TDEB Top Bar

Trending
Semiconductor Industry News — The Digital Electronics Blog
Follow Us

TDEB Masthead

The Semiconductor Hub

Semiconductors.  Chips and Systems.  Innovation.

Indexed by leading Search Engines, AI and referenced by Engineers worldwide 500+ Articles 10M+ Pageviews 30+ Reports 50+ Citations

What is Final (End-of-Line) Silicon Testing? Flow, Bins and Screens

Murugavel Ganesan
by
0
Share Facebook X LinkedIn WhatsApp Email

Final test, also called end-of-line (EOL) or package test, is the last production test a chip goes through before it is shipped. It is run on every fully assembled, packaged device on automatic test equipment (ATE), usually at one or more temperatures, to confirm that the part still works after packaging, meets its data sheet limits, and belongs in the right speed or quality bin. It is the final gate between the factory and the customer.

Packaged integrated circuits at the end-of-line test stage of semiconductor manufacturing

In the context of silicon testing, end-of-line testing is a critical step in manufacturing silicon chips and integrated circuits. It is performed on fully assembled devices that have already been packaged, and its purpose is to verify that they meet their specified requirements and are free of defects before they are shipped to customers.

Key takeaways

  • Final test happens after assembly, so it catches defects that wafer sort cannot: damage from dicing, wire bonding, bumping, molding or substrate faults.
  • It combines functional (mostly structural scan and BIST), parametric and temperature or stress screening, and it sorts parts into bins.
  • Only packaged parts can be tested at full I/O speed and with final thermal behavior, so many at-speed and analog tests live here.
  • Reliability qualification (such as HTOL) is a sample-based activity; final test applies short screens to every unit.
  • A typical flow is: final test, optional burn-in and retest, optional system-level test, then quality assurance sampling and packing.

How final test differs from wafer sort

Production testing has two main insertions. The overview is in What is Silicon Production Testing? The table below compares them.

AspectWafer sort (probe)Final test (end of line)
Device under testBare die on the waferPackaged part in a socket
ContactProbe card on pads or bumpsHandler and test socket on package pins or balls
Main aimRemove bad die before paying for packagingProve the shipped part is good and bin it
Speed limitsProbe parasitics limit high-speed I/O testsFull-speed I/O and SerDes tests possible
TemperatureOften one temperature, sometimes hotCommonly hot and cold, depending on grade
Defects caughtFab defectsFab escapes plus assembly defects

What final test checks, in brief

End-of-line testing typically combines three kinds of test. Functional and structural tests confirm the logic works. Parametric tests measure electrical values such as leakage and supply current against the data sheet. Temperature and stress screens run the part at the hot and cold limits of its grade, and for some products add burn-in. Each part then lands in a bin: sell at top speed, sell at a lower grade, or scrap. The details and a worked binning example are in the deep dive below.

Why end-of-line test matters

End-of-line testing ensures that only high-quality, functional devices are shipped to customers. By identifying and removing defective devices, it reduces the risk of product failures in the field, improves customer satisfaction and minimizes the cost of returns and warranty claims. A part caught at final test costs the price of the die and package; the same part failing on a customer's board, or in a car, costs many times more.

FAQ

Is final test the same as end-of-line test?

In semiconductor manufacturing the terms are used for the same step: the last electrical test on packaged parts before shipping. Some board and system makers also use "end of line" for the last test of their own assembled product.

Who runs final test?

Often an outsourced assembly and test company (OSAT), using a test program written by the chip company's product and test engineers.

How long does final test take?

From well under a second for simple parts to minutes for large SoCs, and longer when system-level test is added. Multi-site testing spreads the cost across several parts tested in parallel.

Does final test check reliability?

It screens for weak parts with temperature, voltage and, where needed, burn-in. Lifetime reliability is demonstrated separately on qualification samples.

Deep dive: the numbers and details

This part holds the test content in detail, the step-by-step flow, a worked binning example and the common mistakes. Skip it if the overview is all you need.

