Page 305 - Semiconductor Manufacturing Handbook

Geng(SMH)_CH19.qxd 04/04/2005 20:00 Page 19.32

INSPECTION, MEASUREMENT, AND TEST

19.32 FINAL MANUFACTURING

devices; they may be made for a variety of reasons including lowering COT, obsolescence, or seek-
ing partnerships advantageous to both parties.

19.3.1 Understand the Test Need

The first thing to do is to understand the end application focus for the IC. Based on that, IC design-
ers will design a chip that will function in that application. The IC designers will know what kind of
source (stimulus) and measure (response) each circuit and pin will require. A device data sheet is a
description of how the IC operates, its end application, and every parameter critical to its operation.
These parameters and the functionality of the part will dictate what kind of resources will be required
in the tester. The market segments of the various common types of devices were described in Sec. 19.2.
The device data sheet specifications are design guidelines for designing the IC into its end applica-
tion. Some of these rules are more critical than others and will be dependent on the design of the IC
and the end application itself. The limits are the not-to-exceed low and high numbers that define the
performance.
Characterization was described in Sec. 19.1, and a decision may be made to characterize the part
using bench equipment where an application mock-up is used, or test equipment can be used for the
application’s stimulus and response. Once a proper characterization is done, then the production
decision of rack and stack versus ATE needs to made.
ATE Versus Rack and Stack. ATE refers to a fully integrated test solution supplied usually by one
manufacturer. The advantage to selecting an integrated test solution is that a single vendor is
accountable for all the engineering, manufacturing, and support of integrated hardware and software,
documentation, spare parts, calibration, and diagnostics.
Rack and stack refers to the practice of building a test solution from stand-alone components.
This would encompass all the responsibilities applied to building the tester architectures and soft-
ware to control them, outlined in Sec. 19.2. When comparing the costs of an integrated solution ver-
sus rack and stack, it is important to factor into this decision the engineering time required and also
assume a certain amount of time-to-market to a rack-and-stack solution—not just the “hardware
costs” of each. The IC manufacturer must weigh all of these factors when making the rack and stack
versus ATE decision.
Different markets have different requirements, but for most manufacturers, cost-of-test and time-
to-market are the two biggest factors determining the success and profitability of an IC product, as
explained in Sec. 19.1.
Guaranteed Versus Typical Limits. On a device data sheet, most parameter limits are tested in pro-
duction and can be guaranteed to be between the minimum and maximum specification. Other test
parameters are not tested in production, but have been characterized to “typically” be at the nominal
specification. There are various reasons that a limit may be typical instead of tested. For example, it
may be difficult to test a particular parameter in production because of fixturing, long test times, or
simply the lack of failure over a large number of devices. Typical limits lower the COT and perhaps
even reduce the requirements that the test solution needs for production. However, in some market
segments, preferences and potential price premiums may be given to manufacturers with guaranteed
limits. The IC manufacturer needs to determine these tradeoffs. If the decision is to go with guaran-
teed limits, then the correlation and repeatability requirements need to be determined. The role that
repeatability plays was described in the six sigma discussion in Sec. 19.2. The end application will
also determine how close the correlation of known good devices and production units needs to be.
Typically the closer the correlation required, the more resolution and accuracy the production solu-
tion needs. The ultimate judgment of how well a test is screening bad parts is from the customer and
the rate of device returns. Sometimes manufacturers even do production correlation by comparing
the tally of good and bad devices.
Test Philosophies. There are different philosophies regarding what IC manufacturers want to
achieve with testing. Largely this is determined by the market norms and the manufacturer’s desired

Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)
Copyright © 2004 The McGraw-Hill Companies. All rights reserved.
Any use is subject to the Terms of Use as given at the website.