Design-for-Test: Product Lifesaver

Imagine creating a product that couldn't be tested before shipping to a customer.

Now, I realize just turning it on or trying it in a real-world application to see if it works is a fundamental practice that all companies use to verify product readiness. After all, if it doesn't turn on, or doesn't fit, or it doesn't open, or it doesn't begin to function at the most basic level as designed, it probably won't make it passed final inspection.

But, as we all know, most products require a multitude of testing operations to verify not only initial functionality, but also conformity to required operational, regulatory, and environmental compliance mandates. Testing provides the assurance that the product is performing as designed and is customer and market worthy.

Early on in the R&D process, the design engineer, working with test engineering and software assurance, should have developed a testing methods document that includes the associated list of relevant test equipment that will be needed to perform each type of the tests.

Guiding principles
This document becomes the fundamental guide in assisting the engineers in determining the test points needed on the printed circuit card, the test processes that can be automated for Automatic Test Equipment, ATE, the connectors to add to the product's design that will be the interface between the Unit Under Test, UUT and the test equipment, and the diagnostic software that needs to be written. This test process is known as “Design for Test,” DFT.

Pay me now (or later): A thorough design-for-test methodology  can save money long term and ensure product success.

Pay me now (or later): A thorough design-for-test methodology
can save money long term and ensure product success.

Each major area of testing requires sub-levels of consideration to ensure as close to 100 percent test coverage on the individual assemblies and ultimately the top level product. At the circuit card level, the designer will want to consider not only what has to be tested, test vectors, but also how much can be tested and how those test sites can be accessed.

For in-circuit testing, ICT at the board assembly house, the designer must leave room for the “bed of nails” used in ICT to contact each pad where the test data is to be gathered. Bare PCB testing can check for open and shorts, but functional ICT testing is performed under power where test data is fed to ATE gear programmed to gather the specific test point parametric or pass/fail status.

Good connections
Connector placements may pick up test points where a probe is unable to reach. This is especially true with Ball Grid Array packages where the IC solder ball connections to the printed circuit card are on the underside of the integrated circuit package. A well-placed connector designed to take signal or bias data from traces to and from specific BGA inputs and outputs will make it possible to determine if the BGA circuit is working or not. If there is no way to isolate and test a 576 ball package, then the technician cannot determine if a problem encountered on the board is due to a BGA problem or some other component issue.

Knowing what circuits can be tested using JTAG will also assist the test fixture design. Adding components like EEPROMS and I2C Buss chips will add more test capabilities. Software scripts embedded in PROMS can run diagnostics or start a test sequence. Often times, testing harnesses and custom test fixtures facilitate the testing processes at various stages of assembly.

If a board or product is not designed with DFT in mind, then trying to play catch up and guarantee maximum test capability coverage is a nightmare.

Technicians are forced to use shotgun repair techniques, meaning they just start swapping components until the board works. Last time I checked, the assembly house charged $50 per BGA to remove and re-ball. Custom SMT solder screens have to be made to replace a BGA component on the board. If there are a number of BGA ICs on the board with no accessible test points, then the nightmare can turn into a full-blown production horror novel. Always include DFT into product concept and design review meetings.

Someone should always be asking, “How do we test it?”

8 comments on “Design-for-Test: Product Lifesaver

  1. elctrnx_lyf
    February 28, 2013

    It's a very good article summarising the importance of considering testing from initial design itself. At least with an year of experience of working in a manufacturing of big electronic board of mobile networks product, watching it live the number of BGA's replaced because of improper ICT testing made huge difference to the final cost incurred for a product.

  2. dalexander
    February 28, 2013

    @elctrnx_lyf… You said it! There is nothing like a real life experience to make a point. With so many BGAs, that must have been a painful lesson. Sometimes just using 2mm pin header dips or 3-row straight SMT connectors with trace connections to key signal and power and ground points on the BGA can save a lot of development time and effort. After the design is proven, you can always opt to not stuff the pin headers at the time of assembly. Then, if you want to troubleshoot later you at least have the option of placing a connector in the footprint provided. Another headache is BGA sockets for test and bring-up. They are usually very bulky and the design has to consider adjacent component placements to avoid interference issues with the BGA socket adapter. That being said, if the circuit with the BGA can be built separately for design concept testing, then a BGA socket adapter is exactly what you want. Later when the BGA circuit design is proven, the whole CAD layout on the test board can be moved to the final larger board. Either way, BGAs need to be DFT to avoid what happened to you. Thanks for the great example.

  3. William K.
    March 1, 2013

    Certainly a product that uses a BGA would need to have those devices chaecked befor attaching them to the board, since they are quite difficult to remove neatly. A large set of test points on the PCB would also be handy, even for determining that the rest of the board would work with the BGA installed.

