design and optimization into circuit simulator SIMPLORER,” Power Conversion and Intelligent Motion Conf. (PCIM) 2000, Nuremberg, June 2000, pp. 533-538.

[5]

P. WALLMEIER: “Automatisierte Optimierung von induktiven Bauelementen für Stromrichteranwendungen,” Ph.D-Thesis, in German, University of Paderborn, 2000.

[6]

P. WALLMEIER, H. GROTSTOLLEN: “Automated optimization of high frequency inductors,” IEEE proc. of ann. conf. of the IEEE Ind. Electr. Society (IECON) 1998, Aachen, Germany, pp. 342-347.

[7]

STB LYNDON B. JOHNSON SPACE CENTRE: CLIPS “User Manual,” Aug. 98, www.ghgcorp.com/clips/CLIPS.html.

[8]

N. FRÖHLEKE, R. PIEPER, H. MUNDINGER, H. GROTSTOLLEN: “Computer Aided Investigation of the Auxiliary Resonant Commutated Pole Converter for Wide Range Applications such as Welding Power Supplies,” IECON 99, vol.2, San Jose , CA, USA, Nov./Dezember, 1999, pp. 891-896.

[9]

N. FRÖHLEKE, H. MUNDINGER, A.O.: Resonant Transition Switching Welding Power Supply. Int. Conf. on Ind. Electr. Contr. & Instr., IECON 97, pp. 615-620.

[10]

F. CANALES, P. M. BARBOSA, F. C. LEE: “A Zero Voltage and Zero Current Switching Three Level DC/DC Converter,” APEC 2000, New Orleans, Louisiana, USA, Feb. 2000, pp. 314-320.

[11]

R. AYYANAR, N. MOHAN: “A Novel Full-Bridge DC-DC Converter for Battery Charging Using Secondaryside Control Combines Soft-Switching over the Full Load Range and Low Magnetics Requirement,” APEC 2000, New Orleans, Louisiana, USA, Feb. 2000, pp. 340-346.

[12]

P. COOKE: “Modeling Average Current Mode Control,” APEC 2000, New Orleans, Louisiana, USA, Feb. 2000, pp. 256-262.

[13]

WALLMEIER P., GROTSTOLLEN H.: Improved modelling of inductive losses in gapped HF-inductors. Proc. of IEEE Industrial Applications Society Annual Meeting (IAS) 1998, S. 913 – 920.

[16] H. MUNDINGER, M. SCHNIEDERMANN: N. FRÖHLEKE „Enhanced Analysis and Design of a 3-Level DC/DC Converter with Zero Voltage and Zero Current Switching“, proposed to be presented at EPE 2001, 27.29. August, Graz, Austria

[17] K.V.DAMME, M.BAELMANS: “Thermal modelling of welding power supplies with high power density in a CAE-environment”, 6th Int. Workshop on Thermal Investigations of IC’s and Systems (Thermonic), Budapest, Sept. 24th-27th 2000

7