Engineering Edge Volume3-Issue1

8/10/2019 Engineering Edge Volume3-Issue1

1/31

E N G I N E E R I N G Vol. 03 / Issue. 01 / 2014

Accelerate Innovationwith CFD & Thermal

CharacterizationEDGE

mentor.com/mechanical

Facebook

Likes ThisThermally Ef cientDatacenter Design Page 16

More Power!Stanley Black &Decker Power ToolDesign Page 48

Seiko EpsonEmpoweringEngineers Since

1989 Page 38


2/31

mentor.com/mechanical mentor.com/mechanical 3

Perspective Vol. 03, Issue. 01

Greetings readers! It is my pleasure to offer you this edition ofEngineering Edge as the new General Manager of the Mechanical

Analysis Division. Many of you will know me from my previous roleas Product Line Director for our FloEFD , FloTHERM and FloVENT products. As I take over from Erich Buergel, who was promoted to VicePresident of sales in Mentor Graphics in September, I feel both humbleand excited. Humble, because Erich leaves big shoes behind him havinggrown the Division substantially during his tenure, and excited becausebeing a product man I believe we have an exciting technology roadmapto execute on to increase our innovation.

I am also very excited because we have two very weighty products that were released to the marketrecently: FloTHERM V10.0 and the brand new MicReD Power Tester 1500A. We devote several pagesof this newsletter to explaining the changes and unique capabilities of these two world class products

so I commend them to you. After our 25th

Anniversary for FloTHERM last year, V10.0 unveiled someexciting product interface changes implemented by Product Management that have been well receivedby you, the users. The Power Tester 1500A is also a signi cant evolution of our MicReD product linewhere we have produced an 'Industrial' solution rather than a laboratory-level product. The 1500A hasall the accuracy and reliability of the standalone T3Ster but in a very robust power cycling and testingstation that can be used by operators in a manufacturing environment. We have put a lot of effort intouser experience for the software that goes with the Test Station 1500A, and its touch-pad screen isboth intuitive and easy to navigate. We genuinely think it will be a game changer in power electronicscycling and testing, especially for IGBTs. If you get the chance, check out the video of the Power Tester1500A on our website; it gives a good feel for what it can do.

Finally, Engineering Edge wouldn't be Engineering Edge without a slew of customer stories. Thecover story from Facebook shows how they use FloTHERM to design custom rack systems fortheir datacenters. In addition, ADA designed combat aircraft fuel systems, Black & Decker's PowerDrill electronics cooling, Seiko's big time savings with FloEFD for projector design, Flanders Driveautomotive electric powertrain cooling, Cofely's revamping of a biomass furnace, and MercurySystems' military avionics thermal design, are all noteworthy in their own right. Do read John Murray'sinteresting survey of Engineering Apps, plus an interview with long-time veteran of the electronicscooling industry, Bill Maltz from ECS. He has some great insights showing his company is still workingat the cutting edge, his colleague Guy Wagner's story on how they tear down and analyze severaltablet computers is well worth reading.

I commend Engineering Edge to you and I trust we will continue to work with you to produce industrialsolutions that meet your current and emerging needs.

Mentor Graphics CorporationPury Hill Business Park,

The Maltings, Towcester, NN12 7TB,

United Kingdom Tel: +44 (0)1327 306000

email: [email protected]

Editor:Keith Hanna

Managing Editor:Natasha Antunes

Copy Editor:Jane Wade

Contributors:Byron Blackmore, Keith Hanna, David Hunt, Andrey Ivanov, Boris Marovic, John Murray,

John Parry, Maxim Popov, Joe Proulx, SvetlanaShtilkind, Steve Streater, Tatiana Trebunskikh,

Jane Wade

With special thanks to: A1 Racing

Aeronautical Development Agency IndiaBromley Technologies

Cofely Fabricom GDF SUEZD2T

Electronics Cooling Solutions, Inc.Facebook

Flanders' DRIVEMercury Systems, Inc.

Seiko Epson CorporationStanley Black & Decker

Voxdale BVBA

2014 Mentor Graphics Corporation,all rights reserved. This document contains

information that is proprietary to MentorGraphics Corporation and may be duplicated

in whole or in part by the original recipientfor internal business purposes only, providedthat this entire notice appears in all copies. In

accepting this document, the recipient agreesto make every reasonable effort to prevent

unauthorized use of this information. All trademarks mentioned in this publication are

the trademarks of their respective owners.

Roland Feldhinkel, General ManagerMechanical Analysis Division, Mentor Graphics


3/31


NewsPROFILE:Roland Feldhinkel

NewsProfle:

Roland FeldhinkelU2U Events Round-UpTheseus-FE CouplerLinkA1 Racing are WorldChampionsNew Release:FloTHERM V10.0

0 New Release:MicReD IndustrialPower Tester 1500A

Contents

Introducing the new General Managerof the Mentor Graphics Mechanical

Analysis Division.

The recent promotion of Roland Feldhinkelfrom Product Line Director to GeneralManager of the Mechanical Analysis Divisionis a milestone in the history of the Division.Roland brings a diverse and eclecticrsum to the role, he started out as apilot of military helicopters and a Captainin the German army, (where he obtained aMasters degree in Aerospace Engineeringfrom the German Military Academy) andended his time in the German army as aninstructor for new recruits.

When he left the army, Roland set up anengineering consulting company, SolidTeam,

in northern Germany with a universitycolleague, offering structural analysis(FEM) services for aerospace companies,using an early version of Cosmos (nowpart of the SolidWorks suite of products). The duo quickly realized that becomingresellers of the Cosmos product linein Germany would be bene cial asa result of the wave of FEM analysishappening in the early to mid 1990s. In1996 Roland met Alexander Sobachkinfrom Moscow at the spring CeBITComputer Expo Fair in Germany. Alexander had a radical CFD code that

needed commercializing. In the sameyear, SolidTeam became resellersof the edgling 3D CAD product,SolidWorks and were one of the

rst distributors in Germany for theproduct that subsequently took theCAD market by storm.

SolidTeam generated enough revenue fromreselling SolidWorks to fund a businesspartnership with Alexanders team inMoscow to allow for the development ofa new type of nite volume, immersedboundary, cartesian mesh CFD softwarethat had originated in the Russianaerospace sector in the late 1980s (seeFigure 1). Roland and Alexander identi edthe opportunity to embed the CFDtechnology inside the SolidWorks MCADproduct to offer both high productivity aswell as upfront design bene ts to bothdesigners and engineers who had neverused CFD before. In 1998 they took anextraordinary risk to go it alone andfounded Nika GmbH, targeting the CFDmarket with a new type of CFD they calledEngineering Fluid Dynamics. The product

morphed into several CAD variants overthe next ve to six years, including what

became the foundational technology forSolidWorks Flow Simulation and all thevariants of FloEFD that Mentor Graphicssells today. It is also the enabling technologyfor FloTHERM XT.

Rolands experience includes all theunpredictability of establishing and runninga start-up business: managing start-upventure capitalists; living hand-to-mouthduring cash ow dif culties; controllingexceptional high growth scenarios (NikaGmbH grew 200% between 2004 and2006); creating and developing a reseller

and direct sales channel network; andthen overseeing the acquisition processwith potential purchasers. Nika GmbH wasacquired by Flomerics Ltd in 2006 andthen in 2008 Mentor Graphics acquiredFlomerics. Through both dislocationsRoland led the transfer of both FloEFDand FloTHERM/FloVENT technologies intoone product strategy and roadmap. Since2008, ever closer EDA integrations withhis product line inside Mentor Graphics,ultimately culminated in the release ofFloTHERM XT last year.

Throughout Rolands 25 year CFD/CAEcareer he has maintained a strong visionfor design-focused engineering productsclosely coupled with CAD. Having lived andworked in the UK, U S and Germany, he builta strong family culture at Nika that stillexists today.

Time with his two sons, a love of classicalmusic and literature lls his spare time.With Roland we will have continuity of ourroadmap, a strong passion for technologyinnovation and an acceleration of innovation

inside the Mechanical Analysis Division .

Development of AS3DKey Technologies in Russia

AeroShape-3D v.1.0

First license of AS3D soldDevelopmentof CADintegrationtechnologies

NIKAfounded

NIKA acquiredby Flomerics

Flomericsacquiredby MentorGraphics

FloWorks releasedat SolidWorks World 99

1988 - 1990 1991 1992 1993 1994 1995 1996

Range of CADembedded &local languageversions

1997 1998 1999 2000 2001 - 06 2008 2009 - 14

First meeting between Alexander Sobachkin,Roland Feldhinkel andFrank Will of SolidTeamat CeBIT 96

At CeBIT 97 the RussianMinistry of Science notesAS3D as one of ten advancedRussian software projects

24

Engineering Edge16 Facebook Datacenter

Design Thermal Ef ciency in Datacenter Server

Design

21 Bromley Post-SochiUpdate

What's next for Bromley Technologies

22 India AeronauticalDevelopment Agency Designing Modern Aircraft underStringent Regulations using Flowmaster

24 Voxdale & Flanders'DRIVE Deliver

AutomotiveInnovation

Electric Powertrain Innovation

Technology &Knowledge Bank13 Power Electronics

Case Study In uence of Power Cycling Strategy on

IGBT Lifetime

28 1D Thermo-FluidSimulations in Real-Time Environments

The Product of the Collaboration between EnginSoft and Mentor Graphics

32 How To Guide: Hot Lumens in Situ from FloEFD's

unique LED module

34 Engineering Apps Phone and Tablet App Review

52 FloEFD for AirCyclone Simulation

48

16

Regular Features0 Ask the CSD Expert

Mentor IDEAS

1 InterviewBill Maltz, CEO of Electronics CoolingSolutions, Inc.

4 Geek HubLondon Skyscraper Scorching Effect

8 Brownian MotionThe random musings of a Fluid Dynamicist

30 Biomass FurnaceUpgrade by CofelyFabricom E&E

A Simulation Driven Approach

36 Getting Heat Out! Mercury Systems Cooling the Next

Generation of Military ComputingProducts

38 Seiko Epson Empowering Engineers since 1989

42 Electronic CoolingSolutions, Inc.Comparing Tablet Natural ConvectionCooling Ef ciency

46 Practical Fan Curvesin Datacenter

Simulations48 More Power! Stanley Black & Decker Power Tool

Designs

50 D2T & FlowmasterPartner Up

Co-Simulation Capabilities from xMOD


4/31


NewsEvents UpdateGlobal U2Us a Success

he Mentor Graphics Mechanicalam enjoyed a busy end to 2013 withwhirlwind trip of U2Us across threentinents.

oduct Managers Robin Bornoff, Ian Clark,ndras Vass-Varnai, and Mike Croegaertured North America, Japan, Germany,ance, India, China, Taiwan, South Korea,

nd Singapore in November and Decembereeting Customers and showcasing newvelopment in the Mechanical products.

