Failure Rates Of IGCTs Due To Cosmic Rays

—A P P L I C AT I O N N OT E 5 S YA 2 0 4 6 - 0 3Failure rates of IGCTs due to cosmic raysIn the early 1990’s a new failure modefor high current, high voltagesemiconductor devices wasdiscovered. The failure mode was ofconsiderable practical significanceand caused a series of equipmentmalfunctions in the field. This failuremode affects all kind of devices likediodes, thyristors, GTOs, IGCTs,IGBTs, etc. It consists of a localisedbreakdown in the bulk of the devicesand is not related to junctiontermination instabilities.1. IntroductionThe location of the breakdown spot on the wafer israndom. The onset of the breakdown occurs withouta precursor within a few nanoseconds and there isno sign of early failures or wear out. The failure rateis, thus, constant in time but strongly dependent onthe applied voltage and shows a small dependenceon temperature.Experiments in a German salt mine 140 m belowground did not show any of these failures, while experiments on the Jungfraujoch (3480 m above sealevel) in the Swiss Alps yielded a much higher failurerate than in laboratories close to sea level. Furthermore, irradiation with heavy energetic particles creates the same failure patterns. All together it wasconcluded that “cosmic rays” are the root cause ofthis kind of failure and this conclusion is now supported by a vast number of experiments done allaround the world.Primary cosmic rays are high-energy particles,mostly protons, that are found in space and thatpenetrate our atmosphere. They come from all directions and have a wide energy range of incidentparticles. Most of these cosmic rays originate fromsupernovae. Originally the Austrian physicist ViktorHess (Nobel Prize 1936) discovered cosmic rays because of the ionization they produce in our atmosphere. In fact, a primary cosmic ray particle usuallydoes not reach the surface of the earth directly butcollides with an atmospheric particle (see frontpage). There it generates a variety of other energy-rich particles, which later collide with other atmospheric particles. The process of a cosmic rayparticle colliding with atmospheric particles anddisintegrating into smaller pions, muons, neutrons,and the like, is called a cosmic-ray shower. Most ofthe generated particles are harmless for semiconductor devices but some, mostly neutrons, may belethal. Occasionally cosmic ray related events areobserved, which do not lead to any perceivable damage but in general, the device is doomed even if fastfuses are used.Today, ABB’s high current, high voltage semiconductors are designed such that the failure rate due tocosmic rays is reduced to an “acceptable” level. Nevertheless, cosmic ray induced failures have to betaken into account for every power electronic circuit. In particular, semiconductors for applicationswith a high utilisation of the device’s blocking capability and for equipment operating at high altitudes

—Table of contents0103030303040405IntroductionModelling the failure ratesVoltage dependenceTemperature dependenceAltitude dependenceFailure rates of the individual IGCT typesCalculation examplesGraphs for 5SHX 26L4520, 5SHY 35L452x, 5SHY 40L4511and 5SHY 55L450005Graphs for 5SHX 19L6020 and 5SHY 50L550005Varying voltages05Revision history

3—A molten channel through a silicon device created by a chargeavalanche triggered by incident cosmic rays during blocking.2. Modeling the failure ratesIn order to provide the user with a simple failure ratecalculation tool, a mathematical model (Eq. 1) wasdeveloped that covers the three most important influences: blocking voltage, junction temperature,and altitude. The failure rate model consists of threemultiplicands:the dependence on the DC-voltage (VDC in volts,VDC C1) at nominal conditions, i.e. 25 C and sealevelthe dependence on the temperature (Tvj in degrees Celsius), term equals unity if Tvj equals 25 Cthe dependence on the altitude (h in metersabove sea level), term equals unity if h equals 0,i.e. sea levelThe multiplicands and equal unity at nominalconditions (25 C and sea level, respectively). Thus,the formula can be simplified for certain cases. If forexample a converter operates only at sea level, multiplicand can be neglected. The formula is onlyvalid for DC blocking conditions. Varying blockingvoltages, blocking duty cycles or overvoltage spikesdue to switching operations should be addressed asdescribed in paragraph 4.Please note: The model delivers failure rates in FIT, i.e. numberof failures within 109 element hours. The formula is only valid if the DC-link voltage VDC islarger than the parameter C1 because the formulahas a pole at C1 . For VDC values below C1 the failurerate is regarded as zero. The failure rate model describes only failures thatare due to cosmic rays. The model does not coverfailures due to other root causes.2.1. Voltage dependenceThe formula for the voltage dependence (multiplicand ) is a pure fit to measured data at DC-voltage.The formula has no physical background but fits thedata almost perfectly. The model’s parameters C1,C 2, and C 3 are, therefore, characteristic values of theindividual devices and can be looked up in the tablein section 3. The parameters have also no physicalmeaning.2.2. Temperature dependenceThe formula for the temperature dependence (multiplicand ) is again a fit to measured data. However,experiments indicate that the failure rates decreaseexponentially with temperature and that this dependence is practically independent of the device type.Therefore, the formula does not require any devicespecific parameters.2.3. Altitude dependenceThe formula for the altitude dependence (multiplicand ) assumes a screening of cosmic rays by theatmosphere and is, thus, based on the barometricformula. This implies that all devices are affectedthe same way, so again the formula does not containany device specific parameters.—Equation 1—ABB Power Grids Switzerland Ltd.SemiconductorsFabrikstrasse 35600 nductors—We reserve the right to make technicalchanges or modify the contents of thisdocument without prior notice. With regard to purchase orders, the agreed particulars shall prevail. ABB does not acceptany responsibility whatsoever for potential errors or possible lack of informationin this document.We reserve all rights in this document andin the subject matter and illustrationscontained therein. Any reproduction, disclosure to third parties or utilization of itscontents – in whole or in parts – is forbidden without prior written consent of ABB.Copyright 2019 ABBAll rights reservedFailure rates of IGCTs due to cosmic rays / Application Note 5SYA 2046-03have to be assessed carefully. This application noteis intended to provide a basis on which the powerelectronics designer can estimate failure rates, adjust parameters such as DC-link voltages or simplyselect the right semiconductor device for a particular application.

