Design of an Inductive Contactless Power System for Multiple Users

Design of an

Inductive Contactiess

Power

System

for

Multiple

Users

Fredrik F. A. Van derPijl, Jan. A. Ferreira, Pavol Bauer, HenkPolinder

ElectricalPowerProcessinggroupof the Faculty of Electrical Engineering, Mathematics and Informatics

Technical University of Delft Delft, The Netherlands

Abstract- The technique of inductive contactless energytransfer

(CET)hasfound itsapplicationin specificindustrialapplications

where an (air) gap between energy source and load ispreferable

or even required. A relative new application area of CET is where CET is part of a general energy distribution network. As anexample, electrical socketsmight give problems in bathrooms forsafety reasons: CETprovides a solution because of inherent electrical isolation between source (grid) and loads (e.g. hair dryer or electrical shaver). For such application this paper reports on thedesign of a contactless power system that is able to supply to multiple users simultaneously, analogously to the regular plug-and-socket network. The proposed contactless system consists of a special energy supply cable where clamps (eachmaximum 240W)canbe mounted on atarbitrarylocations along the cable while electrical isolation is maintained between cable andclamps.Resultis asupplycabledesignand aprototype with 2 clamps, which are compared by means of analytical calculations,simulations andexperimentalresults.

Keywords-contactless; energy transfer; inductive coupling; multipleoutput; powertransfer

I. INTRODUCTION

THE technique of using an inductive magnetic couplingto transfer energy has already found its use in several

specific industrial applications. For example, in the research field ofprecision positioning, energy supply cables whichare attachedto amoving object disturb the positioningprocess of the object [1]. In the latter example the contribution of contactless energy transfer, increasing the precision, is

apparent. Precision positioning and other specific industrial applications benefit from contactlessenergy transfer asbeing theonly alternative [2] [3] [4] [5].

This paper explores a different type ofapplication, where contactless transfer is part ofa general energy supply chain, rather than that it performs an integrated function within a specific electrical device.

Focus is on applications were energy supply is required to be geometrically flexible. Examples form exposition halls and music stadiums, where the position of (electrical) equipment differs per exposition and concert. Inthispaper a contactless

system is proposed which inherently is flexible in use: electrical devicescanbe clamped on(bymeans ofauniversal

clamp) at an arbitrary position along a special electrically isolated supply cable. Thereby we avoid the problem of the currentplug-and-sockets, whereaplugcanonly be connected

to oneofthe(few) sockets.

Also, such electrically isolated supply cable canbe applied

in wet environments, in for example bathrooms or for the underwater-equipment ofprofessional divers.

Objective for this paper is to introduce an energy supply cable, basedon inductive energytransfer, which is accessible everywhere along the cable by multipleusers. The prototype

that has been build contains two clamps which output DC

voltage and current for apurely resistive load (e.g. user). Itis noted that arelatedapplicationarea from literature consists of

contactless planar DC charging platforms for (mobile) electrical devices[6] [7].

Theory behind theDC cabledesign is not new [8] [9], but results will be used for near-future work on anACmulti-user cable design. Also, a (feedback) controller has not been implemented for theprototype, while focus is on steady-state

system functioning at a large load-range. The design makes useofresults from [10], [11] and [12].

ChapterII introduces theconcept. ChapterIII discusses the practical design and a model of the system. In chapter IV

simulation methodology and implementation is explained in

order to compare analytical and simulation results with experimental results in chapter V. Chapter VI concludes the work.

II. CONCEPTDESCRIPTION

Theproposed contactless supply cablecan be compared in

converter | Load

Fig.1. Proposed contactless system withtwo clamps attached and one

ready to be connected. Arrows next to the conductorspiral denote current

direction

Each clamp-cable connection can be seen as a special transformer. Toobtainareasonablemagneticcoresize for the clamp-cable transformer the system operating frequency is requiredto be several orders ofmagnitude larger than 50 Hz.