There is nothing that compares toscussing real-life challenges with ourustomers. At every U2U we see a newplication for our products and meeters who are pushing the boundaries of

engineering simulation. This is all feedbackwe can use as we continue to improveproducts", said Mike Croegaert, who seestremendous value in these events for bothMechanical Analysis and its Customers.

Ivo Weinhold, our User ExperienceManager also conducted face-to-face userexperience testing at many of the events.He questioned over 50 users across severallocations to gain access to informationthat will help to improve the interface anduser experience of upcoming releases ofFloTHERM , FloEFD and Flowmaster .

Customers attending these meetingsextended to some 1,600 attendees andcovered a diverse range of industries

including Lighting, Power Generation, Automotive, Marine, Consumer Electronicsand even Meteorology. We will beshowcasing some of these interestingprojects in upcoming copies ofEngineering Edge.

Each event featured a product updatesession followed by User stories from bothend-users and distributors of Mechanicalproducts. In 2014 there will be U2U eventsin North America, China, India, Germany,

Taiwan, Japan and Korea. These againpromise a full agenda of technology anduser application stories. Dates and venueswill be announced on our website and viaour e-News bulletins.

New Flowmaster - Link for THESEUS-FE Couplerentor and P+Z Engineering have ang history of co-operation, with 2013arking an important milestone in

ur relationship when P+Z became anpenDoor development part ner.

he need for a formal relationship wasiven by P+Zs desire to create a newpability for its THESEUS-FE Thermalmulation software product, speci callylowing users to co-simulate withowmaster. The new FlowmasterLink

is an important element of the new THESEUS-FE Coupler module of Version4.4, as it provides users with an easy tomanage solution for co-simulation betweengeometrically complex 3D Finite Element(FE) and CFD models, and 1D thermo-

uid system networks. The THESEUS-FECoupler can control data communicationand synchronization between two ormore solvers, and it currently supportsFlowmaster, OpenFOAM and another 3DCFD tool as exchange partners.

A typical automotive application for the THESEUS-FE Coupler is where vehiclecabin inlet conditions ( i.e. air temperatureand mass owrate) are supplied by aFlowmaster HVAC network to a 3D CFDcabin model. Heat transfer and humidity forthe solid elements within the cabin model(i.e. manikin physiology and cabin structure)are modeled by THESEUS-FE, and theresulting change in the 3D cabin CFD modelis then fed back as the re-circulated airboundary conditions to the Flowmaster 1Dmodel.

A1 RacingJourney to a ChampionshipIn 2009, eight teenagestudents from schoolsin Victoria, Australia,began an exciting journeythat would lead them tomeet F1 boss, BernieEccelstone as his VIPguests at the American F1Grand Prix as winners ofthe prestigious Formula 1in Schools Program.

The program offers away to learn for Science,

Technology, Engineering,and Mathematics (STEM)subjects in a way that isexciting, to encouragemore students into engineering careers.It requires a team of students to design,develop, manufacture, market, manage,

track. said Ben Marshall, 15, A1 RacingDevelopment Engineer.

The collaboration was an important meansof giving the young students an insightinto the industry, offering them valuableexperience of working with engineeringexperts. It also taught them valuable lifeskills such as presentation and publicspeaking, which gave them a massiveadvantage over other students going intouniversity and into employment. In Texas,we were pitted against 300 students in 38teams, from 22 countries around the world.With the CFD knowledge and help, our carwas able to achieve the fastest time at theWorld Championship, an amazing 1.043seconds. continued Ben Marshall.

The anxious team waited nervouslyduring the awards ceremony at the WorldFinals. It was worth the wait, as they werepresented not only with the award for theFastest Car, but also with the 2013 WorldChampionship.

The World Championship award issupported by F1 Management and theCity University of London. A1 Racingwas invited to the USA Grand Prix at theCircuit of the Americas, and as the WorldChampions were given VIP access to theFormula 1 paddock and pits where t heywere presented with a trophy by BernieEccelstone. It was the best feeling to knowthat after so many years of hard work, wehad achieved our dreams. What we hadworked for every day and night, we hadachieved. At times things were stressful, butit was all worth it. We had a trip of a lifetime.None of it would have been possible withoutthe mentoring and support given to us bypeople throughout the world. We cannot

thank you enough for being involved andhelping a group of kids reach their dreams,Ben Marshall.

as well as running numerous diagnosticsand simulations of their model in MentorGraphics FloEFD CFDsoftware. Their weeklysessions ran through untilthe week of the WorldFinals, held in Texas lastyear and were vital to thesuccess of the team. Thesesimulations were imperativeto the success of our car,and the overall successof the team. Through over500 CFD images we wereable to analyze all 23 designstages of the car, ensuringeach detail was optimizedto peak performance on the

and race a miniature Formula 1 car. Theteams are judged on all these areas butmost importantly, and accounting for a thirdof the overall score, they must race a modelcar that they have created down a 20 metertrack at over 80K/h to a race time of justover a second.

A1 Racing was formed from anamalgamation of two rival national runner-upteams. As high school students they hadlittle experience in areas regarding advancedmathematics and aerodynamics and sosought professional help by contactingMentor Graphics. Boris Marovic, AutomotiveIndustry Manager, offered guidance andsupport to A1 Racing by offering tuitionin Computational Fluid Dynamics (CFD)

Figure 1. A1 Racing Team (from left to right): Dyaln Sexton,Ben Marshall, Beau Gieskiens, Luke Merideth, Sam Young,Jacqui Cunningham.

Figure 2. A rendered illustration and simulation results of the nal race car design.

Figure 3. A1 Racing team with Bernie Eccelstone at the USAGrand Prix in Texas


5/31


NewsNew Release:FloTHERMV10.0

he latest release of FloTHERM V10.0,nnounced in November 2013, is agni cant user experience enhancementr longterm FloTHERM users, withveral functionality and usability

nhancements that continues theoduct lines 25 year history of technicaloduct innovation. FloTHERM V10.0corporated over 40 ideas collectedom Mentor Graphics IDEAS website,here we solicit requests on software

nhancements and features that Mentorustomers recommended and voted on

the most popular.

his next-generation FloTHERM productatures a new native Windows graphicaler interface (GUI) to handle pre-ocessing and large models with ease,rgeting todays most advanced electronicsigns. As well as the new Windows-mpliant GUI, FloTHERM also features a

arallel CFD solver and ef cient handling ofassive models for pre-processing, solving

nd post-processing.

etails of FloTHERM V10.0 productnhancements:

The new Windows-compliant userinterface can handle models withthousands of objects. A query-basedsearch function, together with datacolumns integrated with the modelnode tree, provides critical modelchecking capabilities to de ne data,identify errors and to further enhancethe FloTHERM user experience.

A new parallel solver provides scalabilityand fast performance for multi-processors. This capability was the No.

1 requested functionality from theMentor customer-feedback portal, andmakes this FloTHERM release onaverage two to three times, and up to14 times, faster than previous versions.

The addition of new modeling objectsto represent racks of equipment anddatacenter cooling devices enables

Figure 1. New FloTHERM GUI

As one of the rst FloTHERM users, Ive always been happy with the speed of FloTHERMs solver,specially when handling large models. Nevertheless, improvements are always welcome when

we talk hours of CPU time and hence, I was pleasantly surprised when I tested this new version ofloTHERM. The scale-up that has been achieved was at least a factor of four on my computer withdual processor, reducing the solution time from six hours to one-and-a-half for a speci c project

equiring two million cells."emens Lasance, Philips Research Emeritus

Figure 2. Typical Parallel SolverSpeed-up

data for nite element analysis (FEA) witha wide range of popular structuralsimulation programs thereby enablingusers to conduct multi-disciplinaryanalyses.

The extended FloEDA Bridge offersthe ability to measure x and y distancesbetween object edges, corners, orcenters. There is also the capabilityto move or deactivate components.Deactivated components will be retainedbut ignored from any subsequentsolutions.

Figure 3. Extended Datacenter Applications in FloTHERM V10.0

FloTHERM V10 FloSCRIPT functionalityprovides an action log le recordingall commands issued in Project Manageror Drawing Board. Files are saved in

XML format, human readable, and areamenable to le editing and leauthoring tools.FloSCRIPT les are easily recorded,and XML based and can be playedback. Ideal for considerationfor automated work ows.