4ProductC1 [V]C2 [V]C3 [FIT]5SHX 26L45205SHY 35L45205SHY 35L45215SHY 35L45225SHY 45L45205SHY 55L4500265055001.39E 075SHX 19L60205SHY 50L5500290087004.21E 075SHY 42L65003100168008.52E 073.1. Calculation examplesAssume a 4.5 kV IGCT in L-package (5SHY 35L4520)operated at a DC-link voltage of 3400 V, a temperature of 0 C and at sea level. Because the altitude isat its nominal value the last multiplicand can be ignored. Together with the parameters from the tableabove, the failure rate formula now reads:15400 FIT means 15400 failures within 109 elementhours or an MTTF of 1/λ 65000 h, i.e. 7.4 y. Assuming a converter output stage with six IGCTs, theMTTF reduces to 1.2 y and this is usually not regarded as sufficient reliability. Obviously, the targeted DC-link voltage is too high. Assume again a4.5 kV IGCT in L-package (5SHY 35L4520) that is operated now at a DC-link voltage of 2800 V, a temperature of 25 C and at an altitude of 6000 m. Because the temperature is at its nominal conditionthe multiplicand can be ignored. Together withthe parameters from the table on the left the failurerate formula now reads:—Equation 3In this example the MTTF is 6.1·109 h or 700000 y.Even if the circuit contains a number of devices theoverall reliability will not be affected by cosmic rayinduced failures. Nevertheless, due to the statisticalnature of the effect there might be cosmic ray failures in the field. Furthermore, the assumption of aconstant DC-voltage is not realistic for typical applications. A variation of the DC-voltage due to e.g. input voltage variations or specific operations modes(breaking operation) is to be expected. Even moreimportant is the repetitive over voltage the devicehas to withstand during switching. Dealing with thisarea is explained in more detail in section 4.—Equation 2—ABB Power Grids Switzerland Ltd.SemiconductorsFabrikstrasse 35600 nductors—We reserve the right to make technicalchanges or modify the contents of thisdocument without prior notice. With regard to purchase orders, the agreed particulars shall prevail. ABB does not acceptany responsibility whatsoever for potential errors or possible lack of informationin this document.We reserve all rights in this document andin the subject matter and illustrationscontained therein. Any reproduction, disclosure to third parties or utilization of itscontents – in whole or in parts – is forbidden without prior written consent of ABB.Copyright 2019 ABBAll rights reservedFailure rates of IGCTs due to cosmic rays / Application Note 5SYA 2046-033. Failure rates of the individual IGCT typesThe following table gives the device-specific parameters for the individual IGCT types. The cosmic rayinduced failure rate of the integrated gate unit partis not accounted for, but is assumed to be less dominant for typical applications. The cosmic ray measurements were done with the smallest (corresponding to the D-housing) device type on wafer level. Themodel parameters were afterwards fitted to themeasured failure rates scaled to the device area ofthe respective larger IGCT. All values are typical values and may vary considerably.Section 3.1 gives two examples of how to calculatethe failure rate by using the formula and sections 3.2and 3.3 show some selected graphs for each productlisted above together with the underlying measurements.