Therefore, a frequency-converter is required between the utility grid and the proposed cable (e.g. input converter), which is abundant in the standard system. For the same reason, an output converter in between clamp and load is required.

A. Clamp-cable transformer

The contactless cable consists of a single (primary) conductor that is wound according to Fig. 2. Result is an inner-winding section andan outer-winding section where the current in both sections flows opposite. Ifthe outer-section windings are distributed equally and tight across the outer periphery, zero-magnetic field outside the cable is obtained (by the cancelingcurrents).

Totransfer energy from the cable to a clamp the magnetic coresoftheclampmustencircleoneof bothwinding sections.

Itwasdecidedtopull-out the inner-winding section locallyat which position a clamp is attached. This action can be compared to putting a plug in a socket. Each clamp is composed oftwoseparatable magneticcorehalves andafixed (secondary) winding.

Fig. 3 shows that the standard cable isaparallel system (in

electricalcurrentsense), while the contactless cable is a series system. Theinput currentis distributedacrossthe loadsinthe standard system while being shared by all clamps in the contactless system.

utility gnd

Fig. 3. Parallel plug-and-socket systeminelectrical current sense.Arrows

denote current direction.

B. Inputandoutput converters

The input converter in between utility grid and cable has

twotasks. First task isto converttheutility's50Hzto ahigher frequency to meet a reasonable magnetic core-size for the clamp. Second task is to regulate the powerthroughput from the grid to the variable number ofclamps. In the prototype

each output converter consists of a passive diode bridge rectifier with filter capacitor, togenerate a DC outputvoltage acrossthe resistive loads.

C. Design requirements

The proposed system is intended for domestic and office environments with a power range from W to kW. The

prototype considers two clamps of each 240 W maximum. Nominaloutputvoltage isset at240 V,which is the rms-value of the50 Hzutility grid voltage. Therefore, theoutputcurrent rangesfrom(near)0 to 1 A.

All system components are rated at maximum power

throughput. Inpractice, theaveragenumber ofclamps thatare attached at a time instant will be (much) less than the maximum allowed. Therefore, applicability of the proposed

system depends on the wide-range efficiency. This paper

explores maximum (practical) efficiencysystem conditions for thecomplete loadrange.

III. SYSTEMDESIGN A. Clamp-cable transformer

Ferritemagneticcores arechosentomaximize the inductive coupling factor of the clamp-transformer. A high coupling factor results in a large magnetizing inductance. A coreless design would avoidcoreloss, but this advantage will be small compared to the increase in copper loss. Fig 4 shows the situation wherea single clamp is connected with a small part

of the inner winding section of the cable in Fig. 2. Core

material and dimensions asused intheprototypeare givenin

B

4- C

Gi

Fig.4. Single clamp configuration as used in the prototype, with dimensions A-E anda) inner winding section of cable, b)U-core, c) clamp winding section

TABLE I

PROTOTYPE TRANSFORMER DIMENSIONS FOR TWO-CLAMP CABLEAS REFERREDTOFIG. 4

Symbol Quantity Value

- corematerialgrade 3C80

- coresize U25/20/13

A 46.5[mm]

B 14[mm]

C See Fig. 4 3.16[mm]

D 3.25[mm]

E 30[mm]

GI air gapwidth ofclamp1 0[mm]

G2 air gapwidth ofclamp2 0[mm]

Objective in choosing a suitable core is to find an optimal balance between copper and iron loss to minimize their combined loss. From athermal point of view the transformer

must have sufficient surface to remove the loss by heat convection with thesurrounding air.

