7.3 Test process quality (ICT/FCT & retest limits)

Testing is an appraisal process designed to verify the hardware. A test station does not fix poor quality; it merely identifies functional units while segregating failures. In high-reliability EMS, one of the most concerning behaviors is “Testing Into Compliance”—the act of re-testing a failed unit continuously until it passes. This practice introduces risk. When a PCBA fails an electrical test once, its integrity is suspect. Should it pass on a subsequent attempt without a documented repair, an intermittent defect is likely present. The goal of Test Process Quality is to establish clear procedures and enforce strict retest logic across the factory to prevent field failures.

In-circuit test (ICT): the structural audit

ICT functions as the electrical equivalent of an X-Ray. It verifies that discrete components are present, oriented correctly, and not electrically shorted to their neighbors. It does not prove the complex device fully works; it only proves the device was built matching the CAD schematic.

False Call Management:

ICT mechanical fixtures rely on thousands of spring-loaded “pogo pins” contacting tiny copper test points. Microscopic dust, flux residue, or slightly warped PCB substrates can cause False Fails.

• The Baseline Metric: Station First Pass Yield (FPY) should be sustained > 95%.
• The Engineering Logic: Whenever the ICT FPY drops below 90%, it is necessary to pause the line. The fixture pins should be cleaned, and the electrostatic debris vacuumed. Ensure the strain gauge stress on the PCBA is verified rather than temporarily widening the electrical test limits to ignore mechanical noise.

Functional circuit test (FCT): the behavioral audit

FCT actively simulates the customer’s real-world application. It powers up the board, injects real signals, and executes the compiled firmware.

The “No Trouble Found” (NTF) Paradox:

This scenario occurs when a unit hard-fails the automated FCT on the line, but passes when an engineer tests it manually on the laboratory bench.

• The Risk: The automated FCT pneumatic fixture often imposes mechanical flexing stress or artificial timing constraints that diverge from how the board is situated in its final enclosure.
• The Action: Perform a formal Correlation Analysis. When the manual “Bench Pass” rate is > 90% for units that failed FCT, it suggests the software limits are set too tight, or the fixture itself is mechanically unstable.

Retest limits (the “three strikes” rule)

As an engineering leader, it is beneficial to define and electronically lock a software limit on the number of times a unique serial number can be cycled through a test station. Infinite manual loop testing can hide intermittent failures, such as fractured cold solder joints.

Decision Logic for Retest:

• Pass 1: Unit enters the test fixture for the first time.
  • Result: FAIL.
  • Action: Do not remove the board. Visually check the mechanical seating and cable connections. Retest precisely once (Pass 2).
• Pass 2 (The Only Retest):
  • Scenario A:
    • Result: PASS.
    • Status: “Conditional Pass”.
    • The Risk: There may be an intermittent structural contact or a marginal test pin.
    • The Mandate: This unit must be visually inspected under a microscope for flux flooding or contamination on the failing test pads before it reaches the customer.
  • Scenario B:
    • Result: FAIL.
    • Status: Fail. The unit requires repair. Route it to the Debug area.
• Pass 3:
  • PROHIBITED. A unit failing the exact same test twice is usually a hardware defect, not a software glitch.
  • The Action: The MES (Manufacturing Execution System) should electronically lock the serial number. Tag it as defective and keep it off the active production line until debugged.

Bonepile management

The “Bonepile” is the designated area for Failed units. It is also a highly valuable source of engineering data in the factory.

Rules of Engagement:

1. Aging Limit: Units should not remain in the Bonepile for more than 3 shifts (24 hours) without attention.
2. Disposition: Every board must be formally diagnosed (Debugged to root cause) or Scrapped.
3. WIP Cap: Should the Bonepile inventory exceed 50 total units, production should be paused. This indicates that scrap is being produced faster than the team can resolve it.

Pro-Tip: Ensure a strict “First In, First Out” (FIFO) logic is applied to the bonepile so that complex repairs are not avoided in favor of easier fixes. If a low-value board sits untouched for weeks, it is often more economical to scrap it, as the labor cost of troubleshooting begins to exceed the raw material value of the PCB.

Final Checkout: Test process quality (ICT/FCT & retest limits)

Control Element	Monitored Parameter	Engineering Limit / Rule
Retest Count	Maximum Attempts	2 (Initial Pass + 1 Single Retest).
FPY Trigger	Yield Floor Limit	< 90% requires an investigation.
Bonepile Aging	WIP Time Limit	< 24 Hours in the bin.
Fixture Maint.	Pogo Pin Cleaning	Preventative cleaning every shift.
Correlation	Bench Test vs. FCT	The delta gap should be < 5%.
Golden Unit	Physical Verification	Run the verified Master Golden Unit daily to prove the tester itself is calibrated.