Electronics Guide

What the Hazard Rate Measures

For a time to failure with probability density f(t) and reliability function R(t), the hazard rate is

h(t) = f(t) / R(t)

This is a conditional rate: the instantaneous rate of failure among units that have survived to time t, expressed per unit of time. Two consequences follow, and both matter in practice.

• The denominator shrinks: Because the surviving population declines, the hazard rate can rise even while the absolute number of failures per month falls. A rising wear-out region does not require an increase in the raw failure count.
• It describes a population, not a unit: An individual product does not experience a bathtub curve. The curve summarizes how risk is distributed across a fleet of nominally identical items, which is why field data from a large installed base is required to observe it.

Because the hazard rate has units of inverse time, it is reported in failures per hour, in FIT, or in percent per thousand hours. The reciprocal of a constant hazard rate is the mean time to failure, but that relationship holds only while the rate is constant.

The Weibull Shape Parameter

The Weibull distribution is the standard tool for classifying hazard behavior because its shape parameter maps directly onto the three regions. With shape parameter β and scale parameter η, the hazard rate is

h(t) = (β / η)(t / η)^β−1

• β < 1: Decreasing hazard rate, the signature of infant mortality
• β = 1: Constant hazard rate; the Weibull reduces to the exponential distribution and the useful-life region
• β > 1: Increasing hazard rate, the signature of wear-out; values near 2 suggest a linearly increasing rate, and larger values indicate a sharply defined end of life

Fitting field or test data to a Weibull model and reading off β is therefore a direct way to identify which region a population occupies, without having to guess from failure counts alone.

Why the Curve Is a Superposition

A single two-parameter Weibull distribution cannot produce a bathtub shape. Its hazard rate is monotonic: decreasing, flat, or increasing, but never all three in sequence. The bathtub emerges only when several failure populations act at once and the total hazard rate is the sum of their contributions.

• A defective subpopulation: A small fraction of units carries a latent manufacturing or component defect and fails early, with β well below 1. Once that subpopulation is exhausted, its contribution disappears.
• A stress-driven background: Externally triggered failures, spread thinly and without age dependence, contribute an approximately flat term.
• A degradation population: Every unit accumulates damage, but the resulting hazard is negligible until late life, when it rises steeply with β well above 1.

Modeling the curve therefore means fitting a mixture or competing-risk model rather than a single distribution. This framing also explains why the flat middle region is an artifact of addition rather than a physical law: no mechanism produces a constant hazard rate on its own, but many small mechanisms with different timings sum to something that looks constant over a limited window.

Infant Mortality Region

The early life period shows a decreasing hazard rate:

• High initial failure rate: The hazard rate is at its highest when units first enter service, then falls as operating hours accumulate
• Decreasing hazard: As weak units fail and are removed, the surviving population becomes progressively stronger; survivors are genuinely more reliable than newly manufactured units
• Duration: Anywhere from a few hours to several thousand operating hours, depending on product complexity, process maturity, and how much screening was applied before shipment
• Weibull shape parameter: β below 1.0 indicates the decreasing failure rate characteristic of infant mortality
• Also called: Early failure period, debugging period, or burn-in region; the semiconductor industry quantifies it as the early life failure rate (ELFR), usually reported in parts per million over a defined number of operating hours

Infant mortality failures are not caused by unusual stress. They represent the weakest members of the production population failing under entirely normal operating conditions, which is precisely why they are addressable through quality control rather than through design margin alone. The fraction of units affected is usually small, but the failures are concentrated in the period when the customer is forming a first impression of the product.

Useful Life Region

The middle period shows a relatively constant hazard rate:

• Constant failure rate: Once the defective subpopulation is exhausted, the hazard rate settles at an approximately constant level
• Random failures: Failures show no age dependence, so knowing how long a unit has run reveals nothing about when it will fail
• Exponential distribution: A constant hazard rate implies exponentially distributed time to failure, with β equal to 1 in Weibull terms
• MTBF relevance: Mean time between failures is meaningful in this region and misleading outside it
• Duration: Encompasses most of the product's operational life, and for well-designed equipment usually exceeds the intended service life

The useful life region represents normal operation, where products perform their intended function with acceptable reliability. "Random" here is a statement about predictability, not about causation: every failure in this region still has a physical cause, but the causes are numerous, independent, and individually too rare to forecast for any particular unit.

