issue 5 – august 2012 power ... · of hpt-igct issue 5 – august 2012 also inside this issue...

02_PEE_0512_Layout 20/08/2012 12:50 Page 1

CONTENTS

www.power-mag.com Issue 5 2012 Power Electronics Europe

3

Editor Achim ScharfTel: +49 (0)892865 9794Fax: +49 (0)892800 132Email: [email protected]

Production Editor Chris DavisTel: +44 (0)1732 370340

Financial Clare JacksonTel: +44 (0)1732 370340Fax: +44 (0)1732 360034

Circulation Manager Anne BackersTel: +44 (0)208 647 3133Fax: +44 (0)208 669 8013Email: [email protected]

INTERNATIONAL SALES OFFICESMainland Europe: Victoria Hufmann, Norbert HufmannTel: +49 911 9397 643 Fax: +49 911 9397 6459Email: [email protected]

Armin Wezelphone: +49 (0)30 52689192mobile: +49 (0)172 767 8499Email: [email protected]

Eastern US Karen C Smith-Kerncemail: [email protected] US and CanadaAlan A KerncTel: +1 717 397 7100Fax: +1 717 397 7800email: [email protected]

Italy Ferruccio SilveraTel: +39 022 846 716 Email: [email protected]

Japan:Yoshinori Ikeda, Pacific Business IncTel: 81-(0)3-3661-6138Fax: 81-(0)3-3661-6139Email: [email protected]

TaiwanPrisco Ind. Service Corp.Tel: 886 2 2322 5266 Fax: 886 2 2322 2205

Publisher & UK Sales Ian AtkinsonTel: +44 (0)1732 370340Fax: +44 (0)1732 360034Email: [email protected]

Circulation and subscription: Power ElectronicsEurope is available for the following subscriptioncharges. Power Electronics Europe: annual chargeUK/NI £60, overseas $130, EUR 120; single copiesUK/NI £10, overseas US$32, EUR 25. Contact: DFA Media, Cape House, 60a Priory Road, Tonbridge,Kent TN9 2BL Great Britain. Tel: +44 (0)1732 370340. Fax: +44 (0)1732360034. Refunds on cancelled subscriptions willonly be provided at the Publisher’s discretion, unlessspecifically guaranteed within the terms ofsubscription offer. Editorial information should be sent to The Editor,Power Electronics Europe, PO Box 340131, 80098Munich, Germany. The contents of Power Electronics Europe aresubject to reproduction in information storage andretrieval systems. All rights reserved. No part of thispublication may be reproduced in any form or by anymeans, electronic or mechanical includingphotocopying, recording or any information storageor retrieval system without the express prior writtenconsent of the publisher.Printed by: Garnett Dickinson.ISSN 1748-3530

PAGE 1

Industy NewsThinner Ferrite Magnets Provide High-Performance Alternative toExpensive Neodymium MagnetMaterial

Zero-Voltage Switching in Point ofLoad DevicesPAGE 23

Comparison of 1200V SiC PowerSwitching DevicesSilicon Carbide (SiC) power semiconductors being actually commercialized and

are promising devices for the future. To outline their characteristics the switching

and conducting performance of two types of SiC normally-on JFETs, a SiC

normally-off JFET two types of SiC MOSFETs and a SiC BJT have been analyzed by

means of measurements at exactly the same boundary conditions and compared

to each other. Beside of the switching and conducting behaviors the requirements

for driving these devices have been investigated. W.-Toke Franke, Danfoss

Solar Inverters, Denmark

PAGE 29

Asymmetrical ParasiticInductance Utilized for SwitchingLoss Reduction High efficiency of power conversion circuits is a design goal on its own. The

reduction of switching losses is the basis for higher switching frequencies which

lead to a reduction of the size and weight of the passive components. The

improvement from 96 % to 99 % efficiency will reduce the effort for cooling by

factor 4. With the utilization of parasitic inductance and consequent execution of

basic rules of power electronics is a new power electronics solution based on

standard Si components disclosed which extends the traditional designs. The

presented new power module concept combines a low inductive turn off with the

utilization of the parasitic inductance for a reduction of the turn on losses and the

usage of three level switching circuits with the paralleling of fast switching

components with components with low forward voltage drop.

Michael Frisch and Temesi Ernö, Vincotech Germany and Hungary

PAGE 33

Static Loss MeasurementMethods for QualityImprovementSteady increase of power for electric converter units leads to the constant

enhancement of characteristics and load capacity of power semiconductors. Thus,

requirements to the maximum current load of power thyristors and diodes, limited

by heat-release losses in a semiconductor element and the intensity of heat

removal from the die, are also increasing. Power semiconductor manufacturers do

their best to the maximum reduction of conducting losses and preserving all the

remaining characteristics at the required level by precise measurement of forward

voltage drop. Alexey Poleshchuk, Automation Lab Engineer, Proton-

Electrotex, Orel, Russia

PAGE 37

Website Product Locator

Past, Present andFuture of HPT-IGCTThe integrated gate-commutated thyristor (HPT-IGCT)is a state-of-the-art bipolar turn-off device for highpower applications. The IGCT was introduced about 15years ago and has established itself as a preferredtechnology for handling very high power. Theapplications range from industrial drives and track-sidesupplies to power quality and high-current breakers.The first device in the HPT-IGCT family was theasymmetric 4500 V device 5SHY 55L4500 released in2009. In 2010, 5SHY 42L6500 followed, anasymmetric 6500 V device and in 2012, yet anotherasymmetric device 5SHY 50L5500 rated 5500 V wasreleased. In this article we take a look at the newestmember - the 5.5 kV asymmetric HPT-IGCT and wewill also take a look at one interesting device in thepipeline.

Cover supplied by ABB Switzerland Ltd,Semiconductors

COVER STORY

PAGE 7

Market NewsPEE looks at the latest Market News and company

developments

p03 Contents_p03 Contents 20/08/2012 14:07

2

WWW.COILCRAFT.COM

®

The path of least resistance is through our new high efficiency XAL/ XFL inductors

Compared to competitive parts, current zips right through our new XAL/XFL inductors.

Their DC resistance is signifi-cantly lower than other inductors of the same size. So your batteries last longer and your power supply runs cooler.

They easily handle large peak current, and their soft saturation

characteristics often let you use a smaller size part without fear of overloading during transients.Built from a uniquely formulated material, XAL/XFL parts do not have the same thermal aging prob-lems as some competitive parts.

See all the advantages of these new high efficiency inductors.Visit coilcraft.com/xal.

Electrons prefer Coilcraft

Competitors’ 4.7uH inductors have much higher DCR per mm3

than Coilcraft’s XAL5030.

38% higher DCR

126% higher DCR

48% higher DCR

04_PEE_0512_Layout 20/08/2012 12:51 Page 1

OPINION 5


A suitablequote as alead in to

the editorsopinion

The global power discrete and module market is forecast togrow by $9 billion to reach $26.2 billion in 2016, according toa new report from IMS Research. With revenues growing 9percent from 2010, it is clear that 2011 was, at best, anaverage year for the power semiconductor market. Thisfollows the dramatic market fall in 2009 and spectacularrecovery in 2010, so perhaps the industry was ready for a“normal” year 2011.Although market conditions have remained flat in the first

half of 2012, the longer-term prospects remain positive.According to market researcher Yole Développement themarket for semiconductor devices (discrete, modules andICs) dedicated to the power electronics industry will reach$20 billion in 2012. These figures underline the robustness ofthe power electronics industry in its potential applicationfields, though some regional shift may occur. Withapplications as diversified as hybrid cars, PV inverters, lighting,energy, and voltage ranging from a few volts to a fewthousands volts, power electronics is and will remain one ofthe most attractive branches of the semiconductor industryover the next decade.In example, 2011 global PV inverter shipments grew by

more than 12 percent, despite the excess inventory overhangfrom the prior year, and reached 27 GW, but the Europeanmarket shrank considerably. Europe’s dominance of the PVinverter market is predicted by IMS to continue to wane dueto its two biggest markets, Germany and Italy, facingsignificant reductions in their annual installations. Europe’sshare of PV inverter shipments and revenues was over 80percent in 2010; however this is forecasted to fall to lessthan 40 percent in 2016 and revenues not to return to 2011levels in the next five years. This in itself presents a huge

challenge to suppliers, which are mainly European, with themajority of their facilities and customers located in thatregion. Analysis showed that whilst the European outlook isnot so bright, the global picture for PV inverter suppliers lookssomewhat better; highlighting the fragmenting nature of theindustry that now needs to look to emerging markets forfuture growth. Global shipments are predicted to continuegrowing at a double-digit rate over the next five years, withrevenues exceeding $9 billion by 2016. In 2012 the market isforecasted to grow by just 3 percent to hit $7 billion for thefirst time.Recognizing these trend PCIM 2013 conference focuses on

“Wind and Solar - Integration, Challenges and Solutions”. Bothare growing rapidly and thus provide huge opportunities forpower electronics.Regarding power semiconductors, IGBTs account for $1.6

billion in the medium to high voltage sector. And the SJMOSFET market will reach $567 million by the end of thisyear. GaN and SiC also have a chance to outdo Siliconperformance and enhance inverter capabilities. With accessto lower-cost material, the SiC industry now has the chanceto ramp up and get organized. However, apart from the PFCbusiness, technological capabilities of SiC show that it willsurely be dedicated to high power/voltage applications. Also,SiC companies have appeared in China, which will definitelyprovide competition and tougher access to local markets. TheGaN industry is mostly a US business with some pioneerslike MicroGaN from Germany, NEC and Powdec showingthere is a trend to globalize the GaN manufacturing industry.From MicroGaN’s perspective, today 4-inch GaN-on-Siliconwafers with the appropriate epi layer thickness are available,which make 600V devices feasible. The material qualityimprovement in 6-inch GaN-on-Silicon is closely monitoredand it is a matter of homogeneity progress in the epi qualityand thereby a matter of yield improvement to switch to 6-inch utilization. In order to enable commercialization with asmaller wafer diameter, MicroGaN developed a newtechnological approach, which cuts die size for givenperformance into half. This is only one example for Europeaninnovations which Power Electronics Europe has supported atPCIM Europe 2012 and will continue to support in 2013.Have a good time by reading this issue!

Achim ScharfPEE Editor

Growth Aheadwith RegionalShift

05_PEE_0512_p05 Opinion 20/08/2012 11:43

SEMISTACK® RE Optimized converter for solar and wind

450 kW – 6 MW

Long l i fet ime

-cooled IGBT platformaterW

p-t 3nardauq-d 4n2 a

Power density of up to 11.4 kV

-cooled IGBT platform

sretrevnoe csahp

A/l i trePower density of up to 11.4 kV

06_PEE_0512_Layout 20/08/2012 11:33 Page 1

MARKET NEWS 7


Learn more: http://indium.us/F607

Solder

©2012 Indium Corporation

A S I A C H I N A E U R O P E U S A

www.indium.com/ [email protected]

Over 200 alloys available

Solder Preforms

Solder Ribbon

“In 2012, the market for semiconductor devices (discrete, modules and ICs) dedicated to the powerelectronics industry will reach $20 billion,” confirmed Brice Le Gouic, Activity Leader, Power Electronics atYole Développement. With applications as diversified as hybrid cars, PV inverters, lighting, energy, andvoltage ranging from a few volts to a few thousands volts, power electronics is and will remain one of themost attractive branches of the semiconductor industry over the next decade.Already well-established in the market, IGBTs account for $1.6 billion in the medium to high voltage

sector. In the new report “Status of the Power Electronics Industry”, Yole describe a trend towardsdecreasing voltage range in order to target consumer applications such as TVs, computer adapters andcameras, in order to access more segments. At the same time, SJ MOSFETs present in these applicationsoffer faster switching frequencies and competitive cost. “We estimate the SJ MOSFET market will reach$567 million by the end of the year.”GaN and SiC also have a chance to outdo Silicon performance and enhance inverter capabilities.

However, materials are still expensive and the technology is not yet ready. On the other hand, both of thesematerials can benefit from their developed status in the LED industry, and plenty of LED players payingattention to the opportunity that power electronics represent. GaN and SiC are not mature yet for thepower electronics market: the first one requires technological enhancement of the manufacturing process,especially for epitaxy thickness, and the second one is an expensive material that does not allowimplementation within consumer-like businesses. Segmentation between technology and thepower/voltage range will take place, and some segments will only accept one « best » technology.