    But the products that can not be tested include parachutes, matches, fireworks, and bullets. Those are the ones that I came up with quickly. How many others are there.

    Adding provisions for testing on electronic products only makes sense for products worth repairing or re-working. That leaves a fair number of consumer items out in the cold, doesn't it?

  4. dalexander
    March 2, 2013

    @William K. Interesting examples. As you know, things like bullets cannot be tested for full performance prior to sale, but this is where the practice of AQL or Acceptable Quality Level comes in to play. Someone at the bullet factory decides that they will fire 10,000 bullets. If one bullet is a dud, someone says, ” that's an acceptable yield for a lot of 10,000 bullets.” Then the inspector comes along and on either a random sample or per 1,000,000 bullets, 10,000 bullets are pulled from the line and fired. Now if the AQL is 1 dud per 10K, and the inspector finds 2 or 3 or more duds, he fails the entire lot of 1million bullets or they go into seconds inventory. You can imagine the stringent quality checks on life- impacting goods like parachutes. Matches may have a fairly high AQL as someone can always grab another match from the box without frothing up to a lawsuit rage. Even so, it would be interesting to investigate quality assurance methods for the items you described. You think like an engineer. Are you?

  5. William K.
    March 2, 2013

    Douglas, YES, I am an engineer, one of those “hands-on” types who not only does the design but then is able to make things work. In fact, I make it a point to say that I am an engineer, not just that I do “engineering stuff”. Sometimes it gets to my friends because I see things that are problems or poorly designed and I may even comment on some of the things that are obvious faults. So sometimes being an engineer can be a bit of a burden. But mostly it is a fun profession. And it would be interesting to see just how the quality of some of those non-testable productsis assured. X-Ray inspection and high accuracy weighing come to mind as valid NDT methods. Of course there are several other failure modes for bullets aside from just not firing. Overload is the most dangerous, underlaod is just a nuisance, while a slug balance or weight problem would affect accuracy. 

    Matches may flake or break, and if there is a chemical problem the heads may explode instead of just burn. Or they may tend to absorb moisture to much or too rapidly.

  6. dalexander
    March 3, 2013

    @William K. I love the way you think. You have an engineering gift for sure. I'm learning from you. Please keep writing whenever you can.

  7. William K.
    March 3, 2013

    I intend to continue writing when I can offer some value. I also post on the Design News and EDN blogs, as well as the “connections” blog. I don't know just what it takes to gain access to them, but my comments are found their regularly. Also on the one physics discussion area. That one has a wide variaty of topics.

    One of the things that I do for fun is troubleshoot and repair things, if they can be repaired without spending a lot. Many things can be repaired fairly easily, but there are quite a few products that are clearly not made to be serviced. Fixing some of them is an interesting challenge sometimes.

  8. dalexander
    March 4, 2013

    All, I just had to include this case example that Tim Pidcock, Director Quality Engineering VP wrote in response to this article being posted on Linked-In: What a life lesson this is! Hopefully, Tim's comment will help prevent other companies from not falling into the same pit. THanks Tim!

    Speaking from my personal experience with a poorly thought out price driven non-DFT approach…

    I once knew a company where… in the effort to reduce “cost” in materials, they became way too focused on price management of PCBA. The Procurement Team leadership decided to measure the performance of their team primarily by focusing on reduced Purchase Price Variance (PPV), and one of the things the hardworking people attacked first was the perceived high costs of testing PCBA In Circuit Test (Bed of Nails) and Full Functional Test because it represented a large part of the set-up charges and PPV for each PCBA. Eventually they started going after Flying Probe tests also.

    Now…on less complex, single layer or interconnect PCBA, it made perfect sense to cut out testing, because there was very little if any value add operations between PCB and PCBA and the PCBA were not worth the costs to troubleshoot and repair if the failed; however, for more complex PCBA, cutting out these tests was a huge mistake. I strongly objected to the approach, but the dollars “savings” was hypnotic to those who were being measured by PPV performance and their management.

    In the short term, they appeared to be saving the company huge money by not paying for test fixtures, time and programming any more. The PPV on the PCBA dropped at a steep rate and they were happy with their achievement. This likewise drove a behavior in our Design Teams to stop including superfluous test points and interconnects on the PCBA since they were no longer needed, and they could also reduce the size of the PCBA to get rid of unused real estate…Program and Project Managers also became hypnotized since they could report reduced design costs and management was happy.

    As time passed, the tsunami of costs associated with buying untested RF and Digital design PCBA began to rise. Bone piles of failed PCBA grew, conflict broke out between our manufacturing, procurement orgs and the PCBA sources over who owned the costs of troubleshooting and repair for these untested PCBA, and scrap cost went up dramatically since the simple answer was…we owned it…we didn't design and pay for test.

    So the moral of the story is… DFT = Good

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.