The magnetic core size and material are determined accordingtothe transformerdesign procedurein[12]:

1. Choosea material with thehighest performance factor atthe nominaloperating frequency ofthesystem 2. Minimize total (core and copper) loss for a range of

available core sizes by tuning the core flux density, whichguaranteesoptimaluseof each individualcore 3. Calculate maximum allowed transformer loss, which is

given by the maximum heat removal capacity

4. Choose the minimumcoresize that doesnotexceed the maximum allowed loss

It is remarked that this procedure gives only an indication for a suitable core material and size. Further detailed design improvementsasforexamplein[ 13]are notcovered.

First step isto choose amaterial withhigh performance at thesystemoperating frequency. The expression for iron loss is givenin

(1),

empirically determined and given bythe manufacturer. Their relative values determine a suitable frequency range for the material for a given power throughput. For example, from [5, p. 747]it is clear that ferrite grade 3F3 is best for applications

from40 to 420kHz.New corematerials (lately the amorphous cores) are continuously under development with even higher (local) performance. Bisthe peak flux densityinthecoreand Vcore is the corevolume. The sum ais taken overthe number ofclampsnclamps.

Copperloss isgivenin(2),

nclamps

, PCu,a

a=l

nclamps nclamps p- a2

= EI

RC,

ala2

Pa2

a=l a=l WCU

(2)

with copperresistance RCU and rms-value of the current in

thecopper la. Thesum ais takenoverthe cable withsubscript

0 and all secondary sides, which are equal to the number of clampsnclamps. Resistance is subdividedin the specific copper

resistancep,length ofasingle turn la, number ofturnsnaand the total copper cross sectional area of all windings WCU,a. WCu,a divided by na is the cross sectional area of a single conductor.

Though optimum flux density in the core depends on the number of clamps, for clearness further design steps are

performed for a single clamp. For multiple clamps the procedure is thesame.

Second step is to determine theoptimum flux density for a specificcore sizetoobtain minimal lossattherequiredpower

throughput. Flux density, and thus optimum density, is controlled by varying the absolute number of windings. Starting from the law of Faraday, the number of primary windingsno is foundtobe relatedtopeak value ofthe primary fluxlinkage

Xo,

thepeakcorefluxdensityBand thecore cross sectionAc(perpendiculartothe direction ofB)in(3),

(3)

n =

02Ac

2A

B

where

Xo

is assumedtriangular, because ofthe block voltage acrossthemagnetizing inductanceasintroducedinFig. 7.

Substituting (3)in (2), minimal loss is found by setting the derivative of total loss to core flux density to zero, which implies equal but opposite derivates ofcopperand iron lossin

(4).

nclamps

= Kcf aB Vcore

a=O

(1)

whereKC,aand are material specific constants that are

/Sptot =

0

implies > PCu =_PeFe

SB

SB

SB

(4)

Solving (4) gives the expression for the optimal peak flux densityin(5).

nclamps

p

,/J fe

(~~~~~~~~~~~~~~~~~~~

Bopt

=P21cable

I~

AO~(f)11+

Bo,ot

W,

/3PK

cf(

)

(5)

Substituting this optimal densityinthe loss expressions (1) and (2) gives the total loss

Ploss,tot.

This total loss must be smaller than the heat removal capacity of the transformer as given in (6), where a uniform heat distribution inside the transformer is assumed.

TT-T

Ploss,tot,allowed ( s a) 6heatAs

0^

(6)

T is the maximum allowed surface

temperature

of the core

and Ta is the maximum expected (and thus allowed) ambient

temperature. The heat convection coefficient 0heatfor the

transformer surface to air is assumed 7.5 W/m2 K.

As

is the total transformer-to-airboundary surface.

For each commercially available core

Plss

tot(which is

optimal for that core size) and

PIss

tot allowedcan be determined.

The smallest (and thus cheapest) core that ensures

PlosS

tot <

Ploss

totallowedis then selected. It must be noted that a

largercoremight resultinlessloss, butatlargerexpenses.