Each technology has a particular industry structureWhile SJ MOSFETs welcome new players and foundry service suppliers, the IGBT dies’ industry is becomingconsolidated thanks to the presence of large players involved in many applications, such as Infineon,Mitsubishi Electric and Fuji. However, the IGBT (and SJ MOSFET) modules business is increasing, and wesee new players entering to provide solutions for cooling, interconnections, substrates, packaging, and gel.On another note, the SiC industry, which has been led by CREE, is now an interesting playground for

new players. With access to lower-cost material, the SiC industry now has the chance to ramp up and getorganized. However, apart from the PFC business, technological capabilities of SiC show that it will surelybe dedicated to high power/voltage applications. Also, SiC companies have appeared in China, which willdefinitely provide competition and tougher access to local markets.At this point in time, the GaN industry is mostly a US business. International Rectifier, EPC, Transphorm,

Microsemi and GaN Systems now propose fully off-the-shelf or customized products. However, somepioneers like MicroGaN from Germany, NEC and Powdec are showing there is a trend to globalize the GaNmanufacturing industry. At the same time, the market is still soft and LED players are considering using theirtechnological platform to enter the power electronics field, which will be very much low power/voltage-oriented.Because power semiconductors are “just a piece” of the power electronics industry, the power

semiconductor industry has to answer requirements from a bigger system: the inverter. For example, someexpensive solutions are still not implemented, even if device and module performance are much higherthan current output.Geographical positioning is also critical in the power electronics area, especially with the boom in China

and other emerging countries, but also because several applications (PV, wind, electric vehicles) aresupported by local governments.

www.yole.fr

New PowerTechnologies Ahead

Power semiconductorspositioning forecast 2015Source: YoleDéveloppement

Market News_Layout 1 20/08/2012 11:52

8 MARKET NEWS

The global power discrete and module market isforecast to grow by $9 billion to reach $26.2billion in 2016, according to a new report fromIHS/IMS Research. Following good growth in thefirst half of 2011, the market fell away in thesecond to finish with only 9 percent growthover the whole year. Although market conditionshave remained flat in the first half of 2012, thelonger-term prospects remain positive.

With revenues growing 9 percent from 2010,it is clear that 2011 was, at best, an averageyear for the power semiconductor market. Thisfollows the dramatic market fall in 2009 andspectacular recovery in 2010, so perhaps theindustry was ready for a “normal” year.However, with business confidence uncertainand the recovery stalling, what are theprospects for the future? According to IMSResearch there remains cautious long-termoptimism, with the market projected to growby almost $9 billion to $26 billion from 2011 to2016, a compound annual growth rate of 8percent. The market for power semiconductor

modules again grew faster than that fordiscrete power semiconductors in 2010,increasing by 32 percent to close on $4.6billion. The power module market growth iscontinuing in 2012 despite slowing demand, asreduced business confidence is felt in industrialmarket sectors. “Power module revenues arepredicted to grow by 14 percent in 2012 , andto be 80 percent higher than in 2011 in 2016”stated senior analyst Richard Eden. The power discrete market was worth an

estimated $13 billion in 2011, having grown byjust over 2 percent from 2010. Continued highdemand for discrete IGBTs accounted for nearlyall the growth. “The sudden surge in discreteIGBT demand was fuelled by sales ofdomestic appliances such as room air-conditioning and variable speed washingmachines in the Chinese market,” added Eden.“In contrast, sales of standard power MOSFETsand thyristors actually declined slightly in 2011.” A high proportion of discrete power

semiconductors are used in relatively fast-moving commodity items such as flat-screen

televisions, notebook computers and mobilephone adapters. Sales of these productsdepend heavily on consumer confidence andthe health of the global economy. Sales ofpower semiconductors in those sectors eitherdeclined or achieved negligible growth in 2011,and are not forecast to deliver much growth in2012. Power semiconductor market growth willtherefore be driven by increased content, eitherto improve power conversion efficiency or addfunctionality. In contrast, growth is forecast toaccelerate in the automotive, renewable energyand transportation sectors. Overall discretepower semiconductor market growth ispredicted to remain less than 5 percent in2012; the market is forecast to reach almost$18 billion in 2016.The latest findings and analysis on this

important market can be found in IMSResearch’s “The World Market for PowerSemiconductor Discretes & Modules 2012”report, due for publication in August 2012.

www.ihs.com

Power Semiconductor Market to Grow$9 Billion Over the Next Five Years


MARKET NEWS 9


CTSR LEM commits to renewable energy sources of the future by enabling control and ensuring safety of today’s solar power solutions. CTSR transdu cers combine safety and per formance, accurately measuring small AC and DC leakage currents. Easy installation for single or three phase residual current measurement: CTSR is today’s choice for the energy of tomorrow.

Two nominal current ranges: 0.3 and 0.6 A RMS

Printed circuit mounting Large 20.1 mm diameter aperture Available with primary inserted conductor +5 V single supply Up to 11 mm creepage and clearance distances

+ CTI 600 for high insulation Low offset drift –40 to +105° C operation Reference Voltage access and control Self-test and degauss High overload capability: 3300 A

www.lem.com

At the heart of power electronics.

Solar energy committed to a lifetime of safety and performance

nd aytefa senbiomcrewo pralo ss’yadot

ocg nilbaney be rutufenero ts timmocEM L

RSCT

lyetaurcc aencamrofre pndsrecudsnar tRST C.snoitulo s

oy tefasg nirusned nal ortnoehtf os ecruosy grenee lbawe

lysfe

ylppue slgni +5 V srmair phti welabliavAemaim d1 m0.e 2gra L

ntuo mtuicri cdeintr Pt rnerrul canimoo nwTTw

.owrrom tofoR iST: Ctnemerusaeme olgnir sofn oitallatsni

C al Alamg snirusaemnd aytefa senbiomc

ortucdno cdetrsen iyerutrepr aete

ginSMR6 Ad 0.n3 a: 0.segnat r

ygrene ehr toe fciohs c’yados tnerrul caudisee rsahe perhtr e oysa. Estnerrue cgakaeC ld DnC a

lyetaurcc a,encamrofre pnd

yty

ly

ww

bpaa cloadrev oigh Husage ddn atset-f Sel aceagtol VecnerefRe

arep° C o501o +0 t4 –tfirt desffw oo L

snh igir ho0 f0I 6T+ Cgapeerm c1 mo 1p tU

ylppue slgni +5 V s

com.mel.ww

A003 3:ytilis

olrtno cdn assec acnoita

noitaluss ecnatsie dcnaraeld cne ag

t oraee hht tA

.scinortcelr eeowf pt o

Rising energy costs and power demand, combined with mainstreamadoption of renewable energy sources, are driving adoption of newpower architectures, such as 400 V DC, that enable more efficientnetworks and businesses.Direct and alternating current (AC) technologies have been evolvingsince Tesla and Edison competed in the late 1800s to establish adistribution method for the electric grid. At that time, technologieswere not available to safely and economically use DC power over longdistances and AC power won out on the grid. Today, long-distance DCpower use is cost effective, DC-based renewable ener gy isproliferating, and most technology equipment operates on DC power.The grid still distributes AC power, however, creating growinginefficiencies in our power-hungry world. “Recent advances in powerconversion technologies, combined with rapidly increasing DC-basedequipment growth, have created an environment where DC powerdistribution should be, and is being, actively considered,” said MarkMurri ll, director of 400V DC power initiatives for Emerson NetworkPower. “Global architecture standards required for widespreadimplementation of 400 V DC are well under way, as demonstratedwith the European Telecommunications Standards Institute EN 300132-3-1 standard released in February 2012. As more and morevendors support this standard, the adoption of 400 V DC will begin torapidly increase in vari ous applications and further standards will beclarified and adopted.”Emerson Network Power currently sees four primary applications for

400V DC technology: telecom central offices, data centres, commercialbuildings, and transportation - each with its own drivers for adoption.

Reducing copper and costs in the central officeUnlike the grid, telecommunications networks have long used DCdistribution, la rgely for its high reliability and good signal qualityperformance. As telecom equipment has moved to Silicon-based DCtechnology, alignment between existing DC power distribution andDC-driven equipment is already in place, but cost and efficiency gainsare still possible. For example, 400 V DC is especially suited for highpower delivery over long distances, because it reduces installation andoperational costs and improves cable management versus -48 V DC.These benefits are realised due to at least an 80 percent reduction ofcopper wire used, affecting both the cost of and time needed forinstallation. In addition, reduced line losses typically will increase end-to-end energy efficiency as well, further reducing operating costs.The logic behind 400 V DC power in the data centre is driven by the

need for high availability, high efficiency and lower total costs. Sinceutility AC power ultimately must be converted to DC power for use byall IT equipment and because stored energy systems (batteries,flywheels, etc.) and renewable sources provide DC power, a typical DCpower architecture requires fewer power conversions from grid tochip. Reducing these conversions not only saves energy, but also canincrease critical power availability through simplified distribution andreduction of failure points in the power chain. Furthermore, DC powerdoes not require phase balancing or considerations for harmonics,

Key Applicationsfor 400 V DCPowerTechnology


10 MARKET NEWS

Issue 5 2012 Power Electronics Europe www.power-mag.com

eliminating stranded power due to equipment de-ratings. Additional DC power options for the data centre include those

based on -48V DC, which is best suited for row-based solutions whereloads are in close proximity to the UPS. This -48V topology is notcomparable to a traditional, room-based AC UPS system, but now withthe introduction of 400V DC power systems and components, for thefirst time the benefits of DC power are available in an enterprise-classpower solution for the data centre. The degree to which companiescan exploit savings from DC power varies from data centre to datacentre. However, in theory at least, the advantages of using DC powerin data centres seem obvious: servers use DC power, and the fewerconversion steps from AC power from the grid to servers, the greaterthe efficiency.

Transportation already using DC distributionTransportation vehicles of many types (cars, ships, mass transit,locomotives, construction machinery and others) are integrating DCpower for motor drive and other loads to increase efficiency and savefuel. For example, in the ship and locomotive industries, DC power isincreasingly leveraged for propulsion and onboard loads. Similarly, aselectric and hybrid vehicles such as cars and buses become morecommon, DC power is a logical choice for fast charging and use invehicle systems. In these vehicles and others, DC will be used withcurrent and 400V distribution technologies as they continue to movetoward hybrid and all electric systems.

Renewable energy drives migration and benefitsWhether financ ially or civically motivated, building owners andoperators today are increasingly using intelligent solutions thatreduce energy consumption by leveraging renewable materials andenergy sources. Using on-site, renewable power generation canreduce or even eliminate power draw from the grid. This is significantfor 400 V DC because renewable sources generate DC power, whichmeans using it for AC-powered applications within the buildingrequires multiple conversions from DC to AC, reducing the efficiencyof the renewable power generation. To avoid these extra conversions,architects and engineers are designing facilities with 400 V DC“microgrid” power distribution throughout or in selected areas of abuilding or campus. This solution increases system power efficiency,simplifies the electrical microgr id, and improves reliability whilereducing operating costs.To benefit from the emergence of 400V DC power distribution

technology as a viable alternative to traditional power architecturesEmerson Network Power has leveraged its expertise in the EMergeAlliance, an open industry association developing standards for thewider adoption of DC power distribution.

www.EmersonNetworkPower.com

To receive your own copy of

POWER ELECTRONICS

EUROPE

subscribe today at:

www.power-mag.comhttp://siliconpower.danfoss.com

A cooler way to harness the wind Stacks with customized ShowerPower® modules maximise performanceIt cannot be stressed enough: e� cient cooling is the most important feature in power modules. Danfoss Silicon Power’s cutting-edge ShowerPower® solution is designed to secure an even cooling across base plates. In addition, our modules can be customized to meet your wind power requirements in detail, o� ering high quality, high performance, extended life and very low lifecycle costs. In short, when you choose Danfoss Silicon Power as your supplier you choose a thoroughly tested solution with unsurpassed power density.

Please go to http://siliconpower.danfoss.com for more information.

