TestProNews Q1 - 2006
Understanding Test Performance Analysis
Companies spend a lot of time and money developing their tests stations. They willingly
incur this expense in order to collect data from their tests to improve their product
designs and manufacturing processes. Unfortunately, because many companies do not
have the proper tools, they unknowingly spend too much in labor costs to organize
and access test data and reports.
Companies most often spend too much attempting to:
- Learn which tests are failing the most frequently
- Learn why products are failing tests
-
Determine if there is a particular pattern to the test data
-
Make the data and reports easily accessible to key people in the company
|
|
Figure 1: Multi-dimensional TPA allows test engineers and managers to immediately analyze a test data cube from an unlimited number of perspectives.
Enlarge Image
|
Engineers and managers want to analyze how their tests are performing, which is
the role of test performance analysis (TPA) in a test data management (TDM) system.
Defining TPA
TPA is the multi-dimensional data analysis that gives immediate visualization of
test data from different perspectives. The information gained from TPA allows engineers
and managers to quickly identify problem areas and take immediate action to improve
product capabilities and manufacturing efficiency. By representing the test data
in a multi-dimensional cube, the test engineering team is able to analyze and view
data from an unlimited number of perspectives (Figure 1). Some of the most common
perspectives are combinations of test results by:
- date
- test station
- product
- product line
- operator
- test revision
|
The technology that enables the immediate, multi-dimensional viewing of TPA is online
analytical processing (OLAP). Without TPA, reports take hours and even days. With
TPA, these same reports are visible in seconds, even with gigabytes of data. For
more information on OLAP, see the
An Overview of OLAP Technology
article in this newsletter.
Example of TPA
A commonly used application of TPA best illustrates its capabilities and benefits.
For an example, consider a system of three test stations whose purpose is to characterize
prototype transmitters. The transmitter family in this example consists of three
types: a low cost transmitter, a high performance transmitter, and a government,
special-ordered transmitter for use in military applications.
Test engineering teams commonly want to first determine which product tests are
most frequently failing. For this example, this is accomplished by viewing:
Failure Total
by all transmitter types
Figure 2’s TPA results show in spreadsheet and graphical form that the low cost
transmitters have the highest number of failures. With this information, the test
engineering team may want to perform additional TPA to pinpoint the cause of the
low cost transmitters failures.
|
Figure 2: To determine which product tests are failing most frequently, the engineering team can use one-dimensional TPA to view failures by product.
Enlarge Image
|
A second dimension to add is failures by oper-ator. By adding this second dimension
and narrowing the analysis to the low cost transmitters, the combined TPA becomes:
Failure Total
by low cost transmitters
by operator
Figure 3’s TPA results identify that Cindy and Jeff have a significantly larger
number of failures with low cost transmitters than the other testers.
A third dimension to add is failures by test stations. With this third dimension,
the combined TPA becomes:
Failure Total
by low cost transmitters
by operator
by test station
|
Figure 3: A two-dimensional TPA of failures by product by operator shows operators 'Cindy' and 'Jeff' have a significantly larger number of failures.
Enlarge Image
|
These TPA results, shown in Figure 4, identify that most failures are occurring
on test station 3 when either Cindy or Jeff is the operator.
By adding date as a fourth dimension, the TPA will identify if the failure problem
is occurring consistently across all dates. Other measurements that can help visualize
the test performance include the percentage of failures and the total number of
failures. Furthermore, the TPA can be narrowed to only view failures when Cindy
or Jeff are operating the station to reduce the quantity of information presented.
The combined TPA then becomes:
Failure Total &
Failure Percentage &
Total Number of Executions
by low cost transmitters
by operator
by test station
by
date
|
Figure 4: A three dimensional TPA of failures by product - by operator - by test station shows a potential problem with test station 3.
Enlarge Image
|
Figure 5’s four-dimensional TPA of the three measurements identifies that the failures
are not limited to any specific date.
Using Test Performance Analysis
All the tables and charts shown in the previous example were copied and pasted into
this article from a standard web browser using
Arendar
Enterprise 2006’s new TPA reports. Simply dragging and dropping the dimensions and
measurements from Arendar’s TPA pallet to the workspace created the reports.
These examples used Microsoft SQL Server as the database to store millions of data
measurement. Although this example only showed four TPA dimensions and the measurements
all related to pass/fail data, the dimensions and measurements that can be added
to the reports are unlimited. In addition, Arendar’s TPA is able to narrow the dimensions
so that the engineering team only views the data necessary to make informed decisions.
|
Figure 5: A four-dimensional TPA with additional measurements of failures by product - by operator - by test station - by date shows no correlation to date.
Enlarge Image
|
Before TPA, creating multi-dimensional test reports typically took days. With the
introduction of Arendar 2006, these reports for the first time can be created in
seconds. Because access to TPA reports is through a standard web browser, engineering
teams can grant secure access to view or create these reports from their test data
to anyone in the company at any geographic location with internet access.
Ask the Expert
Email questions concerning this article to:
expert.testpronews@vi-tech.com