Testing 1-2-3: The Importance of Product Design Testing in DFM

by | Feb 21, 2020

Test! Test! Test! Keep Testing!

It is generally understood that testing is important, critical even, to the success of any product launch. Despite this general understanding, many people think about testing as one time step in the development process rather than an ongoing process.

Bill Drislane, Dragon Innovation’s VP of Manufacturing and Engineering, does a great job of succinctly overviewing the two main kinds of product testing and gives tips for how to go about development and manufacturing with quality in mind.

There are two kinds of testing: design verification testing and production testing.

Design verification testing literally “tests the design”. If you claim that your product can do/achieve X, then you would need to run a design verification test to prove it. 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 is designed to verify that every manufactured product has been manufactured correctly, and continues to meet the design intent. As thousands and sometimes even millions of a single product have to be manufactured, production testing requires a much greater volume of tests.

In general, there are 4 different types of test conditions: Normal Use, Abnormal Use, Abuse, and Life & Reliability. When designing tests, it is important to think about how your product may be used in each of these states. When actually designing tests, there are two types of test data that can be collected, attribute data (Go/No-Go or Pass/Fail) and variable data (scalar measured quantities - voltage, pressure, etc). It is more useful to have variable data if a choice exists. Variable data can always be transformed into attribute data if needed.

When it comes to testing coverage, 100% testing is best if it can be done. With consumer products, it is critical as it is very difficult to get production lines running in a condition that is stable enough to do testing on a sample basis. Pretty much everything that you see in Walmart would be 100% tested.

That being said, sometimes only a sample can be tested as is the case for destructive testing (drop tests, high pot tests, over/under temp, etc) or tests that are too slow or expensive. When doing sample testing, you can construct a sample test plan with sample sizes that can give you required levels of statistical confidence (ie - US DoD MIL-STD-105E typically gives you an 80% level of confidence).

It is just as important to test complete system life as it is to test independent component life. For all intents and purposes, the product is guilty until proven innocent. “There is no such thing as a fluke, you must guard against the tendency to label all failures as flukes to avoid doing the right thing.” Although testing must be realistic, component life testing is not sufficient by itself. As is often seen in complex electronics and software, systems can have problems due to interactions of components, and these problems can only be found by testing the system as a whole.

To conclude, you can’t make a design improvement until you know what the defect is. It is important to design tests that maximize the rate in which you are learning about how your product really works. Focusing solely on passing a test can make it hard to see important problems. It is your job to find out what needs to be fixed.

So test it, determine the failures, and fix everything!
Continue to original Article >

Ioterra – Where you go for IoT Development

Find the right solutions and services for your needs from the most comprehensive marketplace for IoT projects.