MAKING MODERN LIVING POSSIBLE


MARKET NEWS 11


The PCIM Europe 2013 (14. - 16.05.2013) Callfor Papers is now open. The deadline forsubmitting abstracts is 15. October 2012. Theevent is the number one meeting place for allfields of power electronics and theirapplications. PCIM 2012 closed with 263exhibitors, 6.874 visitors and 744 conferencedelegates - the highest numbers recorded so

far. Besides new semiconductor technologiesand devices such as Gallium Nitride (GaN)and Silicon Carbide (SiC) new assembly andinterconnect technologies for power modulesgained a lot of interest.

The topics to be covered in 2013 includerecent developments in powersemiconductors, passive components, thermal

management products, energy storage,sensors and new materials as well as systems.This year’s conference focuses on “Wind andSolar - Integration, Challenges and Solutions”.Both are growing rapidly and thus providehuge opportunities for power electronics.

Proposals will be considered for conferencelectures or posters. The accepted papers willbe included in the official PCIM proceedingsCD and the IET Inspec database. Theconference language is English.

The conference directors will select the bestpapers for the nominations list from thesubmissions. The winner of the Best PaperAward (sponsored by Power ElectronicsEurope) will receive 1,000 Euro and thechance to participate at the PCIM AsiaConference in Shanghai, including flights andaccommodation. For outstandingcontributions from young authors not olderthan 35, three “Young Engineer Awards” of1,000 Euros each will be granted. The prizeswill be awarded at the PCIM Europe 2013Conference Opening Ceremony.

www.pcim-europe.com

PCIM 2013 Call for Papers

Worldwide cumulative installed PV power

High Current low RDS(on) Trench MOSFET in ISOPLUS DIL

For more information pleaseemail [email protected] or call Petra Gerson: +49 6206 503249

New DCB based surface mount packagefor automotive applications

Features

Applications

Examples


12 INDUSTRY NEWS


Innovation by Tradition

E FOR TRANTITUTSSUB

FORMERSE FOR TRAN

S � � � � � �SFORMER

LOW OHMIC PRECISION AND POWER RESISTORS

igh pulse loadability (10 J)H

AND OFFER A NUMBER OF ADV SHUNT RESISTORS SASMD

LOW OHMIC PRECISION AND POWER RESISTORS

AGES:ANT ER OF ADV VANTACEVE SP ORS SA V

reas of use:A

ow thermoelectric voltageL_ ery low temperature dependency over a large temperature rangeV_

)Wigh total capacity (7 H_ igh pulse loadability (10 J)_

ery low temperature dependency over a large temperature range

ery low temperature dependency over a large temperature range

5�eg 3Wibacher E

hone +49 (0) 2771 934-0Pillenburg · D35683 ·

nitioadTrtion by ovaInn

ax +49 (0) 2771 23030F·hone +49 (0) 2771 934-0

As manufacturers of renewableenergy generation equipment suchas wind turbines look for ways toincrease their competitiveness, the

development of a new, advancedmagnet material promises to helpreduce the cost of a key componentof these electric power systems.

Today, turbine generators and electricmotors typically use neodymiummagnets to produce a magnetic fieldsurrounding a rotor, since to date

neodymium has offered the bestavailable combination of small sizeand low weight. This article showshow the new ferrite magnets fromTDK Electronics can rival theperformance of neodymiummagnets while freeing turbine andmotor manufacturers from the highcost and uncertain availability ofneodymium magnets.

In recent years, rare-earthneodymium magnets have risen tothe top of the market in terms ofproduction value, but ferrite magnetscontinue to lead the field in terms ofproduction volume. In fact, rising andvolatile prices for neodymiummaterials coupled with uncertainty ina supply chain dominated by Chineseproducers have led many end-product manufacturers to consideralternatives to neodymium magnets.

In contrast with neodymium,ferrite is an abundant and cheap rawmaterial which is extracted in manyparts of the world. Thus, improvingferrite magnet performance can bethe key to addressing many issuesfacing society as a whole, includingenergy-saving requirements,environmental concerns, andresource conservation. Now TDK hasdeveloped a new productionmethod using as its basis the FB12material to create a mass productiontechnique for FB13B/FB14H high-performance thin-shape anisotropicferrite magnets that can be made assmall as 1.0mm thick.

The world’s first ferrite magnet -discovered by accidentThe world’s first ferrite magnet wasdiscovered in 1930 by Dr Yogoro Katoand Dr Takeshi Takei, two professorsat the Tokyo Institute of Technology.Dr Takei, a student of Dr Kato, wasinvestigating how lowering zinc yieldduring the zinc refining processaffected the physical properties offerrite when he happened uponferrite with high magnetism levels.One day in June 1930, Dr. Takeiaccidentally forgot to turn off hislaboratory equipment before goinghome. The next day, the ferrite hadtaken on incredible magneticproperties, disrupting the magnetic

Thinner Ferrite Magnets Provide High-Performance Alternative to ExpensiveNeodymium Magnet Material

Industry News_Layout 1 20/08/2012 11:57

INDUSTRY NEWS 13


balance (the magnetic field coolingeffect). This marked the discovery ofthe world’s first ferrite magnet. Up to that point, magnets had

almost always been composed ofmetal, but Dr Takei made themonumental discovery that like ironrust, ferrite’s primary component wasferric oxide, making it a viablemagnet material. His discovery was aferrite magnet with the spinel-typecrystal structure represented by theMOFe2O3 chemical formula, amagnet later commercialized as the‘OP magnet’. However, the product’slow magnetic force limited its scopeof application. Mass production ofpractical ferrite magnets began afterthe development ofmagnetoplumbite (MO6Fe2O3)ferrite magnets in the 1950s. Now,the main products in the fieldinclude Ba (barium) ferrite magnetsand Sr (strontium) ferrite magnets.Due to the fact that ferrite

magnets are ceramic (metal-oxide)magnets with oxygen atoms that donot affect magnetism, they willalways demonstrate less magneticforce than metal magnets of thesame volume. However, advances inmaterial technology and sinteringtechnology elevated the magneticproperties of ferrite magnets to rivalthose of alnico magnets (commonalloy magnets) by the 1960s, pavingthe way for the use of ferritemagnets in speakers, inductors,motors and other applications.

FB series of high performanceferrite magnets Sr ferrite magnets are often used inapplications that require strongmagnetic force. This is because Srmagnets offer the best residual fluxdensity (Br) and coercive force(HCJ) of all ferrite magnets.Sr ferrite is represented by the

SrO6Fe2O3 chemical formula, butadding trace elements improves itsproperties. Although researchersmade exhaustive searches for a widearray of different material

compositions, leading many tobelieve that ferrite magnets hadreached their potential, in 1997 TDKdeveloped the La-Co (lanthanum-cobalt) type ferrite magnet (FB9series), taking advantage of a newmaterial composition to achieveremarkable improvements overconventional magnets’ properties.Using this new composition, the FB9series succeeded in improvingmaximum energy product byapproximately 30 % over thetraditional FB6 series.The key to unlocking the

maximum potential of the materiallies in TDK’s process technology, oneof the company’s core technologies.FB12, which entered the market in2007, is another new material thatsurpasses the magnetic properties ofthe FB9 series by miniaturizing andhomogenizing ferrite powder and

TDK thin ferritemagnets

Ferrite characteristics improvement over generations

Using high-performance thin-shape anisotropicmaterial (FB13B,FB14H) to reduce

brushed motorsize and weight

Replacing aneodymium

magnet with ahigh-performance

thin-shapeanisotropic ferrite

magnet (FB13B,FB14H)

establishing microstructure controls.FB12 has over 20 % greatermaximum energy product than theFB9 series, makes dramaticimprovements to demagnetizationresistance in low temperatureregions, and maintains stable powerover a broad temperature rangefrom -40°C to 150°C.

Contributing to motorminiaturization Since a ferrite magnet is a kind of

ceramic, it used to be difficult toproduce it in a thinner size than 3mm through conventionalproduction methods. However,motor and turbine manufacturers areunder constant pressure to maketheir products smaller and lighter.This has led TDK to develop a newproduction method for ferritemagnets called ‘NS1’, which applieshigh density filling method with theFB12 material, enabling theproduction of thin-shape anisotropic


14 INDUSTRY NEWS


ferrite magnets with a thickness of1.0 - 2.5 mm. The FB13B andFB14H ferrite magnets are productsdeveloped using the NS1 method.Motor miniaturization can beachieved by using high-performancethin-shape anisotropic hard ferritematerial (FB13B, FB14H) in a multi-polarized motor design.

Achieving characteristics close tothose of the neodymium bondedmagnet typeNeodymium bonded magnets arehighly prized in motor designsbecause they offer highly flexibleconfigurations and thus can beadapted to different applications.When used in harsh

environments, however, neodymiumbond magnets must be coated toimprove their heat resistance andcorrosion resistance. Thin-shapeanisotropic ferrite magnets (FB13B,FB14H) take advantage of theproperties of ferrite - excellent heatresistance and corrosion resistance -to eliminate the need for anyadditional coating. Magnetization canalso be easily applied to the magnets

after assembly as can be seen in theTable. Also a neodymium bondmagnet can be replaced with aFB13B/FB14H magnet, the processinvolves adjusting the thickness ofthe housing/magnet wall, but doesnot require any changes to therotor’s diameter.

Special configuration enablesfurther improvement in torqueBreaking through the limitations ofconventional methods, the NS1method allows a greater degree of

ferrite magnet configuration freedomand ensures compatibility withspecial wall shapes.The method can, for example,

make it possible to use a C-typemotor magnet in a wide-arcconfiguration that wraps around therotor without having to modify thebasic design of the motor or turbine.This maximizes the use of emptyspace in the motor case, and, bytaking full advantage of theFB13B/FB14H’s potential, achievesmotor performance that rivals that of

neodymium magnets. As shown inthe graph below, the new materialprovides a significant boost in startingtorque, but combining it with a wide-arc shape produces an even greaterincrease in torque, comparingfavorably with the performance ofneodymium magnets. By optimizingparticle orientation and thusimproving magnetization orientation,the magnet achieves the best ferritemagnet properties.

www.tdk-europe.com

Comparison of neodymium magnets, conventional ferrite magnets, and high-performance anisotropic material created via new methods

Vicor introduced a new Picor PI33XX Cool-Power®

ZVS buck regulator series that integrates a Zero-Voltage Switching (ZVS) topology, providingconversion efficiency up to 98 % peak. PI33XXbuck regulators can convert input voltages rangingfrom 8 V to 36 V to output voltages from 1 V to16 V and output current up to 10 A for powerdelivery up to 120 W in a 10 mm x 14 mm x2.56 mm LGA System in Package (SiP). Powercan be further increased by interleaving up to sixPI33XX buck regulators using single wire currentsharing without the need of any additionalcomponents.The use of a ZVS topology enables high-

frequency operation that maximizes efficiency byminimizing the significant switching lossesassociated with conventional buck regulators thatuse hard-switching topologies. The high switchingfrequency of the PI33XX series also reduces thesize of the external filtering components, improvingpower density, while enabling fast dynamicresponse to line and load transients. The PI33XXseries sustains high switching frequency all the wayup to the rated input voltage without sacrificingefficiency and, with its 20 ns minimum on-time,supports large step down conversions up to 36 Vinput. PI33XX series buck regulators require only an

external inductor and minimal ceramic capacitorsfor input and output filtering to form a completeregulator (see demo board). No frequency

compensation, parametric settings or incrementalexternal components are required. A wideoperating temperature range of -40°C to 125°Callows for use in almost any environment.For designers challenged by complex

distribution power schemes, Picor PI33XX offer anoptional I2CTM extended fault telemetry capabilityallowing for six distinct types of fault reporting.Additional device-programmable I2CTM featuresinclude margining, enable pin logic polarity, andphase delay. Device programming is performedvia the Cool-Power Development Tool.