The above procedure for usual ferrite-core transformer designresults inanoptimum fluxdensitybetween 0.1 and 0.2

T [1 1,p. 769], that is well below the usualcoresaturation flux

density of around 0.4 T. This interesting result is explained from the fact that iron loss becomesmoredominantatrelative

high operating frequencies. Therefore, iron loss is traded for higher copperloss. Different from usual transformer design is

the large amount of primary copper that is located in the

contactless supply cable. Therefore, optimal balance between

copperand iron loss in theclamp-cable caseresults inahigher

fluxdensity. Fig. 5 illustrates thisprinciple. Two situationsare

plotted, with Busual and Bcable respectively the optimal flux densityforausual transformer and the cable withclamps,with

largecopperloss.

Fig.5. Demonstration ofahigher optimal flux densityin caseoflarger

copperloss andgivencoresize. The intersectionpoints ofthe individual loss derivativesat1 coincide with the fluxdensityatwhich the total losses reach their minimaat2.

Theprocedure can berepeated for a range offrequencies, because the system operating frequency is adesign choice. In general it willturn outthatahigher frequency resultsinlower transformer loss, but this loss reduction is counteractedby the switching loss that increases with higher frequency. For the

prototype design the latter trade-off has been omitted and an

operating frequency around 100 kHz was assumed design

target. To conclude, Table II lists parameter values for the

prototype asusedinthis section.

TABLE I

PARAMETER VALUES FOR TWO-CLAMP CABLE PROTOTYPE AS REFERRED TO SECTIONIII-A

Symbol Quantity Value

p specificcopperresistance 1.724-10-8 [Qm]

cable length ofthe supply cable 2[m]

AO,0

peak value of flux linkage 2.46-10- [V/Hz]

a 1.42

3 corelossspecificconstant 2.2

Kc 2.06

f switchingfrequency 81 [kHz]

Ac ,core crosssectionalarea 33 [mm2]

Wcu'o totalcoppercrosssection in 100 [mm2]

transformer window

A, supply cablecrosssectionalarea 0.065 [m2] nO number ofwindingsinthe cable 7

nclamps number ofclamps 2

B. Supply cable and inputconverter

Two reasons indicate theuse ofresonant circuit operation. Firstreason is thatmagnetic field inside the cableatpositions where there is no clamp attached is considered as leakage inductance. Despite a small distance between inner and outer sections of the cable, a large voltage drop occurs across the latter relative large inductance. This voltage drop can be compensated by a capacitor. Second reason is to reduce switching lossatthe relative high switchingfrequencyof 100

kHz by zero-current and/or zero-voltage switching. Toobtain

a series resonant

circuit,

a capacitor Cres is placed in series with the cable and connected with switches in full-bridge configuration, as shown in Fig. 4. The resonant capacitor is

chosen to obtain a resonant frequency ofnear 100 kHz with thegiven

L,o

asintroducedin(7).Fortheprototype, theutility

grid with rectifier is represented bya DCvoltage supply.

Cres

DC sore Cable

DCsource-- Ii and

Fig.6. MOSFETfull-bridge andresonantcapacitor (Cres)asDC-ACinput

converter. Theutilitygrid and the following rectifierarerepresented byaDC

voltagesource.Position1referstothesamepositioninFig.7.

The leakage inductance of thesupply cable is dependenton

the number ofclamps that are attached at that time instant. The primary leakage inductance is denoted by

Lu0

and calculatedasin(7).

L _ o

no

Acable

CO /I

m,av

Acable ('cable

-nclampsC) (r2

rl

)

(7)

1 1rr1 _ d In(r2)-

ln(rl)

= I

~~dr

=

m,av

r2

I

r,l(r)

2m

f(r2

1l)

whereAcable is the surface area inside the cable perpendicular to the direction of the flux, Im

av

the average

length of the magnetic fluxpath and rt andr2respectivelyare the inner andouterwinding radii.

C. Powertransfer model

System translation into the electrical domain is useful for exploring full system behavior including input and output converters. With two clamps attached to the cable the equivalent electrical circuit isgiveninFig. 7.

