psde mag cover for pdf (page 2) · chris rexer fairchild semiconductor andreas sperner ixys member...

www.powersystemsdesign.com 1

CONTENTS

ViewpointThe Merry-Go-Round, Bodo Arlt Editor-in-Chief, Power Systems Design Europe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

Industry News

eupec increases its technological leadership . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

TTI distributes WIMA capacitors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4

New web-site at Vicor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4

Intersil to join NASDAQ -100 Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

Power Electronics Design Course announced . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

Breakthrough in high-k/metal gate stacks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

Cover Story

Buck Converter Eliminates Capacitive Droppers, Stefan Baurle and Kent Wong, Power Integrations . . . . . . . . . . . . . . . . .12

Features

Technology Review—MOSFETs, IGBTs and Rectifiers, Bodo Arlt Editor-in-Chief PSD Europe . . . . . . . . . . . . . . . . . . . . . . . 8

Power Player—Energy Efficiency: The Never-Ending Quest, Eric Lidow, Chairman and Founder,

International Rectifier Corp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .10

Thermal Management—Ever Seen a Unicorn?, John Broadbent, Thermacore . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .17

Power Semiconductors—FRFET increases reliability of Phase - Shifted ZVS Full Bridge PWM Converter,

Sampat Shekhawat, Praveen Shenoy, and Bob Brockway, Fairchild Semiconductor . . . . . . . . . . . . . . . . . . . . . . . . . . . . .22

DC/DC Conversion—The ACPI advantage and system design challenge, Geroge Lakkas and

Bogdan Duduman, Intersil . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .26

DC/DC Conversion—Power Systems architectures, Robert Marchetti, Vicor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .41

Power Protection—Protection of subscriber line interface cards, Chris Likely, Cooper Electronic Technologies . . . . . . . . .32

Optoelectronics—The cooling of high-power LEDs, Mark de Vos, Bergquist . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .34

Power Management—Power Management is now programmable, Johannes Fottner, Lattice Semiconductor . . . . . . . . . .38

Design and Simulation—Acoustic evaluation of heat sink attachment, Tom Adams, consultant, Sonoscan Inc. . . . . . . . . .44

New Products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .47,48

Member Representing

Eric Carroll ABB SwitzerlandBob Christy Allegro MicroSystemsPaul Löser Analog DevicesArnold Alderman AnagenesisDr. Uwe Knorr Ansoft Todd Hendrix Artesyn TechnologiesHeinz Rüdger CT-Concept TechnologyDr. Reinhold Bayerer eupecChris Rexer Fairchild SemiconductorAndreas Sperner IXYS

Member Representing

Dr. Leo Lorenz Infineon TechnologiesEric Lidow International RectifierDean Hendersen Intersil David Bell Linear TechnologyRüdiger Bürkel LEM ComponentsPaul Greenland National SemiconductorDr. Martin Schlecht SynQorDr. Thomas Stockmeier SemikronUwe Mengelkamp Texas InstrumentsPeter Sontheimer Tyco Electronics

Power Systems Design Europe Steering Committee Members

Power Systems Design Europe February 20042

Welcome and thank you to the industry forthe overwhelming support and encouragementthat you have given me the opportunity to con-tinue to provide you with a focused technicalpower magazine in Europe.

Having the first issue of Power SystemsDesign Europe out in the engineer’s handsmakes us proud. Julia, Jim and I have turnedaround everything within a short time. We rec-ognize that there remains a strong interest with-in the industry to provide a focused publica-tion dedicated to the power sector. As before,I will keep focused on your needs as powersystems designers.

Within the pages of Power Systems DesignEurope you will find I have added a few extrafeatures, one titled “Power Player”. Each issuewill introduce you to one of the members ofour editorial steering committee who will providea commentary on the industry. This issue’s“Power Player” is a good friend of mine andwell-respected colleague, Mr. Eric Lidow,Chairman and Founder of International Rectifierand one of the industry’s foremost pioneers inpower electronics. He provides you with hisinsight and analysis on the never ending questfor energy efficiency.

The March PCIM China Conference andExhibition in Shanghai as well as the Europeanhighlight with the May PCIM Conference andExhibition in Nuremberg will provide continueddirection for our industry. Mark your calendar forthe last week of May to attend the most important European event in power electronics,PCIM Europe—Nuremberg.

My continued association of more than oneand a half decades as a board member of theNuremberg PCIM conference provides additional fuel for the magazine and keeps the

focus strongly set for power technology andtheir advanced applications.

I will be handling the podium discussions atPCIM 2004 in Nuremberg which is open to visitors and is conveniently located on the exhi-bition hall floor. Each day around lunch timewe will have short presentations from expertsfollowed by an open discussion. This year Ichose semiconductors as the focal elements forpower electronics. MOSFETs on Tuesday,IGBTs on Wednesday and Rectifiers onThursday will be the topics.

I have spent more than a quarter of a centuryin power semiconductors and have seen newswitch technologies emerge, being improvedover time and now approaching ideal behaviorin systems.

Power management in systems has nowbecome a major consideration in productdesign. No matter the application, when man-agement of power in design is a consideration,I will cover it.

My time as Editor-in-Chief of the formerPCIM Europe magazine has inspired me to continue with an even stronger spirit. I amlooking forward to continuing to provide youwith the latest information on new power tech-nology, applications and product offerings tostimulate your future system designs.

This is a new year with a new beginning andI feel good that my promises and commitmentsto you are coming true.

I still wear my old New York Yankees baseball cap, a little continuity is one thing we need.

If not before, I will see you at the PCIM showin Nuremberg!

Best regards

Bodo [email protected]

The Merry-Go-Round

VIEWPOINT

AGS Media Group

Katzbek 17a

D-24235 Laboe, Germany

Phone: +49 4343 421790

Fax: +49 4343 421789

[email protected]

www.powersystemsdesign.com

Editor-in-Chief

Bodo Arlt, Dipl.-Ing.

[email protected]

Publisher

Jim Graham

[email protected]

Advertising & Marketing Director

Julia Stocks

[email protected]

Creative Direction & Production

Eyemotive

[email protected]

www.eyemotive.com

Circulation Management

[email protected]

Advertising Rates, Sizes and Digital File

Requirements available at:

www.powersystemsdesign.com

Subscription Request available at:

www.powersystemsdesign.com/psd/

subslogn.htm

Registration of copyright: January 2004

ISSN number: in progress

Issue print run: 20,000

Printing by: Central - Druck Trost GmbH & Co.

Heusenstamm, Germany

AGS Media Group and Power Systems Design Europe

magazine assume and hereby disclaim any liability to

any person for any loss or damage by errors or omis-

sions in the material contained herein regardless of

whether such errors result from negligence, accident

or any other cause whatsoever.

Send address changes to:

[email protected]

Volume 1, Issue 1

Power Systems Design Europe February 2004 www.powersystemsdesign.com4 5

INDUSTRY NEWS

eupec in the future will have a separatedepartment called Technology & Innovation(TI) for the strategic development of innova-tive technologies.

On 1st November, 2003 the companyformed a special team headed by Dr.Reinhold Bayerer (50) for guaranteeingeupec’s technological leadership.

“We intend to essentially strengthen ouractivities with regard to technological andproduct innovations over the next 5-10 years by this means”, explains board mem-ber Erich Wallner.

eupec since its foundation by the compa-nies of Siemens and AEG (1990) continuous-ly has been improving its word market posi-

tion in the high power semiconductor segmentand meanwhile is the 2nd biggest player afterMitsubishi. This was possible only by a highinnovation rate. While 1990 roughly 22% ofall products were younger than 5 years,today 70% of the company’s turn-over areachieved by new products.

eupec increases its technological leadership

New Web-Site at VicorVicor has launched its new Web-site. The

site makes use of cutting-edge data-miningtools to ensure “minimum-click” navigationand easy search and selection. This content-rich, product-centric site will help design engi-neers to find the products and related infor-mation they need, quickly and efficiently.Easy access is also provided to Vicor’s inter-active design and configuration tools.

The site’s new search engine searchescomprehensively across web pages, docu-ments and media, providing results that areeasy to sort due to intelligent and detailedcategorization.

The technical library, the product selector,and two innovative on-line design tools

enable design engineers to learn aboutpower conversion issues and to select orconfigure—in real time—products specificallysuited to their applications. VCAD (VicorComputer Assisted Design) facilitates on-linedesign of configurable VIPAC and VIPACArray AC and DC power systems. Designscan be saved in the user’s account for futurereference and convenience.

VDAC (Vicor Design Assistance Computer)enables design engineers to design DC-DCconverters to fit their specific requirements.They can specify on-line and verify in realtime the attributes and performance of custom DC-DC converter modules.

The Web site features comprehensive infor-mation on Vicor’s Factorized Power Architecturefeaturing the new V•I Chips, and an enhancedfocus on product support and easy access toapplications and technical support.

TTI distributes WIMA capacitors TTI Electronics has

signed a franchise dis-tribution agreementwith European plasticfilm and metallisedpaper capacitor manu-facturer, WIMA.

Comments GeoffBreed, TTI’s European

Marketing Manager: “Continual technicalprogress and expansion have placed WIMAamongst the top names in film capacitors

worldwide. A high degree of automationensures that WIMA can continue to manufac-ture in Europe while successfully meeting thechallenges and competition of a global market.”

WIMA has taken an environmentally-friend-ly attitude – priority is given to avoiding theuse of material containing harmful sub-stances, keeping an eye on the selection ofrecyclable substances and the reduction ofpacking materials.

Important new product lines include:polypropylene GTO MKP devices, designed

to attenuate voltage spikes on GTO thyristorsand IGBTs; and FKP and MKP snubbercapacitors, developed to meet the demandsof high-power converter technology andavailable in many connecting configurations.

Concludes Breed: “WIMA is an importantaddition to our portfolio of leading passivenames, and provides our customers witheven more choice.”

[email protected]

Intersil To Join NASDAQ-100 Index Intersil is being added to the NASDAQ-

100 Index during the 2003 rebalancing thattakes effect December 22, 2003 upon marketopen, according to a NASDAQ-issued pressannouncement on the annual re-ranking.

“We’re very pleased to see IntersilCorporation join the ranks of the top 100NASDAQ-listed companies, which includeleading technology companies such as

Cisco, Dell, Intel, and Microsoft,” said RichBeyer, president and CEO of Intersil. “This isa testament to the great team of dedicatedmen and women working at Intersil officesaround the globe. This will certainly bringadded awareness of Intersil’s outstandingperformance and strat-egy for growth to alarger number of investing institutions andindividuals and will help us in achieving our

goal of becoming a top-tier high performanceanalog semiconductor company.”

Once a year, NASDAQ adjusts the index’smakeup so it consists of the 100 largest non-fi-nancial companies listed. The Nasdaq-100 Index rebalance is based on market cap-

italiza-tion, and is the basis for the Nasdaq-100Tracking Stock—the most actively traded listedU.S. equity security.

www.intersil.com

www.vicorpower.com or www.vicoreurope.com

www.eupec.com


Power Electronics Design CourseSince 1981, e/j BLOOM has organized

and sponsored more educational courses inpower conversion design subjects than anyother company in the world! Now, as a part ofits 23rd anniversary celebration activities, theEducational Division of the company will offerthree presentations of a new two-segmentModern Power Conversion DesignTechniques course in 2004 in the UnitedStates and the United Kingdom.

The first segment, 2 days in length, willdeal with the practical design and develop-ment issues of state-of-the-art power convert-er circuits, including introductory sections onconverter circuit models for stability controland on design approaches for high frequencytransformers and power inductors. Planarand integrated power magnetics are also topics of this part of the course. This segmentwill be taught by Ed Bloom, President of e/jBLOOM associates Inc. He is a design spe-

cialist and a well-known consultant in the fieldof power converter & magnetics design, andhas published many articles on his designmethods. He is the co-author of the populartextbook, Modern DC-DC Switchmode PowerConverter Circuits, now in its ninth printing.Ed is a Senior Member of the IEEE PowerElectronics Society, and has 41 years ofdesign experience in the fields of power electronics.

The second segment of the 2004 course,also 2 days in length, focuses on the properdesign of switch-mode power converter cir-cuits and systems for low Electro-MagneticInterference (EMI) through the use of suitabletopologies, component selection, power mag-netics design and physical construction of thepower conversion system, including printed-circuit board layout techniques. The instructorfor this segment will be Bruce Carsten. Bruceis well known in industry circles for his signifi-

cant contributions to the practical design ofmodern power conversion circuits and sys-tems, and has over 34 years of design expe-rience. Bruce is a Member of the IEEE PowerElectronics Society and has presented morethan 115 design-oriented seminars worldwideon modern power electronics circuit & systemdesign issues.

Dates, cities and country locations plannedfor the three 2004 course presentations are:• May 18-21 in Chicago, Illinois, USA. August• 17-20 in Las Vegas, Nevada, USA.• October 12-15 in Portsmouth, England,

United Kingdom.Persons interested in attending the course,

or any segments thereof, can obtain a FREEbrochure containing all registration, cost andcourse details by simply contacting:

IMEC announces that they have success-fully demonstrated the use of high-kdielectrics and metal gates to values belowone nanometer. Achieving this level of electri-cal performance using materials other thanthe traditional polysilicon-based counterpartsremoves one of the industry’s so-called ‘redbrick wall’ barriers to advancing semiconduc-tor technology.

As part of its industrial affiliation program,an IMEC research team utilized metal gatesto overcome the problems imposed by theinteraction between high-k materials with thecom-monly used polysilicon electrode.

Using TiN or TaN gates and HfO2 asdielectric, aggressive scaling down to an0.8nm equivalent-oxide thickness (EOT) wasdemonstrated in both nMOS (8.2 AngstromsEOT) and pMOS (7.5 Angstroms EOT) tran-sistors. The metal-gated devices outper-formed their polysilicon-based counterparts in terms of electrical performance parame-

ters, including high conductance, low leakageand reduced threshold-voltage instabilities.Besides the elimination of gate depletion, themetal gates enhanced high-k scalability andsignificantly reduced gate-leakage by up tothree orders of magnitude. Transistor drivecurrent also improved signifi-cantly.

To realize sub-1nm EOT scaling, appropri-ate interfacial oxide control both prior to andduring high-k deposition was applied. Thelowest EOT values were typically obtainedusing a ‘minimal interface approach’ (i.e. min-imal EOT contribution). In order to achievethis, scaled chemical oxide interfaces withcontrolled thickness and precisely controlleddeposition and annealing conditions wereused. HfO2 was deposited by atomic layerchemical vapor deposition (ALCVD).