Figure 4. Typical CFD-FEA Analysis inside MpCCI

For more information, contact your MentorGraphics sales associate or visit thecompany website at:mentor.com/products/mechanical/products/

otherm

"Our market-leading FloTHERM technology was established 25years ago and clearly demonstrates our ongoing commitment tocustomers by continuing to innovate and provide solutions thatthey need and want. By providing an intuitive Windows-basedGUI and advanced features, we are delivering a dynamic solutionthat will increase user productivity and enable the developmentof innovative products. This is truly the next generation of theworlds most popular electronics cooling solution."Roland Feldhinkel, General Manager, Mentor Graphics Mechanical Analysis Division

Figure 5. Extended FloEDA Bridge in F loTHERM V10.0

users to simulate and optimizeeverything from a chip to an entire room.

The transient thermostatic controlmodeling functionality allows modelinputs to be varied in time and asa function of the temperature duringa transient simulation. A key bene tfrom this feature is the ability to reducecomponent power dissipation, eitherthe components own temperature,or those from external stimuli.

Customers involved with thermostaticcontrol systems, such as consumerelectronics, computing and hand-

held telephony and tablets nd this newfunctionality helpful.

The FloTHERM product now alsosupports FEA interfacing through amesh-based parallel code couplinginterface (MpCCI) bridge developedat the Fraunhofer Institute SCAI. Nowengineers can export CFD analysis


6/31

0 mentor.com/mechanical mentor.com/mechanical 11

NewsNew Release:MicReD IndustrialPower Tester 1500A

0 mentor.com/mechanical

ntroducing the new MicReD Industrial Power Tester 1500A forpower cycling and thermal testing ofelectronics components to simulate

and measure lifetime performance. TheMicReD Industrial Power Tester 1500Atests the reliability of power electroniccomponents that are increasingly

used in industries such as automotiveand transportation including hybridand electrical vehicles and trains,power generation and converters, andrenewable energy applications such aswind turbines.

It is the only commercially available thermaltesting product that combines both powercycling and thermal transient measurementswith structure function analysis whileproviding data for real-time failure-causediagnostics.

Power electronics components are usedfor applications in which electrical energyis generated, converted, or controlled andwhere very high reliability is required duringmany years of constant operation. Thisnew product is built for industrial electronicmanufacturers to test reliability by examiningthe thermally induced degradations withinthe module stack-up. Both power cyclingand thermal transient measurements areconducted on the MicReD Power Tester1500A, without needing to remove thecomponents from the test environment.

A technician or engineer is able to see thefailure as it progresses and determine th eexact time/cycle and cause.

Reliability is a prime concern in manyindustries that use high-power electronics,so accelerated testing of these modulesthrough a lifetime of cycles is a must for thecomponent supplier, the system s upplier,and the OEM. The MicReD Power Tester1500A can power modules through tens ofthousandspotentially millionsof cycleswhile providing real-time failure-in-progressdata for diagnostics. This signi cantly

Ireduces test and lab diagnosis time andeliminates the need for post-mortem ordestructive failure analysis. Commonthermally-induced mechanical failuresthat the Power Tester 1500A analyzesin real-time include die-attach wire bondseparations, die and package stack-updelamination and cracks, and solder

fatigue.

The ability to pinpoint and quantifydegradation in the thermal stack forall semiconductor devices duringdevelopment will greatly assist in thedevelopment of cost-optimized packagingsolutions currently hampered by package-reliability concerns, said Mark Johnson,Professor of Advanced Power Conversion,Faculty of Engineering, University ofNottingham. Mentors Power Tester1500A should be an invaluable tool forinvestigating thermal path degradation inall types of power modules.

The MicReD Power Tester 1500Ais based on the Mentor Graphics

T3Ster advanced thermal tester usedin industries worldwide for accuratethermal characterization of semiconductordevice packages and LEDs. The Power

Tester 1500A is the rst product in theMicReD Industrial line and it provides fullyautomated power cycling and testing (boththermal and electrical measurement) ofpower modules to provide comprehensive

data for failure cause assessment. Thisenables organizations to make productimprovements for reliability and extendedperformance. MicReD Industrial productsincorporate the laboratory-level accuracy

"Mentors Power Tester1500A should be aninvaluable tool forinvestigating thermalpath degradation in alltypes of power module."Mark Johnson, Professor of AdvancedPower Conversion, Faculty of Engineering,University of NottinghamFigure 1. Mentor Graphics Launches MicReD In dustrial Power Tester 1500A for

Power Cycle Testing of Electronic Components


7/31


Inuence of PowerCycling Strategy on IGBTLifetime - A Case Study

endors are working hardto increase the maximumpower level and current loadcapability of IGBT and other

power devices, while still maintaininghigh quality and reliability. Innovationhas brought new technologies suchas ceramic substrates with improvedthermal conductivity, ribbon bonding toreplace thick bond wires, and solderlessdie-attach technologies to enhance thecycling capability of the modules.

Power modules are also being designedand manufactured by end-users becausethe chips, the required direct bond copper(DBC) substrate, and a variety of differentdie-attach materials are all available onthe open market. This offers increased

exibility in terms of mechanical design,however it raises severe thermal andreliability challenges as their end use istypically where high reliability is criticalsuch as hybrid and electric vehicles. High

junction temperatures and high temperature

gradients during operation inducemechanical stress especially at contactingsurfaces between materials with differentcoef cient of thermal expansion, which maylead to the degradation or the completefailure of these components. To avoidpremature failures, proper thermal designand material selection is n ecessary.

Mentors Power Tester 1500A is designedto automate the process of qualifying thereliability of parts to make a good estimatefor the power modules lifetime duringservice, and to identify weaknesses thatcan be removed during developmentthereby increasing reliability and lifetime.

This case study details the applicationof the Power Tester with four mediumpower IGBT modules containing two

Figure 1. Structure functions of Sample 0 corresponding to control measurements at various time points

By John Parry, Industry Manager, Mentor Graphics

half bridges, demonstrating the rich dataobtained from automated power cycling ofthe components. This article is abstractedfrom the two technical papers given in thereferences. [1,2]

The modules were xed to the liquid cooledcold plate integrated into the Power Testerwith a high-conductivity thermal pad tominimize the interfacial thermal resistance.

The coldplate temperature was maintainedat 25C throughout the whole experimentusing a refrigerated circulator controlled bythe Power Tester. The gates of the deviceswere connected to their drains (the so-calledmagni ed diode setup) with each half bridgepowered using a separate driver circuit.

Two current sources were connected toeach half bridge. A high-current source th atcan be switched on and off very fast wasused to apply stepwise power changes tothe devices. A low current s ource providedcontinuous biasing of the IGBT allowingthe device temperature to be measured

when heating, connecting it to a separatemeasurement channel of the Power Tester.

An initial set of tests on four samples wasconducted using constant heating andcooling times. Heating and cooling timeswere selected to give an ini tial temperatureswing of 100C, for a power of ~200Wwith 3s heating and 10s cooling. This mostclosely mimics the application environment,where degradation of the thermal structureresults in a higher junction temperatureleading to accelerated aging. Of the fourdevices, Sample 3 failed shortly after 10,000cycles, signi cantly earlier than the others.Samples 0, 1 and 2 lasted longer, failingafter 40,660, 41,476 and 43,489 powercycles respectively. Figure 1 illustrates thestructure functions generated from thethermal transients measured on Sample 0after every 5,000 cycles. The at region at0.08 Ws/K corresponds to the die-attach.It can be seen that the structure is stableuntil 15,000 cycles, but after that point

V

Power Electronics

the T3Ster product in robust machinesr operators to use inside manufacturingcilities.

Our MicReD Power Tester 1500A servese growing demand for power electronics

omponents that need to perform underxtreme conditions with high reliability,ated Roland Feldhinkel, General ManagerMentor Graphics Mechanical Analysis

ivision. Were leveraging our expertisethermal characterization and testing to

eliver a product for industrial applications,here we see great potential from electricehicles and railway systems to renewablenergy products.

he MicReD Industrial Power Tester 1500Aan perform power cycling tests of metal-xide semiconductor eld-effect transistorsMOSFETs), insulated-gate bipolaransistors (IGBTs) and power diodes. The

MicReD Power Tester 1500A provides aser-friendly touch-screen interface andan record a broad range of informationuring test, such as current, voltage and diemperature sensing, and detailed structurenction analysis to record changes in the

ackages thermal structure. This makes itn ideal platform for package developmentnd quality checking of incoming partsefore production.

he MicReD Industrial Power Tester 1500Aovides several key bene ts:

Continuous power cycling until failure,which saves time because thecomponent doesnt need to beremoved, taken for lab testing then backto tester for more cycles.Shortens total testing time by up to 10times.Leverages MicReD's industry-proven T3Ster technology with laboratory-precision accuracy.Enables multiple samples to be testedconcurrently.Different powering strategies (constantpower on/off time, constant case

temperature swing, and constant junction temperature rise) can beapplied during operation.

Provides real-time structure functiondiagnostics to show failure in progress,number of cycles, and failure cause.

Eliminates the need for lab post-mortem(x-ray, ultrasonic, visual) or destructivefailure analysis.

Outputstage

Powerstage

Powergrid

3*500A power source3 phase equipment(22kW)

UPS

Highcurrent

COLDPLATE

T3Ster

DAQ

Meas.