Losses and topologyRegulator MOSFET switching losses are the lossesattributed primarily to the high-side MOSFET duringturn-on, Miller gate charge, and body diodeconduction losses. High-side MOSFET turn-on iswhen the power MOSFET has the highest voltageand current switching and thus the highest powerloss attributed to it. These losses further magnifythemselves within a design as higher inputvoltages are converted or regulated.The higher the input voltage, the higher the

voltage across the primary MOSFET, and the higherthe losses at turn-on. These switching lossesprevent dramatic improvements in overall powersystem solutions. For example, within industrialprocess control systems a desired regulation of 24V to 3.3 V typically is achieved by first regulatingfrom 24 V to 12 V followed by a second regulator

converting 12 V to 3.3 V. In contrast, a singleregulation stage can regulate 24 V to 3.3 V with anefficiency level at or higher than the two stagesand thus improve cost, board space, and reliability.Switching losses also limit the switching

frequency of the regulator. The higher theswitching frequency the more time the MOSFETswitches and the more losses occur. The inabilityto switch at a high frequency limits the use ofsmaller passive components, penalizing thedensity of the regulator. The PI33XX with its ZVStopology allows for operation at a higher frequencyand at higher input voltages without sacrificingefficiency. The ZVS topology is a soft switchingtopology in contrast to a hard switching topologydeployed in conventional regulators. The PI33XX devices operate up to and over

1.5 MHz, which is typically 2x to 3x that ofconventional high density regulators. Higherfrequency operation not only reduces the size ofpassive components but also reduces theburden en external filtering components andallows for fast dynamic response to line and loadtransients.

ZVS switching topologyA direct comparison can be made between aconventional buck topology and the PI33XX withZVS buck topology. First, with the typical approachat the start of a complete switching cycle, the high-side MOSFET is commanded to turn on. Just prior

Zero-Voltage Switching in Point of Load Devices


The IGCT is the semiconductor of choice for demanding high power applicationssuch as Medium Voltage Drives, Marine Drives, Co-generation, Wind PowerConverters, STATCOMs and Interties. ABB’s portfolio offers a complete range ofIGCTs and Diodes for all your high power switching needs. For more informationplease visit our website: www.abb.com/semiconductors

Power and productivityfor a better world™

15000000 Watt. An IGCT controls the equivalentof a small power plant during switching.

ABB Switzerland Ltd / ABB s.r.o.www.abb.com/[email protected]

15_PEE_0512_Layout 20/08/2012 11:34 Page 1

16 INDUSTRY NEWS


to turn on, there is inductor current flowing in theoutput inductor and the synchronous MOSFET. Inorder to avoid cross conduction of the MOSFETs,there is a delicate balance between turning off thesynchronous MOSFET and turning on the high-sideMOSFET.

The result of this design trade-off is often bodydiode conduction. Conduction of the body dioderequires that the stored charge accumulated whileit is conducting at some peak current be sweptaway before the diode can support any reverseblocking voltage. As such, the drain-to-sourcevoltage of the MOSFET is practically clamped at theinput voltage (strongly dependent on parasiticinductance) while very high current flows throughthe body diode and through the high-side MOSFETuntil a depletion region forms and the body diodestarts to support reverse voltage.

Until that time, the instantaneous powerdissipated in the high-side MOSFET is quite high.Part of the losses is incurred due to the reverserecovery current and the rest are due todischarging the MOSFET output capacitance. Inaddition, there are also reverse recovery losses inthe low-side synchronous MOSFET body diode.These losses increase as the switching frequencyor input voltage increase. The power curve for Q1in the switching waveforms for the conventional

buck regulator shows the effect of turn-on. Thepeak instantaneous power at turn-on can be adominant loss contributor for the high-sideMOSFET in high-switching frequency applications.There have been attempts to remedy the switchingloss associated with body diode conductionranging from switching faster to adaptive gatedrivers even to MOSFET improvement throughvarious improved Figure of Merit (FOM) devices.

The PI33XX starts its switching cycle with nearlyzero current flowing in the output inductor. Thevoltage across Q1 is nearly zero as a result of ZVSaction. The inductor current ramps up from zero toa peak value followed by Q1 turning off. Theinductor current is commutated to the Q2 bodydiode for under 10 ns and Q2 turns on. The storedenergy in the inductor is delivered to the load. Theturn-off of Q2 is delayed until the inductor currentis driven negative by current from the outputcapacitor, thus storing energy in the inductor.

Next, the clamp switch turns on, allowing thestored energy to circulate until Q1 needs to turnon again and clamping VS to Vout. To facilitate ZVS,the clamp switch opens. The stored energy thenresonates with the output capacitance of Q1 andQ2. Resonant current flows into Q1 from source todrain and into Q2 from drain to source. Thisresonant action discharges Q1 output capacitance

and charges Q2 output capacitance. Then Q1 canturn in a lossless manner.

The PI33XX addresses the high turn-on losses ofthe conventional regulator by eliminating highcurrent body diode conduction prior to turn on ofthe high side MOSFET, bringing the drain/sourcevoltage of the high-side MOSFET to nearly zeroand produces no high current spikes or damagingringing. The ZVS action applied to Q1 removes theMiller effect at turn-on of Q1, allowing the use of asmaller driver and lower gate drive at turn-on.

Regulation from a higher voltage, at a higherefficiency, and in a smaller form factor is realizablewith implementation of an improved switchingtopology. By utilizing the ZVS topology within thePI33XX, a buck regulator is offered thatdemonstrates high performance regulation up to36 V input voltage exceeding the performancefound in conventional hard switching, high densityregulators.

http://www2.vicorpower.com/zvs_buck2

LEFT: Comparison between ZVS switching technology(left) and conventional buck topology

RIGHT: Efficiencycurves for

different inputvoltages at light

and full load

Picor ZVS buck regulator demo board


e

e

ee

eeee

e

eeee

e

e

eee

e e

e

e

e

ee

e

the world needs innovative electronics.they are on display here.

Tickets & Registration: www.electronica.de/en/2012

25th International Trade Fairfor Electronic Components,Systems and ApplicationsMesse MünchenNovember 13–16, 2012www.electronica.de

17_PEE_0512_Layout 20/08/2012 11:37 Page 1

18 POWER SEMICONDUCTORS www.abb.com/semiconductors


Past, Present and Future of HPT-IGCTThe integrated gate-commutated thyristor (HPT-IGCT) is a state-of-the-art bipolar turn-off device for highpower applications. New devices employing this technology have been released during the past few yearsand in this article we take a look at the newest member in ABB’s product portfolio - the 5.5 kV asymmetricHPT-IGCT. The development of the HPT-IGCT continues, however, and in this article we will also take a lookat one interesting device in the pipeline. Tobias Wikström and Björn Backlund, ABB Switzerland Ltd,Semiconductors

corrugated p-base, has been developedrecently.

IGCT basicsThe first device in the HPT-IGCT family wasthe asymmetric 4500 V device 5SHY55L4500 released in 2009. In 2010, 5SHY42L6500 followed, an asymmetric 6500 Vdevice and in 2012, yet anotherasymmetric device 5SHY 50L5500 rated5500 V was released. All these devicesuse the same package (see Figure 1) with

a copper pole piece diameter of 85 mm.The main strengths of the device lie in its

thyristor-like on-state with maximalpossibilities for engineering the on-stateplasma distribution for optimal trade-offbetween on-state and turn-off losses, itsrugged mechanical design and goodthermal coupling to the cooler. Twodisadvantages of the IGCT compared to theIGBT, the only competitor device in thepower range of the IGCT, is the relativelylarge effort needed to control the device -

The IGCT was introduced about 15 yearsago and has established itself as apreferred technology for handling very highpower. The applications range fromindustrial drives and track-side supplies topower quality and high-current breakers.The IGCT is available as reverse conductingdevices with an integrated free-wheelingdiode and as asymmetric devices. The firstdevices, as well as most devices madetoday, use a flat p-base but a new family ofdevices, referred to as HPT-IGCT, using a

Figure 2: Inductive situation (simulated values)of the individual segment rings on an IGCTwafer showing how inductances can bedistributed over the wafer. The rings far awayfrom the gate contact are more heavily loadedby inductance than the rings in the vicinity.Hence, the gate signal will propagate at somefinite speed and disfavor the gate-remoteregions

Figure 1: The HPT-IGCT, consisting ofthe silicon switch in ahermetic ceramichousing on the left,closely coupled to thegate driving circuit onthe right

ABB_Cover story_Layout 1 20/08/2012 12:18

www.abb.com/semiconductors POWER SEMICONDUCTORS 19


tens of Watts for a 3.6kA device - as well asthe inability to control the anode voltageduring turn-on. The former is due to thefact that it is a current-controlled device.The latter is due to it being a thyristor and,as such, it is either off or on and thetransition between those states is onlystable in theory. Hence, implementing it inmost common inverter topologies meansprotecting the antiparallel diode at turn-on.Nevertheless, thanks to its low losses andefficient cooling, it is and continues to bethe preferred choice for manymanufacturers of very large power inverters.Other requirements for applications such asbreakers for large currents can only be metusing the low on-state of the IGCT.The maximum controllable current

(MCC) of the IGCT does not scale linearlywith device area. The reason is theinductive (and resistive) coupling to areasremotely placed from the gate contact. Thearea scales with the square of thediameter, whereas the MCC merely scaleslinearly, using the same technology. Agraphical summary of this situation ispresented in Figure 2. The technology canbe improved by decreasing the totalinductance in the package (i.e. theminimum, 2 nH, in Figure 2.), improvingthe local ruggedness to facilitate morecurrent redistribution and increasing thedriving gate voltage. The high-powertechnology was built using the first two.

High power design elementsThe enablers for very high current turn-offare a combination of improving the localruggedness of the silicon device itself byemploying p-base corrugation, andincreasing the gate’s reach by minimizingthe impedance, mostly in the gate drivercircuit itself.Finding the optimum for the p-base

corrugation used for improving localruggedness means trading off manyparameters, such as blocking capability,thermal budget, process limitations andruggedness. In general, the higher thedevice voltage, the deeper and morehighly doped the p-base has to be made. The improvements to the gate unit

include improved lifetime of the capacitorsused, using a 6- instead of 4-layered PCBsubstrate, increasing the parallelconnection of the turn-off channel byincreasing the number of MOSFETswitches and capacitors, as well asoptimizing the layout of the componentson the gate unit. With careful selection ofthe used components, it is also possible toreduce the losses in the gate unit.A further improvement to the gate unit

is the possibility to equip the IGCT with ananode-voltage sensor feature to improvethe applicability of the device, facilitating

Figure 3: Circuit used in dynamic testing of the IGCTs (parameters: CCLAMP = 8�F, L� = 300nH, RCLAMP = 0.6 �, LCOMM = 6 �H)

Figure 4: Current handling capability of the 5.5 kV HPT IGCT. At 3.3 kV DC-link voltage, more than 5.5kV can be controlled (device testing stopped without destruction). At 3.9 kV, the tests stopped ataround 4.4 kA, as the over-voltages involved exceed 6 kV, and hence the specified device capability bymore than 500 V

early error detection. The IGCT technology can utilize all

commonly used lifetime adjustmenttechniques. Using electron irradiation,proton irradiation, or both, one can tailorthe electron-hole plasma distribution to thebest shape and tune the trade-off betweenstatic and dynamic losses to the best fit forthe application. Thanks to the vast surplusof charge in the on-state, lifetime at-tenuating techniques can be utilized withina broad range. The 5.5kV device wasdesigned using electrons and protonirradiation from the anode side, whichsignificantly reduces the overall losses.

5500 V device capabilitiesDynamic electrical testing was carried outin a circuit displayed in Figure 3. The clampcircuit used to protect the freewheelingdiode is close to the application and

facilitates rapid and reliable testing, asopposed to measuring replicas of inverters. Safe operating area (SOA) and losses




were evaluated in the dynamic circuit,both at single pulse as well as at burstfrequency (10 kHz) for specialapplications. Samples of waveforms fromSOA measurements are presented inFigure 4. Noteworthy is that SOA testingwas interrupted when the maximalvoltage reached 6 kV during testing.Beyond 6 kV, one would risk a blockingfailure in the clamp-circuit dischargefollowing the switching transient, whichwould not add any information, as thiscondition would be far beyond specifiedcapabilities. The burst capability of the device was

tested - five pulses at 10 kHz. As thetemperature coefficient of the MCC isnegative, the failures always occur at thefifth pulse. Due to limitations in the testcircuit, it is not possible to test at constantcurrent and voltage. Instead, the currentincreases and the voltage decreases asthe pulse train progresses. In this mode,the IGCT withstands a current of around 4kA at a voltage between 3 - 3.3 kV. Ofcourse, when switching at this speed, theprocess is more or less adiabatic whichmeans that the wafer temperature issignificantly higher than the allowed125°C after pulse number 5, if thestarting temperature is 112°C. A typicalpulse pattern from the burst tests ispresented in Figure 5.To optimize the losses in the 5.5 kV

HPT-IGCT, both proton and electronirradiation are used. Figure 6 shows thehigh flexibility that is possible usingdifferent irradiation doses.Using these lifetime tailoring

techniques, a profound influence can beseen on the loss trade-off (Figure 7).Before a newly developed device is

released, it must undergo extendedreliability testing. The performance of thedevice in the presence of cosmic rays isespecially of interest with new Siliconspecifications and Thyristor designs.Testing was done in a proton beam, forwhich a sound correlation to actualcosmic rays has been established, withthe obvious advantage that the testing isdone in a matter of hours instead of years(Figure 8).