Wear-Out Region

The late life period shows an increasing hazard rate:

• Increasing failure rate: As products age, the hazard rate rises because accumulated damage and degradation erode the margin between applied stress and remaining strength
• Wear-out mechanisms: Fatigue, corrosion, electrolyte loss, and other time-dependent mechanisms become dominant
• Weibull shape parameter: β above 1.0 indicates the increasing failure rate characteristic of wear-out
• Predictability: Wear-out failures are more predictable than random failures because they follow a physical progression that can be modeled and, in many cases, measured
• End of life: The rise continues until replacement or overhaul becomes more economical than continued repair

Wear-out represents the natural end of product life as accumulated damage exceeds component or system capability. Its predictability is what makes it manageable: because the onset can be estimated in advance, wear-out is the one region in which scheduled replacement reliably reduces the failure rate. Applying scheduled replacement during the useful life region achieves nothing, because a new unit is no better than a survivor when the hazard rate is constant.

Manufacturing Defects

Process variations create weak units:

• Workmanship defects: Assembly errors, contamination, and handling damage during manufacturing
• Process excursions: Temporary process variations outside normal control limits
• Material defects: Flaws in raw materials and components that escape incoming inspection
• Solder joint defects: Cold joints, voids, and insufficient wetting create weak connections
• Wire bond defects: Poor bonds that pass production test but fail under stress

Manufacturing quality improvements directly reduce infant mortality rates.

Component Defects

Weak components contribute to early failures:

• Semiconductor defects: Gate oxide defects, metallization flaws, and contamination in ICs
• Parametric weakness: Components at extreme ends of parameter distributions
• Latent defects: Damage from ESD or handling that weakens but does not immediately fail components
• Counterfeit components: Substandard or remarked parts that do not meet specifications
• Infant mortality in components: Components themselves have bathtub-shaped failure rates

Component screening and qualification reduce defective components reaching production.

Design Marginality

Designs with inadequate margins are vulnerable to variation:

• Insufficient derating: Components operated too close to ratings fail when stressed
• Tight tolerances: Designs requiring tight tolerances fail when components drift
• Environmental sensitivity: Marginal designs fail at temperature or humidity extremes
• Voltage sensitivity: Circuits sensitive to supply voltage variation fail on marginal supplies
• Timing margins: Digital designs with tight timing fail with component variation

Robust design with adequate margins reduces sensitivity to component and process variation.

Application Stress

Initial operation exposes weaknesses:

• First power application: Initial power-up stresses previously untested connections and components
• Thermal cycling: First thermal cycles stress interfaces and joints
• Mechanical stress: Installation and initial handling stress mechanical connections
• Environmental exposure: First exposure to operating environment may reveal vulnerabilities
• Customer usage: Actual usage patterns may differ from design assumptions

Burn-in and environmental stress screening expose weak units before shipment.

Sources of Random Failures

Multiple factors contribute to apparently random failures:

• External events: Power surges, ESD events, and environmental transients cause unpredictable failures
• Overstress: Occasional extreme conditions exceed component capability
• Latent defects: Previously undetected defects that manifest under specific conditions
• Complex interactions: Combinations of stresses and variations produce occasional failures
• Human factors: Operator errors and maintenance mistakes contribute to failures

The superposition of many independent failure mechanisms produces approximately constant overall failure rate.

Exponential Distribution

A constant failure rate implies specific statistical properties:

• Memoryless property: Future failure probability is independent of past survival time, so a used unit is statistically as good as new
• Reliability function: R(t) = e^-λt, where λ is the constant failure rate
• MTTF relationship: Mean time to failure equals 1/λ
• Simplification: The exponential assumption simplifies many reliability calculations, including series-system combination
• Approximation: A good approximation during useful life even when the rate is not exactly constant

The exponential distribution is widely used in reliability analysis because of its mathematical tractability, though its memoryless property does not hold once wear-out begins.