Excellent nMOS performance was demon-strated with performance indicators slightlyex-ceeding those for poly/SiO2-baseddevices and significant improvement of the

pMOS per-formance since the hole mobilityremained constant down to the lowest EOTvalues. In addi-tion, not only the initial per-formance but also the threshold-voltageinstabilities due to trap-ping were stronglyreduced as compared to typical results onhigh-k films with polysilicon electrodes.

Part of this research has been done in col-laboration with IMEC’s high-k industrial affilia-tion program partners, namely InternationalSematech and its member companies,Renesas, Matsushita and Samsung.

Figure-of-merit for both nFET (left) andpFET performance (right) representing gateleakage as a function of performance– thesymbols represent performance trend linesfor poly/SiO2, poly/high-k and metal/high-kgate stacks.

http://www.ejbloom.com/downld.asp

Breakthrough in High-k/metal Gate Stacks

www.imec.be

INDUSTRY NEWS

Power Events

• PCIM China, March 17-19, Shangai, China

• Electronica USA, Power Electronics Conf.

San Francisco March 29th to April 1s

• Hanover Industrial Fair, April 19-24, Hanover,

Germany

• Semicon Europe, April 20-22, Münich,

Germany

• PCIM 2004, May 25-27, Nüremberg, Germany

• Fisita 2004, May 23-27, Barcelona, Spain

, LTC, LT and Burst Mode are registered trade-marks and ThinSOT & SwitcherCAD are trademarksof Linear Technology Corporation,1630 McCarthy Blvd., Milpitas, CA 95035Cellular Phone is for illustration purposes only and isnot intended to represent any actual product.

Teardown This!

Linear Technology ICs enable the creation of high-performance handheld devices. Switching frequencies up to 4MHzminimize the size of inductors and capacitors, while low standby current Burst Mode® technology maximizes batterylife. New products are added monthly to our broad handheld product portfolio – download the latest selection guide atwww.linear.com/ads/handheld.

Features▼

Tiny & Efficient Power Solutions for Handheld Products

• Switching Frequencies up to 4MHz

• High Efficiency Operation: Up to 96%

• Quiescent Currents as low as 10µA

• Very Fast Transient Response

• EMI Control Techniques ReduceInterference

• Minimal External Components

• Small, ≤ 1mm Profile Packages

Download Thiswww.linear.com/ads/handheld

More Information▼

Brochure▼

www.linear.com/lineardirectOnline Store▼

Visit: www.linear.com

Info: +1-408-432-1900

Fax: +1-408-434-0507


TECHNOLOGY REVIEW

MOSFETs and IGBTs haveachieved improvement in siliconresulting in minimized switching

and conduction losses. Low voltageMOSFET applications are benefitingfrom trench-gate technology, which hasimproved conduction characteristics.While the V MOSFET structure hasbeen present for a long time, this tech-nology has provided improved deviceperformance by significantly increasingcell density per unit area. Other parame-ters must be considered in a MOSFETswitch for a design to be successful.Avalanche capability in inductive switch-ing is an important parameter for keep-ing the device within the safe operatingarea. Furthermore, active clamping isused to work successfully on automotivedevices.

Trench design is replacing the tradi-tional V MOSFETs designs by offeringmore active area for the current flow per device size.

DC/DC converters receive the maincontribution from MOSFETs, which incombination with optimized passivecomponents demonstrates an over 90percent efficiency. Synchronous rectifi-cation with suitable MOSFETs and nec-essary control achieves the critical designs.Distributed power architecture usesDC/DC conversion for point of load supply.

Less than 200 volts, the MOSFETcontinues as the dominant switchingelement particularly in automobiles. The42 volt supply and the sophisticated ele-ments of future automobiles, like starteralternator and drive by wire applications,will inspire more tailored devices for thisvolume market. All manufacturers havea 75volt MOSFET product and futuredevelopment will continue to support theautomotive industry at the required volt-age level. Packaging is important for thismarket segment, as this methodology isa way to compress functions requiringless weight and space in modules.

Above 100 volts IGBTs are the switchof choice. Improvement in IGBTs hasbeen achieved by silicon thinningprocesses, which as a result lead to bet-ter thermal management, and reducingconduction losses. Actual IGBT technol-ogy with trench-gate components isbringing significant advantages thanksto the improved conduction characteris-tic at line voltages. Most IGBTs are atline voltages resulting in preferredbreakdown voltages of 600 and 1200volt. IGBT technology is even replacinghistorical GTO sockets by reaching volt-ages up to 6 kilovolt range.

All of these achievements will guideus to more advanced solutions in thepower quality arena of power distribu-tion. In the electric drive technology sec-tor, the most important product categoryis variable speed drives. Modern invert-er technology at line-voltage level is builton IGBTs. It is the preferred switch

because of short circuit capability, speed and improved saturation voltage.

There are two IGBT concepts thePunch Through (PT) and the Non Punch Through (NPT) both have beenimproved over the years. Vertically byinnovations of the buffer structure or lifetime killing process on the PT side.Reducing the wafer thickness in theNPT side, also the transistor cell struc-tures are changed to minimized planarcells or trench cells. A really outstandingbreakthrough in the IGBT developmenthas been achieved with the Field StopIGBT. This is step towards the next IGBTgeneration in terms of low static anddynamic losses. The resulting improveddice allow the packaging engineer to fur-ther improve his design. Most manufac-turers develop a strategy based on endmarket. Modules of all sizes are readilyavailable. In modules for power applica-tions, the elimination of wire bonds isbecoming a reality by using a solderprocess or pressure contact. All modulemanufactures are moving to more dense packages.

Perfect diodes with soft recovery characteristics compliment the IGBTperformance. All major semiconductormanufacturers are working to optimizerectifiers.

AC to DC rectification is one of themost popular duties for diodes. At 50 or 60 Hertz, there are no really criticalswitching losses. Bridge rectifiers areavailable in a variety of packages and configurations in Voltage andCurrent ranges that serve all industryrequirements.

Schottky diodes will still be used inmany low voltage applications as aresult of their extremely low forwardconduction drop.

Inverters at line voltages are usingIGBTs along with an ultra fast rectifier.As compared to the use of a MOSFETwith an inherently slow intrinsic diode,the IGBT diode combination is a moreefficient solution. As a result, in threephase designs at the required voltage,

the independent IGBT diode is the favoredapproach as seen in the marketplace.

A rectifier manufacturer who wants topropose an optimized solution for eachfunction, needs to develop several fami-lies with different trade-offs, mainlybetween forward voltage VF andreverse recovery charges Qrr (orreverse recovery time trr). Guidelineshave been suggested to optimize thecircuit’s limited EMC behavior and theuse of the power diodes, often incorpo-rated in modules containing the com-plete power section of a frequency con-verter for electric drives. Application ori-entated testing generates the resultsthat indicate that at a higher switching,the electromagnetic compatibility is con-firmed as in some cases it is not correctto push switching speed up and thussuffer from electromagnetic interference(EMI). Therefore, the fast and ultrafastdiodes have to have a soft recoverycharacteristic that minimizes the EMI.

The name of the game in producingan efficient diode is in the dopingprocess and each supplier has a propri-etary recipe. The resulting performanceis what matters. Most have shownexcellent results as evidenced by thedata published.

Summarizing we can say that lossreduction and space minimization is akey goal achieved by MOSFETs offeringextreme low on resistance thus replac-ing Schottky diodes while also being themain switch working at the high switch-ing frequency to keep magnetics smallin the converter design. While IGBTsdominate the motion arena with switch-ing frequencies above the human ear forcomfort. Ultrafast IGBT developmentsare targeting Switch mode power supplydesign in the AC to DC applications.

In conclusion all semiconductor manu-factures are moving to more densepackages. Novel chip packs appear toprovide an alternative in the packagingdevelopment. The rectifier and diodetechnology and design are not a design-er’s first choice of pursuit but are asneeded as the salt in a good recipe.

The independent module and hybridmanufactures have the choice of usingany optimized diode that best matchtheir application from a wide variety ofdifferent performances of independentproducers.

www.toshiba-europe.com/electronics

www.fujielectric.de

www.hitachi.com

www.irf.com

www.infineon.com

www.fairchildsemi.com/power

www.siliconix.com

www.fujielectric.de

www.philips.com

www.ixys.net

www.advancedpower.com

www.zetex.com

www.mitsubishichips.com

www.sanrex.com

www.abbsem.com

www.powersem.de

www.eupec.com

www.semikron.com

www.dynexsemi.com

www.westcode.com

www.danfoss.com

www.em.tycoelectronics.com

www.stm.com

The Semiconductor Elements to

Start the DesignPower Semiconductors chip performance continue to make improvements. Mounting and packaging

technology create more efficient devices. The benefits are influencing all electronics, not only evident in the classical areas like power supply and drive application.

By Bodo Arlt, PSD EuropeEditor-in-Chief

TECHNOLOGY REVIEW


POWER PLAYER

As the world’s population grows and our use of electronics expands, neverhas the need to save electricity been asgreat as it is today. For the makers ofthe electric devices that have becomepart and parcel of modern-day life,power management technology hastaken on an increasingly important roleto achieve key energy efficiency mile-stones. For those of us who pioneeredthe power conversion industry, it’s avery exciting time along a journeyunderpinned by a legacy of technologybreakthroughs spanning half a century.

The quest for power efficiency hasbeen the mission of InternationalRectifier since 1947. In the early years,our quest focused on process materialsand energy sources. Soon after openingits doors, International Rectifier becamethe dominant supplier for selenium recti-fiers then moving to germanium andfinally silicon.

The early years also brought the studyof alternate energy sources that still holdpromise today. One was photoelectricpower. In the late 50’s, InternationalRectifier demonstrated its potential byconverting a 1913 Baker electric car thatran purely on solar energy, the first fullysolar-powered automobile in the world.Today, that car is on display at ourTemecula California facility, a testamentto the futuristic thinking still common inour industry today. International Rectifierwent on to commercialize the use ofphotoelectric cells in space, using it topower the first RCA weather satellite.

The late 70’s heralded one of theindustry’s most fundamental technologyshifts, the birth of the first field effectpower transistor (FET) and IR develop-ments that spawned a $4 billion dollarindustry. The new FETs, manufacturedusing vertical DMOS process, deliveredvery high gain and could switch full loadat much greater speeds than the promi-nent bipolar transistor technology foundin most power supplies of the time.Shortly thereafter, IR unveiled theHEXFET power MOSFET whose com-pact six-sided cell structure brought sig-nificant advancements in size and per-formance. These devices now representmore than twice the number of bipolartransistors sold each year.

By the ‘90’s, it was apparent that inte-grated high-voltage technology wasneeded to drive MOSFETs in manyapplications. International Rectifier ledwith the introduction of a driver capableof operating at 1200V, an IC still uniquein the industry.

As it was in the early years, our devel-opments to drive down power consump-tion have revolved around applicationswhere we can make the largest impact.One of our greatest opportunities todaylies in motion control where we caninfluence the adoption of variable speedmotion to dramatically improve the ener-gy efficiency of household appliancesand manufacturing equipment. Similarly,we can apply the same electronicmotion control technology to automotivesystems for electronic braking, electricsteering and integrated starter alternatorsto vastly improve the mileage, safety

and environmental impact of future cars.IR’s pioneering efforts in the develop-ment of electronic light ballasts alsooffers meaningful ways to conserveelectricity on a global scale.

Looking ahead, we will concentrate ontechnological barriers still hinderingpower management advancement incritical, high-impact applications. Tolessen the interconnect losses that havenow surpassed the silicon losses, weare co-developing and co-packaging anarray of power management functionsinto integrated building blocks usingstate-of-the-art packaging technologies.More importantly, we’re tailoring thesebuilding blocks toward specific applica-tions where power management tech-nology can enable the next big leaps inour collective step into the future. As apioneer in the industry, InternationalRectifier remains committed to thenever-ending quest towards greaterenergy efficiency and leading the wayto technology developments that willmake it happen.

by Eric Lidow, Chairman and Founder, International Rectifier Corp.


COVER STORY

Switching Power Conversion IC Displaces Non-Isolated Passive Power Supplies

For the first time, a new power conversion IC family offers the superior performance of switchingpower solutions plus extreme energy efficiency for a growing number of very low power

applications, at a system cost equal to traditional passive solutions.

Stefan Bäurle and Kent Wong, Power Integrations

Many applications such as appli-ance controls, LED drivers,electricity meters and residential

heating controllers usually require powersupplies without galvanic isolation andwith low output currents in the range of20 mA to 100 mA. Passive solutionssuch as capacitive dropper or line cycletransformer power supplies have tradi-tionally served those needs because oftheir simplicity and low cost. Lately,however, added features to appliancessuch as electronic displays requireincreased power capabilities that cannotbe fulfilled cost effectively with capacitivedropper power supplies. Additionally, con-sumers today demand energy efficiencyin their products. Passive power sup-plies are therefore no longer an option.

The new highly integrated off-lineswitcher IC series, LinkSwitch TN, isspecifically designed to address theincreased load current demands andenergy efficiency requirement for non-isolated power supplies. This flexiblenew family supports a variety of differenttopologies such as buck, boost, buck-boost, SEPIC and flyback. Unlike

passive solutions, it enables universalinput voltage range power supplies andprovides system level protection in caseof faults such as thermal overload, outputshort-circuit and open control loop. Powersupplies based on this device typicallyconsume less than 0.2 W during no-loadat 265 VAC input voltage. Additionally,they can be produced entirely using sur-face mount components.

The LinkSwitch TN family consists ofthree family members. The maximumdeliverable output current for each devicein a non-isolated buck converter is out-lined in Table 1. For example, the small-est family member, LNK304, is able todeliver up to 120 mA running in mostlydiscontinuous conduction mode(MDCM). In continuous conduction

mode (CCM) the LNK304 can deliveroutput currents up to 170 mA. All devicesin the family are available either in a DIP8 or SMD 8 package that offers a mini-mum creepage of 1.98 mm between the high voltage pin and adjacent lowvoltage pins.

