3.5 Test strategy: ICT vs. functional vs. burn-in

Testing is the mandatory tax paid to convert a manufacturing gamble into a guaranteed product. Without a consistently enforced, multi-layered test strategy, the factory is not shipping hardware; it is shipping a probability of failure. The goal of manufacturing test is never to debug the core engineering design—that is the R&D team’s job. The primary purpose of factory testing is to verify that the physical unit was assembled correctly and will not fail immediately upon power-up. Relying on just a single test method leaves gaping holes in the quality defense.

ICT (In-Circuit Testing): the “bed of nails”

Scope: Structural Electrical Verification
Speed: Extremely Fast (Seconds)

ICT does not test whether the product actually “works” or runs its software; it verifies that the physical board was built exactly to the schematic. The PCBA is pneumatically pressed onto a heavy mechanical fixture loaded with hundreds of spring-loaded “pogo pins” (the Bed of Nails) that contact dedicated copper Test Points on the bottom of the board. The machine electrically isolates each component and measures its fundamental value.

The engineering reality

ICT is a direct interrogation of the BOM. It asks: “Is R1 actually 10kΩ? Is C4 actually polarized correctly? Is there a microscopic solder bridge shorting the 3.3V rail to Ground?”

• The Catch: When the Pick & Place machine places a 1kΩ resistor instead of a 10kΩ resistor, ICT flags it in milliseconds.
• The Protection: When there is a hidden solder short under a QFN processor, ICT detects it at ultra-low voltage before the main power is ever applied, preventing the board from burning up.
• The Tradeoff: The mechanical fixture is custom-machined and highly expensive ($3,000 to $10,000+). ICT is typically deployed only when production volume comfortably exceeds 1,000 units.

FCT (Functional Testing): the simulation

Scope: System Performance Validation
Speed: Slow (Minutes)

FCT treats the device as an integrated “Black Box.” It fully powers up the unit, flashes the production firmware, and mimics the real-world operating environment. It pushes digital buttons, reads analog sensors, verifies wireless transmission, and checks display outputs. FCT answers the ultimate question: “Does this device actually do what the customer paid for?”

The engineering reality

FCT is the final safety net for logical and system-level failures.

• The Catch: When the physical solder joints are perfect (ICT passes) but the silicon inside the microchip is internally defective from the factory, only FCT will catch the failure.
• The Protection: When the firmware contains a halting bug or the localized sensor calibration is completely off, FCT blocks the unit from reaching the shipping box.
• The Bottleneck: When the FCT sequence takes 4 minutes to run and the factory needs to ship 1,000 units a day, a single test station cannot keep up. Test stations must be duplicated, instantly multiplying capital equipment and labor costs.

Strict Constraint

Never accept binary “Pass/Fail” results from FCT stations. Test software must log raw, quantitative parametric data (e.g. “Output Voltage: 5.01V”). This data allows engineers to spot manufacturing drift—if the baseline voltage drifts from 5.0V to 5.2V over two weeks, the line can be halted and recalibrated before it hits the 5.5V failure limit.

Burn-in: the stress test

Scope: Infant Mortality (Early-Life Failure)
Speed: Extremely Slow (Hours to Days)

Burn-In is the rigorous process of powering the device inside a thermal chamber at elevated temperatures and maximum voltages for an extended, continuous period. It is specifically designed to prematurely filter out weak or marginal silicon components before they ever leave the factory dock.

The bathtub curve

Semiconductor reliability universally follows a “Bathtub Curve.” Failure rates are relatively high at the very beginning of the product’s life (Infant Mortality), drop to near zero in the middle (Useful Life), and spike again at the end (Wear Out).

• The Risk: Shipping directly off the SMT line without Burn-In effectively uses paying customers as the filter to catch Infant Mortality failures.
• The Protection: Running the units in a 40°C Burn-In chamber for 24 hours forces the weakest units to fail safely inside the factory. The units that survive are statistically proven to be in their highly stable “Useful Life” phase.
• The Standard: Burn-In is non-negotiable for medical devices, automotive modules, aerospace, and high-reliability industrial control systems. It is generally skipped for cost-sensitive, high-volume consumer electronics where a low return rate is deemed financially acceptable.

Flying probe: the fixtureless alternative

Scope: Prototype & Low-Volume Structural Testing
Speed: Very Slow (Tens of Minutes)

When building exactly 50 prototypes for an engineering validation run, it is difficult to financially justify a $5,000 ICT fixture. A “Flying Probe” machine solves this by deploying robotic arms carrying highly precise, motorized needles. These needles “fly” around the stationary board, landing on test points to measure components sequentially, one by one.

• The Advantage: Zero custom tooling cost. The CAD data is provided, and the machine can test a completely novel board design on day one.
• The Penalty: Cycle time. Because it measures sequentially rather than simultaneously, testing a single complex PCBA can easily take 10 to 20 minutes.
• The Use Case: New Product Introduction (NPI) builds, aerospace low-volume runs, and deep-dive troubleshooting for complex field returns.

---

Recap: Test Strategy Selection (ICT vs. Functional vs. Burn-In vs. Flying Probe)

Test Method	Scope	Key Requirement / Constraint	Pass/Fail Criteria	Action on Failure
ICT (In-Circuit)	Structural Electrical Verification	Verify BOM/schematic compliance. Fixture cost: $3k-$10k+. Justified for volumes >1,000 units.	Component values correct. No shorts/opens. Polarity correct.	Flag unit. Halt line for process correction.
FCT (Functional)	System Performance Validation	Validate “black box” functionality. Cycle time: minutes. Requires station duplication for high throughput.	Device performs all intended functions. All interfaces operational.	Block unit from shipping. Log quantitative parametric data for drift analysis.
Burn-In	Infant Mortality Screening	Stress test at max voltage & elevated temp (e.g., 40°C for 24h). Mandatory for medical/automotive/aerospace.	Unit operates continuously for duration without failure.	Scrap failed unit. Survivors enter “Useful Life” phase.
Flying Probe	Prototype/Low-Volume Structural Test	Fixtureless alternative for NPI/low volume (<1k units). Cycle time: 10-20 minutes per board.	Component values correct. No shorts/opens. Polarity correct.	Flag unit for rework/debug.