Decoupling Performance Class and Quality Assurance testing?

How much do you know about IPC’s PCB performance documents IPC-6012 or IPC-6013 Class 3 requirements? Sure, everybody wants Class 3, and why not? The performance classification carries with it the highest level of end product reliability, but does everybody need it? Class 3 was developed for end products where down time means human life is endangered (life support equipment, critical weapon systems, etc.). Is Class 3 appropriate for everyone (IPC, for example, is aware of some military end products that are built to Class 2 requirements)? Even though Class 2 would yield acceptable product in many cases, some customers want the end product built to Class 3, end of discussion.

Performance requirements are one thing but what happens, though, when people discover the level of testing that is required by IPC-6012 and IPC-6013 for Class 3? When an understanding of the number of test samples required for evaluation is developed, a common reaction is “There’s no way we can financially handle that level of product verification.” So we find variations on the responses.

1. We have heard of printed board manufacturers advertising that they build end product to IPC-6012 or IPC-6013 Class 3 requirements, but when the end product user asks for the requisite lot acceptance data, no testing was conducted by the manufacturer.

2. A company might offer Class 3 performance with Class 2 quality assurance testing verification. A variation of this would be what some call Class 2.5 – calling out class 2 assurance and performance and then cherry picking some key criteria to be met from class 3.

3. Then there are of course manufacturers that give exactly the performance and the product verification that are required by the IPC standards.

It is clear that a significant number of customers want class 3. It is not completely clear that the end customer wants or needs all of the testing that currently comes with that class. Some need it. But do all of them? IPC is interested in the concept of decoupling end product performance class from quality assurance. The end product user would then need to call out in their procurement documentation a performance class and an assurance category. For example, one could specify end product to be built to Class 3 end product but Quality Assurance Level 3 (out of, say, 4), that would have reduced lot testing from the existing Class 3 lot testing requirements in IPC-6012 and IPC-6013. In this fashion, it may help end users think about what they truly need from their printed board supplier and what it might cost.

IPC is encouraging active debate on this issue. We want to hear from you!


  1. Alex Tom
    Posted December 2, 2009 at 12:42 pm | Permalink

    I think this summary pretty much hits the issues dead on target for me. One of the other questions I had was what is truely the emperical results for boards built to class 2 as opposed to class 3? Are we talking over what type of lifetime of the board? 5 years, 10 years, 15 years? I believe I, as well as others, could use more background on each of the differentiations between class 3 from class 2. Some differentiations seem to have more of a yield impact rather than a reliability impact.

    • Posted December 3, 2009 at 4:01 pm | Permalink

      Hi Alex.

      Honestly I don’t believe there are empirical results for this type testing. Historically, qualification builds were used to prove out material choices; lot testing was done to evaluate manufacturing variation, and periodic testing looked at manufacturing trends. The frequency testing was born out of the requirements of the military specifications, and backed down from there. There was some input from the long life field industries like Telcom, but not enough to change the perceived needs of the class 3 guys. There are reliability tables in documents like IPC-SM-785 that predict life cycles expected by certain environments, but that has never driven back to the lot frequency testing in the PCB specifications. Industry complains that many people just ignore the testing. Others may have statistical data that shows the variation is not there to require them to test. This is why I started this post to generate some reaction. Unfortunately you response was the first I am afraid.

Post a Comment

Required fields are marked *


%d bloggers like this: