DEVELOPING A “NEXT GENERATION” PV INVERTER
The most serious reliability issue for photovoltaics (PV) isthe inverter. Mean time to first failure (MTFF) is estimatedto be about five years. Recent efforts to rapidly expandthe production of grid-tied inverters have not resulted inimproved reliability. At this time the inverter industry isuniquely positioned to develop a ’next generation’ inverterthat has ten-year MTFF, better performance and lowercost. The recent advents of new technologies such asDSP (digital signal processing), the growth in sales to afew hundred thousand inverters per year and theemergence of larger companies with interest in PVinverters make this possible today. At this time theDepartment of Energy is considering the development of anew inverter for use in distributed energy applications .This report summarizes the status of power electronicstoday, identifies technology issues, and identifies invertermanufacturer attributes thought to be essential for theproduction of a ten-year lifetime inverter.
For inverters, low cost and high reliability are conflictinggoals. There are three ways to improve reliability of a newproduct . One can 1) over-design, 2) implementredundancy, or 3) obtain a detailed knowledge of theproduct working-environment, its failure modes and modecauses and then feed this information back into productredesign. Both over-design and redundancy will result inincreased cost. The third option requires large numbers offielded product so that the failure mechanisms have anopportunity to occur, and can be identified and corrected.The proper integration of field data is the only option thatdoes not increase the cost or complexity of the inverter. Itis one of the goals of this development project to identifycommon elements in applications so that the numbers ofinverters (or inverter subsystems) manufactured can beincreased. Poor inverter reliability has been identified asa problem for PV; however, these problems are not aunique PV issue. DER (distributed energy resources)sources, such as fuel cells and microturbines, also haveinverters. Because these technologies are newer than PVand have not been fielded in large quantities, their pendinginverter problems are not yet evident.The PV experience has, however, identified a few keyissues that are expected to limit the reliability of othertechnologies if the approach to providing inverters is notchanged. These include:
· poor requirement definition that results in frequentproduct retrofits,
· immature manufacturing processes which do notincorporate a structured approach to productplanning, quality control, manufacturing, or marketing,and,
· outdated designs that do not include the newestarchitectures, packaging methods and technologies.
Thus, a significant improvement in inverter reliability canbe achieved today while still lowering cost. Furthermore,because of the similarities of all the grid-connected DERtechnologies, the inverters for these applications can bevery nearly identical and could be designed to beinterchangeable or to have interchangeablemodules/subsystems. Such an approach will result inhigher quantities of inverters and/or subsystems and thusprovide the larger numbers of fielded units that would, inturn, provide the feedback necessary to improve inverterreliability. A modular design could further improvereliability through the development of standardizedsubsystems that are utilized for different products, such asPV inverters of small and large size. Such modularcomponents would thus be manufactured in larger quantityand lead to higher reliability.A next generation inverter will have improvedperformance, high reliability (ten-year MTFF), andimproved profitability. What is required is not simply animproved version of what has not worked well; it is anorder-of-magnitude step forward. Such a development willhave risk. Success is not guaranteed. It has beenestimated that the development of a ’next generation’inverter could require approximately 20 man-years of workover an 18- to 24-month time frame. Companies withexisting design and manufacturing capability couldaccomplish the task with fewer resources. Theinvestment, however, is not small for most companiesinterested in the task. For that reason a governmentindustrypartnership will greatly improve the chances ofsuccess. While most companies that are developing orcontemplating the development of PV inverters are notstarting from zero, they must have financial resources andmanagement commitment for a multi-year effort.
Objective of a “Next Generation” inverter.
The objective of the proposed ‘Next Generation’ inverter isto develop a high reliability grid-tied DER inverter whilecontaining cost and improving performance. The
approach would include:
· incorporation of new technologies,
· use of mature manufacturers,
· large product manufacturing volume, and
· failure mode feedback.