Applicability and Limitations

The constant failure rate assumption has important limits:

• Wear-out mechanisms: Components with significant wear-out do not have a constant failure rate
• Complex systems: System failure rate depends on component interactions and on the mix of underlying mechanisms
• Environmental dependence: Failure rate varies with operating conditions such as temperature, humidity, and electrical stress
• Small populations: Statistical variation may obscure the underlying pattern in limited field data
• Mission profiles: Varying stress levels over time change the instantaneous failure rate

Reliability engineers should verify the constant failure rate assumption rather than assume it applies. For many modern, well-screened electronic components, no observable wear-out occurs within the intended service life, so the curve reduces to a declining infant-mortality region followed by a long flat tail. Empirical studies of fielded electronics also report profiles that depart from the textbook shape, including curves that decline throughout life with no clear wear-out rise.

Failure Rate Units

Failure rates are expressed in various units:

• Failures per hour: The basic unit; typical values for electronic components fall between 10^-6 and 10^-9 per hour
• FIT (failures in time): Failures per billion (10⁹) device-hours, so 1 FIT equals 10^-9 failures per hour. Mature integrated circuits are commonly rated on the order of a few to a few hundred FIT
• Percent per thousand hours: Sometimes used for higher-failure-rate items, where 1%/1000 h equals 10^-5 failures per hour
• PPM (parts per million): Often used for one-time manufacturing defect rates rather than time-based failure rates
• MTBF and MTTF: Mean time between failures applies to repairable items and mean time to failure to non-repairable items; for a constant failure rate, both equal the reciprocal of that rate

Consistent units are essential when comparing or combining failure rates from different sources. A frequent error is treating MTBF as a guaranteed service life rather than as the inverse of an average failure rate.

Electronic Wear-Out Mechanisms

Specific mechanisms cause electronic component aging:

• Electromigration: Momentum transfer from conducting electrons drives metal atoms along interconnects under high current density, thinning the line until it opens and piling up material until it shorts to a neighbor
• Hot carrier injection: Carriers accelerated by high channel fields gain enough energy to be injected into the gate dielectric, shifting threshold voltage and degrading transconductance
• Bias temperature instability: Negative BTI in p-channel devices and positive BTI in n-channel devices shift threshold voltage under sustained bias at elevated temperature, eroding timing margin over years of operation
• Time-dependent dielectric breakdown: Defects accumulate in the gate dielectric until a conductive path percolates through it and the oxide fails
• Interconnect aging: Via and contact resistance increases with time and thermal cycling, degrading timing before it causes an outright open

These mechanisms are modeled quantitatively rather than merely described. Electromigration life, for example, is conventionally estimated with Black's equation, in which the mean time to failure varies as A · J⁻ⁿ · exp(E_a / kT), where J is current density, T is absolute temperature, k is the Boltzmann constant, and the current-density exponent n is customarily taken as 2. The activation energy E_a depends on the dominant diffusion path; grain-boundary diffusion in aluminum is commonly assigned a value near 0.7 eV, and copper interconnects tolerate substantially higher current densities than aluminum in part because their activation energy is higher.

Two design levers follow directly from the form of that expression. Reducing current density is a strong lever, because the dependence is roughly quadratic, and reducing junction temperature is a very strong lever, because the dependence is exponential. This is why current-density design rules and thermal management dominate semiconductor wear-out control, and why the same two levers appear again in the strategies discussed below.