In the pipeline - the 150 mm IGCTThe quest for ever higher power ratingsmakes the option of expanding into largerSilicon diameters viable. Since bipolardevices using 150 mm Silicon since havebeen in production for some time now,the means are available. In addition, theimproved scalability of the HPTtechnology is an important enabler fortaking such a step. The first 150 mmreverse conducting 4.5 kV HPT-IGCTprototypes have recently been

Figure 6: Flexibility of on-state voltage tailoring in IGCT technology

Figure 7: Loss trade-off for the 5.5 kA IGCT switching 3.3 kA at 3.3 kV with a Tj of 125°C. Additionalcircuit parameters are listed in the caption of Figure 3

Figure 5: Waveforms from the 10 kHz burst measurement. This example failed at the fifth pulse, at 4.4 kA


CORNELLDUBILIERCDE

Capacitor Solutions for Power Electronics

New! Type 947DHigh Density, DC Link Capacitors

TYPE 947D POWER FILM CAPACITORS

er desigtervtion inat generxNe

ors with higher capacitanclink capacit

ailable in new case sizvA

er the ors o� �er the highest bulk ener�lm capacit

or wind and y f for wind and solar potancxpece

om hundrt frelec. Stionsapplica

able enerenewor rns fer desig

oltage, valuese vors with higher capacitanc

ubiliernell Dor, Ctingsaes and railable in new case siz

ageorgy ster the highest bulk ener

er desigterver inwor wind and solar po

d caeds of standarom hundr

eliable DC gy demand rable ener

. tingsat renr, and curoltage

er wype 947D poTType 947D pos ’’s ubilier

e ing and lif fe eripple �lt, rage

ehicle ic vtrell as elec, as wnser desig

t with onnec, or ctalog listings

ehicle

t with

PE 947D POWER FILM CYT

85, 90, 100 & 116 mm C

UES ALVVALANCE CITTANCE AAPC

O TGE ATTAAPPLIED VOL LT

T AATINGS RIPPLE CURRENT R

ORSCITAAPPE 947D POWER FILM C

ASE SIZES85, 90, 100 & 116 mm C

O 3600 µFTUES

dc1500 VO

O 90 ATTINGS msrelop special desigo devineers tCDE eng

equests or moror sample rF

our ro yns telop special desig

tion, visit wormaal infechnice tequests or mor

.tsemenequirour r

om.cde.cwwtion, visit w CDEor Capacitor Solutions f

DUBILIERCORNELL

onicswer ElectrPo

DUBILIERCORNELL

msr

21_PEE_0512_Layout 20/08/2012 11:35 Page 1



manufactured and a sample is shown inFigure 9.

This device is expected to have a ratedSOA capability in excess of 7 kA against aDC voltage of 2.8 kV at similar conditionsas presented in Figure 3. With this device,it will be possible to make 3-level invertersup to about 20 MW without the need for

series or parallel connection of powersemiconductor devices. With the free-wheeling diode already integrated in theIGCT package, assembly designs can beimproved. Critical to the performance isthe size and uniformity of the inductancefor the segment rings. However, byemploying the design elements from the

existing HPT-IGCTs, the wafer has thetargeted performance in respect touniformity and absolute inductance.

Today, the added challenges lie in thedesign of the housing package and the gateunit. On the wafer design front, the mainchallenge is to provide a soft integrateddiode performance under differentoperational conditions including the bestIGCT versus diode area ratio selection.Hence, there is still some work to be donebefore the device is ready, since extensiveelectrical and environmental qualificationsalso lie ahead before the first devices canleave the factory.

Figure 8: Corresponding failure rates due to cosmic rays measured using biased devices in a high-energy proton beam. The specification can be used arbitrarily; however, the 100 FIT level has becomethe accepted standard where failures due to cosmic rays will become a significant factor in the field

Figure 9: Prototype of 150mm 4.5 kV RC-IGCT

DirectContact:

Ian Atkinsonon +44 (0)1732 370340

[email protected]

�DIRECT MAIL

�LIST RENTAL

�E-CASTING


www.danfoss.com/solar POWER SEMICONDUCTORS 23


Comparison of 1200V SiCPower Switching DevicesSilicon Carbide (SiC) power semiconductors being actually commercialized and are promising devices forthe future. To outline their characteristics the switching and conducting performance of two types of SiCnormally-on JFETs, a SiC normally-off JFET two types of SiC MOSFETs and a SiC BJT have been analyzed bymeans of measurements at exactly the same boundary conditions and compared to each other. Beside ofthe switching and conducting behaviors the requirements for driving these devices have been investigated.W.-Toke Franke, Danfoss Solar Inverters, Denmark

Since 2002 the first silicon carbidediodes (SiC) are available on the market.Already at that time the diodes outperformtheir Silicon counterparts regarding toswitching characteristics and conductingbehavior. Since then a significantdevelopment process in the field of widebandgap semiconductors could beobserved. Today a number of SiC basedswitching power semiconductors in thevoltage range from 600 V to 1700 V andmaximum currents up to 50 A areintroduced to the market or are about tobe launched. Typical devices in the voltage range of

1200 V are EMJFETs, DMJFETs, MOSFETsand BJTs [1, 2, 3]. These powersemiconductors have been investigatedconcerning their conducting and switchingcharacteristics at exactly the sameconditions. Their current capability isbetween 12 A and 25 A. Furthermore therequirements for driving these devices arediscussed as they differ from the well-known Silicon power devices [4].

Properties of Silicon CarbideThe great interest for silicon carbide power

semiconductors is based on the physicalfeatures of the 4H-SIC polytype, which hassome advantages over Silicon in therelevant criteria for power semiconductors.The most important parameter is thebandgap of 4H-SiC that is about three timesas large as for Si. Table 1 shows that thecharge carrier density of SiC is at roomtemperature 19 orders of magnitudesmaller than for Silicon [5, 6]. In Figure 1the intrinsic charge carrier density versus

temperature is shown. It is clearly visiblethat the charge carrier density of SiC issignificantly below typical doping density(1·1015 cm-3) even at high temperatures.However, with Silicon this number is alreadyreached at about 250°C. The impact of thisfor semiconductors is a significantly lowergeneration of charge carriers within thespace-charge region and thus leads to lowerleakage currents in off-state. Theoretically, SiC power semiconductors

can be operated at very high temperaturesdue to this feature. The limiting factors arethe packaging and connection facilities torealize the electrical and thermalinterfaces to the outside. Besides that thelarge band gap enables Schottky barrierswith high potential barrier so that highvoltages with low leakage currents couldbe blocked.

SiC bipolar transistorThe Silicon bipolar junction transistor (BJT)is no longer relevant in power electronics,although its production is simple andinexpensive. However, the effort for drivingthe BJT is high and therefore cost-intensive, since a high base current isrequired. On the other hand is the SiC BJTwith 1200 V blocking voltage, a

Table 1: Physicalproperties of Si andSiC

Figure 1: Intrinsiccharge carrier densityvs. temperature

Danfoss–0512_Layout 1 20/08/2012 12:27

24 POWER SEMICONDUCTORS www.danfoss.com/solar


temperature range of up to 200°C and 12A or 40 A collector current. The requiredbase current is low compared to Sitransistors. The current gain � is defined bythe ratio between collector and basecurrent (� = iC/iB) and can be traced backto the physical characteristics of thetransistor. A gain higher than 1 can be obtained

when the doping concentration of theemitter is significantly above the dopingconcentration of the base and the basewidth is in the range of the diffusion length.The use of SiC as a substrate for the

bipolar transistor has a number ofadvantages compared with the Silicon-based transistors - significantly shorterswitching times and despite the largerband gap lower saturation voltages and higher current gains could be realized.

In order to keep the bipolar transistor inthe conducting state the driver has toprovide a base current. Because of thevoltage drop across the base-emitter pn-junction, power is consumed throughoutthe on-state. A possible driver circuit isshown in Figure 2. The base current islimited by the resistor RB and for fastswitching a capacitor CB is connected inparallel to RB. CB acts like a short circuitduring switching, so that the base regioncould be flooded and cleared out fast withcharge carrier respectively for turning onand off.

SiC depletion mode MOSFETCharacteristic for the depletion mode JFET(DM-JFET, Normally-On JFET) is that it isturned on at a gate voltage of 0 V [7, 8].This characteristic is contrary to allsemiconductors established on the market[9] and leads to considerable efforts forthe driver: If the JFET is directly driven, itmust be ensured that the driver is alsosupplied with power in case of a failure onthe AC or DC side, to hold the JFET inblocking state. If the JFET is directly drivena gate voltage of -15 V is required to block

the device and 0 V to hold the device inconducting state. For further reduction ofthe on-resistance a positive voltage of upto 2 V can be applied.An alternative to drive the JFET directly is

to operate it in a cascode (see Figure 3).

MOSFET.However this topology also has two

disadvantages: Firstly, a positive gate-source voltage cannot be applied to theJFET and therefore the optimumconducting performance cannot beachieved. Secondly, the full dynamicpotential cannot be utilized, because theswitching performance is mainly given bythe Si-MOSFET.However, the leading manufacturers of

these devices Semisouth and Infineonintroduce driver concepts in theirapplication notes that combine theadvantages of the cascode with the directdriven approach. The effort for the driverthus increases considerably.

SiC enhancement mode MOSFETThe enhancement mode JFET (EM-JFET,Normally-Off JFET) differs from the DM-JFET by blocking at a gate voltage of 0 V[10]. Initially it resembles theperformance of the established powersemiconductors. However, the driving isnot comparable. Due to the JFETstructure it starts conducting at a gate-source voltage of 1,5 V. At about 3 V it isfully turned on; however the pn-junctionbetween gate and source startsconducting at about 2 V and lead to agate current that has to be provided bythe driver. This gate current isconsiderably lower than the base currentof bipolar transistors, but due to thevoltage drop it causes power loss. To turnon the EMJFET optimally, a voltage pulseof 12 to 15 V must be supplied to thegate for about 200 ns. This voltage pulsecan be realized as described for thebipolar transistor by a parallel connectedcapacitance to the gate resistor (Figure 4). The gate resistor must be dimensioned

to limit the maximum gate current duringon state and to dissipate the power losses.A more efficient method that also allowsany modulation index is a 2-stage driver.Here the voltage pulse is provided by anadditional driver stage [11] as depicted inFigure 5.

SiC MOSFETThe SiC MOSFET differs from Si MOSFETmainly due to its higher blocking voltage.In principle blocking voltages up to severalkV can be achieved with SiC MOSFET dueto the high critical electric field strength ateven superior dynamic performance. Thegate driver requirements of the actualgeneration differ only slightly from the SiMOSFET. The MOSFET is turned off at 0 V,however it is recommended to bias thegate in blocking state in order to avoidunintentional turn on by electromagneticcoupling. The MOSFET is fully turned on at

Figure 2: Driver forbipolar transistor

Figure 3: Cascode consisting of a SiC JFET and aSi MOSFET

Therefore a Si-MOSFET with lowblocking capability (e.g. 30 V) and a verylow RDSon the gate of the JFET is connectedto the source terminal of the MOSFET andits gate is driven by a conventionalMOSFET driver. When the MOSFET isblocking the potential of the sourceterminal of the JFET is lifted. Since the gateof the JFET is on the source potential ofthe MOSFET, it results in a negative gate-source voltage of the JFET and it blocks aswell. The voltage drop across the MOSFETis limited to the gate voltage that isrequired to block the JFET. The cascode isturned on by turning on the MOSFET. Inthis case the voltage drop across theMOSFET is almost 0 V and hence the gate-source voltage of the JFET is also 0 V, theJFET is conducting.The advantage of the cascade is that the

JFET can be driven as a conventional

Danfoss–0512_Layout 1 20/08/2012 12:27

www.danfoss.com/solar POWER SEMICONDUCTORS 25


20 V. As with all other SiC powersemiconductors, the complete dynamicpotential can only be exploited if the driveris correspondingly fast and the electricalconnection is extremely low inductive, toavoid over-voltages due to the extremelyshort rise and fall times of current andvoltage.

