Guard-banding is the concept of tightening test limits by the amount of your measurement accuracy in the measurement system in order to prevent bad DUTs from being mistakenly passed.

For example, let’s say you are taking a measurement with a specification limit of 5 volts and the measurement system being used to measure a DUT has a measurement accuracy of 1 volt. To make sure that a DUT with a true value greater than 5 volts does not pass, the limit should be guard-banded by the 1 volt measurement accuracy and set at 4 volts.

Here is how this works. You have a DUT with a true value of 5 volts that could be measured anywhere from 4 to 6 volts due to measurement accuracy of 1 volt. See Figure 1.

For example, let’s say you are taking a measurement with a specification limit of 5 volts and the measurement system being used to measure a DUT has a measurement accuracy of 1 volt. To make sure that a DUT with a true value greater than 5 volts does not pass, the limit should be guard-banded by the 1 volt measurement accuracy and set at 4 volts.

Here is how this works. You have a DUT with a true value of 5 volts that could be measured anywhere from 4 to 6 volts due to measurement accuracy of 1 volt. See Figure 1.

If the test limit is set at 5 volts according to the specification, then half of the time this DUT will pass and half of the time this DUT will fail. This is an example of a DUT we would want to pass, but might false fail.

Now let’s say that there was a DUT with a true value of 5.5 volts. We would never want this DUT to pass but it would pass 25% of the time. See Figure 2.

Now let’s say that there was a DUT with a true value of 5.5 volts. We would never want this DUT to pass but it would pass 25% of the time. See Figure 2.

To avoid the potential of a false pass a guard-band can be applied to create a test limit of 4 volts. Now, with the limit at 4 volts the DUT with a true value of 5.5 volts from Figure 2 can never pass. See Figure 3.

It’s pretty easy to see that applying a guard-band will potentially fail more good parts. This highlights the importance of knowing where your measurement population is compared to your test limit and having an accurate measurement system. This brings up the concept of CPK which is a measurement to determine if your measurement population is too close to your limit.