Mechanical Wear-Out

Mechanical components and connections age:

• Solder fatigue: Mismatched thermal expansion between package and board strains the joint on every temperature excursion, accumulating damage until a crack propagates through it
• Wire bond fatigue: Repeated flexure, driven by the same expansion mismatch, fatigues the bond heel until it lifts
• Connector wear: Contact plating is abraded by each mating cycle; once the noble plating is worn through, the base metal beneath it oxidizes and contact resistance climbs
• Bearing wear: Fans, drives, and other rotating assemblies wear with operating hours, and cooling fans are frequently the first wear-out item in otherwise solid-state equipment
• Vibration fatigue: Resonant excitation of boards and leads causes fatigue cracking, particularly where a heavy component is supported only by its own solder joints

Thermal-cycling fatigue is usually modeled with a Coffin–Manson relationship, in which the number of cycles to failure falls as a power of the applied strain range, so a wider temperature swing costs disproportionately more life than a larger number of gentle ones. That relationship is why reducing the magnitude of temperature excursions often buys more solder-joint life than reducing their frequency, and why power cycling is a harsher test than ambient soak.

Mechanical wear-out frequently limits the life of an assembly even when every semiconductor in it remains fully functional. Equipment that will be maintained rather than discarded is therefore usually designed so that the shortest-lived mechanical items are the ones that are easiest to reach and replace.

Environmental Degradation

Environmental exposure causes progressive damage:

• Corrosion: Metal surfaces corrode in humid or contaminated environments
• Oxidation: Surface oxidation degrades contacts and connections
• Polymer aging: Plastics and elastomers degrade with time, temperature, and UV exposure
• Contamination accumulation: Dust and deposits accumulate affecting thermal and electrical performance
• Moisture absorption: Hygroscopic materials absorb moisture affecting properties

Environmental protection and appropriate material selection extend life in challenging environments.

Consumable Components

Some components have inherently limited life, and their wear-out is a specification rather than a defect:

• Aluminum electrolytic capacitors: The liquid electrolyte evaporates through the seal, raising equivalent series resistance and reducing capacitance. Manufacturers publish an endurance figure in hours at the rated upper temperature and an Arrhenius-based derating rule of thumb under which expected life roughly doubles for each 10 °C reduction in core temperature; keeping such a capacitor well below its rated temperature is therefore the single most effective way to extend it
• Batteries: Electrochemical cells lose capacity with both charge-discharge cycles and calendar time, so a battery ages on the shelf as well as in use
• Light sources: LED emitters lose luminous flux gradually rather than failing abruptly, so their life is specified as the time to reach a stated fraction of initial output rather than as a time to catastrophic failure
• Flash memory: Each program-erase cycle degrades the tunnel oxide, and endurance falls sharply as more bits are stored per cell, so single-level-cell devices tolerate far more cycles than multi-level, triple-level, or quad-level parts
• Electromechanical parts: Relays, switches, and connectors are rated for a finite number of actuation or mating cycles, after which contact resistance is no longer guaranteed

Consumables place a hard ceiling on the useful-life region regardless of how good the rest of the design is. Two responses are available: select a part whose rated life exceeds the required service life with margin, or make the part field-replaceable and plan for its replacement. Choosing neither is what produces the early wear-out pattern described below, in which a product's failure rate turns upward well inside its intended service life because one inexpensive component reached the end of its own.

Reducing Infant Mortality

Multiple approaches reduce early failures:

• Burn-in: Operate products under elevated temperature and voltage before shipment so that weak units fail in the factory rather than in the field
• Environmental stress screening: Temperature cycling and random vibration precipitate latent defects that steady-state operation would not reveal
• Improved manufacturing: Better process control reduces defects at the source, which is the only approach that lowers cost and failure rate simultaneously
• Component screening: Additional testing of incoming parts, most valuable where the supply chain is unfamiliar or the consequence of an escape is severe
• Design margin: Robust design tolerates component and process variation, so ordinary variation never produces a marginal unit in the first place

These approaches are not equivalent, and the trade-off between them has shifted over time. Screening removes weak units but does nothing to prevent them, so it treats a symptom at recurring cost per unit shipped. It also consumes life: burn-in expends useful-life hours on every good unit, adds thermal cycles that themselves accumulate fatigue damage, and can introduce handling damage of its own. As semiconductor process control matured and defect densities fell, blanket burn-in became progressively harder to justify, and much of the industry moved toward statistical process control, targeted screening of high-risk populations, and outlier detection at wafer test.