SAFETY BOX

INDUSTRIALPC

Ctrl andanalysis

Thermostate,Process water

DUT

Power grid

gure 2. User-friendly touch-screen interface Figure 3. Device creation in interface

Figure 4. Placing devices on cold plate in interface

Figure 5. Power Tester Con guration

Features touch-screen setup andcontrols enabling use by both specialistsand production personnel.

For more information visit:mentor.com/powertester-1500a


8/31


Power ElectronicsAcross all semiconductor devices the ability to pinpoint and quantify degradationn the thermal stack during development will greatly assist the development ofost-optimized packaging solutions that are currently hampered by packageeliability concerns."ark Johnson, Professor of Advanced Power Conversion, Faculty of Engineering, University of Nottingham

e degradation of the die-attach can beearly noticed as its resistance increasesntinuously until the device fails. Again,e immediate cause of the device failure is

nknown, but we found that a short circuitformed between the gate and the emitter

nd burnt spots could be seen on thehip surface.

second set of tests were performedn an identical set of samples using thefferent powering strategies supported bye Power Tester. In this case we kept the

urrent constant for IGBT1, the heatingower constant for IGBT2, and the junctionmperature change constant for IGBT3.

o ensure a fair comparison, the settingsere chosen to give the same initial junctionmperature rise for all components, withheating and 17s cooling, and ~240W

itial heating per device chosen for the test.he whole heating and cooling transientmeasured for each device in all cycles,ith the following electrical and thermalarameters monitored continuously by theower Tester:

Device voltage with heating currentturned on, V onHeating current applied in the last cycle,ICyclePower step, PDevice voltage after heating currentturned off, V hotDevice voltage before heating currentturned on, V coldHighest junction temperature during thelast power cycle, T hotLowest junction temperature during thelast power cycle, T coldTemperature swing in the last cycle, T Temperature change normalized by theheating power, T/P

dditionally, the full length thermal transientom powered on steady state to poweredf steady state was measured after 250

ycles using a 10A heating current, toeate structure functions to investigate anygradation in the thermal stack. Again, the

xperiment was continued until the failure ofl IGBTs.

s expected, IGBT1 failed rst, as thereno regulation of the supplied power ase part degrades. Interestingly, it showed

no degradation in the thermal structure asshown in Figure 2.

In order to nd the cause of the devicefailure we have to examine evolution ofdevice voltage during the experiment. InFigure 3 the forward voltage of IGBT1 atheating current level can be seen as functionof elapsed power cycles. In the rst threethousand cycles a decreasing tendency canbe seen.

This initial change caused by the slow

Figure 2. Change of the structure function of IGBT1 during the power cycling

change of the average device temperaturethat decreased by almost 5C. Despitethe negative temperature dependenceof the device voltage at low currents,at high current levels th e temperaturedependence of the forward voltage becomepositive. After about 35,000 cycles thistendency changed and the voltage startedto increase slowly. This was followed bystepwise changes in the device voltagewhile the increasing tendency continuouslyaccelerated until the failure of th e device. Asthe structure did not change the increasing

Figure 3. Forward voltage of IGBT1 at heating current level as function of applied power cycles

voltage can be attributed to thedegradation of the bond wires. Thisalso gives an interpretation to thestepwise changes of the voltagewhen a bond wire nally detaches.

The increasing heights of these stepsare caused by the increasing changein the parallel resistance sum of the

bond wire thermal resistance as thenumber of bond wires decreases.If we use constant current strategy,the crack of a bond wire increases thecurrent density in the remaining bonds andaccelerates aging.

Figure 4 shows the same type of curvecorresponding to IGBT3. Here theincreasing tendency of the device voltagestarts even earlier but due to the regulationto keep the junction temperature constant,the heating current was proportionallydecreased. The decrease in current reducedthe load on the bonds and increased themeasured lifetime.

To conclude, the two sets of experimentswere conducted which showed differentfailure modes, illustrating how differentpowering strategies, and possibly electricalsetup, can in uence failure mode. The rstset of measurements at a constant cycletime, that most closely re ects operationaluse, veri ed that the Power Tester is ableto detect immediately the appearance ofdegradation within the devices structure,including the die-attach and othercompromised layers.

The second experiment clearly identi eddegradation of the bond wires as theforward voltage of the device was observedto increase stepwise. While with thesepowering options, (current constant,constant heating power, and constanttemperature rise), the thermal structuredid not change for any of the samplestested. Due to the low number of sampleswe have to be conservative in formulatingconclusions. However the results warnsus that the measurement results candiffer depending on the cycling st rategy,and lifetime predictions based on certainstrategies can overestimate the real lifetimeof power devices.

References:[1]. Zoltan Sarkany, Andras Vass-Varnai,Marta Rencz (2013) Investigation of die-attach degradation using power cyclingtests, Proceedings of 15th IEEE EPTC, pp780 784, Singapore.[2]. Zoltan Sarkany, Andras Vass-Varnai,Sandor Laky, Marta Rencz (2014) Thermal

Transient Analysis of Semiconductor DeviceDegradation in Power Cycling Reliability

Tests with Variable Control Strategies,Proceedings of SEMI-THERM 30, pp. 236-241, San Jose CA.

3.76

3.78

3.8

3.82

3.84

3.86

3.88

3.9

0 10000 20000 30000 40000 50000

Cycle No. [-]

V o n

[ V ]

3.45

3.46

3.47

3.48

3.49

3.5

3.51

3.52

3.53

3.54

3.55

0 10000 20000 30000 40000 50000

V o n

[ V ]

Cycle No. [-]

I . .

Figure 4. Forward voltage of IGBT3 at heating current level as function of applied power cycles


9/31


arge-scale datacenters consumemegawatts in power and costhundreds of millions of dollars toequip. Reducing the energy and

cost footprint of servers can thereforehave substantial impact.

Web, Grid, and Cloud Servers in particularcan be hard to optimize, since they areexpected to operate under a wide rangeof workloads. For its rst datacenterin Prineville, Oregon, Facebook setout to signi cantly improve its poweref ciency, cost, reliability, serviceability,and environmental footprint. To this end,many dimensions of the datacenter andservers were redesigned, using a holisticapproach. This article is abstracted fromthe Facebook paper High-ef ciency ServerDesign, which was presented at the 2011

ACM Conference on Supercomputing, andfocuses on this server design, combiningaspects of power, motherboard, thermal,and mechanical design. In this articlewe have looked at the thermal aspectsin isolation. In the full paper, Facebookcalculated and con rmed experimentallythat its custom-designed servers canreduce power consumption across theentire load spectrum while at the same timelower acquisition and maintenance costs.

The design does not reduce the serversperformance or portability, which wouldotherwise limit its applicability. Importantly,the server design has been made availableto the open source community via theOpen Compute Project, a rapidly growingcommunity of engineers around the worldwhose mission is to design and enable

L

mentor.com/mechanical 176 mentor.com/mechanical

Datacenter

Thermal Efciency:Facebooks

Datacenter Server

DesignBy John Parry, Industry Manager, Mentor Graphics


10/31


e delivery of the most ef cient server,orage and datacenter hardware designsr scalable computing.In the past decade,e have witnessed a fundamental changepersonal computing. Many of the modern

omputer uses such as networking andommunicating; searching; creating andonsuming media; shopping; and gamingcreasingly rely on remote servers for their

xecution.

he computation and storage burdens ofese applications has largely shifted from

ersonal computers to the datacentersservice providers such as Amazon,

acebook, Google, and Microsoft. Theseoviders can thus offer higher-quality andrger-scale services, such as the abilitysearch virtually the entire internet in a

action of a second. It also lets providersene t from the economies of scale andcrease the ef ciency of their services.

s one of these service providers, Facebookased datacenters and lled them with

ommodity servers. This choice madense at small to medium scale, while thelative energy cost is still small and thelative cost of customization outweighse potential bene ts. As the Facebookte grew to become one of the worldsrgest, with a corresponding growth in

omputational requirements, they startedxploring alternative, more ef cient designsr both servers and datacenters.

hermal Designhe goal of server thermal design is to coolown the hot components to their operatingmperatures with a minimal expenditureenergy and component cost. The typical

echanism used to cool servers at theatacenter level is to cool air at large scalend push it through the servers using their

ternal fans. The cool air picks up heatom the server components, exits from therver outlet, and is then pushed back toe atmosphere or chilled and recirculated.

ore ef cient cooling is achieved with airontainment in aisles, with the front (or inlet),de of the server facing the cold aisle ande back facing the hot aisle. Yet anotherchnique to improve cooling ef ciency is toeate an air-pressure differential betweene aisles using large datacenter fans. Inis case the speci c design goal was to

e able to cool the upcoming datacenterithout chilling the outside air almostear round by allowing effective serverooling even with relatively high inlet airmperature and humidity. To achieve thisoal, a more effective design was needed

Figure 1. FloTHERM isometric view of thermal design shows chassis, motherboard (with dual processorsand memory slots side-by-side), fans, and the hard-disk drive (HDD) behind the PSU. The temperature rangehere assumes an inlet temperature of 27C. The air duct on top is elided for visualization purposes.

for heat transfer than currently used in thecommodity servers.

Improving air ow through the server is akey element here: when internal servercomponents impede air ow, more coolingenergy is expended (for example, by fasterfans, cooler inlet air, or higher air pressure).One technique by which improved air owis achieved in the chassis is to widenthe motherboard and spread the hotcomponents side by side, not behind eachother. The hottest componentsprocessorsand memorywere moved to receive thecoldest air rst, by locating them closer tothe air inlet than in the typical back-mountedmotherboard.