What final test checks

End-of-line testing typically includes a combination of functional, parametric and stress testing.

Functional and structural tests

The devices run a series of tests to verify they perform their intended functions correctly. For digital logic, most of the coverage comes from scan-based ATPG patterns (stuck-at and transition-delay faults) and memory BIST, with a smaller set of functional patterns for interfaces, clocking and analog blocks. Speed binning runs selected patterns at several frequencies or voltages to find the grade each part can sell at.

Parametric tests

The devices are tested for key electrical parameters: continuity and opens and shorts, leakage, supply current (including IDDQ), output drive levels, input thresholds, timing, and for analog and mixed-signal parts, gain, offset, noise and linearity. Pin capacitance, which the original post listed, is normally characterized on samples rather than measured on every unit.

Temperature and stress screening

The original post described reliability testing on the devices under temperature, voltage and stress conditions to simulate real-world operation. In practice there are two separate things. Final test applies short screens to every unit: test at the hot and cold limits of the grade, voltage margin tests, and, for products that require it, burn-in followed by retest. Long reliability tests that simulate years of use (for example HTOL to JEDEC JESD47, temperature cycling, humidity tests) are run on qualification sample lots before volume release, not on each shipped part.

A typical end-of-line flow

  1. Final test 1 (FT1), often at hot temperature, full program.
  2. Burn-in (see What is Silicon Burn-In Testing?), where required (common for automotive and some server parts), to trigger early-life failures.
  3. Final test 2 (FT2), often at cold or room temperature, to catch parts damaged or weakened by burn-in and to confirm cold behavior.
  4. System-level test (SLT), for complex SoCs: the part boots an operating system on an application-like board.
  5. Quality assurance (QA) sample: a sample of passing parts is re-tested on a separate setup to confirm the main test is correct.
  6. Visual inspection, marking, tape and reel, and shipment.

Temperature limits depend on the product grade. As a common example, AEC-Q100 Grade 1 automotive parts are specified from -40 to +125 degrees C ambient, so their final test usually covers both ends.

Worked example: binning a batch

The numbers below are illustrative. A lot of 10,000 packaged processors goes through FT1 and FT2:

BinMeaningCountShare
Bin 1Pass at top frequency7,20072.0%
Bin 2Pass at lower frequency grade2,10021.0%
Bin 5Fail, opens or shorts (often assembly)1501.5%
Bin 6Fail, scan or BIST3503.5%
Bin 8Fail, leakage or IDDQ2002.0%

Final test yield is 93 percent, of which 72 percent sells at the top price. A sudden rise in Bin 5 points to the assembly line; a rise in Bin 8 points back to the fab. This feedback is a large part of the value of end-of-line test.

Common mistakes

  • Relying on wafer sort alone. Assembly adds its own defects, so final test is not optional for most products.
  • Poor tester correlation. If two testers or sites disagree, good parts are scrapped or bad parts escape. Correlation units and guard bands keep sites aligned.
  • Socket and contact issues. Worn sockets cause false fails and retests; tracking contact resistance and retest rates is standard practice.
  • Testing only at room temperature. Many timing and leakage failures appear only at the temperature extremes.
Murugavel Ganesan
About the Author
Murugavel Ganesan
Product Director | SerDes · IVN · Edge AI · Compute | Ex Intel · Infineon

A semiconductor product leader working across product architecture, definition, strategy, and silicon and software execution, from concept through volume production. He has worked with teams across Asia, Europe and the USA, including at Intel, Infineon and ArrayComm.

His work covers ASICs, SoCs, FPGAs and chiplets for wireless, wireline, automotive, IoT and AI, including PCIe, CXL and DDR interfaces alongside RF and mixed-signal design.

Share Facebook X LinkedIn WhatsApp Email

Post a Comment

0 Comments

Your comments will be moderated before it appears here.

Post a Comment (0)

#buttons=(Ok, Go it!) #days=(20)

Our website uses cookies to enhance your experience. Learn more
Ok, Go it!