Theprimaryleakage inductance

L.0

depends onthe number ofclamps by (7). This dependence is assumed relative small

(C<<lcable).Theresulting changeinresonantfrequency istobe acteduponby the controller of theinputconverter, which will besubject ofafuturepaper.

La0 1 Lai Lml Lm2 2 (ideal) L62 3 nO:n2(ideal)

Fig. 7. Two-clamp cable equivalent electrical circuit. Position 1 refers to the sameposition in Fig. 6 and positions 2 and 3 refer to the same position in Fig. 8.

Magnetizing inductances LM1 and LM2 are found in (8) by assuminganeffectivemagnetic pathlength le' Secondary side leakage inductance

L.1

andLu2are approximated with (9) from[11, p.780],

L~~g

UlrP *

A*A

LMI1 LM2 =

le

LuJLu2~/10.1j2]

2[1,2]

(r b + (8) (9)

wherep is the number ofsecondary winding layers, hw is the height of the winding window,

bc,

is the width of the

copper, bi is the interwinding insulation thickness andIisthe

length ofa single turn. Brackets denote clamp 1 or clamp2,

which are assumed equal. Table II shows prototype measurements versuscalculated inductances.

TABLEII

CALCULATEDINDUCTANCES FOR TWO-CLAMP CABLE AS REFERRED TO EQUATIONS 10 AND 11

Measured Calculated

inductance inductance

L calculatedmagnetizing 35.0[,uH ] 39.2[,uH ] mnl,calc inductance clamp1

L calculatedmagnetizing 38.6[,uH] 39.2[,uH] m2,calc inductance clamp 2

L calculatedleakage 35.3 [,uH ] 32.7[,uH ] cO,calc inductance cable

L calculatedsecondary 49.0[uH] 30.9[,uH ]

al,calc leakage inductance

clamp1

L calculatedsecondary 23.8[,uH] 30.9[,uH]

a2,calc leakage inductance

clamp2

An output converter as inFig. 8 is connectedtolocations 2

^2 ^"2

n2]cab

Io =len[l ]I[1,2] 2/cable > [1,2] >

2,3 Zload

D3 D4

Fig. 8. Diode bridge rectifier and filter capacitor as AC-DC output

converter

The LLC-type resonant converter topology with output

rectifier, asgiven by the combination of Figs. 6, 7, and 8, isa

well-studied topology. An analytical model can be found in

[10], where in the analysis theoutput stage isrepresented bya

voltage source that is put in parallel to the magnetizing

inductance of the transformer. The voltage source changes

polarity when the current through the diodes crosses zero.

With N clamps there are N output voltage sources, but by

assuming the clampstobe identicaltheycanbe combined into asinglesource, because thecurrentthoughthe diodes reaches zero atthesametime instant forall clamps, and theanalytical

model in [10] is applicable. When input, required output

voltage, switching frequency and the LLC values areknown,

steady-state voltage and current waveforms at all system nodes can be calculated analytically. Following the model in

[10], typical waveforms for the clamp-cable system are

presentedinFig. 10. I-200 vc.ss 100 2rs106 4 106 6 106 8106 1 105 1 210 1 410 1 1 810 prim.ss t lthe D.Systemloss~~.s VprimftQ) 0 0 ILA0t vinpling -100~ -10 -200r -15 6 6 6 6 5 5 ~ 5 5 5 0 2.10 4.10 6.10 8.10 1.10 1.2-10 1.4-10 1.6-10 1.8-10

Fig.9. Steady-state analyticalwaveforms ofDC-DCcable forparameters

as inTable withi1Lsstheresonanttankcurrent, vc.ss thevoltage acrossthe

resonant capacitor, Vin.ss the block-voltageacross theinput converterbridge, andvpnm.ss thesumof the voltagesacrossthe magnetizinginductances of the

twoclamps.