Another modi ed dimension was the serverheight: given a relatively constant rackheight (for servicing purposes), a taller serverreduces cooling energy but also the rackscomputational density. Calculations foundthat the optimal server height to maximizethe compute-capacity per cooling-energyratio to be the uncommon 1.5U heightwith large-surface-area heat sinks. Thisheight also allows for an air duct that sitson top of the motherboard and surgicallydirects air ow to the thermal components inparallel heat tracks, reducing leaks and airrecirculation inside the chassis. Obstructionsto air ow are kept to a minimum, decreasingthe number of fans required to push the airout (Figure 1).

And since the high-ef ciency PSUgenerates less than 20W of waste heatunder load, the HDD remains well withinspeci ed temperature operating range evenbehind the PSU. Contrast this with typicalserver designs that locate the HDD in thefront of the chassis to meet its coolingrequirements. Also reduced is the amountof air ow required through the system tokeep it coolup to half the volume owratecompared to standard 1U servers, for thesame inlet to-outlet temperature difference(Figure 2).

This low requirement, combined with smartfan-speed controllers, results in fans thatspin at their minimum continuous speednearly year-round, depending on ambienttemperature and workload.

An additional advantage of this low speed,continuous operation is a longer expectedfan lifetime compared to the typicalfans start-stop cycles, leading to overallimproved server reliability. It also naturallytranslates to lower power and operatingcosts for server coolingapproximately1% of the total server powercomparedto the more typical 10% in commodityservers. Somewhat surprisingly, even theCAPEX of the servers cooling componentsalone is about 40~60% lower than a typicalserver, depending on OEM componentpricing. The two main reasons for thisimprovement are the use of thinner fans(owing to the reduced air ow) and simplerheatsinks without a heat pipe (owing tothe larger surface area). Closing the cycle,these ef ciency gains carry forward tothe datacenter level as well. The serveris capable of working reliably at air inlettemperatures of 35C and a relativehumidity of 90%, exceeding the most liberal

ASHRAE recommendations for datacenter

equipment. In practice, this allowsFacebooks datacenter to be cooled almostexclusively on free (outside) air, relying oninfrequent evaporative cooling instead ofchillers only on particularly hot days.

Figure 2. FloTHERM CFD simulation of air ow speed at minimum continuous fan speed

MethodologyFacebook have evaluated the power,thermal and performance properties ofa prototype of the new design againsttwo commodity servers. Both commodityservers are a common off-the-shelfproduct from two major OEMs, with dual

Xeon X5650 processors, 12GB DDR3ECC memory, on-board Gigabit Ethernet,and a single 250G SATA HDD in a 1Ustandard con guration. The rst server,Commodity A, is widely deployed inthe leased datacenters for Facebooksmain Web application. The second server,Commodity B, is a three-year-old modelthat was updated to accept the latestgeneration processors. To ensure a faircomparison, the exact same CPUs,DIMMs, and HDD unit are used in turn,moving them from server to server. Theonly differing components between thethree servers were therefore the chassis,motherboard, fans, power supply, andpower source (208V ac/277V ac).

Thermal Ef ciency Thermal ef ciency is another importantelement of the total cost of ownership(TCO), both in terms of cooling energyin the server (fan energy) and in thedatacenter. The thermal design is basedon a spread and unpopulated boardplaced in a 1.5U pitch open chassis,and employs four high-ef ciency custom60 25mm axial fans. In contrast, thecommodity servers use a thermallyshadowed, densely populated 1U chassiswith six off-the-shelf 4025mm fans. Toevaluate the thermal ef ciency, each serverwas placed in a specially-built air owchamber that can isolate and measurethe air ow through the server, expressedin cubic-feet per-minute (CFM). Themeasured CFM value was also con rmedanalytically by measuring the s ervers ACpower and air temperature differencebetween inlet and outlet. The servers areloaded with an arti cial load resembling

Facebooks production power load (around200W, with leakage power at less than10W), while maintaining the constraint thatall components remain within their operatingthermal speci cations. The results for theprototype (Figure 3) show a signi cantimprovement. For a typical 7.5MWdatacenter, this reduced air ow translatesto a reduction of approximately 8~12%of the cooling OPEX. More importantly, itenables free air cooling to be used for thedatacenter.

Conclusions This new server design measurably reduces TCO without reducing performance. Thecustomized server design can:1. Reduce operating and cooling power

(e.g. ef cient power conversions,higher-quality power characteristics,fewer components, thinner and slowerfans, improved air ow).

2. Lower the acquisition cost and serverweight (e.g. fewer and simplercomponents, lower density, fewerexpansion options).

3. Cut costs on supporting infrastructure(e.g. no centralized UPS, no PDUs, nochillers).

4. Increase overall reliability (e.g. fewerand simpler components, distributedand redundant batteries, smooth normal /backup transitions, staggered HDDstartup, slower fans).

5. Improve serviceability (e.g. all-frontservice access, simpler cablemanagement, no extraneous plasticsor covers).

At large scale, this design translates tosubstantial savings. Facebook calculate thatover a three year period, these servers alonewill deliver at least 19% more throughput,cost approximately 10% less, and use

several tons less raw materials to build thana comparable datacenter of the same powerbudget, populated with commodity servers.When matched with a correspondingdatacenter design (including all aspects ofcooling, power distribution, backup power,and rack design), the power savings growto 38% and the cost savings to 24%, witha corresponding power usage effectiveness(PUE) of 1.07.

Reference:[1]. Eitan Frachtenberg, Ali Heydari, HarryLi, Amir Michael, Jacob Na, Avery Nisbet,Pierluigi Sarti, Facebook. High-Ef ciencyServer Design, 2011

Figure 3. Air ow comparison (in CFM) at 200W

Datacenter


11/31


entor Graphics is committedto improving product userexperience. Naturally,listening to user feedback

a vital part of this, so we really valueour ideas for new product featuresnd enhancements. Wit h this in mind,entor IDEAS provides users with theility to provide direct feedback for allechanical Analysis products, allowing

ou to post your suggestions for productmprovements, and vote on other users

ggestions: entor.brightidea.com/mechanical

he suggestions with the most votes areose that our engineering and developmentams will look at rst. While it may not beactical to implement every top vote-getter,will help us to understand your priorities.nderstandably most people want this to a two-way process so we do provideedback on your top suggestions at the

BLOG HOME" link.

ome users may think that categorizinge suggestions on Mentor IDEAS is an art,

wed like to reassure you that there isience behind the statuses.

o here is what you can expect from Mentorraphics on responding to top ideas:

ECEIVED: This is the initial status of anyea. Once the idea garners enough voteshe number of votes may vary from productproduct), a Product Manager will reviewe idea for inclusion into an upcominglease.

RCHIVE: This status is used forousekeeping on the Mentor IDEAS site.is important that ideas with few votese periodically removed so ideas remain

urrent and the site does not become tooumbersome to navigate.

M

Your IDEAS to meetYour Needs

Ask The CSD Expert

OUT OF SCOPE: The idea may havereached a vote threshold, however it isoutside the product roadmap.

UNDER CONSIDERATION: The idea hasreached a vote threshold and is beingevaluated for inclusion in the productrelease.

PLANNED NOT SCHEDULED: This is anew status to indicate that the idea is inthe products longer term planning, but notyet committed to a release. These will bereviewed during the normal planning cyclefor future release assignment.

PLANNED : The idea is scheduled into anupcoming release and the planned releaseis added as a tag to the idea.

IMPLEMENTED: The idea is currentlyavailable in a release and the tag indicatesthe release name. The release is available onSupportNet for download.Finally, we should point out that MentorIDEAS is not a replacement for techni calsupport or SupportNet which can beaccessed through http://supportnet.mentor.com. We look forward to hearing your ideas!

Sports Engineering

Bromley Post-SochiUpdate: You dont learn ifyoure standing still

or athletes, there are fewcomparable stages to theOlympics in terms of eitherprestige or exposure. Like

the Football (Soccer) World Cup orWimbledon, it is a spectacle that drawsin a wide audience, one that maynot engage with any of the events inquestion at any other time. For thoseat the sharp end of the competition,the pressure to deliver upon years ofpreparation is therefore immense.

Having participated at Sochi 2014 as bothathlete and supplier to many of the otherteams in the Skeleton, Kristan Bromleyunderstands the nature of this pressure alltoo well. Post-Sochi, what are the mainlessons learnt? What we do works, saysKristan There were consistent indicationsthat our sleds are performing as weneeded them to. These indications hardnumbers from reams of split time data allow the team at Bromley Technologiesto unpick how their sleds are performingindependently of the idiosyncrasies of eachindividual rider.

That the approach worked wasnt asurprise, the methodical approach takenby Bromley in designing and manufacturingtheir sleds not only required that they

employed cutting edge tools, it alsonecessitated that the approach itself beconsistently scrutinised and questioned.

They werent just learning, they werelearning about learning as they went along.

F

Such an approachis essential in anenvironment whereperformance is underconstant and mercilessscrutiny. Skeleton is asmall community, meaningthat underperformingequipment is not onlydiscovered quickly, itsrapidly communicatedthrough the ranks.

Little wonder that Kristanregards knowledge as one ofBromley Technologies keyassets. However, withoutaction, knowledge is notonly academic but quicklyreaches its sell by date.In order to harness andrefresh it, Bromley integrateadvanced tools andmethods throughout thedesign and manufactureprocess. Laser metrologyof athletes is used inconjunction with FloEFDin order to producesimulations upon whichdesign decisions cancon dently be drawn.