The objective is to make reliable inverters available for PV,fuel cells, energy storage, microturbines, etc. Toaccomplish this, the development effort will concentrate onthe grid-tied inverter that has the largest expected shorttermmarket, i.e. PV residential inverters. The intendedinverter market could include foreign markets, and fuel cellor other DER sources as well as photovoltaics.The decision to include storage has not yet been made.Inverters with storage (batteries) do offer the advantage ofproviding backup power. This would be an essentialconsideration for individuals concerned with powerreliability; however, the resulting inverter system is morecomplicated, less reliable, and also has higher cost for thebatteries and power electronics for battery charging.
Present Reliability. Because the number of fieldedDER inverters is small and because designs are rapidlychanging, it is difficult to benchmark the reliability of powerelectronics in general or DER inverters in particular.
There are however a couple of data points. In 2000Sandia commissioned Trace Engineering (presentlyXantrex) to quantify the reliability of their PV inverters.The result of that study  indicated that the MTFF of PVinverters was about 4.7 years. Another snapshot wasprovided by an article that indicated that 2% of computerpower supplies fail in the first year . This industry ismuch more mature than DER and typically provides awarranty of one year. It is clear that the entire powerelectronics industry an benefit from improved reliability.
PATH TO HIGH RELIABILITY
The following three sections discuss the author’sperception of essential elements in developing a high
reliability inverter. No single approach for the inverterconfiguration stands out. The need to increase
manufactured volume, however, suggests that somestandardization should be employed. This has promptedthe use of the term ’universal’ inverter. The configurationcould be a single box that performs all tasks, a singlepower section with various, interchangeable, dc-to-dcconverters, or a building block concept similar to that usedfor personal computers. In the building block conceptcertain black boxes (e.g. a controller, a power converter,and so on) will be identical for various inverterconfigurations. Just as a computer can have varyingamounts of memory, so the inverter could mix and matchcapabilities as required.
1. Essential Elements of the Inverter Manufacturer
It is clear that many of the past problems with PV invertersresult from the absence of mature manufacturingprocesses in the PV inverter business. The low quantity ofinverters produced has limited the market to smallmanufacturers. This is beginning to change with currentinverter output of a few hundred thousand per year for alltypes of inverters. Some of the smaller companies haveinstalled quality programs and are currently ISO(International Organization for Standardization) registered.Others have formed agreements with larger companiesthat do have some or all of the requisite features in place.Developing inverter manufacturers who have the desiredcapability and who can demonstrate a long-termcommitment to the production of DER inverters is the keyto developing a next generation inverter.Those features, that maximize the probability of asuccessful ‘high reliability’ inverter, include anunderstanding of how the inverter product fits with theoverall business plan. For example, it is essential thatmanagement have commitment to the development, thusthe product must be important to future corporatedevelopment plans. The corporation must have performedmarket research and have a firm understanding of theprospective market. Key market drivers  such as qualityof product, reliability, price, standards, flexibility,performance/features, size/packaging/weight, and systemintegration, must also be clearly understood. A productdistribution and service network must either exist or beclearly attainable.Essential mature manufacturing processes include ISOcertification, quality programs, a systems engineeringdesign approach, structured documentation, and access toexperienced assemblers. Automation is desirablebecause it can enhance reliability. Experience in thedesign and manufacture of power electronics is essential.
2. Definition of Requirements
A systems engineering design approach focuses ondefining requirements; the first step in definingrequirements is an understanding of the environment. ForDER sources the environment includes nearby lightning,line voltage perturbations, overloads, number oftemperature cycles, ambient temperature, dust, humidity,vibration etc. Additional requirements include compliancewith regulations such as IEEE 929, IEEE 1547, IEEE-519(US), IEC-1000 (European), FCC Part 15, and familiaritywith utility interconnect requirements.