D. Systemloss

System loss is subdivided into conduction loss of the input

converter switches, copper loss in the supply cable, iron loss

inthemagneticcores and conduction lossintheoutputdiodes.

Copper loss inthe secondary windings isneglected in (13) as

motivated by (10) and (1 1), where (10) assumes ideal

coupling.

pn-p

2?

21 '-2 2 2

PCU/O

= P - Io >>

PCu[1,

2] I[1,2] (1 1)

WCU,O wu[1,2]

Also, switching loss is neglected in (14) because

above-resonance or resonant operation is assumed, which (ideally)

implies zero-voltage and zero-current turn-on and lossless-snubber turn-off. Together with (1) and (2) the system loss model is complemented with switches conduction loss in (11) and diodes conduction lossin(12),

p =n *v Io

Sw,condloss,tot switches sw

JIo

Pdiode,oss,tot =ndiodesvdiodeI[1,2 ndiodesvdiode

n p assumed >o CU[1,2] p assumed 0 SW,SWoss

(11)

(12) (13) (14)

wherenswitches and

ndiodes

are the number of switches and

diodes respectively that are conducting at the same time instant,V5M is the voltage drop across a conducting switch, Vdiodeis the conduction resistance of a diode and n is the transformerwinding ratio.

TABLE III

PARAMETERVALUES FOR TWO-CLAMP PROTOTYPE ASREFERREDTO SECTIONS

III-B, III-CANDIII-D

Symbol Quantity Value

n primary-secondary windingratio 6 [

1:n

ri radius of innerwindingsection 7.5[mm]

r2 radius ofouterwindingsection 40[mm]

(from thecentreofthecable)

bi. interwinding insulation thickness -0 [mm]

nj,n2 number ofsecondary windingsfor 42 [

clamp 1 and2

nswitches number of switches 4[-]

ndiodes number of diodes 8 [-]

Vdiode forwardvoltageacrossdiode 1.05[V] VsW forwardvoltageacross switch 1.4 [V]

cable length of the supply cable 2[m] bc. widthoftotalcoppercrosssection 30[mm]

inwindingwindow

PI, P2 number of secondary winding 3 layers forclamp1and2

h. height of thewinding window 3 [mm]

fo magneticpermeabilityinvacuum 4n-10-7[H/m]

Pr relativepermeability ofthe 1900 [-]

magneticcore

output.

IV. SYSTEMSIMULATION OFDC-DCCABLE

The transformer core has been chosen large enough for a 240 W throughput at 240 V. For a future 50 Hz AC system,

theoutputvoltage willvary. Also, the required loadpower per

clamp willvaryfromnear-zero to240 W. Therefore, objective of the simulations is to quantify the system efficiency and

component ratings at various load resistances and output

voltages.

The analytical model from section Ill-C and [10] is informative when the output voltage is known exactly. A

difficulty arises when theoutputvoltage isnotknown exactly and theswitching frequency isnear resonance. Inthis situation

a slight error in output voltage causes a large change in resonantcurrent amplitude because only voltagesarespecified

in the model. In practice this will not occur, because the resistive load will damp the system, which is clear from the voltage-current relation ofaresistor.

To overcome the difficulty, the system topology has been simulated to find power throughput (with variable load resistances) versus efficiency. The topology was simulated lossless to find the peak resonant current, which was then insertedin (1),

(2),

(11) and (12)tocalculate system loss. The resulting loss can be added to the simulated model as

resistances and simulatedagaintoconverge to moreprecision. Simulated power throughput versus efficiency is compared withmeasurementsinsectionV.

V. COMPARISONBETWEENMEASUREMENTS AND MODEL

Objective of this section istorelate simulation results with

measurements from the prototype as presented in Fig. 10.