Comparison of SiC powersemiconductorsWhile IGBTs are dominating the 1200 Vvoltage range of Silicon based switchestoday, older switching device structures asBJT and JFETs are relevant for SiC again.Since these structures are relatively simpleto manufacture they are applicable for SiC,as the experiences in the productionprocess are still minor in comparison toSilicon. To be able to select the rightcomponent for a specific application itsproperties must be well known. In generalit is a fact that components of the sametype but from different manufacturerscannot just be replaced since they differ inthe physical internal structure and thusalso differ regarding driving, switching andconducting behavior.

Driving a SiC MOSFET is similar todriving a Si IGBT. The handling of the

bipolar transistor is safe due to its basecurrent, however it requires a non-negligible driving power which has on theone hand to be provided by the driver andon the other hand leads to additionallosses in power semiconductors.

The EMJFET can be driven also rathersafely, but the driving losses are muchsmaller compared to the BJT. However, thecomplexity of the driver increases due tothe different gate voltage levels (peakswitch-on voltage, conducting voltage, andnegative reverse voltage). For the DMJFETit can be chosen between driving the JFETdirectly or in a cascode that leads basicallyto a Si MOSFET driver. They both requirecomplex driver structures if maximumperformance of the JFET should be utilizedand at the same time safe operation in allcircumstances are applied.

The DC-current capacities of theanalyzed power semiconductors at roomtemperature are stated in Table 2according to the datasheets in order toevaluate the measurement results. Exceptfor the bipolar transistor all powersemiconductors are in the same powerrange.

The conducting behavior is illustrated inFigure 6. The first three graphs show the

U-I characteristic of the powersemiconductors at 25°C, 80°C and 150°Cjunction temperature. Since the CreeMOSFET is only specified up to 125°C, the150°C curve was extrapolated for bettercomparability. It can be seen that theMOSFET has low temperature dependenceand thus also at typical operatingtemperatures it can carry almost the fullrated current. For the JFETs and the bipolartransistor a halving of the current byincreasing the temperature is observed.Regarding the bipolar transistor it isremarkable that in contrast to Silicon BJTand IGBT no collector-emitter saturationvoltage is visible, but shows a resistive V-I-characteristic as known from unipolarcomponents.

The other two curves show the reverseconducting performance while turned on.Since the BJT cannot be operated withnegative current, it was not considered.The temperature dependence is similar tothat in forward conducting mode.Noticeable is the current waveform of theMOSFET from Rohm which has V-I-characteristic similar to a bipolarcomponent. This could be explained bythe existence of an antiparallel diode.

The total switching losses are shown at

RIGHT Figure 4:EMJFET driver

BELOW Figure 5:EMJFET Driver withadditional voltagepulse for optimizedperformance

Danfoss–0512_Layout 1 20/08/2012 12:27



650 V and different currents at 25 and150°C in Figure 7. It was decided not todifferentiate between turn-on and turn-offlosses since due to the very shortswitching times the losses could be movedbetween turn-on and off depending on the

parasitic inductance and capacities in thecommutation path. For all components thesame SiC Schottky barrier diode was used.For the Cree MOSFET the losses at 150°Cwere extrapolated. All powersemiconductors have been investigated

with an identical test setup (double pulseexperiment [4]). The drivers were as far aspossible designed identical and they havebeen configured to allow the fastestswitching times without instable ringing sothat the lowest switching losses could beachieved.

Regarding the dynamic performance theJFETs and the bipolar transistor showsignificantly lower losses than theMOSFETs. Especially for the bipolartransistor but also for the JFETs the testsetup and the use of discrete components,with their parasitic effects turned out to bethe limiting factor. Therefore the switchingtimes had to be artificially slowed down byapplying ferrite beads.

The temperature dependence of allinvestigated power semiconductors is lowand some even show a slight reduction inswitching losses at high temperatures.

Regarding cost, the processing of theraw material Silicon Carbide is significantlymore expensive than Silicon. Firstly, themanufacturing process itself is morecomplicated as it requires many processsteps at higher temperatures and longertime. Secondly, the number of defects perwafer is much higher, so that only a smallproportion of the raw material can be usedfor the production of semiconductors [12].This is also the reason why SiC switchesare manufactured with rather low currentcapacities, because the probability ofproducing a defect-free chip is better thesmaller the chip area is and thus thecurrent capacity. At present the cost perchip varies and should therefore berequested directly from the manufacturerat interest.

ConclusionAll relevant SiC power semiconductorswere investigated in terms of theirelectrical features. A final assessment,which of these power semiconductors isthe best cannot be given because thisdepends very much on the finalapplication. In fact each SiC power

Figure 6: Conductingcharacteristic of theSiC powersemiconductors atdifferenttemperatures andpositive/negativecurrent

Table 2: Revieweddevices in accordancewith the datasheets

Figure 7: Switchinglosses at 25°C and150°C versus currentat 650 V

Danfoss–0512_Layout 1 20/08/2012 12:27

27_PEE_0512_Layout 20/08/2012 11:36 Page 1



semiconductor has its specific advantages.Therefore a universally validrecommendation for one of these powersemiconductors cannot be given. This work was supported by the project

“SUNRISE - Opbygning afinnovationskapaciteten vedrørede solcelle-anlæg” and partly funded by RegionSyddanmark (journal 09/12927) and theEU. It was first presented at PCIM Europe2012 under the title “Comparison of SixDifferent SiC Power Switching Devices inthe 1200 V Range”.

Literature[1] P. Friedrichs and R. Rupp, “Silicon

carbide power devices - currentdevelopments and poten-tialapplications,” in Power Electronics andApplications, 2005 EuropeanConference on, 0-0 2005, pp. 11 pp.-P.11.[2] P. Friedrichs, “Sic power devices -

recent and upcoming developments,” inIndustrial Electron-ics, 2006 IEEEInternational Symposium on, vol. 2, July2006, pp. 993-997.[3] S. Araujo and P. Zacharias,

“Analysis on the potential of siliconcarbide mosfets and other innovativesemiconductor technologies in the

photovoltaic branch,” in PowerElectronics and Applica-tions, 2009. EPE‘09. 13th European Conference on,Sept. 2009, pp. 1-10.[4] W.-T. Franke and F. Fuchs,

“Comparison of switching andconducting performance of sic-jfet andsic-bjt with a state of the art igbt,” inPower Electronics and Applications,2009. EPE ‘09. 13th Eu-ropeanConference on, sept. 2009, pp. 1 -10.[5] D. Schröder,

Leistungselektronische Bauelemente,2nd ed. Springer, 2006.[6] G. Davis, L. Casey, M. Prestero, B.

Jordan, J. Scofield, K. Keller, J. Sheahan,J. Roach, and R. Singh, “Predictiveprognostic controller for wide band gap(silicon carbide) power conversion,” inAerospace Conference, 2007 IEEE, mar.2007, pp. 1 -17.[7] P. Friedrichs, H. Mitlehner, K.

Dohnke, D. Peters, R. Schorner, U.Weinert, E. Baudelot, and D. Stephani,“Sic power devices with low on-resistance for fast switchingapplications,” in Power Sem-iconductorDevices and ICs, 2000. Proceedings.The 12th International Symposium on,May 2000, pp. 213 -216.[8] P. Friedrichs, H. Mitlehner, R.

Schörner, K. O. Dohnke, R. Elpelt, and D.Stephani, “Applica-tion-orientedunipolar switching sic devices,” inMaterials Science Forum, January 2001,pp. 1185-1190.[9] A. Knop, W.-T. Franke, and F.

Fuchs, “Switching and conductingperformance of sic-jfet and esbt againstmosfet and igbt,” in Power Electronicsand Motion Control Conference, 2008.EPE-PEMC 2008. 13th, Sept. 2008, pp.69-75.[10] R. Kelley, M. Mazzola, S.

Morrison, W. Draper, I. Sankin, D.Sheridan, and J. Casady, “Power factorcorrection using an enhancement-modesic jfet,” in Power Electronics SpecialistsConference, 2008. PESC 2008. IEEE,June 2008, pp. 4766-4769.[11] R. Kelley, A. Ritenour, D.

Sheridan, and J. Casady, “Improvedtwo-stage dc-coupled gate driver forenhancement-mode sic jfet,” in AppliedPower Electronics Conference andExposition (APEC), 2010 Twenty-FifthAnnual IEEE, feb. 2010, pp. 1838 -1841.[12] B. Thomas, C. Hecht, R. Stein,

and P. Friedrichs, “Challenges in large-area multi-wafer sic epitaxy forproduction needs,” Materials ScienceForum, pp. 527-529, 2006

Name:

Company Name:

Address:

Post Code:

Tel: Total Number of Copies @ £ p+p Total £

Drives S & S Hyd H/B Pne H/B Ind Mot CompH/B H/B Air

DFA MEDIA LTD, Cape House, 60a Priory Road, Tonbridge, Kent TN9 2BL

QUANTITY QUANTITY

Hydraulics&Pneumatics There are now 6 of these handy

reference books from the publishers ofthe Drives & Controls and

Hydraulics & Pneumatics magazines.

Published in an easily readable styleand designed to help answer basicquestions and everyday problems

without the need to refer to weightytextbooks.

We believe you’ll find them invaluableitems to have within arms reach.

From the publishers of

QUANTITY QUANTITY QUANTITY

If you would like to obtain additional copies of the handbooks, please completethe form below and either fax it on 01732 360034 or post your order to:Engineers Handbook, DFA MEDIA LTD,Cape House, 60a Priory Road, Tonbridge, Kent TN9 2BLYou may also telephone your order on 01732 370340 Cheques should be made payable to DFA MEDIA LTD and crossed A/C Payee.Copies of the handbooks are available at £4.99 per copy.Discounts are available for multiple copies. 2-5 copies £4.30, 6-20 copies £4.10, 20+ copies £3.75.Postage and Packaging:1-3 copies £2.49 4 copies and over £3.49

QUANTITY

PRACTICAL ENGINEER’S HANDBOOKSPRACTICAL ENGINEER’S HANDBOOKS

HYDRAULICS

INDUSTRIALMOTORS

SERVOSAND STEPPERS

PNEUMATICS

COMPRESSED AIRINDUSTRIAL

ELECTRIC DRIVES

PLEASE ALLOW UPTO 28 DAYS FOR DELIVERY

Danfoss–0512_Layout 1 20/08/2012 12:27

www.vincotech.com POWER MODULES 29


Asymmetrical ParasiticInductance Utilized forSwitching Loss Reduction High efficiency of power conversion circuits is a design goal on its own. The reduction of switching losses isthe basis for higher switching frequencies which lead to a reduction of the size and weight of the passivecomponents. The improvement from 96 % to 99 % efficiency will reduce the effort for cooling by factor 4.With the utilization of parasitic inductance and consequent execution of basic rules of power electronics is anew power electronics solution based on standard Si components disclosed which extends the traditionaldesigns. The presented new power module concept combines a low inductive turn off with the utilization ofthe parasitic inductance for a reduction of the turn on losses and the usage of three level switching circuitswith the paralleling of fast switching components with components with low forward voltage drop. Michael Frisch and Temesi Ernö, Vincotech Germany and Hungary

The goal of this development project isthe reduction of switching losses in powerapplications >100 kW with screw typemodules. The limitations in suchapplications are mainly parasitic effects asstray inductance [7, 8] and reverse recoverybehavior of the diodes [5]. The over-voltagespike caused by the parasitic inductancewill limit the turn-off switching speed. Thelosses and the increased electromagneticinfluence (EMI) caused by the reverserecovery behavior of the freewheelingdiode is the drawback for increased turn-onswitching speed. In a first step, the parasiticinductance is reduced to a minimum tosolve the issue with turn-off. In a secondstep are the turn-on losses reduced. This isachieved with the utilization of the parasiticinductance for turn-on by keeping the lowinductive turn-off behavior. The third activityfor a switching loss reduction is theintroduction of a neutral point clamped(NPC) inverter topology. Finally a specialtopology for paralleling MOSFET with IGBTis introduced to show the advancedprospects of the idea. The feasibility of thenew concept is proven with a powermodule concept incorporating all thediscussed arrangements.