IC DescriptionLinkSwitch-TN devices integrate a

700 V power MOSFET with a low volt-age controller that includes simpleON/OFF control, an oscillator, frequencyjittering, a fully-integrated auto-restartcircuit, a high voltage switched currentsource, cycle-by-cycle current limit, andthermal shutdown circuitry. Figure 1illustrates the internal block diagram ofthe LinkSwitch-TN. The simple ON/OFFcontrol scheme enables the powerMOSFET to switch when a current ofless than 49 A is delivered into theFEEDBACK (FB) pin, representing avoltage of less than 1.65 V. The typicaloscillator frequency is set to 66 kHz witha frequency jitter of 4 kHz peak-to-peakto minimize EMI emissions. In order tooptimize EMI reduction for both averageand quasi-peak emissions, the modula-

Figure 2. Universal input 12 V, 120 mA off-line buck converter with LinkSwitch-TNTable 1. LinkSwitch-TN Output Currents

tion rate of the frequency jitter is set to 1 kHz. In the event of a fault conditionsuch as output overload or into auto-restart mode, which disables the powerMOSFET for 800 ms if no feedback fromthe output is received for 50 ms.Becausef the extremely low power con-sumption of the internal control circuitryof LinkSwitch TN, the chip can be pow-ered through the integrated high voltageMOSFET. The high voltage switched mode current source charges a capaci-tor connected to the BYPASS (BP) pinto 5.8 V directly from the drain whenever

the drain is off. A bypass capacitor of 0.1F is sufficient for energy storage and for high frequency decoupling. TheBYPASS pin under-voltage circuitryenables (turns on) the power MOSFETafter the BYPASS pin reaches 5.8 V atstart-up. If the BYPASS pin voltage islowered below 4.85 V, the power MOS-FET will be disabled; it must then beraised back to 5.8 V to re-enable thepower MOSFET. A 6.3 V shunt regulatorclamps the BYPASS pin when current isprovided through an external resistor, tofurther decrease no-load consumption.

The current limit circuitry senses thedrain current of the power MOSFET andterminates the switching cycle when thedrain current exceeds an internal thresh-old. The current limit circuitry also incor-porates a leading edge blanking timecircuit, which inhibits the current limitcomparator for a short time after thepower MOSFET is turned on so that cur-rent spikes caused by rectifier reverserecovery time and primary-side capaci-tance will not prematurely terminate thecycle. The over-temperature protectioncircuit senses the die temperature and istriggered when the temperature exceeds142 C. his disables all switching until thedie temperature drops by 75 C, where-upon it is re-enabled.

Application ExampleThe circuit in Figure 2 depicts a typical

implementation of a 12 V, 120 mA non-isolated buck converter employingLNK304 that operates from a worldwidemains input. The input stage comprisesfusible resistor RF1, rectifiers D3 andD4, DC rail electrolytic capacitors C4and C5, and inductor L2. Thanks to theIC's integrated frequency jitter, the sim-ple pi-filter consisting of C4, L2, and C5provides enough attenuation to meet therequirements for conducted emissionsaccording to EN 55022, class B.Resistor RF1 is a flameproof, fusible,wire wound resistor that accomplishesthree functions. First, it limits the inrush

Figure 1. Functional Block Diagram

COVER STORY


current limit to a safe level for the inputrectifiers. Secondly, it attenuates differ-ential mode emissions. Finally, it acts asan input fuse in case any other compo-nent fails short-circuit.

The power stage is formed byLNK304, the freewheeling diode D1,output choke L1, and the output capaci-tor C2. In this particular case the IC andL1 have been selected to set a mostlydiscontinuous mode of operation. ChokeL1 is a standard off-the-shelf inductorwith an appropriate RMS current rating.The minimum required inductance todeliver the desired output power can becalculated as follows (MDCM operation):

where VMIN is the minimum DC busvoltage, VDS is the LinkSwitch TNDrain-to-Source voltage (typically 10 V),ILIMIT(MIN) is the specified minimuminternal current limit of the respectivedevice (0.24 A for the LNK304),fOSC(MIN) is the specified minimumoscillator frequency (62 kHz), and ? isthe estimated efficiency (e.g. 0.7). Thefactor (1+2?)/3 takes into account thatapproximately 66% of the losses in theconverter are dissipated in L1 and D1.Those losses are processed by the buckchoke as well. For a proper selection ofthe inductor, the inductance tolerance(typically ±10%) and the drop of induc-tance at elevated component currents(typically 10%) need to be factored in.Therefore, L1(MIN) is multiplied by afactor of 1.1 to 1.2 (using 1.15 as a typi-cal value) to find the data sheet value ofinductor L1.

Electrolytic capacitor C2 is the outputfilter capacitor. It is primarily selected todetermine the desired maximum outputvoltage noise, which is generally astronger function of the ESR than thecapacitance value of the capacitor. Theoutput regulation is performed with abootstrap circuit formed by the feedbackdiode D2, tracking capacitor C3, and theresistor divider R1 and R3 connected toLNK304's FB pin. To a first order, theforward voltage drops of D1 and D2 areidentical. Thus, the voltage across C3tracks the output voltage. Resistor R3effectively increases the ON/OFF cur-rent threshold compared to the internalFB pin threshold (typically 49 µA) byestablishing a bias current for the feed-back resistor divider. A bias currentvalue of 0.75 mA is recommended. Thefeedback resistor network can thereforebe calculated as follows:

COVER STORY


THERMAL MANAGEMENT

No one has ever seen a Unicorn,nor are likely to, it’s an imaginarycreature. Everyone however will

be able to describe a white horse withsingle long spiralled horn. Unlike theUnicorn, Heat Pipes do exist, andalthough commercially available sincethe mid 1960’s they are little known andlittle understood.

In the past decade, the electronicsindustry seems to have defied the lawsof economics by providing increasingperformance at less cost. However theindustry has been unable to suspendthe laws of physics, higher performancehas been accompanied by higher heatgeneration.

As a result thermal management iscritical and finding thermal solutions willbecome important for the cost reductionand time-to-market pressures governingsuccess in electronics sales.

Heat Pipe cooling of electronics isincreasingly popular due to thermaladvantages, reliability and cost.

Heat Pipe operationHeat pipes are efficient, two-phase

heat transfer devices with effective ther-mal conductivities several thousandtimes that of solid copper. They are ineffect heat movers or spreaders; theyacquire heat, and move or spread it towhere it can be more readily rejected.

Construction usually begins as a tube,typically copper Figure 1. The inside ofthe tube is lined with a wick structure,which provides surface area for the

evaporation/condensation cycle andcapillary capability. The core of the tubeis open for vapour flow. Before beingsealed the tube is evacuated and asmall quantity of working fluid justenough to saturate the wick is added.

The fluid, typically water, is the onlydynamic component in the heat pipe;the pressure inside the pipe is equal to the saturation pressure associatedwith ist temperature. As heat enters theequilibrium is upset generating vapourat a higher pressure and temperature.The higher pressure causes vapour totravel to the condenser where the lowertemperature causes the vapour tocondense onto the wick structure givingup its latent heat of vaporisation. Thecondensed fluid is returned back to theevaporator by gravity and/or the capil-lary forces in the wick structure. Thiscycle can transfer large quantities ofheat with very low thermal gradients.

Applications:Embedded Heat Pipes

As heat fluxes increase solid metalextrusions exhibit inherent limitations.The geometry and material result in highspreading resistance and in order tooffer enough surface area these sinkstend to be large and heavy.

When embedded into extrusionsFigure 2. The high thermal conductivityof heat pipes spread the heat through-out the base of the sink minimisingspreading resistance making the sink far more efficient.

Heat Pipes do existHeat pipes are efficient, two-phase heat transfer devices with effective thermal conductivities sev-

eral thousand times that of solid copper. They are in effect heat movers or spreaders.

By John Broadbent, Thermacore

This simple feedback approachrequires a constant current flow throughinductor L1, allowing capacitor C3 toaccurately sample the output. ResistorR4 therefore establishes a pre-load ofapproximately 3.6 mA, keeping the out-put in regulation even at very light loads.

Figure 4 shows that the load regula-tion of LinkSwitch-TN at 240 VAC inputis much better than the load regulationof the capacitive dropper exampledepicted in Figure 3.

The increasing output voltage withsmaller load currents can be controlledwith the pre-load resistor R4 as described

above. For a minimum system load of10 mA, the pre-load resistor can beeliminated and the feedback capacitorC3 can have a smaller value (1 µF, for instance).

Figure 5 shows that the full load effi-ciency of the buck converter example istypically 72% compared to 45% for thecapacitive dropper at its maximum cur-rent capability of 80 mA. It is interestingto note, that even at light load currentssuch as 20 mA, the LinkSwitch-TN buckconverter still offers an efficiency of 57%,whereas the efficiency of the capacitivedropper decreased sharply to 11%.

The no-load input power of the LNK304based buck converter is typically below180 mW at 265 VAC input voltage. Thiscompares to 2.6 W for the capacitivedropper at 230 VAC input voltage.

ConclusionThe LinkSwitch TN IC family is a

universal, cost effective and extremelyenergy efficient switching alternative tothe traditional highly inefficient passivetype non-isolated power supply currentlybeing used. The highly integrated deviceis optimized for a simple low-cost buckor buck-boost converter topology that iswell suited to handling the electroniccontrol requirements for major appliancesand any other non-isolated low outputcurrent power supplies. Please visitwww.powerint.com for more information.

Figure 3. Capacitive dropper example delivering 12 V, 65 mA (230 VAC input only)

Figure 4. Load Regulation LNK304 Buck Converter vs. Capacitive Dropper Example

Figure 5. Efficiency LNK304 Buck Converter vs. Capacitive Dropper Example

COVER STORY

Figure 1. Heat Pipe Operation

Figure 2. Embedded Heat Pipe Heat Sink


This reduction of spreading resistanceis particularly beneficial in applicationswhere the heat source or heat sourcesare concentrated. Even higher perform-ance in reducing base spreading resist-ance would be to fabricate the entirebase from a vapour chamber (Therma-Base). The vapour chamber utilizes theeffective thermal conductivity of two-phase heat transfer to virtually eliminatespreading resistance in the heat sinkbase, increasing the effectiveness offins located at the perimeter of the heatsink, offering the possibility to reduce itssize and weight.

Figure 3 compares the relative per-formance of three heat sinks. All had the same geometry and a 75 watt heatsource placed 1.3cm from the right handedge. The sink on the left is an aluminumextrusion. The middle profile is the samewith three embedded heat pipes show-ing a 36% improvement. The profile onthe right is a sink with a vapour chamberbase showing a 44% improvement.

Remote fin stack (Therma-Sink)Many custom situations are improved

by heat pipe assemblies. Constraintssuch as height restrictions, which limitheat sink profile options and componentdistances from available air source, arecommon problems that are solved.

Therma-Sink designs typically consistof an evaporator block (heat input) theheat pipe(s) (heat transport) and fins(heat dissipation). The heat pipe tech-nology moves the heat from the heatsource to a location where enough airvolume exists for adequate heatremoval. Fins are stacked onto the heatpipes to provide the surface arearequired for heat dissipation.

The relatively small size (a 6mm heatpipe can transfer 30-40 watts with littleor no ?T) ensure maximum flexibility inwhere to site the remote fin stack Figure 4.

Wick structuresThe main distinction of heat pipes,

besides working fluid, is wick structures.Several types of wick structure are usedin copper/water heat pipes including,screen, grooves and sintered metalpowder. Wicks are primarily used to provide capillary pumping to return con-densed liquid to the evaporator. Theselection depends on the power handlingrequirements and working orientation.Groove and screen mesh wicks havelimited capillary force capability, theycan not overcome significant gravitation-al forces, and dry out can occur and aretherefore limited to gravity aided andhorizontal orientations.

Sintered powder wicks are capable ofreturning working fluid to the evaporatoragainst gravity by using capillary pump-ing action and are effective in againstgravity applications.

The finer the pore radius of the wickstructure, the higher against gravity theheat pipe can operate.

Sintered powder wicks provide a largesurface area available for evaporationand provide a mechanism to accommo-date vapour and liquid simultaneously.Typical sinter powder wicks handle50W/cm≈, and have been tested to250W/cm≈. In comparison a groovewick will nominally handle 5W/cm≈ anda screen wick 10W/cm≈ Figure 5.

Orientation & Pipe GeometryThe efficiency of the heat pipe will

decrease as the angle of operationagainst gravity increases, thereforethe orientation of heat pipe designs arecritical. When possible the heat sourceshould be below or at the same eleva-tion as the cooling section. This allowsgravity to aid the capillary action andresults in greater heat carrying capability.If this orientation is unacceptable, a sin-tered powder wick will be required.Custom geometry is possible by thebending and flattening of the heat pipes.

LimitationsHeat pipes can be manufactured in

many shapes and sizes and designed tocarry a few watts or several kilowatts.For a given temperature gradient, heatpipes can transfer significantly moreheat than even the best metal conduc-tors. There are two primary limitationsthat must be considered:

Thermal ConductivityWhen driven beyond its rating the

thermal conductivity of the heat pipe willbe reduced. It is important to design aheat pipe to safely transport the requiredheat load.

THERMAL MANAGEMENT

Figure 3. Comparison of Three Heat Sinks.

Figure 4. Therma-Sink™


The maximum heat transport capabili-ty of the heat pipe is governed by sever-al limits. Viscous, sonic, capillary pump-ing, entrainment and boiling Table 1.

Thermal ResistanceAnother primary heat pipe design con-

sideration is the effective heat pipe ther-mal resistance or overall ?T at a givendesign power. As the heat pipe is a two-phase heat transfer device, a constanteffective thermal resistance value can-not be assigned. The effective thermalresistance is a function of a number ofvariables, such as heat pipe geometry,evaporator length, condenser length,wick structure and working fluid. Thetotal thermal resistance is the sum of theresistances due to conduction throughthe wall, conduction through the wick,evaporation or boiling, axial vapour flow,condensation and conduction lossesback through the condenser sectionwick and wall.

Detailed thermal analysis of heatpipes is complex. There are however a

few guidelines that can be used for firstpass design considerations.

A rough guide for a copper/water heatpipe with a powder metal wick structureis to use 0.2°C/W/cm≈ for thermal resist-ance at the evaporator and condenserand 0.02°C/W/cm≈for axial resistance.The evaporator and condenser resist-ances are based on the outer surfacearea of the heat pipe. The axial resist-ance is based on the cross-sectionalarea of the vapour space. This designguide is only useful for powers at orbelow the design power for the givenheat pipe.