Technical Issues. The prospective inverter manufacturermust also be cognizant of known failure modes in powerelectronics (not limited to inverters). Some of the more
salient issues are identified below.One of the principal problems in inverters is heat thatresults from losses in the semiconductor switches. Heatgeneration can be reduced by decreasing switching timeand by selecting devices with smaller conduction losses.Advances in power electronic switches, such as theCOOLMOS™ , are continuing to improve theseparameters. Additional issues related to excessive heatare junction over temperature, inadequate heat transfer,and fans. Computer analysis of heat flow and advancedbonding materials can improve heat removal. Fans are alimited lifetime component and special considerationshould be given to removal of heat to minimize this failuremode. Wire bond breaks in semiconductor switches resultfrom frequent power cycling  and the resultant changein switch temperature. This is less a problem in DERinverters than automotive inverters since grid-connectedDER inverters tend to operate at slowly changing power;they do not undergo no-load to full load transitionfrequently.Optimum selection of the switching device is essentialfor high reliability. Faster control responses, frequentlyaided by on-device control, have ignificantly improvedfault tolerance. Inductive turn-off of large currents must bemitigated to prevent high dv/dt stresses across deviceterminals that can exceed device voltage specifications.Selection of a switching scheme is equally important.Soft switching offers lower heat generation and lower rfi(radio frequency interference); however, circuit parasiticsthat cause high frequency ringing are difficult to suppress.Hard switching has the understood problems withinterrupting large currents at high voltages, generatesundesirable heat, and is limited in the flexibility of controlschemes.
Capacitor failure is another area where failures are wellunderstood, but unlike semiconductors, research isproceeding slowly. US government research in capacitorshas been limited . Manufacturers for the largestmarkets (except power correction capacitors) have beendominated by foreign companies. Further, the capacitormarket of $25 to $50 million generates marginal IR&Dinvestment.A further inverter requirement is a reasonable cost anduser-friendly design. The cost of existing residentialsized inverters (.5 to 5 kW) is approximately $1/watt withquasi-sine inverter cost of as low as $.5/watt. The targetcost for the ‘next generation’ inverter is less than $.5/watt.The user-riendly design of a residential sized invertershould start with consideration for the installer.Minimizing installation cost would indicate that the inverterweight should be less than 70 pounds and installable by asingle individual in one hour. All connections should beeasy to make; all individual components should be ULapproved. Communication, while not a requirement, is ahighly desirable feature that will help ensure properoperation, ease installation of latest software, and provideuseful reliability data.
3. Analyze Potential New Technologies, PackagingMethods, and Control Schemes
There are areas where significant improvements ininverter design can result from emerging technologies.Each of these must be evaluated to determine itssuitability for an improved inverter.The adaptation of flexible architecture that leads tosubcomponent standardization and eases assembly is oneof the desirable features of a next generation inverter.The design and implementation of such architecture forboth software and hardware will reduce design time andfacilitate interchangeability of modules for varyingapplications.
Interconnection must be easy, inexpensive, and reliable.Standardization is essential to odularization. “Pushingthe standardization of key power electronic interfaces tonear optimum configuration is a lesson learned from theworld of small signal/logic signal processing .”
Modularity is extremely desirable for both hardware andsoftware. Modularity will decrease time to market, simplifyrepairs, reduce need for new design, and improvereliability. In the initial stages of this inverter developmentit is more likely that inverter subsystems will be more likelyto be manufactured in larger volume than a completeintegrated inverter of a single design.
