|
Louis Y. Ungar, Editor, BestTest Newsletter
Each issue of The BestTest Newsletter focuses on a
test product or service, and we have vendors explain to our readers
how to select such a product We usually have no problem finding
one or more test vendors to tackle the subject and submit a 600 or so
word "watch outs," making our readers the most
informed buyers of test products and services.
For some reason none of the vendors for Integrated
Circuit (IC) Automatic Test Equipment (ATE) stepped up to the plate
this time to explain to our readers what they should look for when they buy
an IC ATE. A friend of mine who sells IC ATEs told me a while
back that when he sells the 6 figure priced ATE, he doesn't even talk
to the test engineers. He feels that they are not the ones who
would make a decision involving so many zeros in the price, so why
bother? He did, however, admit that test engineers have a
vetoing power in deciding against a particular ATE, so
as a sales professional, he wants to make sure that the ultimate users
of the ATE do not have any objections on a technical
level.
I will not attempt here to outline detailed
specifications of such complex testers for which somebody (hopefully a
test professional) will make a purchasing decision that goes far
beyond the $1 million+ price tag. When you consider the test
programming costs, the test fixturing costs, test simulation and
design for testability costs associated with the IC or System-on-chip
(SOC), you can easily triple the ATE price tag to estimate the
five-year ownership cost. Rather, I want to explore alternatives
that have been tried by those who just cannot or do not want to come up with the
investment - regardless of any expected return on that
investment.
Some of the alternatives we have seen lately are
worth exploring:
-
Lower priced ATE compromising high precision
parametrics
-
Used (refurbished) ATE
-
Reduced ATE requirements through sampling
-
Improved Design for Testability to reduce ATE
requirements
-
Built-In Self Test to reduce ATE requirements
-
Foregoing IC Tests altogether
Lower Priced ATE - We do not live under an
illusion that a $50,000 tester can do the same job as a $1,000,000
one, but the very first thing we have to do is review the parameters of
the IC device under test (DUT). Clearly note what parameters cannot
be accurately measured or stimulated by a) any ATE, b) by a lower
priced ATE. Once you identify these parameters there are a
number of economic models you can create to calculate the frequency of
such occurrences, the cost impact of each occurrence, and any other
factors that can mitigate the situation (such as Design for Testability,
which is discussed later). You need to be aware of the penalty costs
associated with foregoing functional capabilities in the ATE, but if
this is something that you can live with, the savings may be
substantial.
Used (refurbished) ATE - There is certainly
a great deal of prestige in owning one or more million dollar ATEs,
and all of this prestige is lost if you buy the equivalent ATE on the
secondary market. How much is prestige worth to your
organization? We have ran previous issues dealing with used test
equipment of all kinds, and it is often possible to buy
full support from the original ATE manufacturer. (Naturally, the
time to negotiate such support is not when the ATE manufacturer holds
out hope that you will buy a new one.) Test engineers should be
cautious in making this decision, however. While it may appear
that the purchase price of the used ATE is lower, the company's
financial people may rightfully choose to lease a new
ATE. Their decision would involve tax consequences the test
engineers may not understand.
Sampling - Using an ATE for sampling (that
is testing only some of the DUTs) may not be cost effective at all in
some organizations. The major costs of the ATE - purchase price,
test program development, fixturing, etc. - are required regardless of
the number of units you test. So, in fact, the larger the number
of units you test, the greater the cost spreads and becomes less
expensive to test per chip. The only time sampling makes sense is when
you have exceeded your ATE capacity. At that point, you may
consider buying your second ATE as a used ATE.
Design for Testability - One of the
best ways to reduce the functional testing requirements that ATE has
to deal with is to design controllability and observability features
within the chip. Scan techniques are quite effective in
accessing internal points, though they do not always assist in
diagnosis. For that reason, it is important for the chip
designers to consider exactly how the circuit will be tested with the
ATE. Parametric tests and test vector depth, which are the
greatest contributors to the high price of IC ATEs, can be tapered
with testable-friendly designs. It may be possible that with DFT
+ Used ATE, you can achieve all the features that your new ATE would
have provided - but at substantial savings.
Built-In Self Test - I had an interesting
discussion last week with a BIST tool maker about who benefits most
from BIST - the IC manufacturer, or the board and system manufacturer
end user who buys the chip equipped with BIST? He felt it was
the end user, and I felt it was the IC vendor. Hopefully, we are
both correct, but for the purpose of this article, I urge your IC's BIST
designers to consider the ATE that will (or now can) be used to test
your chips. There are substantial opportunities to save with
BIST at every level of construction.
Forego Testing ICs - This is the most
dangerous, and probably the most popular when it comes to incoming
inspection. Certainly, IC quality has increased, and as a
purchaser of ICs, you are looking for less than 100 parts per million
(ppm) failing. Of course, the only way you could be sure that this
number is close to what you are getting is to test - at a high
cost. So most board and system manufacturers will take the IC
vendor's word for it that the chips were thoroughly tested.
(Note, BIST could be used to find bad chips - so would IC vendors want
you to do that?)
For an IC vendor, it is inconceivable to sell chips that have not been
tested. First, yields are usually not that high, and even with a
95%+ yield they would be selling chips with a 50,000 ppm failing.
But taking
this example further, assume that some test is done that will detect
30,000 of the bad chips - resulting in 20,000 bad
chips/million sold. With no incoming IC test performed by the
end user, these 2% bad chips may not be discovered until board or even
system test and only about half of them would be discovered at
all. Most, if not all faults would be considered to have failed
during board assembly. (The rest would be discovered by the system end
user). Based on such flimsy evidence, could the IC vendors be
blamed for systems failing after manufacturing and after fielding? Probably
not.
Conclusion
We have offered some alternatives to high priced IC ATEs, not because
we want to compromise on the tests they perform, but because it may not
be economical in all instances to buy the latest and the greatest ATE. From a technical perspective there are a number of
other ways we could eliminate faulty chips from winding up in
electronic systems. The technical aspects will undoubtedly be
compromised. But we must remember that test is ultimately a
technical solution to an economic problem.
|
Newsletter