Implementation details are shown in Fig. 11. Component

values as used for theprototype are listed in previous tables. The prototype was build with already available materials in

the laboratory. Therefore, non-optimal components were implemented. Especially, core material and size are far from optimal unfortunately. Typical waveforms measured at full load are presented in Fig. 12, where the almost perfect sinusoidal waveforms show that switching happens at the

resonantfrequency.

Fig. 10. Prototype contactless system, with 1) cable, 2) clamp, 3) input

converter,4)connectors for MOSFET drivercircuitry, 5) connectorsfor DC voltage supplyand6)connectorsfor load

Fig. 11. Details ofprototype system, with 1) MOSFET full-bridge with

heatsink, 2) resonant capacitor Cres, 3) filter capacitor between DC voltage

supplyand MOSFET bridgetocompensatefor reactive power, 4) magnetic

core half, 5) diode full-bridge, 6) pulled-out inner winding section and 7)

outerwinding section remains within thecable

2006/06/14 17:04:54 25ok Normnal Stopped z 805 500MS/s 2[ivsAh CH1 1:1 1.00U/div DC Full CHZ 1:1 0.500U/div DC Full CH3 1:1 0.500U/div DC Full EdgeCH1 F Auto 0.80U Xi -ZU0.Ouns Y1 -708 .333nV X2 12.300us Y2 0.00000U AX 12.320us IY 708.333nU 1/AX 81.16883kHz

Fig. 12. Typical waveformsatfull 480 Woutputpowerobserved from the

prototype systemforacomparison with the analytical waveforms in Fig. 9.

The resonant current on channel 1 at 1 V/div is measured with a 0.1 V/A

probe. The voltageacrosstheresonantcapacitoronchannel 2at0.5V/div is

measured witha200x attenuation factor.Channel 3 showsadriving signal for onephase-arm of the inputconverterbridge.

Measurement results are compared with simulation and

calculation results in Figs. 13-15. For Figs. 13 and 14 two

clampswereconnected to thecable, while forFig. 15 only a

single clampwasused in ordertoreduce thepowerthroughput

of thesystem.

First aspecttonotice is the relativelarge contribution of the

core loss to the total loss, which was expected from the

inappropriate 3C80 core material and size. This loss can be

reduced substantially with the implementation of a proper core. Furthermore, efficiency is relative flat in the complete power range covered by Figs. 13-15. Fig. 13 shows a total

outputpower of 480 W decreasingto 280 W with an almost

constant efficiency. ForFig. 14 the input voltageis decreased from 80 Vto 10 V, where the output power decreases from

400 W to less than 7 W. Also in this case the measured

efficiency curve is rather flat. It must be noted that the

(extremely high)measured efficiencies for small input voltage lackprecision,because ofanalogmeterreadingerrors. Fig. 15

shows results for a single clamp with variable load resistance

connected. Efficiency for the rated power of 240 W is little

lessthan for two clamps, becauseprimary loss in this case is

notsharedacrossmultiple clamps.

Perhaps most interesting is the fact that with suboptimal components an average efficiency of near 9000 was

established,which showsfeasibilityof the concept inpractice.

\. .1M

70 304 pm 20 15 10 PFe 56 P{,? 40 I0 Psw >diodeI 250 300 350 400 Total load resistance(Ohm)

Fig. 15. Calculated and measured system loss and efficiency as a function of output resistance of asingle clamp. The leftfigureshows calculated iron, copper, conduction switching and diode loss. The right figure relates measured and calculated total loss. Measured system efficiency is shown on theright.

I 1,/ \ Pmeas

Pcalc

250 300 350 400 Total load resistance(Ohm)

1]

Fig.13. Calculated and measuredsystemloss andefficiencyas afunction of combined output resistance oftwo clamps. The left figure shows calculated iron,copper,conductionswitching and diode loss. The right figure relates measured and calculated total loss. Measured system efficiency is shownontheright.

