iSixSigma

MSA for PCBA

Six Sigma – iSixSigma Forums Old Forums General MSA for PCBA

Viewing 13 posts - 1 through 13 (of 13 total)
  • Author
    Posts
  • #52015

    Sinnicks
    Participant

    Hi
    Does anyone has experience how to  apply MSA in PCBA testing? I am familiar with use of MSA in machine part measurements but PCBA is different thing. Or can you give any good references? Thanks!

    0
    #182272

    tcm
    Member

    Hi,
    Does PCBA = Printed Circuit Board Analysis?
     
    Tcm

    0
    #182287

    Gary Cone
    Participant

    Printed Circuit Board Assembly

    0
    #182289

    Gary Cone
    Participant

    Several things you should look at with your product -1) There is a lot of visual judgement, certainly attribute studies
    should be conducted.2) Usually there are multiples of any test equipment, they should
    be treated as different operators to make sure the SYSTEM gives
    the same results across the board.3) As part of MSA, great scrutiny should be given to any kind of
    functional testing as great assumptions are often made by the test
    folks.4) Go look at any parts the tester cannot get to directly in any kind
    of probe testing. There will either be holes in the test strategy or
    again great assumptions are sometimes made.Don’t assume that the test is good because it’s automated, part of
    your MSA should be an analysis of test coverage and a FMEA strictly
    from the test point of view.

    0
    #182291

    Ron
    Member

    There should be no difference. You are using a test stand to verify certain elements of the PBCA. You will be performing an MSA on the inspection/Testing system.
    Breakdown this system into it finitie elementsand perfomr the analsysi as you would if you were measuring a machined surfce.
    Is there something you are not telling us that would make this any different?

    0
    #182298

    Sinnicks
    Participant

    The things I am thinking are such these.
    Should MSA be conducted for each new PCB design? When the renew it (e.g, when adapter in the tester is changed)?
    How to choose samples so that they represent true process variation if variation depends on variation of individual components on each PCB? Should we just collect golden samples from several PCB batches to represent  the long term variation range, after first seeing what is the  real range of variation (that would mean doing MSA kind of afterwards)?

    0
    #182338

    MBBinWI
    Participant

    I work at a company that also does PCB manufacturing and what I have found is that the test stands are generally configured for pass/fail tests.  Yes, you can do an attribute MSA, but it is much easier to improve if you have continuous data.  This is a concept that seems incredibly difficult for electronics oriented companies to understand (or at least the two for which I have now worked).
    If you’re looking at physical placement, measure distance from some reference point.  If current, voltage or resistance is key, then measure the parameter on a continuous scale.  Same thing if response time or heat build up are critical.
    I’ve also found that the testing operation is one subject to high variation.  In fact, it is not uncommon for the board that is failed in the test stand to be taken out and “slammed” back in and retested.  If it passes, it’s “good”.  Need to stabilize the process of doing the testing so that it is consistent.
    Just my humble opinion.

    0
    #182342

    Sinnicks
    Participant

    Yes, this is one of the problems: fail and retest and then pass.
    The  idea of MSA in PCBA testing is indeed unknown to many and it seems that no articles are availabe. MBBinWI, would you be willing to share more of your knowledge in this field?

    0
    #182343

    Ron
    Member

    Go a couple of levels deeper in to the test stand.. You must program into the test stand the continuous variable that makethe little green or red light come on.
    Come on people use some analystic skills
     

    0
    #182349

    Gary Cone
    Participant

    Mark,Go read what has been said already. You have plenty to work with
    if that is what you desire. The only thing more you could get from
    Ron or me or MBBinwi is for us to come do it for you.The job is tedious, but to ship product that fails once, passes once
    is dumb. To handle electronics more than the minimum necessary
    degrades the product. To test and repair rather than run a better
    process degrades reliability. The key to that is to have good data.
    The good data starts with a test that people trust.And yes, it is a pain in the butt, but it is key to better product. This
    problem has been addressed and solved 20 years ago so you are
    not in new territory – just go do it.

    0
    #182413

    Jonathon Andell
    Participant

    In some instances you can invest time to uncover one or more underlying continuous variables. If it has to do with signal alignment (1’s and 0’s arriving as expected), you might do some work with counting number of misalignments. If you can do that, it could be feasible to do an MSA.If your response is strictly pass-fail, you could try some attribute tests, but as Gary already mentioned, the PCBs will degrade with testing – so those units should not be shipped to customers afterward.And yes, each time the board is reconfigured or revised, another round of tests may be warranted unless costs are prohibitive.

    0
    #182431

    Sinnicks
    Participant

    Hi Jonathon
    Thanks for your kind comment. Workload or importance of continuous data vs. attriubute was never a problem for me. Also, I was not looking for shortcut to pass components once already rejected, on the contrary, I was worried about it. Maybe I communicated my needs not clearly enough. But happily you understood anyway what I was looking for: when another round of tests is  needed. Would you say that change of individual component bacth (used for specific PCBA design)  or   would also trigger another round of tests? Are there possibly  any traceable references dedicated for this kind of application?  Or would you suggest to create any golden samples of your own here? I’m still new in this and studying also the process and any advice here is appreciated.
     

    0
    #182453

    Jonathon Andell
    Participant

    I’m not a big believer in the “golden-ness” of supposedly golden samples. If your shop is like the one I worked at before becoming a consultant, there are too many models and too many revisions to have a prayer at keeping up with everything. There are some things you could do. I’d be willing to chat a bit about some options – there’s no way to do justice to the topic via these posts. If you are interested, contact me at [email protected].

    0
Viewing 13 posts - 1 through 13 (of 13 total)

The forum ‘General’ is closed to new topics and replies.