A trip through Quality - Part 2

-July 10, 2012

In Part 2 (see Part 1) we will look at a technique that will actually uncover quality issues early on and allow them to be fixed before production starts.

What I am referring to is commonly called: HALT for "Highly Accelerated Life Testing" (and a whole bunch of other names like: HALT, HASS, STRIFE, etc [1]).

I'm not sure where the name(s) originally came from, but the concept was popularized mostly by HP in the 1980's [2] as a way of quickly inducing failures in products.

The basic premise is to do the usual qualification tests like temperature, voltage, shake and shock, ESD, etc. Then just keep upping temperature, voltage, shake and shock and ESD until a one or more failures occur. Each failure is analyzed, fixed and more torture testing is performed until you get to a point where the part or product is just about melted or shaken apart.

At this point the failures are analyzed and a decision can be made as to if they need to be fixed or not, but at least one has data to make a decision on. Simple things or the "Low Hanging Fruit" are almost always fixed because they are so easy to deal with.

These sorts of testing made a huge improvement in HP's instruments, and as I recall the field failure rate essentially was cut by two in ten years even as the instruments complexity was growing exponentially during the 1980's and 1990's.

This method can be applied to any company of any size - I have even found useful data by thermal cycling alone and most companies have thermal temperature chambers for testing (even a small one can be used). Thermal cycling it turns out is very effective at simulating real "field aging" as the expansion and contraction tends to uncover all sorts of manufacturing defects quite well.

When I was involved with modular instrumentation and power supplies I would program my temp chambers to slew back and forth as fast a possible with just enough dwell to stabilize the part under test at temperature. Likewise I used to bake full products and sub assembly parts to qualify them, like the ferrite transformers mentioned last time. This too proved very instructive and I think gave us a real leg up in the quality race.

The point is that this did indeed uncover flaws that would be seen in production products after years of field use, but the important part is that we had a chance of finding them in the factory before the product was released for shipping and it didn't take a lot of money or equipment or time do do this. It took the will and the passion to do an outstanding job.

Once qualification data has been taken it can be used as a baseline to judge future production line processes changes on.

For Instance: When surface-mount ICs first came out we were worried that using these in potted modules would result in long-term field failures - those little itty-bitty pads just didn't seem to be very strong compared to a through-hole IC lead. Well HALT testing with thermal cycling proved to everyone that that was not an issue. Without this data we would have been just guessing.

Also in my previously mentioned "Case of the Glued Ferrite Transformers Gone Bad", when we discovered that there were lots of ways for this process to go bad, I used to do head to head tests of a known good transformer and a new or improved one and test them together to make sure that new and improved glue really was and would go the distance with one that was known good [3].

Another useful application for thermal cycling where aggravated failures aren't the desired result is to screen higher reliability parts out of a production process (HASS - Highly Accelerated Stress Screening). From time to time I get clients that need a high degree of certainty that there are no obvious latent manufacturing defects in a part. I have found that simple thermal cycling can find all sorts of bad solder joints, etc. This not only then provides long term product data that can be analyzed to improve ones processes but it can be used to screen early production parts to be sure that the processes are in place and working properly which gives a real boost to keeping the early return rate low.

HALT testing can be taken too far just like everything else. I once had a project to supply some aerospace-type quality modules, where full MIL-SPEC hybrids were used before. This was more like a Space Shuttle Cargo Bay science experiment program, so it was pretty benign actually.

The folks qualifying the system took our part and subjected it to the full MIL-SPEC thermal cycling which was like heating the part to the surface temperature of Venus and then immediately dipping it into liquid Nitrogen (just kidding, but you get the idea). This super-fast slew rate just ripped the epoxy module apart, I don't think that it provided any useful commercial field like failure information as I have some of those potted modules still working 25 years later in test equipment and they don't have any signs of wear any where near that.

So that was taking the whole concept too far and like I said you don't need any more equipment than you probably have already to really make a difference with HALT testing.

There are some standard test procedures [4] and the bigger companies probably have standard profiles that they have found useful.

I have found that if you just keep it simple and analyze the data you too can make a real difference and learn a lot with HALT / HASS testing.

[1] Wikipedia: http://en.wikipedia.org/wiki/Highly_accelerated_life_test

[2] "A Unifying Approach to Designing for Reliability", by Kenneth F. Watts, HP Journal July 1981

[3] A failure like this caused one large data terminal operation to go belly up due to a seemingly small change that wasn't fully qualified and lead to nearly all the product dying in the field.

[4] See IPC-9592 Appendix D / Stress Testing

Loading comments...

Write a Comment

To comment please Log In