ManufacturingEvery heat pipe should be subjected

high temperature and high internal pres-sure to be aged. After this aging everypipe must pass a thermal response testto check its leakage and ensure its nor-mal function. If the heat pipe is to bebent into some custom geometry theminimum bending radius is 3 x the pipediameter. Tighter bends increase the

risk of the pipe damage. Bending 2.5 xthe diameter may be possible depend-ing upon pipe hardness, wall thicknessand the process used. Heat pipes canalso be flattened, calculating the minimumflattened thickness for sintered pipescan be straightforward if the productionmandrel diameter is known. Thermacorecan change the pipe specification byadjusting the mandrel diameter andreducing the pipe wall thickness in orderto achieve the flattening requirements.

ConclusionsWith increasing power levels in smaller

packages heat pipes provide effectivesolutions.

These passive, efficient and versatileheat transfer devices can be applieddirectly to a heat source(s) and attachedto a heat sink such as a fin stack, extru-sion, thermoelectric devise, cold plate orenclosure wall. Heat pipes can also beused to enhance other thermal solutionssuch as extrusions.

Thermacore are investing a large por-tion of earnings in continues R & D todevelop advanced wick structures andin the introduction of new and advancedmaterials into the production environ-ment. Flexible, laminated and pulsatingheat pipes are just a few of the newtechnologies on the Thermacore roadmap.

THERMAL MANAGEMENT

Figure 5. Three kinds of wicks.

Heat Transport Limit

Viscous

Sonic

Entrainment Flooding

Capillary

Boiling

Description

Viscous forces prevent vapor flow in theheat pipe

Vapor flow reaches sonic velocity when exiting heat pipe evaporator resulting in aconstant heat pipe transport power andlarge temperature gradients

High velocity vapor flow prevents conden-sate from returning to evaporator

Sum of gravitational, liquid and vapor flowpressure drops exceed the capillary pump-ing head of the heat pipe wick structureFilm boiling in heat pipe evaporator typicallyinitiates at 5-10 W/cm2 for screen wicksand 20-30 W/cm2 for powder metal wicks

Cause

Heat pipe operating below recom-mended operating temperature

Power/temperature combination,too much power at low operatingtemperature

Heat pipe operating abovedesigned power input or at too low an operating temperatureHeat pipe input power exceedsthe design heat transport capacityof the heat pipeHigh radial heat flux causes filmboiling resulting in heat pipe dryout and large thermal resistances

Potential Solution

Increase heat pipe operatingtemperature or find alternativeworking fluidThis is typically only a problemat start-up. The heat pipe willcarry a set power and the large^T will self correct as the heatpipe warms upIncrease vapor space diameteror operating temperature

Modify heat pipe wick structuredesign or reduce power input

Use a wick with a higher heatflux capacity or spread out theheat load

www.thermacore-europe.com

Table 1.


POWER SEMICONDUCTORS

AbstractHigh reliability is very important for

today’s Switched Mode Power Supply(SMPS) market. High switching frequen-cy allows reduction of power supplysize. The zero-voltage-switching (ZVS)or zero-current-switching (ZCS) topolo-gies which allow for high frequencyswitching while minimizing the switchingloss are of interest. ZVS topology oper-ating at high frequency can improve theefficiency as well as reduce the size ofthe power supply. ZVS also reduces thestress on the semiconductor switch,improving the reliability. These advan-tages have made the high voltage highpower Phase-Shifted Full Bridge ZVSPWM converter a very popular topology.The conventional power MOSFET bodydiode can cause device failure duringswitching due to un-recovered minority

carriers. The new fast body diode MOSFET (FRFET MOSFET) is designedto address the diode reverse recoveryproblem.

IntroductionThe SMPS industry is experiencing

revolutionary developments in increasedefficiency, and reliability due to improvedpower devices for the circuit topologies.The Phase-Shifted, Pulse WidthModulated, Full-Bridge Zero-Voltage-Switched (PS-PWM-FB-ZVS) topologyutilizing the FRFET as power switchescan achieve the above goals. The poorreverse recovery (Trr) characteristics ofthe inherent body diode of high voltageMOSFET can reduce the reliabilityunder certain conditions. These condi-tions can occur even during light loadswhere switch with high output capaci-

tance can enter a hard-switching condi-tion and increase switching losses. If thebody diode is not properly designed thepower switch may lose gate control. Thenew FRFET overcomes the reliabilityissue for this topology by incorporating a robust, low Trr, soft recovery diode.

Topology descriptionThe PS-PWM-FB-ZVS converter as

shown in Figure 1 usually operates atswitching frequencies (fs) in excess of100 kHz. In the conventional H-bridgetopology the switches are turned on andoff under hard switching conditions. Aconventional H-bridge topology can bemodified to obtain the PS-PWM-FB-ZVStopology. First, modulation is achievedby phase shifting two overlapping con-stant frequency square waves by usingleading-leg and lagging-leg. Second, it

is important to achieve resonant ZVS tominimize the switching losses that resultfrom hard switching. The primary differ-ence between this topology and themore traditional H-bridge topology is theswitching method to achieve ZVS.

Instead of turning on the diagonallyopposite switches of the bridge simulta-neously (i.e. Q2 & Q3, Q1 & Q4), aphase shift is introduced between theswitches in the left leg (leading-leg Q1 &Q2) and those in the right leg (lagging-leg Q3 & Q4) as shown in Figure 2. Thisphase shift determines the operatingduty cycle of the converter. Zero-voltageturn-on is achieved by using the energystored in the leakage and series induc-tance of the transformer to dischargethe output capacitance of the switchesthrough resonant action. The resonanceforces the body diode into forward con-duction prior to gating on the switch.

ZVS process Two different mechanisms exist which

provide ZVS [3] for the lagging-leg andleading-leg.

From Figure 1 and Figure 2, at t=t1,Q1 is turned-on just after Q2 hasturned–off. The current starts to freewheel through D1 and Q3. At t=t2, thestored energy in the inductor may not be adequate to turn-off Q3 under a ZVScondition. The energy stored in L1 isused to charge the output capacitanceof Q3 and discharge the output capaci-tance of Q4. The energy stored in theoutput filter inductor was freewheelingthrough the output rectifier diodes anddoes not affect the lagging-leg switches.When the output capacitance of Q4

discharges, D4 turns-on and then Q4can be turned-on at ZVS. Once D4 con-ducts, positive bus voltage is applied tothe primary winding of the transformer.Using a switch with low effective outputcapacitance eliminates hard turn-off andturn-on improving the efficiency at lightloads. To achieve soft turn-on and turn-off of the lagging leg, the followingequation should be satisfied.

ZVS topology operating at high frequencycan improve the efficiency

If the body diode is not properly designed the power switch may lose gate control. The newFRFET overcomes the reliability issue for this topology by incorporating a robust, low Trr, soft

recovery diode.

Sampat Shekhawat, Praveen Shenoy and Bob Brockway, Fairchild Semiconductor

POWER SEMICONDUCTORS

Figure 1: PS-PWM-FB-ZVS topology


Where Coe3 & Coe4 = Effective outputcapacitance of the power switches Q3 &Q4, Ip1 = Primary transformer current atturn-on and turn-off for Q3 and Q4 andCTRW = Transformer winding capacitance.

For the leading-leg switches (Q1&Q2), the ZVS occurs as explained below.

At t=t5, when Q1 is turned-off, theenergy stored in the output filter inductorL2 and primary side inductor L1 is avail-able to charge Q1 output capacitanceand discharge Q2 output capacitance.The output filter inductor value is high.So, even at light load, the energy storedin L2 is high enough to turn-off Q1 andturn-on Q2 and vice versa under a ZVScondition. Once the output capacitanceof Q1 is charged and the output capaci-tance of Q2 is discharged, there is stillenough energy available to forward biasthe diode D2. Once D2 is on, the lead-ing-leg switch can be gated-on. If theswitch Q2 is now commanded to turn-on, the diode D2 continues to carry cur-rent for a long period of time due to theenergy stored in L1 and output induc-tance L2. The stored energy in the filterinductor is large compared to that

required to charge and discharge thecapacitances in the primary. So theswitch capacitance is charged at a linearrate. Accordingly, for soft turn-on andturn-off of the leading leg the followingequation should satisfy.

Where: Coe1 & Coe2 = Effective outputcapacitance of the power switches Q1 &Q2. L2p is the output filter inductancereferred to primary.

Importance of free wheeling diode forthis topology

The MOSFET body diode must not betaken lightly in the ZVS topology. Evenwith the reduced switching stress on theMOSFET, failure may occur because ofa poor diode recovery. The failures havebeen reported [1, 2] at no-load or light-load conditions. These failures resultfrom the lagging-leg losing ZVS at turn-on and turn-off forcing a hard switchingcondition. One potential failure mecha-nism is due to the CGD x dVDS/dt cur-rent. The resulting current can causeVGS to charge above VTH. A shootthrough condition will exist where bothof the MOSFETs in the lagging-leg canturn-on and cause the leg to fail.

To achieve ZVS at turn-on, the bodydiode of the on coming MOSFET isallowed to conduct current. For exam-ple, body diode (D2) of Q2 will conductcurrent when Q1 is turned-off. When theQ2 MOSFET is gated-on, channel willstart carrying current in the third quad-rant (same direction as diode). Currentwill free-wheel through D2 and Q4.

When Q4 is turned-off, diode D2 anddiode D3 will carry current. When Q3 isturned-on at t6, the current in diode D2decays to zero. The MOSFET Q2 willstart to conduct in forward direction. Thebody diode D2 starts to recover with avery low reverse voltage. The bodydiode Trr increases significantly due tolow reverse voltage. Figures 3 & 4 illus-trate the influence of Vrm on dioderecovery. Figure 3 shows that whendiode recovers with di/dt= 150A/mA and20V reverse voltage, the body diode Trris 616ns. The Trr increases further whenVrm is reduced to 10V as shown in fig-ure 4. The di/dt is less in this case whichshould have reduced Trr but it increasedinstead due to reduced Vrm. It is clearthat the Trr increases as the reversevoltage is reduced.

If the MOSFET Q2 is turned-off andits body diode has not fully recoveredbefore the complementary switch Q1turns on, the body diode of Q2 can create shoot-through. The dVDS/dt can cause a current, made up of the un-recovered charge in thediode and the charge inCDB, to flow through the par-asitic NPN Rb base resist-ance. This current can biason the parasitic NPN BJTcausing loss of Q2 MOSFETgate control and failure. Thediode should have a low Qrr,Trr and Irrm to overcome theproblem discussed above.The Qrr and Trr of the bodydiode should be low enoughto provide complete minoritycarrier removal before thedevice turns-off. If this doesnot occur, the turn-offdVDS/dt of this device couldturn-on the parasitic NPNtransistor and forcing thetransistor into secondarybreakdown. The possibility of this failure occurringincreases as the frequency is increased. The life time ofthe parasitic body diode hasbeen killed by Platinum diffu-sion and that has reduced theQrr, Trr and Irrm.

ConclusionThe reliability problems of the PS-PWM-FB-ZVS topology have been

documented. These problems can beover-come by using the FRFET series

MOSFET. The CGD x dVDS/dtimmunity of these MOSFETs has been increased by reducing QGD, the QGD/QGS ratio, and the internalgate resistance (ESR) and increasingVGS(TH). The intrinsic body diode of the FRFET MOSFET has beenimproved by reducing the Qrr and Trr as shown in Figure 3. The Qrr of the FCS fast body diode MOSFEThas been reduced by 10X compared tostandard MOSFET and Trr has beenreduced by 5X. This take care of thereliability issues. The Trr and ofFDxxxN50F series Qrr has beenreduced compared to our previous generation FRFET (FQxxxN50F series).For more information, go to:

www.fairchildsemi.com/power/

References:[1] H. Aigner, et al., “Improving the

Full-Bridge Phase –shift ZVT Converterfor Failure-free Operation underExtreme Conditions in Welding andSimilar Applications” IEEE proceedingsof IAS Society Annual Meeting, St.Louis, 1998.

[2] L. Saro, et al., “High-Voltage MOSFET Behavior in Soft-SwitchingConverter: Analysis and ReliabilityImprovements,” International Tel-com-munication Conference, San Francisco,1998.

[3] J. A. Sabate, et al., “Design con-siderations for high-voltage high powerfull-bridge zero-voltage-switched PWMconverter,” in Proc. IEEE APEC, 1990,pp. 275-284

[4] K. Shenai, et al., “Soft-Switched,Phase-Shifted Topology Cuts MOSFETSwitching Stress in FBCs,” PCIMMagazine, May 2001.

POWER SEMICONDUCTORS POWER SEMICONDUCTORS

Figure 2: ZVS Waveforms

0.5 (L2p+L1) x (Ip)2 > 0.5(Coe1+Coe2) x (Vin)2 + 0.5CTRW x (Vin)2

Figure 3: Diode reverse recovery comparison @ Vrm = 20V, IF=10A, di/dt = 150A/us. Figure 4. Diode reverse recovery comparison @ Vrm = 10V, IF=9A, di/dt = 50A/us.

0.5L1(Ip1)2 > 0.5(Coe3+Coe4) x

(Vin)2 +0.5CTRW x (Vin)2


DC/DC CONVERSION

With minimal setup and intervention,these computers will be able to wake upto perform routine maintenance, down-load software updates, send and receivefax transmissions, answer the telephone,download audio and video streams and,when the tasks are complete, go back toa low power sleep state where they con-sume less than a night light. This is theworld of Instantly Available PCs (IAPC).

Picture the following scenario in anoffice setting: a computer system will bepowered up in the morning, and theCPU and all other system componentvoltages will receive regulated powerfrom the on-board dc/dc regulators oper-ating off the AC/DC supply (also knownas the “silver box”). During the day theuser may leave the computer unattend-ed for coffee break or meetings. Upon apre-determined lack of user input, thecomputer transitions into a stand-by(sleep) mode during which it consumesminimal power, while still monitoring all input ports for incoming data.Depending on the stand-by mode, the PC responds to inputs such as anincoming fax through the PCI slot, a

keystroke on the keyboard, a mouseclick, or an incoming photo of a USB-connected digital camera, by resumingnormal operation in a timely manner.IAPC systems may be able to modifytheir performance by clock throttling and changing the CPU core voltage onthe fly, to match user and applicationdemands. These, and other advancedmodes of operation, enable the comput-er systems to consume energy in themost efficient manner, while providingan enhanced user experience.