Blanket screening still makes economic sense in two situations: when the cost of a field failure vastly exceeds the cost of the screen, as in implantable, aerospace, or undersea equipment, and when a process is genuinely immature, as during the first production of a new design. The goal in every case is the same, namely to ship products that have already passed through infant mortality, but the cheapest way to achieve it is to build fewer defective units rather than to find more of them.

Extending Useful Life

Maximize the constant failure rate period:

• Derating: Operate components below ratings to reduce stress and extend life
• Thermal management: Lower operating temperatures slow degradation mechanisms
• Protective measures: Conformal coating, filtering, and shielding protect against environmental stress
• Redundancy: Redundant components or systems maintain function despite individual failures
• Condition monitoring: Detect degradation before failure enables preventive action

Extending useful life maximizes return on investment in electronic systems.

Managing Wear-Out

Address wear-out proactively:

• Life prediction: Physics of failure models predict when wear-out will occur
• Preventive maintenance: Replace wear-out limited components before failure
• Condition-based maintenance: Monitor degradation and replace when needed
• Design life matching: Select components with wear-out life exceeding system requirements
• Graceful degradation: Design systems to maintain function with reduced capability as components age

Proactive wear-out management prevents unexpected failures and extends system life.

Warranty Period Selection

Bathtub curve informs warranty decisions:

• Coverage timing: Warranty should cover infant mortality period to capture manufacturing defects
• Cost implications: Longer warranties increase exposure to random and wear-out failures
• Competitive factors: Market expectations may require warranties extending into useful life
• Product positioning: Extended warranties can differentiate premium products
• Risk assessment: Understand failure rate in each region to assess warranty cost exposure

Warranty terms balance customer protection, competitive positioning, and cost management, and the shape of the curve determines how that balance behaves as the term is extended. A warranty that ends inside the flat region has a cost that grows roughly in proportion to its length, which makes it straightforward to price. A warranty that reaches into wear-out has a cost that grows faster than its length, so the last year of an extended term can cost more than all the preceding years combined. Knowing where wear-out begins is therefore a commercial question as much as an engineering one, and warranty return data is often the earliest and cheapest evidence a manufacturer has about the real shape of its own curve.

Flat Bathtub

Products with minimal infant mortality and distant wear-out:

• Mature products: Well-established designs built on optimized, statistically controlled processes
• Effective screening: Thorough screening has already removed the weak units before shipment
• Long-life components: Every part has a wear-out life far exceeding the intended product life
• Appears constant: The hazard rate looks nearly constant across the entire observation window
• Analysis implication: The exponential distribution is an adequate approximation, and MTBF is a legitimate summary statistic

This is the most common profile for well-engineered modern electronics, and it is a success condition rather than an exception. The important caveat is that a flat observed curve may simply mean the observation window ended before wear-out began. Concluding that a product has no wear-out region, when the fleet has only accumulated a fraction of its design life, is a frequent error in field-data analysis.

High Infant Mortality

Products with pronounced early failure period:

• New products: First production of new designs may have higher infant mortality
• Complex assemblies: More components and processes create more opportunities for defects
• Immature processes: Manufacturing processes not yet optimized
• No screening: Products shipped without burn-in or stress screening
• Improvement opportunity: High infant mortality indicates opportunity for quality improvement

High infant mortality directly impacts customer satisfaction and warranty costs.

Early Wear-Out

Products with wear-out beginning during expected life:

• Limited life components: Batteries, capacitors, or bearings with life shorter than system
• Harsh environments: Accelerated degradation from extreme conditions
• Insufficient margin: Design life margins inadequate for actual conditions
• Maintenance implications: May require scheduled replacement of wear-out items
• Design review need: Early wear-out indicates design or component selection issues

Early wear-out requires either design changes or maintenance programs to address.