15 P m 10 Psw / i PFe i 1 40 0 p 2 /. Pcu 4/_diode 7 Pmeas 20 40 60 so 20 40 60 so Vil Vil

Fig.14. Calculated and measuredsystemloss andefficiencyas afunction of the system input voltage. The left figure shows calculated iron, copper,

conduction switching and diode loss. The right figure relates measured and calculated total loss. Measuredsystemefficiencyis shown on theright.

206 16B 16 14 g 12 P Po~M f 18 l SW

8MLkPl1

4 VI. CONCLUSION

A DC-DC system for transferring energy with electrical isolation between primary and secondary has been introduced

anddesigned. Practicalresults from a prototype are backed-up by simulation results. Measured efficiencies of almost 90°0

withnon-optimal system componentsin alarge loadrange are

promising. Futurework comprises extension of theconcept to an AC-ACmulti-user system.

REFERENCES

[1] A. V.Lebedev,E.A. Lomonova, P. G. Van Leuven, J.Steinberg,D.A. H. Laro,"Analysis and initial synthesis ofanovel linearactuatorwith activemagnetic suspension", ConferenceRecord of the 39thIndustry Applications Conference Annual Meeting, Vol.3, 3-7 Oct. 2004, pp. 2111-2118

[2] K. W.Klontz,D. W.Novotny,"Contactlesspowerdeliverysystemfor mining applications",IEEETransactionsonIndustry Applications, Vol.

31, No. 1,January/February1995

[3] A. W.Kelley,W. R.Owens,"Connectorlesspowersupply foran

aircraft-passenger entertainment system",IEEETransactionsonPower

Electronics, Vol.4, Issue 3,July1989, pp. 348-354

[4] H.Matsuki,T.Matsuzaki, "Simulations oftemperatureriseon transcutaneousenergytransmissionbynon-contactenergytransmitting coils",IEEETransactionsonMagnetics, Vol. 29,No.6, November1993 [5] J. M.Kim,D. H.Kang, S. J. Jung, D. J. Bang,"Designof the transverse

flux linearmotorwith theintegrated contactlesspowersupply", unpublished.

[6] S. Y. R.Hui,W. W.C.Ho,"A newgeneration ofuniversal contactless battery charging platform for portableconsumerelectronicequipment",

IEEETransactionson PowerElectronics, Vol. 20,Issue3,May2005,

pp. 620-627

[7] C. L. W.Sonntag,E.A.Lomonova,J. L.Duarte,A. J. A.Vandenput andA. J. M.Pemen, "Contactless EnergyTransfer forOfficeand

DomesticApplications", unpublished

[8] S.Valtchev,B. V.Borges,K.Brandisky andJ. B.Klaassens, "Efficient

ResonantInductiveCouplingEnergyTransferUsingNewMagnetic and Design Criteria,"36thIEEEConferenceon PowerElectronics

Specialists,pp.1293-1298,June12,2005

[9] J. M.Barnard,J. A.Ferreira,J. D.vanWyk, "Slidingtransformers for

linear contactlesspowerdelivery",IEEETransactionsonIndustrial Electronics, Vol. 44,Issue6, December 1997,pp.774-779

[10] J. F.Lazar, R.Martinelli, "Steady-state analysis oftheLLCseries

resonantconverter", 16thAnnualIEEEAppliedPowerElectronics Conference andExposition, Vol. 2, March 4-8, 2001,pp. 728-735

[11] N.Mohan,T. M.Undeland,W. P.Robbins,PowerElectronics: Converters,Applications and Design, 3rd ed.,NewYork:Wiley,2003

[12] R. W.Erickson, Fundamentals ofPowerElectronics,1st ed.,NewYork: Chapman and Hall, May1997

[13] C. R.Sullivan, "Optimal choice for number of strandsin alitz-wire transformerwinding",IEEETransactionsonPowerElectronics, Vol. 14,

No.2, March1999

OI -- fle

400 600 800 10001200 400 600 800 1000 1200 Loadresistance(Ohm) Loadresistance(Ohm)