Hardware and software featuresThe restriction of power management

algorithms by the information availableto the BIOS that implements them, thelack of a well-specified power manage-ment and configuration mechanism toaddress advanced system architectures,such as Universal Plug and Play, andthe existence of legacy hardware suchas the ISA slots, serial and parallelports, limit the functionality and flexibilityof many of the computer systems stillavailable today and ultimately deprivethem of becoming truly intelligentmachines.

The Advanced Configuration andPower Interface (ACPI) Specificationprovides a platform independent, industrystandard approach for operating systembased power management. The ACPIspecification is the key constituent inOperating System-Directed PowerManagement (OSPM). OSPM and ACPIapply to all classes of computers includ-ing handheld, notebook, desktop, andserver machines. In ACPI enabled sys-tems, the BIOS, hardware, and powerarchitecture must utilize a standardapproach that enables the operatingsystem to manage the entire system inall operational situations. The ACPI OStakes over the power management andPlug and Play functions from the legacyBIOS interfaces (such as the APM BIOSand PNPBIOS) and allows inexpensivepower management hardware to sup-port very elaborate power state transi-tions, maximizing the computer's powerconsumption and resource utilizationefficiencies.

The ACPI OS can put the whole computer or parts of it, such as the theprocessor and motherboard devices, in

The Late and future-generation of computers will benefit

Today's computers represent a quantum leap in performance and complexity compared to theavailable technology of only a few years ago. Nevertheless, a paramount goal of these late and

future designs is to make this complexity transparent to the user.

By George Lakkas and Bogdan Duduman, Intersil

www.powersystemsdesign.com28 29

and out of various sleep states, basedon user settings and applicationrequests. It has knowledge of whichdevices will be affected prior to dockingor undocking, device insertion andremoval, or thermal events. The OS cal-culates the battery remaining capacityand battery remaining life, and deter-mines the various battery warning levelsin mobile PCs. It allows other driverswithin the system to communicate and use the resources of systemembedded controllers.

Hardware implementation challengesWhat does ACPI power management

mean from a computer power system

designer's viewpoint? A multitude ofvoltages on the motherboard and risercards that enable the processing ofaudio, video, and data streams have tobe generated and managed with no userintervention. ACPI-compliant computersrequire the generation of these multiplevoltages at various current ratings asthe system transitions between differentsleep states. There are six discrete sys-tem operating states possible as definedin the ACPI Specification. These statesare referred to S0 to S5 in order of high-est to lower power consumption (Fig. 1).All states correspond to some level ofpower management. In the S0 state, theCPU may toggle between C0 and C1 or

C2 states, and the power system mustbe able to respond quickly to resultantcurrent transients. Other system compo-nents have their own correspondingpower states (D0 to D3) that best matchtheir usage profile based on user andapplication needs. As the system transi-tions into lower power states (S1 andS2), the power consumption decreases.

Lowest power consumption comes inthe sleep states of S3 through S4,where the computer system slips into avirtual shutdown. A typical system willsupply standby power to system memo-ry, USB and PS/2 ports, and LAN cardor modem circuitry in an S3 state. Thesame system will typically removestandby power to the system memoryand modem when entering an S4 state,and all components except the powerbutton when entering an S5 state.

The deep, power saving, sleep statesdo not come without drawbacks. Themore circuitry that is shut down to con-serve energy, the longer the time that isrequired to restore the system to anoperational status. A multitude of powermanagement challenges arise whenthese systems transition between thesevarious states. A system in an S3 statethat wants to go into an S4 state mustfirst go to an S0 state, perform therequired tasks, then enter the lowerpower S4 state. The on-board powerconversion and power management cir-cuitry must generate and enable theappropriate voltage and current levelsas the system transitions into the vari-ous states completely glitch free. Here,hardware design and implementationquality are important, as a power anom-aly at any transition point could causesevere problems, including loss of user data.

The computer “dual” power supply concept

The concept of “dual supply” wasintroduced by Intel in the “Power Supply98” initiative. The initiative is targeted atdealing with the different power supplydemands of the computer subsystemsduring their operation in various states.A supply powering a computer sub-sys-tem in an ACPI environment must beable to supply the full active current, yet

drop into a low-power mode when thesub-system itself slips into a low-powersleep state. To maximize power deliveryefficiency, it makes sense to have adedicated power supply for each groupof active and sleep states. Thus, thisconcept leads to a high-power supplyand a low-power supply, and matchingthe power needs of the subsystem withthe appropriate power supply (Fig. 2)

Dual supplies can be realized eitherwithin the AC/DC power supply itself ordirectly on the computer motherboard.However, traditional AC/DC supplies arenot user-configurable. Thus, if computermanufacturers want to use a dual-modesilver box, they must pay for the cost offull ACPI support in every power supplypurchased, regardless of whether thecomputer or the motherboard is capableof full ACPI support.

Implementing the dual supplies on themotherboard allows for application flexi-bility and custom configurations, as wellas cost advantages if the project callsfor supporting only a limited set of ACPIfeatures (Fig. 3). In this implementationmodel, a series of on-board regulatorsand switches are required to generatethese various voltages and to switchbetween supply rails as the varioussub-systems change operating modes.

On-board ACPI power management: discrete vs. integrated approach

A fully discrete implementation mayinitially seem as the most simple andcost-effective method of regulating thevarious ACPI power states and the tran-sitions between them. The various on-board regulators for system memory,PCI slots, Universal Serial Bus, keyboard,and mouse can be realized with voltage

references, resistors, capacitors, op amps,diodes, bipolar and MOS transistors.Such discrete implementations havemany drawbacks, the most important ofwhich were the subject of many techni-cal articles. Noise coupled into the cir-cuits leads to "illegal" ACPI state transi-tions and power perturbations that canlead to loss of critical data. Switchingbetween supplies while they have notfully stabilized, or switching severalloads into the same supply line leads topower overloads, responsible formomentary brownouts, leading to outputregulation falling outside design limits.Interfacing the various discrete regula-tors to the South Bridge additional gluelogic that is not immune to systemnoise. Glitches may trick the regulatorsto enter a fault state or latch off. Thetypical result of data loss is a systemcrash, followed by a much needed reboot.

Good, discrete implementations canbe realized, but, typically, a discreteACPI power management circuit takesup precious board area, and lacks sys-tem monitoring and protection functionsthat could significantly reduce the TotalCost of Ownership (TCO) and prolongthe life of an ACPI-compliant computer.The surface mount placement costassociated with these discrete compo-nents also contributes to a higher imple-mentation cost. Additionally, the reliabili-ty of the computer is potentially reduced,as compared to an integrated solution,as there are many more components on the board that could fail over time(Table 1).

An integrated circuit that combines allthe ACPI regulators and system moni-toring and protection functions neededto properly manage the operating statetransitions and the power demanded, isthe most economical, simple, and reli-able method of ACPI-compliant mother-board power management.

The ACPI Power ControllersIntersil's HIP6501A was the industry's

first integrated device targeted at theACPI power management initiative. TheHIP6501A regulates/controls 3 voltageplanes. The output voltages are fixedeliminating the need for external resistors.

DC/DC CONVERSION DC/DC CONVERSION

Figure 1. There are six discrete systems operating states possible.

Figure 2. This concept leads to a high-power supply and low-power supply.


Figure 3. If the project calls for supporting only a limited set of ACPI features.

Table 1. Comparison integrated to discrete solution.


The HIP6501A interfaces with the com-puter system through four digital lines,which allow it to be configured in foursupport configurations, as well as becontrolled directly by the Southbridgechip. Up-integration and intelligentdesign allows the HIP6501A to employonly eight (8) external components while operating from within a small 16-pin narrow-body SOIC package.TheHIP6501A integrates three linear controllers:a 5VDUAL for USB/Keyboard/Mouse, a3.3VDUAL for PCI/Auxiliary/LAN, and aselectable 2.5V/3.3VMEM for systemswith either RDRAM, SDRAM, or DDR

SDRAM memory. The HIP6501A alsoprovides adjustable soft start, internalcompensation, over-temperature andunder-voltage protection with centralizedFAULT reporting. The HIP6501A hasbuilt-in filters for noise immunity andrequires no glue logic (Fig. 5).

The small pin count package and thesmall number of external componentsdo not impede on the chip's functionalityor ease of use. Wherever possible, thechip was designed to be compatible withbipolar transistors, reducing system costwhen compared to implementations

using entirely MOSFETs. Thus, a bipolarPNP transistor can replace the PMOSswitch employed on the 5VDUAL output.For systems employing RDRAM or DDRSDRAM memory, the HIP6501A uses abipolar NPN as a pass element of thecorresponding output. For SDRAM-based systems, the NPN pass elementis to be replaced by an NMOS transistor.A HIP6501A-based ACPI circuit elimi-nates, typically, a minimum of twenty-nine (29) external components, as usedin a fully discrete ACPI power manage-ment implementation.

Intersil's HIP6500B, HIP6502B, andHIP6503, each regulate and control five(5) voltage planes. In addition to the5VDUAL, 3.3VDUAL, and 2.5V/3.3VMEM, theHIP6500B also regulates the 3.3VSB

(3.3V standby), and 2.5VCLK (2.5Vclock). The 3.3VSB output is used topower the chipset. The 2.5VCLK

output powers the motherboard clockgenerator in the active states only.

The HIP6502B differs from theHIP6500B in that it combines the3.3VDUAL/3.3VSB voltage planes in oneand provides two individual memory out-puts (2.5VMEM and 3.3VMEM) for mother-boards that utilize both RDRAM andSDRAM (separate video and system)memories. The HIP6500B andHIP6502B eliminate, typically, a mini-mum of thirty-five (35) external compo-nents, as used in a fully discrete ACPIpower management implementation.

The HIP6503 addresses Intel ICH2-based chipset architectures and in addi-tion to the 5VDUAL, 3.3VDUAL/3.3VSB,2.5V/3.3VMEM, and 2.5VCLK voltages, itprovides a fixed 1.8VMCH output for thechipset (Fig. 6). The HIP6503 elimi-nates, typically, a minimum of thirty-five(35) external components, as used in afully discrete ACPI power managementimplementation.

Intersil’s two-power IC solution forcomplete PC motherboard voltageregulation

The HIP6500B, HIP6501A, HIP6502B,and HIP6503 utilize an ATX AC/DC sup-ply (“silver box”) and work in conjunctionwith the HIP6020/21 and ISL6523/24four-in-one PWM (Pulse-Width-Modulated)

controllers to simplify the power supplydesign on ACPI-compliant motherboards.

An Intersil two-regulator chipset con-sisting of the HIP6020/21 and HIP6501Atarget VRM8.4-compliant Intel Pentium(R) III motherboards employing the Inteli810, i810e, i815, Via Apollo ProMedia133, and SiS SiS620/5595 and SiS630chipsets (Fig. 7).

By substituting the HIP6501A with the HIP6500B, Intersil's dc/dc regulatorcombination addresses VRM8.4 IntelPentium (R) III motherboards with theIntel i820, Via Apollo Pro 133A, ApolloPro 133, and Apollo KX133 K7 chipsets.

The HIP6502B ACPI linear powercontroller is intended for VRM8.4 IntelPentium (R) III-based workstation andserver motherboards utilizing Intel’s i840and Serverworks’ LE/HE chipsets.

An ISL6523/24 and HIP6503 chipsetregulates all active and sleep-state volt-ages on VRM8.5-compliant mother-

boards featuring Intel's newTualatin microprocessor,i815e/i815ep Memory ControllerHub (MCH) and I/O Controller Hub (ICH2/ICH3) (Fig. 8).

ConclusionPragmatically, IAPC power

management using integratedapproaches, such as Intersil’s dualdc/dc regulator solutions, providethe computer and motherboardmanufacturers with the lowest billof materials cost and highest relia-bility, leading to a longer mother-board and computer system life.

ACPI-compliant computers provide a rich experience for theconsumer and corporate user, asonce set up properly, these sys-tems require far less maintenance,and are one step closer to trulybecoming the intelligent appli-ances we have come to expect.

DC/DC CONVERSION DC/DC CONVERSION

Figure 5. The HIP6501A has build-in filters for noise immunity.

Figure 6. The HIP6503 addresses Intel ICH2-based chipset architecture.

Figure 7: An Intersil two-regulator chipset.

Figure 8: An ISL6523/24 and HIP6503 chipset.regulates all active and sleepstate voltage.


POWER PROTECTION

There are two main over-voltagethreats to subscriber line interface cards(SLICs): lightning and short circuit to theAC utility supply. The protection mecha-nism for these two types of over-voltageevent need to be different, after a light-ning strike the protection circuit needs to reset and the SLIC must continue tooperate. However, mains crossingevents are less common than lightningstrikes and potentially more destructive,under these circumstances the protec-tion circuit is allowed to fail in order toprevent damage to the SLIC. In order to meet these requirements the protec-tion circuit components need to be carefully selected.

Subscriber lines can be divided intotwo main types, analogue and digital.Analogue lines work from a nominal–48V supply but also have ringing volt-ages of up to 150Vrms. Digital lines onthe other hand tend to operate with amuch lower supply voltage, as low as3.3V, and use digital signals to producering tones in the terminal equipment.Both types of subscriber lines use SLICs and both require over-voltageprotection, the main difference being inthe selection of the transient voltagesuppresser (TVS).

In analogue systems during ringing avoltage of 150Vrms is superimposed ontop of the -48V supply giving a peakvoltage of up to 270V. When the tele-phone is answered the loop currentincreases and as a result the ringing is

stopped and the call is connected. Callconnection normally occurs when theloop current exceeds 3mA. This mode of operation sets the minimum require-ments for the TVS: it must not operateat less than 270V and at this voltagelevel the leakage current must be lessthan 3mA.

When the telephone is ‘on-hook’ thehook switch will be open, as shown infigure 1, any transient voltage thatoccurs in this state will appear directlyacross the hook switch.

For adequately rated electromechani-cal devices this is not a problem but alarge number of hook switches are nowsemiconductor devices such as FETs.Semiconductor switches have a maxi-mum operating voltage which, to avoiddamaging the device, should not beexceeded. Hence the TVS must operateat a voltage below the hook switch volt-age rating. When the hook switch isclosed it must be able to withstand the

surge current associated with an over-voltage event, this is often achieved byusing series resistors to limit the current.

Protection from lightning strikes isnormally achieved by using two or threestages of protection, as shown in figure 2.

The first protection stage is known asthe ‘primary protection’, this generallyconsists of a gas discharge tube or aspark gap with current limiting seriesresistors and slow acting high currentfusing. This stage of protection isdesigned to take the initial lightning volt-age of >20kV down to between 1-4kVand reduce the transient period fromsomething over 50ms to less than 1ms.