Theory about switching losses withinductive loadThe power dissipation in power electronicsis caused by conductive losses andswitching losses. The conductive losses aredefined by the forward voltage drop in thesemiconductor. The switching losses aredependent on the switching speed of thetransistor, the reverse recovery behavior of

the diode, serial inductance, and additionalparasitic effects.In a system with inductive load (see

Figure 1) is the freewheeling diodeconducting at turn on and the outputvoltage equals the negative DC-bus voltage(DC-). The transistor starts now to conduct.As soon the transistor takes over thecomplete output current, the output voltagewill develop to the positive DC-bus voltage

(DC+). The diode faces reverse voltageand will conduct in reverse direction. Thiswill cause losses in the diode (EREC) and itwill increase the current in the transistor.This current peak is often the root cause ofEMI in the system. After the diode iscompletely recovered and blocking, thecurrent in the transistor will fall back to thelevel of the output current. The turn-onprocess is now complete. An increasedserial inductance with the transistor willreduce the turn-on losses. The energystored in the serial inductance is calculatedaccording [9] to EL = 1⁄2 x L x I2.At turn off will the voltage at the

transistor develop up to the DC-bus voltagelevel. The diode will take over the outputcurrent. The over-voltage will causeadditional losses and it endangers thetransistor to be destroyed. The usage of fasttransistors is limited on the inductance andthe maximum current, as the voltage peakis dependent on the turn-off speed. The

Figure 1: Switching circuit with inductive load

Figure 2: Asymmetrical inductance in a switchingcircuit with feedback to the main DC-link

Figure 3: Asymmetrical inductance in a switchingcircuit with regeneration of the stored energy

Vincotech_Feature_Layout 1 20/08/2012 12:33

30 POWER MODULES www.vincotech.com


stored energy in the serial inductance of the DC-input causes avoltage over-shooting according [1] to VCE(peak) = VCE + L x di/dt.

Low inductive module technologyWith the new low inductive moduletechnology [3] we achieve:� Fast and reliable turn-off of high currentpower modules and

� switching loss reduction (turn-off).The reduced voltage overshot at turn offallows the usage of fast components.The reduced inductance will not reduce theswitching losses at turn-on. The turn-offlosses will decrease but the turn-on lossesof the transistor and the reverse recoverylosses in the diode will increase even more[4]. The efficiency of low inductive circuitswill increase with the utilization of fastcomponents. The goal of low turn-onlosses without increasing EMI requires ultrafast freewheeling diodes.

Asymmetrical inductanceIncreased inductance at turn on and ultralow inductance at turn off is the new goal.This approach is named “AsymmetricalInductance”. The way to achieve the newswitching behavior, is the utilization of theparasitic inductance Lparasitic at turn-on andbypassing it during turn-off. The diode Dtran(see Figure 2) assigns the stored energy ofthe parasitic inductance during turn-off tothe integrated capacitor Ctran.The stored energy circulates in Lparasitic, Dtran

and Rtran until it is dissipated in the parasiticresistor. With this circuit we are able torelease the semiconductor from switchinglosses but some energy has to bedissipated in passive components. Theregeneration of the stored energy with aDC-DC circuit (see Figure 3) is an option toincrease the efficiency.

Verification of the asymmetricalinductanceThe idea is verified with the comparison of different parasitic inductanceon a traditional power module setup andasymmetrical setup with integrated snubbercapacitors (see Figures 4 - 6).Test conditions were RG = 2� (+/-

15V), VDC = 600V, IOUT = 400A with InfineonHS 3 / 1200V / 400A. Results: EON=16,92 mJ, EOFF=27,78 mJ,

EREC=31,78 mJ. The switching losses arefine but more important is the voltage over-shooting at turn-off. The most critical case isturn-off at low temperatures. At 25°C is anover-voltage of 180 % measured whichlimits the usage to 650 V. Safe turn-off atover-current conditions is not anymorepossible. In our test the module failed at720 A / 600V / TJ=25°C.The identical measurement is now

performed with asymmetrical inductance of50 nH at turn-on and 5 nH at turn-off(Figures 7 - 9).Results: EON=15,487 mJ, EOFF=25,66 mJ,

EREC= 28,27 mJ. The switching losses of thenew asymmetrical setup are lower.Surprisingly are not only the turn off lossesreduced, all switching losses are reduced.The reason for the turn-on loss reduction isthe reverse recovery behavior of the circuit.

At turn-on is the current of the transistor T1increased by the reverse recovery currentthrough the diode D1. After completedrecovery is the current reduced but in theparasitic inductance Lparasitic is the additionalenergy stored which causes an over-voltageat the collector of the transistor comparedwith the positive voltage in the transientcapacitor Ctran.The energy will flow into the capacitor.

Figure 4: IGBT turn-on characteristicswith symmetricalinductance L[ON] =L[OFF] = 50 nH

Figure 5: Diodecharacteristics withsymmetricalinductance L[ON] =L[OFF] = 50 nH

Figure 6: IGBT turn-off characteristicswith symmetricalinductance L[ON] =L[OFF] = 50nH

Figure 7: IGBT turn-on characteristicswith asymmetricalinductance L[ON] =50 nH, L[OFF] = 5 nH


www.vincotech.com POWER MODULES 31


This will reduce the reverse current in thediode and the voltage drop in the transistorwhich results in a significant reduction ofthe switching losses.

With the asymmetrical inductance circuit,it is possible to increase the inductancefurther to take advantage out of the turn-onloss reduction.

In a setup with Lparasitic[ON] = 90 nH,Lparasitic[OFF] = 5 nH we get EON=12,44 mJ,EOFF=25,77 mJ, EREC=26,70 mJ.

Decoupling of upper and lowertransistorThe circuit in Figure 10 is a solution forreducing the turn-on losses due to theoutput capacitance of the complementaryswitching device and the elimination ofcross conduction during turn-on. The goalof this idea is the decoupling of the outputtransistors. The output is divided into twobranches and the parasitic inductance ofthe output connection is utilized todecouple the components. The

corresponding diodes are direct connectedfor achieving a low inductive commutationloop.

Multi level topologyThe usage of multi level topology as neutralpoint clamped inverter (NPC) or mixedvoltage NPC inverter cuts the voltage at theswitching transistor into half which reducesthe switching losses accordingly [2]. On topon this loss reduction have thefreewheeling diodes to be sized only forhalf of the voltage e.g. 600V instead of1200V. But for 600V are manycomponents with ultra fast reverse recoverybehavior available.

The target of “Advanced Paralleling” is tomerge the advantages of standard NPCwith mixed voltage NPC [6] (see Figure11). Here the output current faces only onejunction, this leads to reduced static losses.

Parallel switch is the paralleling of a fastcomponent (e.g. MOSFET) and acomponent with low voltage drop (e.g.

IGBT). The loss reduction is achieved byrendering the static losses of the switch tothe IGBT and the dynamic losses to theMOSFET. Even smarter is to use a so-calledmixed voltage NPC topology and to parallelwith just 600V MOSFETs a 1200V IGBT(see Figure 12).

In the advanced paralleled NPC is theoriginal idea of paralleling MOSFET withIGBT maintained (see Figure 13). TheMOSFET and the IGBT are turned onsimultaneously. The MOSFET as the fasterdevice will take over the current at turn-on.The IGBT will turn on with zero voltage. Thevoltage drop in the IGBT is lower so theIGBT will take over the major share of thecurrent. At turn-off is the gate signal of theMOSFET delayed. The IGBT will turn off, theMOSFET will take over the current and willturn off with a delay of some 100 ns.

The measurements are showing that theadvanced paralleled NPC topology is ableto cut the switching losses into half.

The inverter reaches now more than

Figure 9: IGBT turn-off characteristicswith asymmetricalinductance L[ON] =50 nH, L[OFF] = 5 nH

Figure 8: Diodecharacteristics withasymmetricalinductance L[ON] =50 nH, L[OFF] = 5 nH

Figure 10: Pseudo half-bridge topology with decoupled branches for improved switching behavior

ABOVE Figure 11: Mixed voltage NPC circuit

ABOVE Figure 12: Advanced paralleled NPC circuit


32 POWER MODULES www.vincotech.com


99% efficiency with a PWM switchingfrequency of 16 kHz. At 64 kHz switchingfrequency is the efficiency still above 98%(see Figure 14)!

Power module conceptThe combination of all this ideas leads tothe following power module specification:� 200 kVA output power at 20 kHz� Asymmetrical parasitic inductance withenergy regeneration

� Decoupling of low side and high sideswitches

� Three-level topology� Paralleling of fast MOSFET or IGBT _swith components having low static losses.

With these ideas in mind we developed apower module concept:� 1200V/500A power rating� Asymmetrical parasitic inductance withonboard snubber capacitors (5nH turn-off inductance) and DC-DC regenerationcircuit

� Separated outputs for high- and low-sidecircuit for decoupling of thecorresponding switches

� 3-phase advanced paralleled mixedvoltage NPC topology.The result is shown in Figure 15 and

Figure 16. The inductance shown in theschematics is the parasitic inductance ofthe power module setup.

ConclusionBeside the advantages of new technologiesas SiC or GaN switching components it isstill possible to increase the efficiency withstandard Si Components. To extendtraditional power designs we have to

additional reduction of the turn-on losses.The parallel switch technology achieveshighest efficiency at elevated switchingfrequencies of 50 kHz and more.

Literature[1] Siemens: “IGBT Fundamentals” ,

May 1997[2] Michael Frisch and Temesi Ernö:

“Advantages of NPC Inverter Topologieswith Power Modules”, VincotechGermany and Hungary 2009[3] Michael Frisch and Temesi Ernö:

“Power Module with Additional LowInductive Current Path”, VincotechGermany and Hungary 2009[4] Wilhelm Rusche and Marco

Bässler: “Influence of Stray Inductanceon High-Efficiency IGBT Based InverterDesigns”, Infineon Technologies,Warstein, Germany 2010[5] Peter Haaf, Jon Harper: “Diode

Reverse Recovery and its Effect onSwitching Losses”, November 2006[6] Akira Nabae, Isao Takahasi,

Hirofumi Akagi: “A New Neutral-Point-Clamped PWM Inverter”,September/October 1981[7] Dr.-Ing. Paul Chr. Mourick:

“Parasitic Inductivities and ParasiticOscillations an Overview”, 24. Feb. 2011[8] Dr.-Ing. Eckart Hoene: “ Parasitic

Effects - An overview”, ECPE, Feb. 2011[9] Helmut Lindner, Dr. Ing Harry

Brauer, Dr. Ing. Constans Lehmann“Elektrotechnik”, 1982 (page 82)

remember the basics of power electronics.The low inductive module concept

ensures the fast and reliable turn-off in highcurrent power modules and reduces thevoltage over-shoot. The power modulesetup with very low internal parasiticelements it is able to utilize the externalstray inductance for a reduction ofswitching losses without investing inexpensive high-speed semiconductortechnology. The asymmetrical inductanceleads to lower switching losses, reducedEMC and minimized effort for the invertermechanics. Low inductive bus bars are notneeded anymore. A flexible low cost cableconnection in the DC link can be used. Theincreased serial inductance will just cause

Figure 13: Currentsharing in the parallelswitch system (redcurrent in theMOSFET, blue currentin the IGBT)

Figure 14: Advanced paralleled NPC - efficiencyvs. switching frequency in steps from 2 kHzto 128 kHz doubling in each step - 2 (blue), 4, 8,16, 32 (green), 64 (yellow), 128 kHz (orange)

ABOVE Figure 15: Power module conceptfor a 3-phase advanced paralleled NPCwith asymmetrical inductance andregeneration circuit

Figure 16: Power module withintegrated snubber capacitors andasymmetrical inductance


www.proton-electrotex.com POWER SEMICONDUCTORS 33


Static Loss MeasurementMethods for QualityImprovementSteady increase of power for electric converter units leads to the constant enhancement of characteristicsand load capacity of power semiconductors. Thus, requirements to the maximum current load of powerthyristors and diodes, limited by heat-release losses in a semiconductor element and the intensity of heatremoval from the die, are also increasing. Power semiconductor manufacturers do their best to themaximum reduction of conducting losses and preserving all the remaining characteristics at the requiredlevel by precise measurement of forward voltage drop. Alexey Poleshchuk, Automation Lab Engineer,Proton-Electrotex, Orel, Russia

Sufficiently small spread of parameterswithin 10-15% range can have a keyinfluence on comparative load capacity ofsemiconductors due to the restrictions ofmaximum heat removal from the die. Onthe other side, it’s often necessary tocommute the currents higher than current-carrying capacity of large diameters dies

(90 mm and more). In this case it’snecessary to connect powersemiconductors in parallel in a single unitwhere the devices should be preciselyselected according to their characteristics inorder to provide symmetric semiconductorload and annealing of heat release. The above-mentioned tasks determine

the necessity of maximum precisemeasurements of semiconductorcharacteristics, responsible for the staticlosses of conductivity (in this case - peakon-stage voltage (VTM), for their correctclassification and selection in groups forparallel connection. In order to measureVTM at the operating currents insemiconductors with large diameters it’snecessary to have the equipment able toform current pulses of desired amplitudeand form with high accuracy andfrequency. Thus not only current pulseamplitude is significant for characteristicsmeasurement, but also its length, formand further treatment of the received dataare important as well.