Engineering and competitive experience issystematically and continuously captured inthe design process. Ultimately, this will beharnessed in the sleds that will compete inPyeongChang in 2018.

FloEFD is an essential part of the process.Its not just the accuracy we know fromwind tunnel testing that the results areaccurate its the ability to turn out theseresults quickly on my laptop. This gives methe freedom to virtually tweak and revisedesigns throughout the design and testingphase. (Figure 2)

So much for those at the cutting edge ofsliding sports; what about us mere mortalswho are never likely to go near anOlympic track?

By John Murray, Industry Manager, Mentor Graphics

The answer is the Baseboard (Figure 3).Launched at the Whistler resort in Canadalast year with over 1,000 people trialling itin six days, it returns to the World Ski andSnowboard Festival this year with its owndedicated track.

Serving such diverse markets effectively isonly possible because of this knowledgebased ethos in combination with toolsthat can add value to the design process.Distilling the data from testing, simulationand design down to core concepts meansthat Bromley Technologies can hit verydifferent design requirements withoutcompromising on either.For more information visit:bromleysports.com

Figure 1. Bromley Sports at Whistler Blackcomb, Canada

Figure 3. Bromley Baseboard

Figure 2. Windtunnel results vs. FloEFD simulation data


12/31


Model Based SystemEngineering at theAeronautical DevelopmentAgency

he design of modern aircraftremains an extremely challengingundertaking. A combination ofstringent and often con icting

requirements, expensive equipment withlong lead times and highly co-dependentand integrated systems presentengineers with problems of almostunimaginable complexity. Consider, forexample, how a change to the designof the fuel system has implicationsfor the ight dynamics of the aircraft;its structure; hydraulic systems andpotentially even cooling and avionics.

gure 1. Squadron of Light Combat Aircraft LCA) Tejas Series

y V Krishna Prasad, Deputy Project Director, General Systems

T

These constraints are felt all themore keenly for those designingmodern combat aircraft. Not onlyare the packaging requirementsmore stringent, both the rangeof operating conditions and the

rate at which they are coveredduring operations is much moreextreme.

The Aeronautical Development Agency (ADA) of India wasestablished in 1984 to wrestlewith such challenges duringdevelopment of the Light Combat Aircraft(LCA), an extremely ambitious project aimedat producing an indigenously designed andbuilt aircraft.

The task before the fuel systems group wasa daunting one: while the most obviousrequirement is to ensure reliable fuel feedto the engine, this must be met whileensuring that the stability of the aircraft isntcompromised as tanks begin to empty.Meaning that the fuel tank pressurizationand venting lines must be adequately sizedand routed and the dynamic response of thesystem is well understood across a range ofchallenging mission pro les.In order to meet these challenges, Mentor's

Flowmaster 1D CFD software wasbrought in by ADA as a complementarytool to existing theoretical and physicaltest procedures. Flowmaster allowedthe group to achieve more accuratepredictions of system performance at allstages of the design loop. This increasein accuracy translates to less uncertaintyand consequently the ability to nalize onan appropriate con guration sooner. As aconsequence, the overall design cycle canbe signi cantly shortened.

However, while signi cant, the bene ts ofusing system simulation extended furtherthan allowing the group to accomplish thesame tasks to a higher degree of delity.

A key bene t offered by Flowmasterwas the ability to interrogate the virtualsystem in regions not easily accessible byinstrumentation in physical prototypes. Thisability streamlines the trouble-shooting ofthe snags that inevitably arise during thedesign of such complex systems. Thisin turn further reduces the physical testburden.

These high level bene ts are the result ofincremental improvements at each stageof the design process. For example, anaccurate characterization of the drop tankcircuit enables you to arrive at the pressurerange of the Pressure Transducer intended

Figure 2. Tejas cockpit

Figure 3. Drop Tank characterization in Flowmaster

to be used for monitoring the health ofthe drop tank fuel transfer system. Thisallows component speci cation to begin,not only for the Pressure Transducer itself,but a range of up and downstream systemcomponents.

Perhaps not as obvious, but equallysigni cant, is the fact that the impact ofany mid-design changes can be readilyassessed and accounted for. Where anywork must be revisited, options can beexplored and assessed quickly and withcon dence.

Fourteen squadrons of the LCA, now inproduction as the Tejas (radiance), willultimately enter service with the Indian AirForce (IAF), while a further 40-50 aircraft willenter service with the Navy as part of theircarrier based air arm. Such a commitmentunderlines the capabilities of the Tejas andindicates that the design will continue toevolve as operational demands change.Flowmaster will continue to play a rolethroughout this evolution at ADA.

Designing modernaircraft under stringentregulations for themilitary is a challengethe ADA in Indiaface head-on withFlowmaster

Defense


13/31


14/31


rm effects on the cell packaging due toe cooling liquid meant that the Dry Moduleas our recommendation for the EV Batteryeing designed. (Figure 6)

ased on our work and that of the otheronsortia partners, Flanders DRIVE set up aress and media day in April 2013 showinge new EV Powertrain design inside a

wo-wheel drive Land Rover Evoquesowertrain (Figure 1). It included fourwitched reluctance motors, one poweringach wheel separately, thus eliminatinge need for a differential between wh eels.owering each wheel with an independentotor should also lead to enhanced safety,

peed and handling of the electric vehicle.ornering would also be improved becausee cars outer wheels would spin fasteran the inner wheels as the car turns. Thisturn improves the cars overall torque

ecause the power generated by the

gure 6. Geometry and detailed FloEFD thermalredictions for parts of the Dry Module

Figure 7. Voxdale concept electric vehicle chassis

gure 8. Aquilo concept car battery pack coolingmulation in FloEFD

Figure 9. Aquilo concept car design in PTC Creo with external aerodynamics pressure predictions fromFloEFD Simulations

batteries would be sentto where it is neededmost. The FlandersDRIVE motors also didnot use magnets thuspotentially reducingthe price and helpingto generate additionalpower for the enginethrough a regenerativebraking system. The next step for FlandersDRIVE researchers will be to optimize thepowertrain technology in conjunction withits European Partners which include JaguarLand Rover and Skoda so that they canbe applied to all hybrid and electric vehiclecombinations.

Finally, my colleague at Voxdale, WouterRemmerie, and I decided to see if wecould think through what we had learntfrom the initiative to facilitate some bluesky thinking related to electric vehicleswith novel battery locations, a patented 3Dprinted chassis monocoque concept, andthermal management (via FloEFD), all insidePTC Creo. We called our Project Aquiloand decided to look at a fundamental newconceptual aerodynamic design processwith CFD simulation prototyping relatedto cooling and heating of battery packs,cooling of powertrain (Figure 8), coolingof electronics, and simulation of battery

cooling media, plus lightweight casings fora generic two door EV sports car chassisdesign (Figure 9). The battery casings weredesigned with FloEFD and fabricated in thelaboratory to validate them. Our nal sports car design with novelmonocoque, FloEFD aerodynamic stylingand battery driveline was visualized in PTCCreo (Figure 9). A very interesting exerciseand something that illustrates the power oftools we have available to us insidePTC Creo. For more information visit:http://bit.ly/1kWdC9M

* The Flanders' DRIVE Electric Powertrainproject was a consortium of 12 companiesincluding Arteco, CTS, DANA, EIAElectronics, Imec, Inverto, LMS, NXP, PunchPowertrain, Triphase, Umicore and Voxdale:

andersdrive.be/envoxdale.be/en


15/31


ontrol Systems are beingincreasingly used in complexengineering systems. Wheretesting cannot easily or safely

e done in a real operating environmenten development of these control

ystems requires robust, portable, real-me models of the engineering systemsfacilitate:

Failure mode analysisTrade off studiesDesign validationWhat-if studiesHardware-in-the-loop (HIL) testingIn operation simulators

owmaster is able to accurately model aide range of thermo- uid systems butese simulations are often not able to run

atively in real-time, making them unsuitabler coupling with an HIL environment.owever, Flowmaster can generateesponse-Surfaces that characterize theehavior of its models. Mentor Graphicsollaborated with EnginSoft to implementis capability and the Response-Surfaces

an be exported as C++ code or asMATLAB S-Functions.

hese Response-Surface Models (RSM)ovide robust, portable, real-time modelshich are suitable for Control Systemevelopment, including HIL testing.

dditionally, RSMs can be used in anyodel based environment; they can be

assed to design and operational teamsnd used by non-experts to review thesults of a simulation model analysis and

nderstand a systems behavior more easily.he main application areas for hardware--the-loop simulation are in the design oflectronic Control Units where the controllerconnected to a real-time simulator. Thisovides a way of testing control systems

ver the full range of operating conditionsncluding failure modes) both cost effectivelynd safely.

his article will look at a simpli edutomotive engine cooling system (Figure.

and how a robust, portable, real-timeesponse-Surface Model (RSM) can be

C

generated using Flowmaster. In this examplethe engine is represented by a heat sourcetransmitted into the cooling system via athermal bridge component. The ow passesthrough a heat exchanger and there is abypass line controlled by a set of globevalves that represent the thermostat.

The primary circuit consists of a pump, aheat source, a set of globe valves, a cross-

ow heat exchanger and a pressure sourcethat pressurizes the system. The bypassline includes a globe valve, componentC4, which controls the amount of uid thatpasses through the heat exchanger bymodulating between position 0 and 1. If thevalve is in position 0, then a quantity of ow

delity in each Response-Surface, thenthey can be exported to a Response-Surface Model (C++ code or MATLABS-Functions) suitable for use in a real-timesimulation or as a portable model for otherdesigners or operators.