The second stage is refereed to as‘secondary protection’ or ‘line side’ pro-tection. This level includes protectionagainst short circuit to the utility supplyand generally consists of a fuse or posi-tive temperature coefficient resistor(PTC) a TVS and possibly a resistor in

Withstand the surge currentA large number of hook switches are now semiconductor devices such as FETs. Semiconductor

switches have a maximum operating voltage which, to avoid their damage, should not be exceeded.

Chris Likely, Cooper Electronic Technologies

each line. The effect of this stage is toreduce the 1-4kV transient down to lessthan 50V. The final stage is normallyonly require in digital applications inorder to protect the low voltage CMOScircuitry, it reduces the 50V transient toless than 10V.

The line side protection is designed toprotect against the two main types oflightening induced over-voltage event.These are metallic or transverse protec-tion that is required to protect againstdirect strikes on the overhead lines andlongitudinal that protects againstinduced over-voltages due to ground

strikes effecting under ground cables.Figure 3 shows the protection circuitsrequired for these types of over-voltagetransient and figure 4 shows a typicalline side protection circuit.

The use of series resistors in eachline is common practice when using afuse for protection from short circuit to

the utility supply. The resistors limitthe peak current during over-voltageevents allowing the use of a lower cur-rent rated TVS and less robust fuse. Asmentioned previously this technique canalso be used to protect the hook switch.In order to save space and reduce costit is common practice to replace the fuseand resistors with PTCs which provide acurrent limiting impedance and act as are-settable fuse.

Although PTCs are suitable for manysubscriber line applications they dohave some problems, both the specifiedtrip current and holding current valuesvary significantly with temperature. Thetrip current is the current level at whichthe PTC becomes high impedance andthe holding current is the current levelbelow which the PTC resets after anover-current event. Both these valuescan vary by as much as ±50% over thedevice operating temperature range.PTCs also tend to be slower to operatethan fuses thus allowing more, potential-ly damaging, energy through to theinterface card.

More significantly, for subscriber lineapplications, the PTC will tend to resetfrom an over-current to a slightly differ-ent resistance each time, hence any ini-tial line balancing will be upset after aPTC has tripped and reset. Also in digi-tal circuits, such as E1 & T1, where theoperating impedance is only 100 Ohms,introducing the additional resistance of aPTC in to the line can significantly loweroutput levels. In this type of applicationrobust ‘telecom circuit protection’ fuses(TCP’s) are the ideal solution. For moreinformation:

Cooper (UK) LimitedBurton-on-the-WoldsLeicestershire LE12 5TH UKTel +44 (0) 1509 882 737Fax +44 (0) 1509 882 786www.cooperet.com

POWER PROTECTION

Figure 1: Hook Switch Circuit

Figure 2: Stage of Lightning Protection

Figure 3: Types of Lightning Protection

Figure 4: Typical Line Side Protection Circuit


OPTOELECTRONICS

Here, we look at how The BergquistCompany’s Thermal Clad InsulatedMetal Substrate (IMS) material hashelped lighting specialist Osram indeveloping its latest family of LED-based traffic light solutions.

Thanks to the potential for both ener-gy savings and reduced levels of main-tenance, highways authorities aroundthe world are rushing to adopt trafficlights based on LED arrays rather thanconventional incandescent lamps.Around 10% of all traffic lights in the US,for example, were converted to LED-based solutions in both 2001 and 2002,and the same rate of change is takingplace in Europe. In China, Beijing isupgrading its traffic lights to LED-basedunits in time for the 2008 Olympics.Osram Opto Semiconductors is at theforefront of the development of LEDs for traffic light applications.

Benefits of LED replacementIn the USA, where a conventional

traffic light can consume up to 150W,energy savings offered by LED arrayreplacements are particularly significant.Assuming an average of eight lights perjunction, for example, a large US cityhaving several thousand intersectionscould expect to save over $1 million per

year in energy costs alone by fitting LEDarrays that consume between 15W and20W. Conventional European trafficlights only consume around 40-60W,making the potential energy savingsless dramatic, but the longer lifeexpectancy of LEDs delivers additionalbenefits. At 50,000 hours typical lifetime,a coloured LED used in a traffic signalshould last more than five years, com-pared to approximately nine months fora conventional incandescent bulb. Thisnot only reduces purchasing costs butmore significantly, saves the expense of maintenance crews, who frequentlyhave to close a lane simply to changeone bulb.

Osram traffic signalsThe latest additions to the Osram

Opto Semiconductor family of modularTRAFFICsignal units are designed tocompletely replace the incandescentlamp, reflector, lamp socket and frontdisk of conventional 210mm and 300mmtraffic lights. To allow easy retrofitting,these LED lamps will fit into all standardsignal housings and are compatible withnearly all symbol masks, such asarrows, X indicators for lane closures,and STOP or WALK/DON’T WALK leg-ends. They are also resistant to mois-ture and particle ingress to IP65. Lights

are available in red, green and yellow,and each achieves high luminous effi-ciency. Each unit comprises three mod-ules: the LED light source array, anadvanced optical system that minimisesphantom effects and presents a uniform-ly illuminated front lens to the viewer,and an electrical driving unit.

The LED arrays used in the OsramTRAFFICsignal products are based onthe company’s Power TOPLED family ofsurface mount power LEDs. These high-intensity red, yellow and green InGaAlPand InGaN devices are housed in com-pact P-LCC-4 packages, have forwardcurrent ratings (IF) of 70mA, and powerdissipation ratings of up to 180mW.Optical efficiencies ranging from 15lumen/W to 30 lumens/W, and the needto comply with light intensity regulationssuch as those covered by the EuropeanEN12368 standard (see Figure 1), dictate that the number of LEDs used in each light is between 60 and 120depending on the colour (fewer arerequired in the case of higher brightnessred LEDs, while more are required for inthe case of lower brightness green). Andit is this combination of power dissipationand mounting density—plus the need tooperate in some very high ambient tem-peratures - that meant Osram had to

Effective solution for rapidly dissipating heat

Osram chooses Bergquist insulated metal substrate (IMS) technology for traffic signal coolingThe use of high-power LEDs in applications that have conventionally used incandescent lamps

is fuelling demand for high-efficiency cooling solutions.

Mark de Vos, Bergquist

Get your own copy of Get your own copy of

Sign up today online at www.powersystemsdesign.comSign up today online at www.powersystemsdesign.com

your trade journal for the power engineer!your trade journal for the power engineer!


find an effective solution for rapidly dis-sipating heat away from the LED arrays.

Heat dissipationThe cooling of LEDs has rarely, until

now, been a key design criteria. Indeed,to date in consumer electronics, officeequipment and most other applicationsemploying LEDs, product life cycles aresuch that there is a significantly higherchance that the product will fail or reachthe end of its useful life before the LEDssuffer any degradation. However, whenit comes to the high-density mounting ofsurface mount power LEDs such as theTOPLED devices used by Osram usesin its OS-TR family of traffic signal units,cooling becomes a critical factor.

The main issue is that, while LEDshave typically been seen as ‘cool run-ning’ devices, the fact is that they actu-ally create a significant amount of heatin relation to their size. Not only can thisheat build-up damage the electricalintegrity of the LED (eventually leadingto lamp failure), it can also affect thelight intensity over time. In applicationssuch as traffic lights where there areclear rules and regulations governinglight intensity (see side panel) andwhere long-term reliability is essentialfor both cost and safety reasons, it

therefore becomes important to choosematerials that provide the most efficientmethod of heat dissipation.

As surface mount devices, the majori-ty of heat generated by the TOPLEDLED lamp is transferred to the printedcircuit board through the base metal tothe dielectric layer (see Figure 2).

However, during the traffic signaldevelopment, Osram found that conven-tional FR4 substrates did not deliver theappropriate level of heat dissipationbecause the dielectric layer that FR4employs did not offer a high enoughlevel of thermal conductivity. The com-pany also experimented with aluminiumsubstrate materials which, again, couldnot deliver the requisite levels of thermalconductivity. Eventually the search foralternative substrate materials led theOsram team to The BergquistCompany’s Thermal Clad InsulatedMetal Substrate (IMS) technology.

Thermal Clad Insulated MetalSubstrate (IMS)

Designed originally to provide thermalmanagement solutions for higher Watt-density surface mount applicationswhere die size is reduced and heatissues are a concern, Thermal Clad is

perfectly suited to long-life, high-powerLED applications. Thermal Clad min-imises thermal impedance and conductsheat more effectively and efficiently thantraditional PCB materials, ensuring thelower operating temperatures neededfor long-life, consistent LED operation.

Thermal Clad IMS comprises a circuitlayer, a thermally enhanced dielectriclayer and a metal substrate (see Figure 2).

The dielectric layer offers electricalisolation with high thermal conductivityand bonds the base metal and circuit foiltogether. By adjusting the compositionand thickness of the dielectric, as wellas the type of base material, a range ofThermal Clad solutions is available tomeet a variety of power dissipationrequirements. When developing its high-intensity LEDs, Osram selected a verythin dielectric material offering a highthermal performance. In comparison tocommon dielectrics such as reinforcedfibreglass or pressure sensitive adhe-sive (PSA), the Thermal Clad enhanceddielectric minimises thermal resistance,resulting in more effective and efficientheat conduction. A comparison of thethermal resistance of the OsramThermal Clad material and alternativetechnologies is shown in Figure 3. Thislower resistance is key to delivering thehigh thermal conductivity necessary tomeet Osram’s traffic signal requirements.

SummaryLED technology in traffic signals and

other types of high intensity signageoffers persuasive energy and mainte-nance savings. The enormous numberof such signals in use globally presentsa valuable opportunity for manufacturerswho can meet emerging national specifi-cations. The key is to effectively dissi-pate the significant heat generated bysuch large arrays of densely clusteredLEDs. Practical experience has shownthat LEDs mounted very densely onstandard FR4 or aluminium substrateswill not meet minimum standardsthroughout the lifetime of the signal.Failure to dissipate this heat also com-promises safety and reduces the lifetimeof the lamp itself. Osram’s experiencehas shown that the enhanced dielectricof an IMS such as Bergquist’s Thermal

Clad delivers a cost-effective solution tothe thermal management challenge pre-sented by high intensity LED traffic signals.

Finally, it should be noted that, inaddition to traffic signals, there are anumber of other applications that useLEDs where good heat dissipation is

required. The ‘Golden Dragon’ LED fromOsram, for example, can be driven withcurrents in excess of 300mA and is usedin applications where high brightness is required from a very small area.Dissipating the heat away from this LED is a key design consideration and,again, the Bergquist Thermal Clad IMS

offers a high-efficiency solution.

For more company information visitthe Bergquist web site at:www.bergquistcompany.com

For more information on OSRAMOpto Semiconductors please visit:www.osram-os.com

Contact details for publication:Mark de VosThe Bergquist CompanyEuropean HeadquartersPO Box 276 1250 AG LarenThe NetherlandsTel: +31 35 538 06 84e-mail: [email protected]

OPTOELECTRONICS OPTOELECTRONICS

Figure 3: Thermal Clad IMS comprises a circuit layer, a thermally enhanced dielectric layer and a metal substrate.

Figure 2: The majority of heat generated by the TOPLED LED lamp is transferred to the printed circuit board through the basemetal to the dielectric layer.

Figure 1: Thermal resistance by substrate.


POWER MANAGEMENT

Advanced integrated ciruits such ascommunication processors achieveincreased performance, added function-ality, and reduced power consumption inpart by being fabricated using the latestsub-micron technologies. The reductionin process geometries also results inreduced operating core voltages. Inter-device communication standards, how-ever, often dictate the use of specific I/Osupply voltages, which results in devicesthat require multiple power supplies.Many of these devices also require thattheir power supplies be turned-on orsequenced in a particular order andadditionally, some require that the I/Ovoltage be within a given voltage rangeof the core voltage during turn on.

While the individual functions may beeasy to implement in isolation, satisfyingthe complete set of power supplyrequirements for all the devices on thecircuit board present a significant chal-lenge for today's board designers. ThePower Manager series from LatticeSemiconductor is capable of performingall these power supply managementfunctions in a single chip.

Power Manager ArchitectureBeing the first mixed signal PLD in the

world the Power1208 and Power604,packaged in 44-pin TQFP packages,provide the following programmablemodules to implement sequencing andmonitoring functions (figure 1):

• On-board ispMACH sequence controller CPLD

• Delay timers (for sequence delaysand watchdog timers)

• Internal oscillator for precision timingand clocking the CPLD

• Multiple analog inputs with program-mable precision analog threshold

• Buffered comparator outputs• Programmable N-channel MOSFET

drivers (Power1208 only)• General-purpose open-drain digital

logic outputs• General-purpose logic inputs

Programmability reduces prototype-debugging time

The power manager family provides the complete collection of functions needed for sequencingand monitoring power supplies in all 5 phases of power supply cycle; it is a highly integrated

single-chip solution with all sequencing and monitoring functions in one central place.

Johannes Fottner, Lattice Semiconductor

The Power Managers' embeddedCPLD is derived from that of theispMACH4000 CPLD.

Inputs for the CPLD block are derivedfrom the comparator outputs, the logicinputs, the internal clock generator andthe delay timers, which enable thePower Manager to realize any combina-tion of voltage sequencing and monitor-ing. Programmability of both analog anddigital functions makes it easy to adaptto changes during the design phase.

For clarity the differences of Power1208and Power604 are shown in table 1.

Each of the 12 analog inputs providesa precision programmable thresholdcomparator for logic supply voltagesranging from 1.03 V (1.2 V- 14 %) to 5.7V (5 V + 14 %) with a typical step widthof 1 %. Using external resistive dividerscan extend this input range and alltwelve voltages can be monitored simul-taneously. The comparator outputs assuch are fed to the CPLD block. In thePower1208, 8 out of the 12 comparatoroutputs are also fed to an open drainoutput, in the Power604, 6 out of 6 com-parator outputs also feed open drainpins generating supervisory signals, andcontrolling the delay timer blocks.

Timing for the device is provided byan internal clock generator, which oscil-lates at 250 kHz, or by an external clocksignal. In both cases the signal passesthrough a prescaler block with divisionratios of 20 to 28. Synchronizing multi-ple power managers is easy by makingone of them the master which then pro-vides the clock signal which is then inturn fed to the slave devices. Depending

on the use of an internal or an externalclock the related pin is either an opendrain output pin or an input pin.