Current pulse formationThe developed equipment uses the parallelbuck-converter topology operating as acurrent supply to form a current pulse withvariable form. A set of capacitor banks(batteries) commuted by high-speed IGBTmodules serves as the energy supply forthe pulse and is aimed to form currentpulse with the desired form and amplitudeafter filtering. Module construction withparallel connection of cells allows sizingunit power within broad limits of 1 to 9 kAand maintaining independent automaticdiagnostics of the condition of each battery.By using the adaptive circuit of signal

digitalization and digital control of powerkeys it’s possible to obtain almost any formof current pulse, including trapezoidal, halfsinusoidal, step trapezoidal and S-shaped.The main requirement to the control

circuit is the maximum precise generationof the desired current pulse with lengthsmall enough, as well as provision of itsrepeatability within the series of tests.

Figure1: Functional scheme of measuring module

Figure 2: Simplified schematic of current regulator

Proton_Feature_Layout 1 20/08/2012 12:39

34 POWER SEMICONDUCTORS www.proton-electrotex.com


Moreover, the characteristics of the circuitruns are rather stable and do not changeduring the measurements. This allowsrealizing current regulator that performscurrent control in the circuit on the base ofpreviously formed mathematical model(model following control). The scheme inFigure 2 consists of* ID(t) - desired form of current pulse;* WM - model of the object controlled,current circuit;

* FM - filter of measured current value;* I(t) - actual current value in a circuit;* g(t) - external perturbation;* LQR-regulator - regulator, minimizing anerror of current pulse formation.

With the more or less identified systemmodel it’s possible not only to increase theprocessing speed of the desired signal, but tosensitize optimal filter and regulator toincrease stability of the control circuit towardsexternal clutter and noises. The latter isextremely important taking into considerationrather high level of electromagnetic noise,generated by the power keys. Automatic identification of the circuit

characteristics is based on determinationof such parameters as ohmic resistance(R) of circuit, inductivity (L), effectivecapacitance of energy storage units (C) byproviding the series of test pulses andaftertreatment of results for the systemself-adjustment according to the presentcircuit characteristics. Besides, diagnosticsof the capacitors batteries condition isperformed to exclude their ageing effectand further system unbalance. The above-mentioned opportunities

provide deviation of actual current fromthe desired in dynamics less than 2-3%and repeatability of measurement resultsat the level of 0,6 - 0,9% (Figures 3, 4).

Treatment of digital measurementresults Main aim of the performed measurementsis to receive more adequate VTM (FM)results of bipolar power semiconductors atthe desired value of current and pulseform. Actual value of current fluctuatesaround target value during themeasurements due to the pulse methodof current regulation that provides a certainscatter into the measured values whichappears to be the noised externaldisturbances. In order to get a more exactvoltage value it’s necessary to make anadditional mathematical treatment of themeasurement results. The known equationapproximates the type of voltage-currentcharacteristics of the power unit [1]:

While small deviations from themeasurement point the formula can be

Figure 3: Sinusoidal pulse 3.3 kA max

Figure 4: Deviation between measured and desired results

Figure 5: Trapezoidal pulse duration of 2000 microseconds


www.proton-electrotex.com POWER SEMICONDUCTORS 35


successfully linearized. Method of thesmallest squires can be further used to thereceived sum of measurement results inorder to obtain the most reliable estimateof the measured parameter. Measurementis performed not less than in 10-20 pointsof the neighborhood of the desired currentvalue for the successful realization of thedescribed method.

Measurement at different forms oftest pulsesVTM value may vary depending on the currentpulse form due to the different heating-up ofa semiconductor element as well as due todynamic processes in the device. Main taskis to provide equal complete opening of apower semiconductor for the runningcurrent, together with the minimization of itscharacteristics variation due to the heating bythe running current. That’s why it may bereasonable to use different test pulsesforms: from sine-shaped to S-shaped(Figures 6 and 7).

ConclusionMeasurement of power semiconductorconducting losses allows for assembling ofparallel stacks that have advancedoperational reliability and more symmetricload of each device. For large diameterelements that have rather large turn-ontime, the mentioned equipment allows toselect optimal test mode that will minimizeself-heating of an element together withprovision of complete turn-on of a die intothe conducting state across the surface.The described measuring unit is a part of

the complex for semiconductor staticcharacteristics measurements. The complexensures testing of thyristor and diodeelements produced by JSC “Proton-Electrotex”. Further developments will targetat test equipment for large currents (20-30kA) and test equipment for IGBT modules.

Literature[1] Thyristors. Information material of

ABB Semiconductors AG. - ABBSemiconductors AG, 1999

Figure 6: Different pulse forms for VTM measurement at ITM 4 kA

Figure 7: Scatter of measurement results for one device at ITM = 1570A, 25°C. Average value ofselection is M = 1235.53 mV, � = 4.74 mV, that makes less than 0.7 % of the measured value

To receive your own copy of

Power Electronics Europe

subscribe today at:

www.power-mag.com


Your Gateway to…

the South AmericanPower Electronics Market.

International Conference and Exhibition for Power Electronics, Intelligent Motion,Renewable Energy and Energy Management

www.mesago.de/en/pesa/home.htm

Power Electronics South America – sponsored by PCIM

São Paulo, 11 – 13 September 2012

36_PEE_0512_Layout 20/08/2012 11:40 Page 1

WEBSITE LOCATOR 37


AC/DC Connverters

www.irf.comInternational Rectifier Co. (GB) LtdTel: +44 (0)1737 227200

Diodes

Discrete Semiconductors

Drivers ICS

www.microsemi.comMicrosemiTel: 001 541 382 8028

Fuses

GTO/Triacs

Hall Current Sensors

IGBTs

DC/DC Connverters

DC/DC Connverters


www.power.ti.comTexas InstrumentsTel: +44 (0)1604 663399


www.neutronltd.co.ukNeutron LtdTel: +44 (0)1460 242200

Harmonic Filters

www.murata-europe.comMurata Electronics (UK) LtdTel: +44 (0)1252 811666

Direct Bonded Copper (DPC Substrates)

www.curamik.co.ukcuramik� electronics GmbHTel: +49 9645 9222 0


www.mark5.comMark 5 LtdTel: +44 (0)2392 618616


www.dgseals.comdgseals.comTel: 001 972 931 8463



www.digikey.com/europeDigi-KeyTel: +31 (0)53 484 9584


www.mark5.comMark 5 LtdTel: +44 (0)2392 618616 www.irf.com

International Rectifier Co. (GB) LtdTel: +44 (0)1737 227200

www.protocol-power.comProtocol Power ProductsTel: +44 (0)1582 477737

Busbars

www.auxel.comAuxel FTGTel: +44 (0)7714 699967

Capacitors

Certification

www.powersemiconductors.co.ukPower Semiconductors LtdTel: +44 (0)1727 811110

www.powersemiconductors.co.ukPower Semiconductors LtdTel: +44 (0)1727 811110

Connectors & Terminal Blocks


www.productapprovals.co.ukProduct Approvals LtdTel: +44 (0)1588 620192

www.proton-electrotex.com/ Proton-Electrotex JSC/Tel: +7 4862 440642;

www.voltagemultipliers.com Voltage Multipliers, Inc.Tel: 001 559 651 1402




Website Locator_p41-42 Website Locator 20/08/2012 12:45

38 WEBSITE LOCATOR


Power Modules

Power Protection Products

Power Semiconductors

Power Substrates

Resistors & Potentiometers

RF & Microwave TestEquipment.

Simulation Software

Thyristors

Smartpower Devices

Voltage References

Power ICs

Switches & Relays

Switched Mode PowerSupplies

Switched Mode PowerSupplies

Thermal Management &Heatsinks


www.rubadue.comRubadue Wire Co., Inc.Tel. 001 970-351-6100











www.ar-europe.ieAR EuropeTel: 353-61-504300



www.universal-science.comUniversal Science LtdTel: +44 (0)1908 222211

www.power.ti.comTexas InstrummentsTel: +44 (0)1604 663399


www.dau-at.comDau GmbH & Co KGTel: +43 3143 23510

www.lairdtech.comLaird Technologies LtdTel: 00 44 1342 315044

www.power.ti.comTexas InstrummentsTel: +44 (0)1604 663399



www.isabellenhuette.deIsabellenhütte Heusler GmbH KGTel: +49/(27 71) 9 34 2 82

ADVERTISERS INDEX

ABB 15

Advanced Power Electronics 22

Coilcraft 4

Cornell Dubillier 21

CT Concepts 27

Danfoss 10

DFA Media 22/26

Electronica 2012 17

Fuji IBC

HKR 19

Indium Corporation 7

International Rectifier OBC

Isabellenhutte 12

Ixys 11

LEM 9

Power Electronics South America 36

Powerex 8

Proton-Electrolux IFC

Semikron 6

Mosfets

Magnetic Materials/Products

Optoelectronic Devices


Packaging & Packaging Materials






www.biaspower.comBias Power, LLCTel: 001 847 215 2427

www.digikey.com/europeDigi-Key Tel: +31 (0)53 484 9584


www.digikey.com/europeDigi-Key Tel: +31 (0)53 484 9584






www.abl-heatsinks.co.ukABL Components LtdTel: +44 (0) 121 789 8686

Website Locator_p41-42 Website Locator 20/08/2012 12:48

39_PEE_0512_Layout 20/08/2012 11:39 Page 1

Ultra-fast Switching, Rugged 600V High Frequency IGBTs

Automotive COOLiRIGBT™ Gen 1

COOLiRIGBT™

35

30

25

20

15

10

5

01 kHz 10 kHz 100 kHz 1000 kHz

Switching Frequency (kHz)

Max

Inpu

t Cur

rent

(Arm

s)

CooliRIGBT™

Gen 1Tj = 175°C

Super JunctionMOSFET

Tj = 150°C

50% Higher currentat 100kHz

Tj max = 175°C

Tj Max

Manufacturability

Switching Frequency

Losses At High Currents

150˚C

Super Junction MOSFET

Complex

High

High

175˚C

COOLiRIGBT™

Gen1

Simple

High

Low

Super Junction Super Junction Super Junction Super Junction Super Junction Super Junction MOSFETMOSFETMOSFETMOSFETMOSFETMOSFET

COOLCOOLCOOLCOOLCOOLiRiRIGBIGBIGBTTT™™

Gen1Gen1Gen1

CooliRIGBT™ offers 50% higher current than super junction MOSFETs COOLiRIGBT™ Gen 1 are designed to be used in a wide range of fast switching applications for electric (EV) and hybrid electric vehicles (HEV) including on-board DC-DC converters, and battery chargers.

Features:

THE POWER MANAGEMENT LEADER For more information call +49 (0) 6102 884 311

or visit us at www.irf.com

40_PEE_0512_Layout 20/08/2012 11:42 Page 1

issue 5 – august 2012 power ... · of hpt-igct issue 5 – august 2012 also inside this issue...

Documents