The product of the collaboration betweenEnginSoft and Mentor Graphics in the latestversion of Flowmaster V7 now allows forrobust, portable, real-time models to begenerated using a Design of Experimentsapproach. These Response-SurfaceModels can be exported as C++ code oras MATLAB S-Functions suitable as thebackend code to a runtime model of thesystem or for use in an HIL environment.

The ability to characterize a systemsbehavior in exported code opens up a widerange of possibilities, such as creating asimple dashboard that allows non-expertusers to understand and predict systemperformances in Flowmaster, inserting thecode into a hardware-in-the-loop logic, orembedding the code into other codes forco-simulations.

Figure 1. Simple automotive cooling system

Figure 2. Experiment input values

Accurate 1DThermo - Fluid Simulationin Real - Time Environments

Technical Article

Here a Latin Square of ten levels isconsidered which generates 100 simulationsi.e. n2 simulations, where n is the number ofdifferent levels. Using four discrete values forthe valve component C4 of 0, 0.3, 0.6 and1 will produce a total of 400 steady statesimulations. More levels, n, leads to moresimulations and means that the Response-Surface has a higher delity but takeslonger to generate.

In uid systems, discrete values (like valveposition) can lead to radically different

ows and a single response-surface wouldbe inaccurate. So, separate Response-Surfaces are created for each combinationof discrete values to increase accuracy ofthe RSM. Once the 400 simulations arecompleted, response surfaces for eachoutput variable can be created and reviewedby the user. The Response-Surfaces aregenerated using Radial Basis Functions(RBF). Flowmaster offers different types ofRBF but will select the best t while offeringthe user the exibility to customize theRBF. The result of applying a RBF to th esimulation results is shown in Figure 3.

The Deviation Details tool provides an

immediate and simple evaluation of thegoodness-of- t of each response surfaceon the basis of its deviation. As shownin Figures 4 and 5 the best responsesurfaces for the ow through the pumpand through the heat exchanger are thosecomputed with Gaussian RBF while the bestResponse-Surface for the temperature inthe primary circuit is the one computed withHardys MultiQuadrics.Once the user is happy with the level of

will pass through the bypass line and if thevalve is in position 1, then the entire owpasses through the heat exchanger.In this study, we need to model the impacton the cooling system of varying pumpspeeds, air ows over the radiator andengine heat outputs. We also want toinclude the effects of various valve positionsfor the bypass line from fully closed to fullyopen.

The following four input parameters arede ned for the network: [Pump Speed] Mixed Flow Pump, C16 [Air Flow] Flow Source, C14 [Engine Heat Output] Heat Flow

Source, C1 [Valve_C4] Valve Opening, C4

The output parameters are de ned as: Top Hose Temperature (Thermal Bridge

C2, Node 2) Pump Flow Rate (Mixed Flow Pump,

C16)

To generate the Response-Surface,Flowmaster runs simulations for this modelover the range of the input parameters.It is not possible to run every possiblecombination of input parameters. Instead,a set of parameters are chosen which givea uniform distribution of combinations. Theuser sets up the input parameters using theDesign of Experiments feature in Figure 2.

Figure 4. Response Surface view

Figure 3. Deviations of Response Surface for the ow rate through the pump

References:[1]. McKay, B., and Rogoyski, E. 1995, LatinSquares of Order 10, Electronic Journal ofCombinatorics, 2(3), 1[2]. Fisher, R., and Yates, F. 1934, The6 x 6 Latin Squares, Proceedings of theCambridge Philosophical Society, 30, 492[3]. Jacobson, M., and Matthews, P. 1996,Generating Uniformly Distributed LatinSquares, Journal of Combinatorial Designs,4(6), 405 6[4]. Buhmann, M. D., Radial BasisFunctions: Theory and Implementations,Cambridge University Press, 2003.[5]. Rippa, S., An algorithm for selectinga good value for the parameter c in radialbasis function interpolation, Adv. in Comp.Math., 11, pp. 193-210.

Figure 5. Deviations of Response Surface for temperature downstream of heat source

y David Hunt, Flowmaster RDI Manager & Product Architect, Mentor Graphics


16/31


ProcessUpgrading a BiomassFurnace: A SimulationDriven ApproachBy D. Hasevoets, Cofely Fabricom E&E & Wouter Remmerie, Patrick Vlieger, Voxdale

ofely Fabricom GDFSUEZ is a leading multi-disciplined engineering,project management and

onstruction organization in the oil,

as, power and allied industries. Partthe global energy group GDF SUEZ,

ofey Fabricom employ over 5,500eople worldwide who strive for themprovement of energy ef ciency,urable development and quality of life.

he Cofely Fabricom, Energy &nvironment Division specializes in turnkeyngineering, Procurement, Constructionnd Commissioning (EPC) projects in thenewable energy sector emphasizing on

alance-of-plants, ue gas treatments, andogeneration plants.

April 2013, Cofely Fabricom E&E waswarded a turnkey revamping contract tomprove the reliability and performance ofe biomass furnace of the cogeneration

nit in Sart-Tilman, Lige, Belgium, whichpart of the central heat production andstribution across the Universit de Lige

ULG) and the CHU University Hospital.he project was centered around the

mprovement of a number of failuresxperienced by the two institutions,peci cally:

Insuf cient instrumentation for propercombustion control;Reduced performance due to a lack ofrecirculation of smoke gases in nominaloperation;Incomplete coverage of grid with pellets,leading to primary air bypass non- uniform combustion; andRefractory damage caused by extremelyhigh local temperatures.

order to address these issues, aompletely new combustion controlhilosophy was investigated.

he approach would involve:A mixture of 50% re-circulated smokegases into primary and secondary air toimprove performance;

C Optimizing, by means of Computational

Fluid Dynamics (CFD) calculations,secondary air injections to respect legalemission limits;

Installing new fans, frequency drives and

electrical cabinets; and Implementing superior instrumentation,

including ultrasonic ow meters andinfrared cameras.

Cofely Fabricom E&E rst performedthe necessary process calculations foroptimized combustion, accounting for owrate, temperature, oxygen content andpressure drop, which sized the primary,secondary and recirculation fans as well asthe duct diameter and valves.

The next step was to ensure the optimalcon guration of the secondary air injections.Since the main objective of the project wasto achieve an increase in performance, theoxygen concentration at the furnace outlethad to be reduced from its current valueof 13% while respecting the legal emissionlimits. The secondary air injection thereforehad to ensure that all produced CO wasrecombined to CO 2. This would be achievedby obtaining an optimal level of turbulenceinside the narrowest section of the furnace,referred to as the furnace neck.

To gain suf ciently high velocities at the tubeoutlet, the secondary air injection nozzleswere divided into two rows: an upper andlower ramp of 2 x 14 tubes each at the frontand back end of the furnace, 112 tubes

in total. At lower ow rates, the two lowerrows were isolated in order to maintain aminimum velocity of 20m/s at the remainingtube outlets.

Physically building and testing differentprototypes of an installation of this sizewould be time consuming and prohibitivelyexpensive. In order to understand theresponse of the system and arrive atthe optimum con guration of nozzles,Computational Fluid Dynamics (CFD)simulations were carried out using MentorGraphics FloEFD by project partners

Voxdale.

The rst step was to create a model of theexisting furnace and simulate its behaviors.

This approach would allow simulationparameters and boundary conditions to beadjusted to obtain a good match betweena well-known system and the simulationmodel. During this phase con dence in thesimulation approach is established and the

rst insights into the system are generated.

Combustion processes werent directlymodeled in the simulation, instead, air

temperatures and ow rateswere adjusted to obtaincomparable behavior. Thisreduced the time requiredfor each simulation withoutappreciably affecting accuracy,which in turn allowed forCFD to cover a large numberof design iterations in areasonable time.Figure 1 illustrates the originalsetup of the furnace withwood pellets that are fed intothe system, forming a layeron the moving grid. Below

this grid, ve distinct supplies of primaryair, provide the oxygen required for thecombustion process. In addition, secondaryair is supplied through nozzles on both thefront and lateral walls. At the furnace neck,

exhaust gases are injected into the furnace,providing the system with recirculation air.

In the second phase, different designproposals were set up and simulated inFloEFD. Three-dimensional turbulence plotsand velocity patterns were used to assessthe various con gurations. One of the earliersetups is illustrated in Figure 2, in which thesecondary air nozzles were moved closerto the furnace neck and the number ofnozzles was greatly increased. Although thissolution offered some bene ts, it resulted ininsuf cient turbulence intensity across thefurnace neck.

Based on the insight offered by suchsimulation and analysis, further designiterations could intelligently evolve. Forexample, it became apparent that offsettingthe recirculation nozzles slightly with respectto each other created the desired amount ofturbulence in the furnace neck.

The positioning and layout of the secondaryair jets were similarly optimized in order toensure suf cient turbulence in the adiabaticchamber (Figure 3). As the simulations wereperformed without combustion, it couldreasonably be assumed that an even higherlevel of turbulence, higher velocities andbetter mixing would occur in practice. Theresulting design is shown in Figure 4.

