Electricity: measuring and testing – Fault detecting in electric circuits and of electric components – Of individual circuit component or element
Reexamination Certificate
1997-10-08
2002-01-22
Karlsen, Ernest (Department: 2858)
Electricity: measuring and testing
Fault detecting in electric circuits and of electric components
Of individual circuit component or element
C324S757020
Reexamination Certificate
active
06340894
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
This invention relates to electrical test equipment for semiconductor devices. More specifically, the invention relates to an apparatus and method, which are used to perform dynamic burn-in and full electrical/performance/speed testing on discrete nonpackaged or semi-packaged dies.
2. Background of the Invention
Semiconductor devices are subjected to a series of test procedures in order to assure quality and reliability. This testing procedure conventionally includes “probe testing”, in which individual dies, while still on a wafer, are initially tested to determine functionality and speed. Probe cards are used to electrically test die at that level. The electrical connection interfaces with only a single die at a time in wafer; not discrete die.
If the wafer has a yield of functional dies which indicates that quality of the functional dies is likely to be good, each individual die is assembled in a package to form a semiconductor device. Conventionally, the packaging includes a lead frame and a plastic or ceramic housing.
The packaged devices are then subjected to another series of tests, which include burn-in and discrete testing. Discrete testing permits the devices to be tested for speed and for errors which may occur after assembly and after burn-in. Burn-in accelerates failure mechanisms by electrically exercising the devices (DUT) at elevated temperatures, thus eliminating potential failures which would not otherwise be apparent at nominal test conditions.
Variations on these procedures permit devices assembled onto circuit arrangements, such as memory boards, to be burned-in, along with the memory board in order to assure reliability of the circuit, as populated with devices. This closed assembly testing assumes that the devices are discretely packaged in order that it can then be performed more readily.
If the wafer has a yield of grossly functional die, it indicates that a good quantity of die from the wafer are likely to be fully operative. The die are separated with a die saw, and the nonfunctional die are scrapped, while the rest are individually encapsulated in plastic packages or mounted in ceramic packages with one die in each package. After the die are packaged they are rigorously electrically tested. Components which turn out to be nonfunctional, or which operate at questionable specifications, are scrapped or devoted to special uses.
Packaging unusable die, only to scrap them after testing, is a waste of time and materials, and is therefore costly. Given the relatively low profit margins of commodity semiconductor components such as dynamic random access memories (DRAMs) and static random access memories (SRAMs), this practice is uneconomical. However, no thorough and cost effective method of testing an unpackaged die is available which would prevent this unnecessary packaging of nonfunctional and marginally functional die. Secondly, the packaging may have other limitations which are aggravated by burn-in stress conditions, so that the packaging becomes a limitation for burn-in testing.
It is proposed that multiple integrated circuit devices be packaged as a single unit, known as a multi chip module (MCM). This can be accomplished with or without conventional lead frames. This creates two problems when using conventional test methods. Firstly, discrete testing is more difficult because a conventional lead frame package is not used. Furthermore, when multiple devices are assembled into a single package, the performance of the package is reduced to that of the die with the lowest performance. Therefore, such dies are tested on an individual basis at probe, using ambient and “hot chuck” test techniques, while still in wafer form. In other words, the ability to presort the individual dice is limited to that obtained through probe testing.
In addition, there is an increased interest in providing parts which are fully characterized prior to packaging. This is desired not only because of the cost of the package, but also because there is demand for multi-chip modules (MCMs), in which multiple parts in die form are tested and assembled into a single unit. While there are various techniques proposed for testing, burning in and characterizing a singulated die, it would be advantageous to be able to “wafer map” the die prior to assembly with as many performance characteristics as possible. Ideally, one would want to be able to map the wafer with full device characterization.
MCMs create a particular need for testing prior to assembly, as contrasted to the economics of testing parts which are discretely packaged as singulated parts. For discretely packaged parts, if the product yield of good parts from preliminary testing to final shipment (probe-to-ship) is, for example, 95%, one would not be particularly concerned with packaging costs for the failed parts, if packaging costs are 10% of the product manufacturing costs. Even where packaging costs are considerably higher, as in ceramic encapsulated parts, testing unpackaged die is economical for discretely packaged parts when the added costs approximates that of cost of packaging divided by yield:
C
DIE
×
C
PACKAGE
Yield
=
C
DIE
×
C
ADDL
.
KGD
where
C
=
cost
C
DIE
=
manufacturing
cost
of
functional
die
C
ADDL
.
KGD
=
additional
cost
of
testing
unpackaged
die
in
order
to
produce
known
good
die
(
KGD
)
Note that in the case of discretely packaged parts, the cost of the die (C
DIE
) is essentially not a factor. This changes in the case of MCMs:
(
C
DIE
)
×
(
number
of
die
)
Yield
×
C
_
PACKAGE
=
C
DIE
×
C
ADDL
.
KGD
Note that again C
DIE
is not a factor in modules having identical part types; however, the equation must be modified to account for varied costs and yields of die in modules with mixed part types.
With MCMs, the cost of packaging a failed part is proportional to the number of die in the module. In the case of a x16 memory array module, where probe-to-ship yield of the die is 95%, the costs are:
16
0.95
×
C
PACKAGE
=
C
ADDL
.
KGD
so the additional costs of testing for known good die (KGD) may be 16 times the cost of testing an unrepairable module and still be economical. This, of course, is modified by the ability to repair failed modules.
Testing of unpackaged die before packaging into multi-chip modules would be desirable as it would result in reduced material waste, increased profits, and increased throughput. Using only known good die in MCMs would increase MCM yields significantly.
Testing unpackaged die requires a significant amount of handling. Since the test package must be separated from the die, the temporary packaging may be more complicated than either standard discrete packaging or multichip module (MCM) packaging. The package must be compatible with test and burn-in procedures, while securing the die without damaging the die at the bondpads or elsewhere during the process.
In U.S. Pat. No. 4,899,107, commonly assigned, a reusable burn-in/test fixture for discrete TAB die is taught. The fixture consists of two halves, one of which is a die cavity plate for receiving semiconductor dies as the units under test (UUT); and the other half establishes electrical contact with the dies and with a burn-in oven.
The first half of the test fixture contains cavities in which die are inserted circuit side up. The die will rest on a floating platform. The second half has a rigid high temperature rated substrate, on which are mounted probes for each corresponding die pad. Each of a plurality of probes is connected to an electrical trace on the substrate (similar to a P.C. board) so that each die pad of each die is electrically isolated from one another for high speed fu
Farnworth Warren M.
Hembree David R.
Wood Alan G.
Gratton Stephen A.
Karlsen Ernest
Protigal Stanley N.
LandOfFree
Semiconductor testing apparatus including substrate with... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Semiconductor testing apparatus including substrate with..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Semiconductor testing apparatus including substrate with... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2852145