DSP (digital signal processing). Digital signalprocessing is now available on a chip and easilyincorporated into an inverter design. In fact a new inverterdesign has been accomplished entirely by DSP andincludes no microprocessors . The reason DSP isimportant is that inverter control functions that havetraditionally been accomplished with hardware can now beaccomplished with software. Thus by accessing differentsoftware packages the power converter can accomplishdifferent tasks with fewer hardware changes. This resultsin more applications for a single piece of hardware andthus more quantity of product. New chips, such as theTMS320F2810 having a speed of 150 MHz provide widerbandwidth for processing data. This results in the ability toprovide better power quality with very rapid response tochanging loads. Historically DSP had limited peripheralI/O capability; however, new chips come with serialperipheral interface (SPI), serial communicationsinterfaces (SCIs), standard UART, enhanced controllerarea network (eCAN), and multichannel buffered serial port (McBSP) with SPI mode . These features allowDSP to perform both inverter control functions and userinterface functions. DSP also offers the opportunity forreducing the parts count, thus improving reliability directly.Over the past 20 years improved semiconductors haveresulted in very significant gains in inverter performance;however, further reductions in on-resistance are reachingthe theoretical limit for silicon and faster switch rates maybe reaching their practical limit, as designers limit v/dt’sto reduce rfi .In general, “faster semiconductor switching speeds aredesirable but will by definition result in higher di/dt acrossthe semiconductor die. You can’t have one without theother in hard switching topologies. High dv/dt is not thepower switch killer (as long as the absolute rated voltageof the device is not exceeded). High di/dt, however,causes current crowding on the die resulting in localizedhot spots. High di/dt is desirable in hard switchingtopologies to achieve high conversion efficiencies.Designing for high di/dt is a matter of proper deviceselection, low inductance dc bus design, and appropriatesnubber and EMI filter design. Also, basically all powertopologies are based on switches that interrupt largeinductive currents .”A subset of semiconductor switch design is made-toorderpower electronics . Made-to-order powerelectronics include the semiconductor designer in theinverter design. Thus the resulting inverter will have asemiconductor switch that is optimized for the applicationand includes many control and protection features onboard.Packaging technologies include system layout, bonding,interconnection, and laminated bus bars. System layoutcan be optimized with computer tools including heat flowanalysis. In medium voltage applications (< 500 Vdc)traditional bonding methods using thermal grease orpolymer pads can be replaced with nodized films placed on aluminum substrates. The anodized layermakes a uniform connection (35 to 50 microns thick) to thechassis or heat sink permitting maximum heat transfer.Potential components should also be analyzed for thereliability of their interconnects. Wire bonding technologyhas been the standard interconnect method in powerdevice and module fabrication . Some advancesinclude ThinPak technology (Silicon Power Corp), PowerOverlay technique (GE), Bottom-less SO-8 package usingflip chip technique (Fairchild semiconductor),PowerConnect technology (Vishay Siliconix) and Dimple-Array Interconnect (CPES).Communications and control may be important forestablishing that the product works, as expected, in thefield. Furthermore field data is essential for theenhancement of reliability in future products and can beobtained from an inverter with communications.These elements must be carefully evaluated during thedevelopment process. The need for a high reliabilityinverter is inherent in the goals of all DER programs in thatthey all require inverters with low cost and dependablepower
Given the increasingly robust design of most invertercomponents, manufacturing and design problems may beresponsible for the largest number of field failures. Animportant new approach in product testing is highlyaccelerated lifetime testing (HALT). This test approachprovides multiple stresses to an operating product,stressing the product beyond specification to identifyfailure modes. The rationale is that, if the product can bedamaged in the laboratory, it will eventually experience thesame failure in the field. After a failure mode is identifiedthe failure mode is eliminated and the product is stressedfurther. Typical stresses include rapid changes intemperature and shock while the inverter is exercised withvarying output power levels. Humidity is also an importantparameter; however, humidity chambers that can containan entire inverter are limited in number.In addition to HALT, the product developer must performdesign verification testing (DVT). The purpose of thistesting is to ensure that the product meets all designrequirements. Third party laboratories often accomplishthis testing so that an impartial assessment is obtained.Sandia National Laboratories power electronics testlaboratory has performed this type of testing for many ofthe PV inverter manufacturers.