Sequencing functions often requiremultiple long-duration delays. There are4 programmable timers onboard thePower1208 and 2 on Power604. Theycan be individually configured to gener-ate timing delays of up to 500 ms usingthe internal 250 kHz clock oscillator.

Timing delays can be extended to anydesired duration through the use of anexternal clock.

Complete Power-Supply ManagementUsing Power1208

The block diagram fig. 2 is an exam-ple of a power supply distribution sys-tem in a circuit board with multi-voltagedevices. The Power1208 implements allpower sequencing and monitoring func-tions. Each one of the Power1208’sanalog inputs is connected to a dedicat-ed threshold comparator. These com-parators offer precise, independentlyprogrammable thresholds that allowthem to monitor the supply voltages forover voltage or under voltage conditions.Power up sequencing entails progress-ing through a series of phases, whichare detailed below.

Phase 1: Input Voltage monitoringensures that all external power supplies

Figure 2: Block diagram of a typical Power1208 Application

POWER MANAGEMENT

Figure1. Power1208 and Power604 block diagram.

Feature Power1208 Power604

Power Supply Sence Inputs 12 6

Supervisory Control Outputs 4 4

High Voltage FET Drivers Digital Outputs 4 —

Reprogramable Timers 4 2

Embedded CPLD Macrocoils 16 8

Operating Voltage 2.25V to 5.5V 2.25V to 5.5V

Package 44-Pin TQFP 44-Pin TQFP

Odering Part Number ispPAC-POWER1208-01T441 ispPAC-POWER504-01T441

Table 1: Comparison of Power1208 and Power604 functions.


are stable before the onboard powersupplies (Brick, LDO, and FET) areturned on. This makes sure that the sys-tem start-up phase won't start while sup-ply voltages are out of the specified range.

Phase 2: The voltages on the deviceside (V1, V2, V3) are monitored toensure reliable power is applied to thedevices. In multi-voltage systemssequencing specifications can be quitecomplex, e.g. core voltage first, thencore voltage and then auxiliary voltages.

Phase 3. All power supplynodes (input and device side)and digital input signals(Reset_In& FPGA_Load_Complete)are monitored to generate logiccontrol signals (CPU_Reset andPower_Good) for the system tostart up. The power manager'sprecise delay capabilities allowthe control of delays ranging froma few microseconds to seconds!

Phase 4. All power supplynodes (input and device side)are monitored for power supplyfaults and to facilitate generationof supervisory signals:Brown_Out_NMI, CPU_Reset, and Power_Good. Also, theManual_Shut_Down signal ismonitored to facilitate the controllogic to jump to phase 5.

Phase 5. Monitor supplies toensure controlled shut down.Integration and ProgrammabilityBenefits with Power Manager

The power manager family ofLattice Semiconductor greatlysimplifies implementation ofcomplex power-supply sequenc-ing and monitoring systems andreduces design time.

Programmability reduces pro-totype-debugging time, and byprogramming the threshold volt-ages, power supply marginingtests can be performed even onthe production line.

The power manager family providesthe complete collection of functionsneeded for sequencing and monitoringpower supplies in all 5 phases of powersupply cycle; it is a highly integrated sin-gle-chip solution with all sequencing andmonitoring functions in one central place.Its robustness ensures reliable start upof the board; even under noisy powersupply start up conditions and the changein supply sequencing and monitoringrequirements due to component substi

tutions on the circuit board can be easilyaddressed through re-programmability.

Lattice SemiconductorJohannes FottnerSenior Analog Field Application EngineerLilienthalstrasse 27, D-85399 HallbergmoosPhone +49 811 550 [email protected]

What’s in? What’s out? What is state of the art?

The Factorized Power Architecture (FPA), in concert with IC-style chip devices, provides powersystem designers with high performance at low cost. FPA is enabled by the power conversionchips, which efficiently process over 200 Watts of power in a power Ball Grid Array with power

densities over 800 Watt/ cubic inch.

Robert Marchetti, Vicor

Distributed power has been a viabledesign approach for decades, albeitalmost exclusively in the telecommuni-cations domain. Board-mounted DC-DCconverters, for example, were construct-ed with discrete components and usedto convert power from a centralized sup-ply to more usable power, typically at abackplane. Subsequently, the advent ofhigh-density DC-DC converter modulespaved the way to more widespread use

of the idea in EDP, industrial, military,and medical applications. Since thattime, the chorus of articles in the tradepress extolling the virtues of DistributedPower Architecture has been increasing-ly loud and clear: centralized architec-ture is going or gone and DPA is king.The structure and operation of each ofthese power architectures is summa-rized in Figure 1, and they are dis-cussed in subsequent paragraphs.

Centralized architectures The evolution of power architectures

began with the centralized architecture.A centralized power supply contains theentire power supply in one housing æfrom the front end through the DC-DCconversion stages. It converts the linevoltage to the number of DC voltagesneeded in the system and buses eachvoltage to the appropriate load. It’s verycost effective. It doesn’t consume pre-cious board real estate at the point ofload with the power conversion function.It’s relatively simple to manage thermalsand EMI. It is fairly efficient because itavoids serial power transformations. Inthe past, the centralized system, usuallya custom design constructed of discretecomponents, was often chosen becauseit was less expensive. These systems,in general, work well when the powerrequirements, once defined, are not like-ly to change and space is not an issue.

The central supply should be locatednear the load to minimize I2R losses,and it should also be located as close aspossible to the AC entry point to reducenoise radiated from the unshielded AClines. This is often a difficult trade-offwith the input cables requiring shieldingto minimize common and differentialmode currents that produce noise.

Centralized power works well in manyrespects, but the most obvious problemis how to distribute hundreds of ampscommon with low output voltages.Centralized power also lacks scalability.Many systems can be configured with

POWER MANAGEMENT DC/DC CONVERSION

Figure 1: A comparison of the structure and operation of major power architectures.


varying numbers of PC cards represent-ing widely varying loads (e.g., line cardsin a PBX). With centralized power, thepower supply must be sized to handlethe maximum configured system, whichcould put the small configurations at acost disadvantage.

What’s more, the remoteness of thesupply from the load can negativelyimpact the ability of the supply to reactto a rapidly changing load (i.e., transientresponse). Also, thermal managementcan be a special challenge in a central-ized architecture, where excess heatcould amount to hundreds of watts all inone concentrated area. Large heat sinksand fans are often needed to keep thepower supply from becoming overheat-ed. System hotspots are a source ofreduced reliability.

Distributed Power Architecture Distributed power architecture (DPA)

addresses some of the shortcomings ofa centralized architecture. Distributedpower is a decentralized power architec-ture usually consisting of an AC-DC con-verter at the AC mains serving DC-DCconverters located elsewhere. The AC-DC converter might provide regulation,isolation, noise suppression, and powerfactor correction as well as an intermedi-ate DC voltage, frequently 48V. Thisintermediate voltage is converted byDC-DC converters located at the pointof the load they serve. Typically, powerin a telecom system is distributed on a48 Volt bus. On-board isolated DC-to-DC converters are matched to the loadrequirement. This helps with dynamicresponse and eliminates the problemsassociated with distributing low voltagesaround the system. DPA was actuallyenabled by the development of high-density bricks.

A distributed approach spreads theheat throughout the system, greatlyreducing or eliminating the need for heatsinks or high velocity airflow. With tem-perature more evenly maintainedthroughout the system, reliability specifi-cations are easier to meet.

DPA, however, can be more costly ina number of ways. For example, isola-

tion, regulation, transformation, EMI fil-tering, and input protection are all doneat every brick. And coming from anoffline source, rectification and conver-sion to the distributed bus voltage areneeded so that’s an additional power-processing step that reduces overallsystem efficiency.

Furthermore, if a single DC-DC con-verter cannot provide adequate poweror fault tolerance for a particular outputvoltage, multiple DC-DC converters willneed to be paralleled, creating addition-al complexity owing to the need to con-nect remote sense leads from each par-alleled converter to a single, common,point and the need for additional circuit-ry within each paralleled converter toforce power sharing among the units.

Intermediate Bus Architecture The Intermediate Bus Architecture

was first to separate the DC-DC con-verter functions of isolation, transforma-tion, and regulation and allocate them totwo devices. The IBC (Intermediate BusConverter) provides intermediate voltagetransformation and isolation and theniPOL (non isolated Point of Load) con-verter provides final transformation andregulation.

IBA can be a very cost-effective solu-tion because point-of-load convertersdon’t require any isolation and tend tobe a lot less expensive. Non-isolatedPOL converters within the IntermediateBus Architecture forego isolation andhigh ratio voltage transformation toimprove cost-effectiveness, but theydepend upon a nearby bus converter tosupply power at a low input voltage. Onthe negative side, the lack of isolation inniPOL converters make over-voltagesensitive loads vulnerable to deadlyfaults and the entire system to potentialground loop problems.

The intermediate bus converter intro-duces a power-processing step to gofrom, say, the 48 to 12Volts that intrinsi-cally reduces efficiency of the system.Also, the bus converter really does needto be located close to the load, becauseeven at 12Volts, fairly high currentsneed to be moved around the board so

large traces or short runs are needed.The 12Volt bus itself is a bit low for effi-cient distribution of a lot of power. Butit’s too high to step down to a very lowvoltage because of the very low dutycycle on the switch. As a result, it is difficult to make a highly IBA system.

Factorized Power Architecture Factorized Power Architecture (FPA)

also separates conventional converterfunctionality into two power buildingblocks. This new architecture, in concertwith IC-style chip devices (see Figure1), provides power system designerswith high performance at low cost. FPAis enabled by the power conversionchips, which efficiently process over 200Watts of power in a small (less than0.25 cubic inch) and light (less than 13grams) power Ball Grid Array (BGA) orJ-level package, with power densitiesover 800 Watt/ cubic inch. These func-tional building blocks are deployed assurface mount (SMD) components tocreate a flexible Factorized Power system.

One building block, the Pre-RegulatorModule (PRM), is designed to accept awide-range supply voltage and convert itto a Factorized Bus æ a controlled volt-age source æ with 97% to 99% efficien-cy. Another building block, the VoltageTransformation Module (VTM), isdesigned to convert the Factorized Busto the voltage levels required by the loadwith efficiencies as high as 97%. TheVTM will also provide input to outputgalvanic isolation.

The combination of FPA and IC chipsgive the power designer the flexibility touse only what is needed where it isneeded. The minimal complement ofPRMs and VTMs depends upon themultiplicity of outputs, power levels, indi-vidual regulation and power system faulttolerance requirements. VTMs andPRMs may be paralleled with accuratesharing for higher power or redundancy;in fact, VTMs inherently current sharewhen inputs and outputs are paralleled.This avoids the need for a power-shar-ing protocol, interface signals and a mul-tiplicity of remote sense connections.

By exploiting a zero-voltage switchingand zero-current switching topology, theVTMs limit the common-mode and dif-ferential-mode noise at the point of load.For example, the output of a VTM con-figured to convert 48V to 12V exhibitsabout 12mV p-p of high-frequency ripplewith just 1 _F of bypass capacitance.That noise voltage amounts to only0.1% of the DC output.

Michelle Norris, VICOR UKColiseum Business CentreRiverside WayCamberley GU15 3YLUKTel: +44 1276 678222Web Site: www.vicoreurope.com

Email: [email protected] website at www.vicorpower.com

DC/DC CONVERSIONDC/DC CONVERSION

Figure 2: The Factorized Power Architecture converter.

Figure 3: Function blocks of the Factorized Power Architecture converter.


DESIGN AND SIMULATION

Since the thermal efficiency requiredin low-power systems is not generallyvery high, defects such as voids may bepresent in the die attach layer withouthaving a significant impact on the heatsink function.

In high-power devices, the subtractionof heat from the device is much morecritical, and the thermal efficiencyrequired from the heat sink system ismuch greater. For example, the dieattach layer between a very low-powerIC and its die paddle might contain voids(trapped air bubbles) over as much as40% of the area of the die attach withoutcausing device failure from overheating.In a high-power device, the presence ofany voids might be enough to lead tothermally-caused failure.

How ultrasound images heat sinkattachment

To ensure the integrity of the heat sinkfunction, acoustic micro imaging can beused to image the attachment layerbetween the heat sink and the device. Inmany devices, this layer consists of sol-der, but it may also consist of a thermalgrease or another type of adhesive. Ifthe attachment layer is homogeneousand without anomalies, the acousticimage of the layer will show no features.But if the attachment layer includes avoid (or any other type of gap), thisanomaly will show up distinctly in theacoustic image.

Figure 1: Acoustic image through a heatsink attached to the back of a PC board.The transducer scanned the surface ofthe heat sink, and the return echoeswere gated at the attach layer depth.Two components are visible becausecutouts through the board attach themdirectly to the heat sink. The numerousred areas are voids in the solder attach layer.

Acoustic micro imaging uses a scan-ning transducer to pulse ultrasound intothe heat sink and to collect the returnechoes (Figure 1). Pulsing and signalcollection both take place several thou-sand times a second as the transducerscans the heat sink. Each return echobecomes a single pixel in the completedacoustic image. The velocity of ultra-sound through most production materi-als is quite high, and the elapsed timefor the pulse-echo round trip is usuallyon the order of a few nanoseconds.

Despite the high speed of transmis-sion of ultrasound, return echoes fromvarious depths arrive back at the trans-ducer at slightly different times. It is theusual practice to gate the return echoesvia a time window, with the result thatonly those echoes from a desired depthare used to assemble the acousticimage. In the case of heat sinks, thisdepth is usually the attachment layerbetween the heat sink and the deviceitself. Ultrasound is reflected only fromthe interfaces between materials, andthe heat sink itself is presumed to bestructurally homogeneous and thereforelacking in acoustic features.

Finding voids in heat sink attachmentFigure 2 is the acoustic image made

through a large-area heat sink that hasbeen applied to the back side of a PCboard. This is a low-power application,but one that demonstrates nicely theacoustic imaging of heat sink integrity.The return echo signals were gated onthe solder layer attaching the metal heatsink to the PC board. The weave patternof the board is visible, as are irregulari-ties in the heat sink attach layer. Theoutlines of two active components arevisible because cutouts were made inthe PC board in order to attach thesecomponents directly to the heat sink.

The key features in Figure 2 are thered regions, which are voids in the sol-der layer attaching the heat sink to the

Ensure the integrity of the heat sink functionMany electronic devices employ some type of heat sink to remove heat and prevent the device

from failing. Low-power integrated circuits (ICs) often depend on the conduction of heat through adie attach layer onto the die paddle.