Multi-Modal Distributions

Complex systems may show multiple failure populations:

• Multiple mechanisms: Different failure mechanisms peak at different times
• Component variation: Different components wear out at different rates
• Usage variation: Different usage patterns create different failure populations
• Manufacturing lots: Lot-to-lot variation creates distinct populations
• Analysis approach: May need to analyze as mixture of distributions

Recognizing multi-modal behavior enables more accurate modeling and targeted improvements. Aggregating distinct populations into a single fit is actively misleading, because the combined data can produce a plausible-looking Weibull fit whose shape parameter describes no real mechanism at all.

Humped and Roller-Coaster Profiles

Field data from complex electronic assemblies often shows a hazard rate that falls, rises to one or more intermediate humps, and only then settles, rather than descending smoothly into a flat plateau. This profile is sometimes described as a roller coaster, and it has a straightforward explanation within the superposition view of the curve.

• Staggered mechanisms: A defect population whose onset is delayed, such as a marginal solder joint that fails only after a few hundred thermal cycles, contributes a bump partway into life rather than at time zero
• Screening artifacts: A screen that removes one defect population but not another leaves the survivors of the second to fail later, displacing rather than eliminating the hump
• Deployment effects: Fleets built and installed over months mix units of different ages and different manufacturing vintages, blurring the population boundaries

The practical lesson is that an intermediate rise in field failure rate should be investigated as a distinct failure population rather than dismissed as statistical noise or accepted as the beginning of wear-out. Reliability literature has long included pointed criticism of the idealized bathtub shape on exactly these grounds, arguing that the flat middle region is assumed far more often than it is demonstrated.

Burn-In and Screening Decisions

Determine appropriate screening based on infant mortality:

• Cost-benefit analysis: Balance screening cost against warranty and customer satisfaction costs
• Screen duration: Long enough to pass through infant mortality but not waste life
• Screen conditions: Accelerated conditions to compress infant mortality period
• Pass/fail criteria: Define criteria for units failing during screening
• Process feedback: Track failures during screening to improve manufacturing

Well-designed screening programs maximize defect detection while minimizing cost and life consumption.

Maintenance Planning

Match maintenance strategy to failure pattern:

• Random failures: Corrective maintenance appropriate when failures are random and unpredictable
• Wear-out failures: Preventive maintenance replaces components before wear-out failure
• Condition monitoring: Monitor degradation to enable predictive maintenance
• Maintenance intervals: Set intervals based on wear-out characteristics
• Spare parts: Stock spares based on expected failure rates and lead times

Maintenance strategy should match the actual failure rate pattern, not assume one approach fits all. This is the central insight of reliability-centered maintenance, which grew out of the observation that scheduled overhaul was being applied to equipment whose failure rate did not increase with age. The studies behind that discipline found that only a minority of failure modes in complex systems exhibited a clear age-related wear-out region, while most followed patterns dominated by early-life or age-independent failures. For those modes, scheduled replacement is not merely wasteful; by returning units to the start of the curve, it can raise the failure rate rather than lower it.

Reliability Demonstration Testing

Plan tests accounting for bathtub curve effects:

• Pre-conditioning: Consider burn-in before reliability testing to remove infant mortality
• Test duration: Test long enough to observe intended failure region behavior
• Accelerated testing: Acceleration factors may differ for different failure mechanisms
• Failure analysis: Analyze test failures to determine which bathtub region they represent
• Data interpretation: Account for bathtub curve when extrapolating test results

Test design should explicitly address which bathtub curve region is being characterized.

Field Data Analysis

Interpret field data in bathtub curve context:

• Failure timing: Early failures likely infant mortality; later failures may indicate wear-out
• Failure rate trends: Decreasing, constant, or increasing rate indicates curve region
• Weibull analysis: Shape parameter identifies failure rate behavior
• Population effects: Aggregate data may obscure distinct failure populations
• Corrective action targeting: Different regions require different improvement approaches

Understanding which bathtub region failures come from guides appropriate corrective action.