Inverter failure continues to plague the PV industry and islikely to also plague emerging DER industries, such asmicroturbines and fuel cells. At this time there is anopportunity to double the inverter MTFF to ten years. Thiscan result from implementation of new devicetechnologies, increased sales, and the use of moresophisticated design and manufacturing processes.Given good design and manufacturing processes, largenumbers of product manufactured may be the single mostimportant issue related to reliability. Large quantities offielded product make possible the feedback of failuremode data that, over time, results in a highly reliableproduct. For that reason, the first of the ‘next generation’inverters that is developed is likely to be a small(residential sized) PV inverter with alternative applicationsin fuel cells.Examination of failures of fielded inverters indicates thatthe following factors contribute to poor reliability:
· poor requirement definition,
· immature manufacturing processes, and
· outdated designs.
A significant improvement in inverter reliability can beachieved by correcting them.The path to high reliability inverters is to incorporatemature manufacturing processes with solid designs thatresult in an inverter that can be marketed in largequantities. The design and lessons learned can then beextrapolated to other sizes and applications. Three issuesare thought to be critical to achieve the ‘next generation’inverter. These are
1. inclusion of an inverter manufacturer who has mature
manufacturing processes in place
2. definition of adequate inverter requirements and
3. incorporation of new technologies, packaging
techniques, and control schemes.
Finally, an extensive testing regimen must be developed.Testing is a part of each phase of development as well as20part of the validation of the final product. 
 Rick West and Konrad Mauch, Xantrex PowerEngineering, Burnaby, BC, Canada, Yu Chin Qin,Millenium Technologies, Troy, Michigan, Ned Mohan, U ofMinnesota, Minneapolis, MN, Russell Bonn, SandiaNational Laboratories, Albuquerque, NM, “Status andNeeds of Power Electronics for Photovoltaic Inverters:Summary Document”, SAND2002-1085. Michael Ropp, Ph.D., Personal Correspondence,South Dakota State University, Brookings SD, January2002. R. Pitt, "Improving Inverter Quality," Proceedings,NCPV Program Review Meeting, April 16-19, 2000,Denver, CO, pp. 19-20. Krishna Shenai, “Made-to-order Power ”, University ofChicago, IEEE Spectrum , July 2000. Ken Phillips, “Power Electronics; Will Our CurrentTechnical Vision Take us to the Next Level of ac driveProduct Performance?”, Rockwell Automation, Mequion,WI. K Shenai, “High-power robust emiconductorelectronics technologies in the new Millennium”,Microelectronics Journal 32 (2001) 397-408. A Morozumi, K Yamada, T Miyaska, Y Seki, “Reliabilityof Power Cycling for IGBT Power semiconductorModules”, Fuji Electric Co. LtD Matsumoto Factory, Japan,2001 IEEE. W Sarjeant,“A Report on Packaging implications ofAdvances in Capacitor Technologies”, State University ofNY at Buffalo, NY, and D Staffiere, Staffiere ConsultingServices, Amherst, NH, IEEE 2000. Yu Chin Xu,R&D development of a Voltage Stabilizerby Millennium Technology, Detroit, MI, project still inprogress. TMS320F2810 Product Review, Texas Instruments,Houston, Texas. Rick West,personal communication, Xantrex, SanLuis Obispo, CA . Tom Morris, International Resistive, Co., CorpusChristi, TX, Shaun Martin, “Anodized Al Substrates forImproved Thermal Performance”,Welwyn Electrics,Bedlington, Northumberland, UK, PCIM 2001 Proceeding,Sept 2001. Simon Wen Daniel Huff, and Guo-Quan Lu, “Dimple-Array Interconnect Technique for Packaging PowerSemiconductor Devices and Modules”, Power ElectronicsPackaging Lab Center for PowerElectronics Systems,Proceeding of 2001 International Symposium on PowerSemiconductor Devices & Ics, Osaka, Japan. Sandia is a Multiprogram laboratory operated bySandia Corporation, a Lockheed Martin Company, for theU.S. Department of Energy under contract DE-AC04-94AL85000.
Russell H. Bonn, Sandia National Laboratories, Albuquerque, NM, 87185-0753