In the 20 February 2008 issue, Dataweek ran the first part of an article outlining HALT (highly accelerated life tests) and HASS (highly accelerated stress screens) technology and some of its application successes. What follows is the second and final part of this article.
Precipitation and detection screens
Correctly done stress screening is a closed-loop, six-step process consisting at least: precipitation, detection, failure analysis, corrective action, corrective action verification and database maintenance.
Precipitation here means changing some flaw in the product from latent (undeveloped or dormant and usually undetectable) to patent (evident or detectable). An example would be to break a nicked lead on a component or to fracture a defective bond or solder joint.
Detection here means to observe in some manner that an abnormality exists, either electrically, visually or by any other means. Note that an abnormality may be intermittent in nature and may only be observable under particular conditions such as low temperature and/or low-level impact vibration. Proven high coverage in the test system is mandatory.
Failure analysis here means to determine the origin or root cause of the flaw. In a practical application, this would determine where in the production process and why a lead had been nicked, why a bond was insufficient or why a solder joint had not been properly formed.
Corrective action here means to implement a change intended to eliminate the source of the flaw in future production. The nicked lead might be prevented by using a correct forming die, the bond might be corrected by using a different pressure or perhaps better cleaning and the solder joint might be corrected by using a different solder or a different temperature.
Corrective action verification means to verify that the corrective action taken did indeed solve the problem. Verification is done by repeating the conditions that caused the problem to be exposed in the first place, as well as any other appropriate conditions or tests, and verifying that the flaw is no longer present.
Database maintenance means to collect all of the data from the HALTs in terms of what the weaknesses were and what the corrections were. This last step is extremely important as, without it, the same mistakes will continue to occur. With the knowledge gained by several HALTs, a company can design products that sail right through HALT with no relevant failures; that is, with no weaknesses.
Each of these steps in conjunction with the others is necessary for a comprehensive screening programme. Any less than all six will not suffice to provide a truly successful screening programme with the entire attendant benefits that all steps, rigorously performed, would provide. Specifically, just breaking and then fixing the bad ones, while surely being better than doing nothing, is just the first step in a comprehensive screening programme. Unfortunately, many efforts at screening stop here, and therefore attain only small gains in quality, but entail the majority of the costs.
It must be borne in mind that screening is quite expensive and, while very cost effective if done correctly, the costs are mostly there even if done incorrectly. In all cases, the obsolete techniques of using single-axis vibration instead of all-axis vibration and using slow thermal cycling - with the attendant high number of cycles required - instead of very high rate with only a few cycles required, will be much more expensive than the more effective modern approaches which use the more sophisticated techniques and equipment.
Precipitation screens
A precipitation screen is intended to convert a relevant defect from latent to patent. Precipitation screens tend to be more stressful than detection screens. An example of a precipitation screen would be high level all-axis vibration, which accumulates fatigue damage extremely rapidly, particularly in areas at a relevant flaw, where stress concentrations usually exist, combined with high rate, broad range thermal cycling, which is intended to create low cycle fatigue in the most highly stressed areas, which, fortunately, are usually found (if the design is proper) at a flaw and finally, combined with power on-off switching, which is intended to generate electromigration at areas of very high current density, usually at a flaw, and to generate rapid temperature swings which force low cycle fatigue in areas of high stress, usually near a flaw.
By using HASS correctly, one uses the highest possible stresses that will leave non-defective hardware with a comfortable margin of fatigue life above that damage which would be done by remaining screens and the shipping and in-use environments. This approach demands the application of HALT techniques and design ruggedisation in order to be able to rapidly and effectively precipitate flaws. Without using these techniques, the application of HASS is usually not possible due to weaknesses that will not allow the high stresses.
Note that precipitation screens may well be run at above an upper design operating limit (or below a lower design operating limit) where the system cannot perform normally and therefore cannot be tested during stimulation. In this case, more than 90% of the defects could be expected to be missed when tested under quiescent conditions ie, without any stimulation at all. This is where the detection screen comes in.
