TechArchive

This article was originally written in the period 1995-2000

A major in-service failure which necessitates the return of all systems for rectification is expensive and embarrassing. Vibration screening is one way to improve reliability.

On average, up to about 40% of all screenable defects are the rsult of other than component or device failures. The earlier a defect is detected, the lower will be the costs of repair and test. The application of effective screening techniques may result in an initial net increase of in-house failure occurrence of between 30-75%, while in-field failures may be decreased by between 20-90%.

Before ever considering the effects of vibration on a PCB it is important to understand that the printed circuit board at resonance will have a bending mode. With the use of surface mount components the actual centre of board deflection must be restricted otherwise failure will occur.

Fitting a large device will cause changes in local stiffness and will modify the first bending mode. The resonance characteristics will be affected by the type of edge conditions – or if pillars are utilised, the number and position of them.

Often overlooked or left until too late in the design cycle, the vibration screening fixture is as important to the production screening process as the vibration controller or the shaker system. The effect of a badly designed fixture can be an over test condition, resulting in unnecessary failures.

There is a need for a vibration assessment to be conducted as early as possible in the design cycle. This enables modifications and redesigns to be conducted before the design is frozen or the start of production when costs are lower.

A vibration survey is done at ambient and using swept sine. By sweeping over the required frequency range at a level high enough to excite the board and evaluating the responses from the auxiliary or monitor accelerometers, it is possible to assess the survivability. The vibration survey is used to:

• verify the resonant frequency of the module
• assess the Q factors at resonance
• ensure the displacements do not exceed the recommendations for the chosen technology

Results are generally obtained by controlling on the vibration fixture and attaching auxiliary accelerometers to the PCB.

The importance of a vibration survey over the full temperature range is often overlooked but it has a number of rather important functions.

• It is used to identify the changes in vibrational characteristics of the materials due the changes in stiffness
• it is also used to assess the effectiveness of the PCB mounting method.

Stepped stressing is generally performed utilising random vibration to determine

• the operating limit, beyond which the test item will not function but if the test item is brought below this level it will recover
• the upper and lower design limits beyond which damage will be done to the test item. This is generally not what the designer has designed too but is the actual achievable limits
• any design weaknesses uncovered during stepped stressing must be evaluated and corrected

With the application of the random profile at a low level, the level may be incremented in 1dB steps and using a dwell period of 5 min at each level until the upper operating limit is identified. The level is further increased until failure occurs. This may be a design weakness or a device defect. Having effected a repair the level is further increased until

• a multiple failure mode occurs
• further increase would destroy the test item.

There are major benefits of following this type of procedure as

• design weaknesses are removed
• a more mature product will reach production
• fewer engineering hours re wasted during the early stages of productionisation.

MIL-STD-781B type of vibration

The screen strength even at a reasonably high level of 6g is very low and at best does not exceed a screen strength of 0.2 after a duration of 100 min.

While this specification with its fixed acceleration level of 2g and fixed frequency (between 20 and 60Hz) fulfilled a useful function in the evolution of stress screening techniques, it has been evident for some time that the fixed frequency vibration method:

• is ineffective in detecting workmanship defects
• will highlight only the most obvious defects
• is likely to cause unwarranted failure through fatigue
• is not a realistic screening technique as it bears little relationship to ‘real life’ operational stresses in that the frequency of the vibration is generally selected to avoid resonances.

As this is the result of a spectrum analysis of a flat bed table it clearly demonstrates that there are three major areas of concern

• the extremely narrow bandwidth of the Power Spectral Density
• the limited frequency range between which the actual vibration can be set (nominally 25 to 60Hz but not on a resonance)
• as the frequency is low the displacements are generally reasonably high and can induce fatigue.

Despite these failures, MIL-STD-781B is still used extensively because this method of applying vibration is cheap and reliable using only an electric motor and adjustable out of balance rotors to produce a very basic sine wave on the vibration table.

But reports from the American defence industries confirm the failure of MIL-STD-781B’s fixed level frequency as an effective screen, because it only exposes about 7% of the most obvious workmanship or manufacturing defects. On at least one occasion an unsoldered joint survived 9 thermal cycles and 7.5 hours of 2.2g at 25Hz before failure occurred because there was a reasonable mechanical contact. Utilising a sensible random spectrum, this type of defect should be found in seconds.

Realising that random vibration is less understood by engineers than any other test method it is advantageous to invite them to both hear and feel different levels on the table so they have an appreciation of what they are specifying or designing for, starting with 810D common transport and working up through tests like the NAVMAT-P9492.

Stressing levels should be high enough to precipitate infant mortalities (such as loose particles in hybrids, poorly bonded substrate, manufacturing defects, chattering relays, etc), and yet not consume a large proportion of the useful life of the equipment.

NAVMAT-P-9492 is an advisory document and not a specification. When optimising a random screen, the Acceleration or Power Spectral Density (ASD or PSD), Bandwidth and the test duration must be considered together with the axis of applied excitation.

The profile is shaped with a +3dB/oct slope between 20 and 80Hz that helps diminish the bending stresses applied to the boards. However, some modules or assemblies may require the low frequency stimulus to precipitate certain defect modes. It may therefore be necessary to run a flat response between 20 and 350Hz. If this is the case, then testing should be conducted at a much lower level and only incremented to the NAVMAT values if no damage occurs. Note that there is evidence to show hat running a lower level profile with a flat response between 20 and 350Hz is more effective than running the NAVMAT.

The -3dB/oct slope between 350 and 2000Hz was introduced to reduce the stresses applied to small components with high natural resonances.

Care must be given to the selection of a random screen because levels, profiles and durations are dependent upon the mechanical design of both modules and units. Larger units over 4 cubic feet (0.113 cubic meters) have shown that random excitation levels as low as 2g can cause serious damage while some boxes up to ONE ATR (Air Transport Racking System) sizes used on aircraft can accept 6g recommended by the NAVMAT-P-9492 without any significant deterioration or the precipitation of latent defects into detectable failures.

Caution must also be exercised when tailoring any screening process where thermal barriers, acoustic barriers, or vibration isolators are an inherent design feature. There is usually some confusion when engineers have to determine the vibration levels and durations for a production screening programme to be applied to their equipment.

• Brumby Systems
• Tel: 01453 548881
• Fax: 01453 545810
• Contact: Mark Ashley