Figure 1. FloEFD simulation geometry

Figure 2. Secondary air nozzle detail results

Figure 3. Turbulence intensity in the adiabatic chamber

Figure 4. Furnace neck resulting design

Figure 5. FloEFD visualizations

Disassembly of the oldfurnace began in May2013, with installationworks running from Juneto September 2013.Start-up and combustionoptimization took place inOctober - November 2013.During cold commissioning

of the new installation the simulatedvelocity distribution was veri ed inside the

furnace and comparable ow patterns wereobserved.Integrating CFD in to the design processhelped Cofely Fabricom E&E deliver theproject successfully and within a shorttime frame. Virtually prototyping variousdesigns in FloEFD provided a reliable andinexpensive means by which differentparameters could be assessed andadjusted. Ultimately, all project objectiveshave been ful lled.

Performance was increased by 8.4%(actual ef ciency 89.3%);

The furnace represents a considerableimprovement over the previous design(see Table 1) and operates well belowlegal limits, e.g. for both carbonmonoxide and nitrous oxides.

The process followed demonstrated thevalue of uid simulation for such projects. Itis therefore an approach which will be usedfor future projects, where appropriate. Theinstallation has been in reliable operationsince November 2013.


17/31


How To...Get Hot Lumens in situ from FloEFDs unique LED Module

y Joe Proulx, Application Engineer, Mentor Graphics

Just occasionally a softwareproduct or module comes alongthat stimulates interest in a newtechnical area and can even

ove to be a game changer. The LEDodule for the 3D CFD product, FloEFD,looking like such a capability.

emerged two years ago from two separatechnical strands within the Mechanical

nalysis Division; the MicReD T3Ster &eraLED hardware for thermal and optical/ diometric characterization of LED lights,ong with the 3D CFD FloEFD software thatermits accurate thermal simulation of LEDsithin complex geometries easily.he big problem with simulating LEDs in a

dependence by way of R-C StructureFunctions, Diode Characteristics, OpticalPower (mW), Radiant Ef ciency (Popt/ Pelect), Luminuous Flux (lm), Ef cacy(lm/W), Scotopic Flux (lm), and Optical ColorCoordinates (X,Y,Z tristimulus values). Asa result of the combined thermal/opticalmeasurement, this data can be displayedbased on the LED junction temperature.

This process is illustrated in Figure 4, wherethe Temperature Sensitive Parameter (TSP),i.e. the diode forward voltage in the caseof an LED, is measured to calibrate theLED. It is then powered up to steady state,allowing the temperature, light output, andpower draw to stabilize. Measurement ismade by suddenly powering down the LEDto a very small measurement current that isused to record the TSP as the part cools.

The T3Ster software is used to convertthe temperature vs. time response into acumulative thermal structure function agraph of thermal capacitance vs. thermalresistance to reveal the thermal structure ofthe LED package.

aditional CFD approach, however, is thattemperature increases inside an LED,

s forward voltage changes leading to heatssipation changes and ultimately the

uality of its light output changing. Hence,odeling an LED as a simple heat source isst not accurate enough. This can be a bit

ke a dog chasing its tail! Figure 2 illustratesis complex interconnectivity beautifully.o how does the LED module help solve

this problem? Lets consider that we havean LED geometry (Figure 3) with a datasheetof manufacturers properties (boundaryconditions) and we have access to Mentorstried and proven combination of T3Ster& TeraLED to characterize the LED (usingJEDEC JESD51-1 and CIE 127-2007compliant techniques). We could measurethe LED under a range of conditions to

produce a set of data for the FloEFD LEDModule. The LED geometry itself could beinput into FloEFD via your favorite MCADtool (FloEFD is embedded within most of thepopular CAD tools available today) and theluminaire it is attached to can also be addedso that the LED is modeled in situ.

T3Ster/TeraLED measurements will yield fora given LED, the current and temperature

gure 1. Typical LED geometry inside the FloEFD LEDodule

gure 2. Interaction between temperature, forward voltage,wer, current and light output for LEDs

Figure 3. Inputs for an LED analysis in the FloEFD LED module utilizing the T3Ster/TeraLED thermalcharacterization and optical measurement devices

Figure 4. Four step process to create the cumulative thermal structure function.

Figure 5. Two Clicks to enter all the LED char acteristic data.

Figure 6. One click entry of the cumulative structurefunction.

FloEFDs LED Module has a standard,premeasured set of typical LEDs in i tsEngineering Database. Alternatively, you canadd a characterized LED yourself from yourown hardware measurements (Figure 5).

An LED Package Cumulative ThermalStructure Function can also be used to

create a Resistance-Capacitance (RCLadder) Model asshown in Figure 4step 4, and thenimported into FloEFDsLED module (Figure6). These modelsdetermine junctiontemperature accurately,without the overhead of requiring the LEDpackage internal geometry to be modeled,and can be used in transient situations. Lessaccurate, 2-Resistor (2-R) models can alsobe used but have the limitations of beingsteady-state only, use an assumed heatingpower as input, and give no informationabout the light output.

The LED Module user can select NativeCAD faces and bodies to apply LEDboundary conditions within their favoriteMCAD packages; PTC Pro/ENGINEER andCreo Parametric, CATIA V5, Siemens NX orstandalone (Figure 7).

The net effect of this LED Module approachis to allow the user to use FloEFDs accurateCFD solver to calculate the actual thermalimpact of the LED inside its luminairegeometry, yielding real-world useful datafor the LED Junction Temperature, the LEDHeat Generation Rate and, uniquely, theLEDs light output, or hot lumens in situ(Figure 8). This allows for the most accuratepredictions of an LEDs operational thermalperformance in any given product andapplication environment.

Figure 7. Applying the LED characteristics to thegeometric boundary conditions

Figure 8. Accurate in situ performance simulation of the LED


18/31


Theres an App for that!

Top Apps for Engineers

Wolfram AlphaReviews Customerratings SeescreenshotsLearnmore about WolframAlpha ontheAppStoreDownload WolframAlpha .

Smart ToolsReviews Customerratings SeescreenshotsLearnmore about SmartTools ontheAppStoreDownload SmartTools .

Spirax Sarco Steam ToolsReviews Customerratings SeescreenshotsLearnmore about SpiraxSarco ontheAppStoreDownload SpiraxSarco .

TopApps forEngineers

WolframAlphaReviewsCustomerratingsSeescreenshotsLearnmoreabout WolframAlpha onthe AppStoreDownload WolframAlpha .

Smart ToolsReviewsCustomerratingsSeescreenshotsLearnmoreabout SmartTools onthe AppStoreDownload SmartTools .

Top Apps for Engineers

Wolfram AlphaReviews Customerratings SeescreenshotsLearnmoreabout WolframAlpha ontheAppStoreDownload WolframAlpha .

Smart ToolsReviews Customerratings SeescreenshotsLearnmoreabout SmartTools ontheAppStoreDownload SmartTools .

Spirax Sarco Steam Tools i u m rr in r n h L rn m r u a a n h r nl a a .

Now that our phones are to the guidance computer used on the NASA Apollo missions, as a ick-book is to a high de nition television, it follows that there are more and more mobile applicationswith the potential to make the life of the average engineer easier. From an extensive trawl of theoptions out there, Ive selected the following apps over which to cast a sceptical eye. Given thatengineering is such a broad church, Ive tried to select apps that are as general as possible andgiven them a roadtest and rating.

Smart ToolsSmart Tools is a complete package of ve app sets; Smart Ruler Pro (Length, Angle,Slope, Level, Thread), Smart Measure Pro (Distance, Height, Width, Area), Smart CompassPro (Compass, Metal detector, GPS), Sound Meter Pro (Sound level meter, Vibrometer),and Smart Light Pro (Flashlight, Magni er, Mirror). I loved this as it gives users many useful

rule of thumb tools all in a handy App for your Smartphone to use with its camera. A mustfor every engineer!

Cost: Most are free and some cost $1 Rating:

Wolfram AlphaDespite comparing itself to the computer from Star Trek, Wolfram Alpha doesnt seemto have made any great strides on the road to quantum teleportation. It is, instead, acomputational knowledge engine a title with the air of a 19th Century curiosity about it. Butwhat does that actually mean? The easiest way to sum it up is that were you to ask it What isteleportation? Youd get a fairly unremarkable dictionary entry. So far, so sub-par Wikipedia.However, if you ask What is the integral of one over e to the minus x between one andtwo?, you get something a whole lot more impressive: it only goes and gives you the answerexpressed algebraically, as a number and an x-y visualization of the result. Very impressive.It also does a range of more applied problems such as material properties and statistics. Ifyoure not convinced, you could always try the web page rst, wolframalpha.com.

Cost: $3 Rating:

Engineering Unit Converter No-one could argue that this isnt a function thats likely to get many excited. If we wantedexcitement wed have become accountants, not engineers, right? So lets get down to thedetails: to be useful, this app is going to need to cover a wide range of parameters quicklyand easily.

On this basis, its job done here. The front screen lets you select from a huge range ofpotential parameters, from which you nd yourself looking at a rotating bezel of potential units.

Very comprehensive, very easy to use. You can even set favorites. Highly recommended.

Cost: Free Rating:

Process Engineering Tools This is a decent little app that calculates some basic properties for common pipelinecomponents. My only question would be, how often engineers nd themselves needingto do these kind of calculations on an app, rather than a more sophisticated program orspreadsheet? Someone obviously thought the answer was frequently enough to make itworth writing one, so who am I to argue?

Cost: $5.50 Rating:

Spirax Sarco Steam Tools This one is a little more parochial to those dealing with steam. It is a nicely designed app,so if you do happen to nd yourself needing to know what the saturation temperatureis for steam at a given temperature or pressure, you could do a lot worse. It also givesspeci c volume and enthalpies at the derive

Documents

Engineering Edge Volume3-Issue1