"Post-production testing is a necessary, costly evil. Yet an evil far less costly than a logo associated with unreliability and non-dependability"

Methodologies is often regarded as bordering on the profane. However, there are times it is a perfectly acceptable word, and this is when used with "test". When used on this context it should not be viewed with the same contempt as with "design methodologies" and other seriously useless wastes of time.

Test methodologies encompass test methods. Although one can have the latter without the former, the methods have a high chance of failing the product if there is no "helicopter view" of the need for the testing. Such a view, from 15+ years of product design, is given below and was used during a successful interview for a Test Manager's role. It is not, by any means, the only view. It is here to serve as an introduction for those entering the test arena.

Test Engineering: A View

After assuming the design of a product being manufactured is fault free, testing done during and after manufacture is to ensure the product has not been subjected to processes that have modified or devalued the original design. This testing does not, in any way, enhance or improve the product. It only ensures reliability and dependability, this in turn protecting the reputation of a company and its logo.

With this in mind it is imperative that such testing be done with the least impact on manufacturing time, and at the lowest possible cost both in capital outlay and ongoing man-hours. The cost limitations are, however, to be achieved such that the testing still aims at a 0% fault tolerance.

Testing should consider not only ensuring the product does what it is called to do, but also to ensure the product does not do anything it should not e.g. inadvertently turning on an output that should stay off.

As miniaturisation becomes more pronounced in modern designs, with surface mount technology becoming more complex and occupying more printed circuit area, test points are becoming more difficult to accommodate on the layout. However, as more products are now microprocessor based, "soft" test points can be made available with the integration of carefully designed software such that the hardware is used to perform tests upon itself and reporting any abnormalities to external test equipment.

A carefully constructed test function flow chart would incorporate the microprocessor testing all the on-chip I/O structure ensuring no pins were shorted together from stray solder during assembly. If this test passes then all peripheral devices would be interrogated to ensure all connections to these were solid ensuring no unsoldered connections. From here all external connections can be tested with the aid of suitable Automatic Test Equipment which now receives instructions from the Device Under Test.

To assist with such testing it is imperative to know exactly what every active device used in the product is capable of, including features not initially designed in but nevertheless have become available to the testing software.

What would also assist greatly would be to have a method of 'talking' to the microprocessor that uses a maximum of one other component and I/O dedicated to this component that should this test fail the area where the error exists is small and and the fault quickly located. An example of this would be a serial port and a RS232 driver (e.g. MAX232). This port being the same RS232 port as would be on most microprocessor products thus is not adding to production costs but rather using what is produced to do the testing.

The use of Automatic Test Equipment must also be carefully planned. All too often tests are carried out in a "singular" method e.g. testing an analogue input that requires stabilising before taking readings. This is a severe waste of time yet nothing prohibits the testing of other areas of the circuitry while the analogue input settles to a new test voltage or level.

Further to this argument the level of ATE introduced must be weighed against the natures of failures of the product as a whole. If the product is likely to never have a short occur on a printed circuit board, e.g. these are tested by the supplier before shipment, then introducing such board testing equipment is of no direct benefit. If it is proved that such a failure becomes apparent it is then an issue to be raised with the supplier of the offending component to rectify the situation. This will ensure the timely implementation and minimum outlay.

Calibration of any portion of a circuit should also encompass the testing of that portion. In a well designed and well manufactured product the testing should prove successful therefore testing and then calibrating a portion of a circuit requiring such calibration would effectively be "double work". If the portion of the circuit will not calibrate only then backtrack and discover why it has failed. It must be borne in mind that a "base test" was already performed to ensure there were no shorts and that everything is connected as should be.

Record keeping is imperative in all areas of the manufacturing process to assist in the improvement of the product during its life cycle. Effectively the "design verification" phase is never ending, although it is not always dealing with the product directly as the design may well be the manufacturing process, or even the testing itself.

There are many facets to testing, and many ways to accomplish it. What must not be forgotten is the reason testing is done in the first place - to keep the product in the marketplace!

Marc Dekenah

© 01.09.02