Detection screens
Detection screens are usually less stressful than precipitation screens and are aimed at making the patent defects detectable. It has been found in the author's investigations that many patent defects are not observable under full screening levels of excitation even when the screen is within the operational limits of the equipment. What is required is modulated excitation, which subjects the article under test to a search pattern in temperature and all-axis vibration looking for the conditions under which the product will exhibit intermittent defects. Screen optimisation results in a minimum cost screen regimen that is safe and effective.
For example, it has been found on several products that plated through-hole solder joint cracks could only be detected by a modulated excitation. In an experiment utilising 13 samples, all 13 exhibited intermittent defects at some combination of stresses but at no others - and each one different. This implies that no defects at all would have been found if modulated excitation were not used.
Detection screens should be used on equipment returned from the field as defective, as one must assume that a patent defect is present or the equipment would not have been returned. It is noted in passing that non-defectives are frequently returned from the field for various reasons caused usually by the press of time to 'get it running ASAP'. Field repair people are inclined to replace whole sets of boards or boxes, when maybe only one of the set truly has a problem.
A full blown precipitation screen may not be necessary on field returns as the patent defect(s) present may be exposed by a much more gentle detection screen. If the detection screens do not suffice, then a precipitation screen followed by a detection screen would be in order. In the case of field returns, it may be prudent to simulate the field conditions under which the failure occurred if these could be ascertained. These conditions might include temperature, vibration, voltage, frequency, humidity and any other relevant conditions. The military, airlines, auto manufacturers and others too, would be well advised to follow this course of action as 'no defects found' account for about 50% of field returns. Simulation is not necessarily called for in this case, as simulation and/or detection screens are probably the more effective approach on field returns.
Summary
Every weakness found in HALT offers an opportunity for improvement. Large margins translate into high reliability. Today, HALT is required on an ever-increasing number of commercial and military programmes. Many of the leading companies are using HALT and HASS techniques successfully; however, most are being quiet about it because of the phenomenal improvements in reliability and vast cost savings attained. The basic philosophy is, simply stated, 'find the weak spots however we can and then make them more robust' - a new paradigm.
Correct application of the techniques is essential to success and there are many incorrect sources of information on the techniques today. Several cases have been observed wherein companies tried to use the methods with incomplete or incorrect training and the result was that essentially all of their mission-critical hardware failed very early in field service due to damage done during screens or due to major design defects missed by an improperly performed HALT. Consistently, completely and correctly used HALT and HASS always work to the benefit of the manufacturer and to the benefit of the end user.
Further to the three application successes revealed in part one of this article are the following:
1) In 1993, Storage Technology Corporation reported 'savings of hundreds of millions of dollars' in the first two and a half years of employing HALT and HASS. This was without the benefit of precipitation and detection screens and before modulated excitation was invented. These advanced techniques have added several orders of magnitude to the effectiveness of HALT and HASS.
2) A large farm equipment manufacturer put a new product into its normal verification tests. About 75% of the way to completion, a failure occurred. A fix was implemented and the test started over. After about 75% of the test, another failure occurred. A second fix was implemented and the test again started over. Again, after about 75% of the test, a third failure occurred. At about this time, the company was introduced to HALT and HASS; it then took an original model of the equipment, that is, without fixes, and ran HALT on it. Within hours, all of the weaknesses that had been discovered in the prior days of testing and one additional weakness were found. After fixes were in place for the four weaknesses, the test ran to completion without any failures.
3) Thermo King, a manufacturer of airconditioning for trucks carrying perishables, compared a programme with HALT to one without. The program without HALT took twice as long to enter production, had many more field failures and cost approximately twice as much in terms of engineering development and field failures.
For more information contact Lambda Consulting, +27 (0)82 344 0345, +27 (0)82 344 0345, [email protected], www.lambdaconsulting.co.za
© Technews Publishing (Pty) Ltd | All Rights Reserved
printer friendly version