DRAGON INNOVATION, INC. !
CLASS 10: TEST DESIGN & STRATEGY !
SEPT 8, 2014 !!!!!!!
WILLIAM DRISLANE | VP ENGINEERING | @DRAGONINNOVATE | WWW.DRAGONINNOVATION.COM
The Purpose of Testing
• Design Verification Testing à Verify that the product design is sound, that the intent of the design is achieved, and that the product is manufacturable.
• Because design represents the ultimate constraint for the performance of a product, a small number of successful tests can demonstrate a successful design.
• Production Testing à Verify that every manufactured product has been manufactured correctly, and continues to meet the design intent.
• In production, thousands or millions of products all have to be manufactured correctly.
Types of Test Conditions
• Normal Use
• Abnormal Use
• Abuse
• Life and Reliability
Types of Test Data
• There are two types of data that can be collected when you perform a test:
• Attribute Data – Go/No-Go or Pass/Fail
• Variable Data – Scalar measured quantities – voltage, pressure, etc.
• Which is more useful? Why? Which can be transformed into the other?
Test Coverage
• 100% Testing is best when it can be done
• Sometimes only a sample can be tested
• Destructive testing
• Too expensive
• Too slow
• Constructing a sample test plan – sample size depends on the level of statistical confidence required.
• Example – Manufacturing QC sampling often done using US DoD MIL-STD-105E.
Component Life Test and Redesign
• Important components are life-tested and the budget is updated.
• Testing must be realistic. Even if there are problems in this area, a weak point is almost always still a weak point.
• You must guard against the tendency to label all failures as flukes to avoid doing the right thing.
• The product is guilty until proven innocent.
System Life Testing• Component life testing is not sufficient by itself
• Systems can have problems due to interactions of components, and these problems can only be found by testing the system as a whole.
• Sometimes this effect dominates
• Software
• Electrical noise problems
• Once the components become fairly reliable, it is typically more effective to switch over to systems testing.
Maximize the Learning
• Focusing solely on passing a test can lead to ‘pencil-whipping’ the data and failing to see important problems.
• Early testing especially must focus on gaining a solid understanding of what is going on with the product.
• Once you understand the problems sufficiently well, the solutions are frequently no-brainers.