 Sophisticated
device modeling is key to optimizing automatic test
generation
By Adam Ley
Chief Technologist
How a boundary-scan test-generation
engine handles the non-boundary-scan devices in a design
is critical to the time it takes to develop a test (time-to-test),
the amount of coverage the test will achieve, and the
safety of the board. Over the years, ScanWorks’
test-generation engine has been improved repeatedly
with an eye toward reducing time-to-test with optimal
coverage and impeccable board safety. The way ScanWorks
handles cluster or non-boundary-scan device models plays
a major role in this.
A critical goal of ScanWorks is to
require as little manual involvement as possible while
generating tests with an acceptable level of coverage.
After a test has been generated, the user can decide
how much effort should be expended to improve the test
coverage or modify other aspects of the test’s
operation.
ScanWorks addresses this goal from
the outset – the building of the design description.
During this preliminary step, many of the cluster model
and BSDL (Boundary Scan Description Language) files
can be accessed and automatically downloaded from the
device libraries on ScanWorks’ maintenance benefit
web pages. In addition, in-house model libraries can
be automatically incorporated into the test. It is always
a good practice for the test engineer or technician
to make sure that all of the passive devices, such as
inductors, capacitors, resistors and others have been
identified and their models included in the design.
As the test developer is reviewing the design, he or
she should note other types of cluster devices but need
not go to the trouble of modeling them during the initial
design description. Should additional coverage be needed
later, the developer will already know which device
types to consider modeling in order to achieve greater
coverage.
ScanWorks’ First Pass
Once the test engineer or technician
is ready for ScanWorks to take its first pass at generating
a test, the test-generation engine will begin its analysis
of the design. ScanWorks examines the cluster models
for all non-boundary-scan devices to determine each
one’s operating modes. For example, a typical
“245” eight-bit transceiver would have three
modes, isolation where no data is being transferred
because all outputs have been turned off, A-to-B mode
where A pins are enabled as inputs and B pins are enabled
as outputs, and B-to-A mode where B pins are enabled
as inputs and A pins are enabled as outputs.
Subsequently, the ScanWorks test generator
begins to build a test by selecting each device’s
operating mode so that, first, the safety of the printed
circuit board will be ensured and, second, test coverage
for shorts will be maximized. To do this, ScanWorks
must examine each device within the context of the design.
For example, if ScanWorks has access
to a non-boundary-scan cluster device from one direction
but does not have enough information about the device
and the design to know what is beyond the device, then
it cannot safely drive test patterns beyond the device.
To do so would be to jeopardize the safety of the board
by setting up a situation where bus contention might
arise. In this case, boundary-scan test coverage would
extend to the  cluster
device that is accessible from a boundary-scan device
but not beyond it. Test coverage can be extended later,
at the discretion of the test engineer or technician
who is developing the test.
This is a complex process because the
operating mode selected for each device has a ripple
effect on other devices in the design. The algorithms
used by ScanWorks’ test-generation engine are
continually fine-tuned to optimize this process. Once
operating modes are selected for all devices, the test
vectors for shorts are generated.
Seeking Opens
Following this first pass, the test
generator automatically performs additional passes which
seek to identify the operating modes that devices can
be placed in so opens can be tested where this test
coverage was not gained during the first pass. Once
this information has been processed, additional test
vectors for opens are generated and included in the
test set.
Of course, during any test-generation
pass, the test generator identifies any cluster device
operating mode that simply must be taken into account.
These types of modes are often hard-wired into the design.
When the test generator encounters these conditions,
it will constrain the test so that the hard-wired operating
modes are maintained.
Iterative Passes
After the initial test generation,
the test developer may continue to invoke the ScanWorks
test generator to make additional passes to further
maximize test coverage. At this point, the engineer
or technician who is developing the test can decide
to provide additional information to ScanWorks, such
as cluster models for those devices that have not been
modeled. This added information is used by the test
generator in these successive passes to obtain higher
and higher levels of test coverage.
 This
type of iterative process gives the test developer the
option to invest his or her development time where it
will have the biggest payoff in terms of test coverage
or other test performance criteria. Quite often, ScanWorks
will generate a test with the level of coverage required
by the developer. Later, when the developer has time
to devote to it, the test can be passed through the
ScanWorks test-generation engine again with additional
modeling information and the test coverage can be increased.
Throughout this process, ScanWorks
continues to evaluate the operating modes selected for
cluster devices in light of any possible bus conflicts
which may ensue. Should the developer manually dictate
an operating mode that might establish a hazardous bus
contention, ScanWorks would notify the user of the possibility
for board damage. At that point, the developer has the
option of continuing with the operating mode originally
selected or revising the mode to avoid the possibility
of bus contention.
Optimizing Automatic Test Generation
There are many facets to automatic
test generation, including time-to-test, test coverage,
board safety and the requirements for manual developer
involvement. Of course, the most effective test-generation
engine simultaneously optimizes all of these factors
to provide the best balance for the inevitable tradeoffs
that arise. For example, a test with high coverage might
endanger the safety of the board. The algorithms of
ScanWorks’ test-generation engine provide an efficient
way to quickly generate high-coverage and safe tests
while giving test developers the information they need
to further refine the tests generated and decide how
best to invest their time.
|
