Data processing: measuring – calibrating – or testing – Testing system – Of circuit
Reexamination Certificate
2001-07-25
2004-03-23
Hoff, Marc S. (Department: 2857)
Data processing: measuring, calibrating, or testing
Testing system
Of circuit
C324S765010
Reexamination Certificate
active
06711519
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
An object of the present invention is a method for the testing of electronic components taking into account the drift in the mean of the output response. It can be used especially in the field of the testing of semiconductor electronic components, especially components such as integrated circuits. In the prior art, there is a known method for testing electronic components that reduces the unit testing time for each of the components. The value of the invention is that it proposes a test method that, first of all, reduces the unit testing time and, secondly, can be used to monitor drifts in responses of these tests, these drifts being caused, for example, by an as yet tolerable but possibly disturbing modification of the tested components.
A component to be tested is therefore subjected to a series of unit tests for the testing of each of its functions. A unit test of a function generally comprises the following elementary steps:
at an initial date D
0
, a terminal of the component, considered to be an input terminal with respect to the test, is subjected to an electrical potential V
e
,
there is a wait, for a specified time period, for the appearance and stabilization of a response at a second terminal of this component, this second terminal being considered to be an output terminal with respect to this test,
at a nominal measurement date Dm
0
at the end of the time period (Dm
0
−D
0
), this response is measured. In one example, this response is a value V
s
of a potential at this output terminal.
Then, to ascertain that the component has given an acceptable or unacceptable response to this unit test,
the measured response is compared with fixed acceptance criteria. Namely, in this example, the value of the potential VS is compared with lower and upper acceptance limits. These limits are generally given in manufacturing specifications.
Since the electronic component has a known structure, the time period at the end of which it is possible to observe an expected response with respect to the test is determined theoretically. This theoretical time period (Dm
th
−D
0
) is for example computed by the designer of the electronic component. Furthermore, in order to avoid having an excessively high failure rate for this test, a safety margin is taken into account by generally choosing the nominal measurement date Dm
0
which is later than the theoretical measurement date Dm
th
of this test.
Then, the nominal measurement date Dm
0
is used as a measurement date for all the components to be tested. This choice, which ensures a response in conformity with the test, has the consequence of considerably increasing the total execution time of a series of tests.
2. Description of the Prior Art
The document WO-A-97/45748 describes a known test method used to reduce the duration of each unit test of a series of tests. To this end, the test method comprises:
a first “learning phase” carried out on a population of acceptable components.
The nominal measurement date Dm
0
is used to perform tests on the first components of the batch. Only the components that have given an appropriate response (within the acceptance limits) are incorporated into the population of acceptable components of this batch, this population constituting the learning population. Then, as far as possible, a measurement date Dm
i
earlier than the nominal measurement date Dm
0
is determined. The tests performed with this measurement date Dm
i
must give a still acceptable response with respect to the acceptance limits,
a second “application phase” during which all the other components of the batch are tested, by using the earliest measurement date Dm
i
determined beforehand.
To determine the earliest measurement date Dm
i
, the elementary steps of the test defined here above are reiterated on each of the components of the learning population, the measurement date applied being reduced at each iteration, either dichotomously or step by step. To choose the measurement date Dm
i
among the intermediate measurement dates tested, Dm
i-TEST1
, Dm
i-TEST2
, the following are compared by means of a criterion:
a statistical image of the responses obtained with at least one tested intermediate measurement date, for example Dm
i-TEST1
, with
a statistical image of the responses of this very same learning population, obtained at the nominal measurement date Dm
0
.
A statistical image comprises all the responses, of the VS type, obtained during the test and especially the computation, on the basis of these values, of the mean M and the standard deviation S, and of all the other statistical values as a function of this mean M and this standard deviation S.
For example, for each statistical image, a statistical value referenced CP is obtained. This value is equal to a ratio between a difference of acceptance limits and the standard deviation S of this statistical image. The difference between the acceptance limits is generally given by a manufacturer's tolerance.
Another statistical value can also be defined to characterize this statistical image. This other value referenced CPK is then equal to the ratio between an absolute value of a difference between the mean and an acceptance limit and the standard deviation S of this statistical image.
To compare the statistical images with one another, their respective statistical values CP or CPK are compared. The earliest possible measurement date Dm
i
is chosen from the tested intermediate measurement dates such that its corresponding statistical value CP
i
obtained with the measurement date Dm
i
remains within a specified proportion of the appreciation criterion CP
o
, where CP
o
is the statistical value characterizing the statistical image obtained with the nominal measurement date Dm
0
.
In a first example, if a wide range of responses is observed for a tested intermediate date Dm
i-TEST1
without any predominance of any value, it means that the behavior of the electronic component under this test is not reliably reproducible on this date. According to the method of the invention, the standard deviation S
i-TEST1
of the statistical image is great and therefore the corresponding value CP
i-TEST1
will be low, and will certainly no longer be within the set criterion of the specified proportion of CP
o
. Then, this measurement date Dm
i-TEST1
will no longer be chosen as the earliest possible measurement date Dm
i
, and Dm
i
will necessarily be greater than this tested date Dm
i-TEST
.
The method of determining the earliest measurement date Dm
i
may include another acceptance condition, for example a condition by which this date is necessarily greater than a theoretical minimum measurement date Dm
min
.
The solution proposed in this document WO 97/45748 is problematic because the learning phase preliminary to the execution of the application phase is excessively lengthy. Furthermore, the earliest possible measurement date determined during this learning phase may result in a greater discard rate for the electronic components tested because it may cause these components to be unjustifiably rejected from the good quality production. Indeed, since the learning phase is long, it cannot be carried out too frequently, and then an inappropriate measurement date may be used to test a part of the batch.
From the teaching of the document FR 99 09055, there is a known method also comprising a learning phase and an application phase; and further adding an adjusting phase.
However, this method provides that the learning phase used to define the earliest possible measurement date Dm
i
is obtained by comparing:
a statistical image between the responses of the learning population obtained with the nominal measurement Dm
0
and
a cumulated statistical image, this image corresponding to the image of all the responses obtained on this same population with the tested intermediate measurement date, for example Dm
i-TEST1
, and with the responses obtained on this population on the nominal measurement date Dm
0
.
To compare the two statist
Baran Mary Catherine
Bealach no bo Finne teo/TA Galaxy
Hoff Marc S.
Patterson Thuente Skaar & Christensen P.A.
LandOfFree
Method for the testing of electronic components taking the... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method for the testing of electronic components taking the..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method for the testing of electronic components taking the... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3196353