By Tom Adams, consultant, Sonoscan, Inc

PC board. Their bright color corre-sponds (in the color map used here) tohigh amplitude in the ultrasonic echosignals. The voids are numerous, andwill to some extent compromise theintended heat sink function. Note thatthere are even small voids along theedges of the two components attacheddirectly to the heat sink.

Reflection of ultrasound from internalinterfaces

Ultrasound pulsed into a sample isreflected only from material interfaces,and not from the bulk of a material. Mostinternal interfaces involve two solidmaterials, but the presence of a voidmeans that the arriving ultrasoundencounters an interface where one ofthe materials is air. Air (or any gas) dif-fers so sharply in its density and in itsacoustic velocity from the overlying solidmaterial that the return echo signal fromthis point has very high amplitude.

The relatively low acoustic reflectionfrom a well bonded interface betweentwo solids means that much of the ultra-sound crosses the interface and travelsdeeper—and this is why an acousticmicro imaging system can “reach down”through several layers of material toimage a single interface of interest. Avoid or other gap-type defect reflectsultrasound so efficiently that essentiallyno ultrasound penetrates through thegap to reach deeper layers.

The vertical dimension of the void isrelatively unimportant for acoustic reflec-tion. Recent tests show that gaps whosevertical dimension is on the order of100Å to 1000Å reflect ultrasound withabout the same efficiency as muchthicker gaps.

Heat sinks in high-power applicationsFigure 3 is the acoustic image of a

thermally critical application—the attachof a very thin laser diode to its heat sinksubstrate. Because of the high thermaloutput of the diode, no voids can be tolerated in the attachment layer. Thebright spot (circle) in Figure 3 is a void,conspicuous because of the high ampli-tude of the echo signals being reflectedfrom the void.

DESIGN AND SIMULATION

Figure 2: Thermally, a void acts as an insulator and prevents heat dissipation.Acoustically, the great differences in density and in acoustic velocity between theair-filled void and adjacent solid material causes the void to reflect ultrasound athigh amplitude.

Figure 3: A critical application: the acoustic image of the attachment of a laser diodeto its heat sink. The white area (circle) is a heat-blocking void.

Figure 4: The attachmentof a thermo-electric cool-er, whose vertical vanes(white in this image)enhance heat dissipation.Black rectangles areareas where the cooler isbonded to the substrate,but these areas containmany small white voids.The diagram at bottomshows a side view.


The Easy1 housing offers a well suit-ed module concept for a compact and

cost-effective integration of power

semiconductors with a footprint of only33 x 45 mm.

The Easy module series devicesdemonstrate excellent performancewhich was now transmitted into a newtopology. This newest available topolo-gy, called EasyBRIDGE1, extends theexisting Easy family with an uncontrolledthree phase bridge rectifier including anintegrated brake arm. Such a topologyprovides best adaptation to EasyPACKmodules and even Econo modules, dueto the common height.

The product range for EasyBRIDGE

modules covers the voltage rangeVRRM (repetitive peak reverse voltage)for 800V and 1600V at an output currentfor Id = 50A resp. 40A.

Further addressed to market require-ments, EasyBRIDGE modules will beavailable additionally in versions withimproved surge forward current ofapprox. 60%.

With these extensions, the productportfolio of Easy modules, realizing inputrectifier, brake chopper and inverter, areapplicable for motor ratings up to 5,5kW.

NEW PRODUCTSDESIGN AND SIMULATION

Because laser diodes are themselvesvery thin (typically between 75 and 100microns), earlier acoustic imaging sys-tems could not easily gate on the veryrestricted depth represented by theattachment layer. This problem wassolved by the development of higheracoustic frequencies (230 MHz, 300MHz) that make such precision a routine matter.

Figure 4 is the acoustic image of athermo-electric cooler. Although thestructure of the thermo-electric cooler ismore complex than the structure of mostheat sinks, the basic principle is thesame, and voids at the attachment layerwill cause a reduction in heat subtrac-tion function. The black rectangles inFigure 4 are the solder attach of the ver-tical cooling vanes that increase the sur-face area. Small white areas are pres-ent within the outline of the solderbonds. These features correspond tovoids and show the same reflectionlevel as the non-bonded areas betweenthe vane bonds.

Measuring the thickness of theattachment layer

In some applications the absence ofvoids in the attachment layer is not initself enough to ensure efficient heat

removal. The precise thickness of theattachment layer from one side of theheat sink to the other is also important.If the attachment layer is too thick or toothin, the device may fail, even if novoids are present.

A method developed at Sonoscan(www.sonoscan.com) uses acousticmicro imaging techniques to measurethe thickness of the attachment layernondestructively and with great preci-sion. Measurements are performed atSonoscan's applications laboratory, andare useful in establishing processparameters for the successful attach ofa heat sink to a device under development.

Variations in the thickness of theattach layer can also be demonstratedgraphically. At each pixel point, the timedifference across the thickness of theattach layer is measured, and colors areassigned to various thicknesses.

Figure 5 shows the thickness profileof the solder attach layer of a directbond copper sub-strate to its heatsink. Thicknesscorresponds tothe color map atthe left, where red indicates the

thinnest attach layer. The thickestregions of the attach layer (green) are at the center because the substrate wasslightly warped. There are also numer-ous voids of various sizes in the attachlayer. Although the substrate is warped,the voids are all at the same level andtherefore all have the same color.

Contact information: Sonoscan, Inc.2149 E. Pratt Blvd.Elk Grove Village IL USA 60007Phone: 847 437-6400Fax: 847 437-1550E-mail: [email protected]: www.sonoscan.com

Figure 5: Color visualization of the thickness of the attach-ment layer. In this acoustic image of the attachment of adirect bond copper substrate to its heat sink, red areas arethinner, and green areas thicker.

Rectifier Bridge

www.eupec.com

LEM has introduced an “Advancedfunction” for its MEMOBOX PowerQuality Analyser family. With thisenhancement, all aspects of PQ analy-

sis, network optimisation and distur-bance analysis (previously only avail-able in separate instruments) are cov-ered in a single instrument. The newversion now covers all measuring tasksneeded for the operation and mainte-nance of an electricity supply networkfor a utility or in large industrial facilities.

The new function enables the user tomeasure and store 500 relevant param-eters simultaneously over a period of upto four weeks. The MEMOBOX can beused in low- and medium-voltage net-works up to 830 V peak-to-peak andcaptures up to four currents. Its compact

size allows it to be installed into thesmallest cabinets. The isolated caseand accessories provide reliable protec-tion against electrical shock.

The MEMOBOX analyses supplyquality fully in accordance with EN50160, including voltage and currentharmonics up to the 50th. For analysis,the package includes the new Version4.0 "Codam plus" software, which isaccepted as the leading PQ softwareproviding the user with clear conclusionswithout extensive training.

Power Quality Analyser

www.lem.com

Intersil announced that the very popu-lar HIP2100 and HIP2101 high-voltagehalf-bridge MOSFET driver ICs are nowavailable in new space-saving and ther-mally efficient packages. The new 8-lead exposed-pad (EP or E-pad) SOICpackage offers enhanced thermal effi-ciency and the 4 X 4 mm dual flat no-

lead (DFN) package yields the world’ssmallest 100-V half-bridge MOSFETdriver. Applications for these thermallyenhanced drivers include, but are notlimited to, telecom and datacom powersupplies, avionics DC/DC converters,two-switch forward converters andactive-clamp forward converters.

Intersil’s exposed-pad device packageis a thermally enhanced eight-leadsmall-outline integrated circuit (EPSOIC)that enables power supplies to run cool-er and more efficiently. This enhancedpackage features a copper pad connec-tion to the printed circuit board ground,which serves as a heat sink.Transferring heat from the MOSFETdriver integrated circuit (IC) through thecopper pad reduces thermal resistance

by up to 50 percent and enables theMOSFET driver to operate cooler andmore efficiently.

The 4 X 4 mm DFN version is sam-pling to customers now and will bereleased to full production in about six-teen weeks. It is the smallest 100-V half-bridge MOSFET driver available thatallows engineers to ad here to IPC-2221design standards for printed circuitboard layout. The guidelines call for 0.6mm of spacing between high voltagenodes in order to assure long-term sys-tem reliability.

Smallest 100-V Half-Bridge Driver ICs

http://www.intersil.com/HIP210X

48 Power Systems Design Europe February 200448

NEW PRODUCTS

Texas Instruments announced it isadding an Enable pin to its widely adopt-ed line of high efficiency synchronousbuck drivers for non-isolated, low outputvoltage DC/DC converters. Providing anew driver with Enable gives systempower designers improved flexibility and

control when managing low output volt-ages in central processing units (CPUs),computing, telecom, datacom and gen-eral merchant power supply systems.

The new UCC27223 driver withEnable, and the previously announcedUCC27221 and UCC27222 devices produce ±3 A of output drive current forefficient power MOSFET switching atcritical Miller plateau thresholds. TheUCC2722x family yields two to four percent better efficiency and up to 40percent improved heat dissipation com-pared to competing ICs that support lowoutput voltages.

The UCC27223 device’s Enable pinand proprietary Predictive Gate Drivedigital control technology reduce diode

conduction and reverse recovery lossesin synchronous rectifier MOSFETs.Employing a closed loop feedback sys-tem that detects body-diode conduction,Predictive Gate Drive adjusts dead timedelays in the next switching cycle tominimize the conduction time interval insynchronous rectifiers. In addition, theoverall converter efficiency improves up to two to four percent.

The new driver incorporates TI’shybrid TrueDrive output stage with paralleled bipolar and complementarymetal-oxide semiconductor (CMOS)transistors to allow efficient currentdelivery at low supply voltages.

Drivers for Non-Isolated DC/DC

www.ti.com

Ricoh announces the launch ofR1115Z LDO Regulator. Ricoh R1115Zcomes in a tiny WL-CSP-4P4 packageconsuming 0.94mm in PCB area and

measuring L 0.97mm x W 0.97mm x H0.59mm.

WL-CSP-4P4 package is its 0.3mmbump size, making it easily recognisablefor PCB assembly machines. RicohR1115Z uses small ceramic capacitorskeeping total PCB area consumption toa minimum. Ricoh R1115Z’s high ripplerejection keeps noise to a minimum andthe combination of low dropout voltageand low supply current ensure lengthybattery life. Ricoh R1115Z is ideal forcreating low power consuming applica-tions for small portable battery operatedaudio or telecommunication devices.

Ricoh R1115Z uses small externalceramic capacitors of minimum 1.0µF atVout <2.5V and minimum of 0.47µF atVout ≥2.5V. Ceramic capacitors arefavourable for minimum PCB area con-sumption. Ricoh R1115Z output noise iskept to a bare minimum with a ripplerejection of 70dB at f=1kHz and 60dB atf=10kHz. Long battery life is guaranteedby a combination of low dropout voltageof 0.22V at 150mA and low supply cur-rent of 75µA. The product’s maximumoutput current is 150mA.

Low noise 150mA LDO

www.ricoh-semiconductor.com

Companies in the issusCompany Page Company Page Company Page

ABB Switzerland . . . . . . . . . . . . . . . . . . .1Advanced Power Technology . . . .23,46Allegro MicroSystems . . . . . . . . . . . . . . .1Anagenesis . . . . . . . . . . . . . . . . . . . . . . .1Analog Devices . . . . . . . . . . . . . . . . . . . .1Ansoft . . . . . . . . . . . . . . . . . . . . . . . . . .11Ansoft . . . . . . . . . . . . . . . . . . . . . . . . . . .1Artesyn Technologies . . . . . . . . . . . . . . .1Bergquist . . . . . . . . . . . . . . . . . . . . . . . .34Cooper Electronic Technologies . . . . . .32CT-Concept Technology . . . . . . . . . . .19CT-Concept Technology . . . . . . . . . . . . .1Curamik Electronics . . . . . . . . . . . . . . .1Danfoss . . . . . . . . . . . . . . . . . . . . . . . .31e/j Bloom . . . . . . . . . . . . . . . . . . . . . . . . .6eupec . . . . . . . . . . . . . . . . . . . . . . . . . .15

eupec . . . . . . . . . . . . . . . . . . . . . . . .1,4,47Fairchild Semiconductor . . . . . . . . . .C2Fairchild Semiconductor . . . . . . . . . . .1,22IMEC . . . . . . . . . . . . . . . . . . . . . . . . . . . .6Infineon Technologies . . . . . . . . . . . . .21Infineon Technologies . . . . . . . . . . . . . . .1International Rectifier . . . . . . . . . . . . .C4International Rectifier . . . . . . . . . . . . .1,10Intersil . . . . . . . . . . . . . . . . . . . . . . . . . . .5Intersil . . . . . . . . . . . . . . . . . . . . . .4,26,47IXYS . . . . . . . . . . . . . . . . . . . . . . . . . . . .1Lattice Semiconductor . . . . . . . . . . . . . .38LEM Components . . . . . . . . . . . . . . . . .3LEM Components . . . . . . . . . . . . . . . . .47Linear Technology . . . . . . . . . . . . . . . . .7Linear Technology . . . . . . . . . . . . . . . . . .1

Mesago-PCIM Europe Nuremberg 43,C3National Semiconductor . . . . . . . . . . . . .1Power Integrations . . . . . . . . . . . . . . .40Power Integrations . . . . . . . . . . . . . . . . .12Power Systems Design . . . . . . . . . . . . .35Ricoh . . . . . . . . . . . . . . . . . . . . . . . . . . .48Semikron . . . . . . . . . . . . . . . . . . . . . . .27Semikron . . . . . . . . . . . . . . . . . . . . . . . . .1Sonoscan . . . . . . . . . . . . . . . . . . . . . . . .44SynQor . . . . . . . . . . . . . . . . . . . . . . . . . .1Texas Instruments . . . . . . . . . . . . . . .1,48Thermacore . . . . . . . . . . . . . . . . . . . . . .14Thermacore . . . . . . . . . . . . . . . . . . . . . .17TTI Electronics . . . . . . . . . . . . . . . . . . . . .4Tyco Electronics . . . . . . . . . . . . . . . . . . .1Vicor . . . . . . . . . . . . . . . . . . . . . . . . . .4,41

Please note: Bold—companies advertising in this issue

psde mag cover for pdf (page 2) · chris rexer fairchild semiconductor andreas sperner ixys member...

Documents