Appendix A I Appendix B I Appendix C

Both above described methods have merit and have critics. Most criticism is found in regard to HALT/HASS testing. First of all, every company has their own way of defining HALT/HASS testing and none of them are the same. Even the terminology varies depending on the organization or manufacturer of test equipment. For more information on the validity of HALT/HASS, see Appendix C.

HALT/HASS for a manufacturer of avionics or military equipment can make perfect sense. Those products have to be very rugged because of the conditions of a very high stress environment.

The most common argument against HALT is that this test is going too far because it takes the product into its destruct limit. For example if the product fails at -45 ºC, is this an actual failure or did we just pushed it too far? It can be assumed that the product in question will never see these temperatures during it operational life.

Another good example is, when designing a power supply for IT that will be used in a 19" rack at a constant controlled ambient temperature of 18 ºC ± 0.5 ºC and receives constant conditioned power. HALT typically will go to extremes like +70 ºC and -30 ºC. If the product fails near or at those limits, engineering time and money is spent to fix it. Proponents of HALT will say that this exposed the weak or unreliable components. Opponents of HALT will say who cares, because the product will never be used at those extremes and you are wasting your resources and driving up the cost.

The truth is to know your operating environment and the expected or warranty life of your product first. Then an accurate model can be derived that dictates the required tests.

For example an Alcon surgical system typically will operate in a temperature range from 15 ºC to 40 ºC and be exposed to occasional controlled transportation and often long periods of storage (non-operating). On the other hand, an US handpiece has a quite different environment, from normal use in surgery to 300 autoclave cycles. During surgery the handpiece can be dropped accidentally and is then subject to substantial mechanical stress in the form of many g's.

By using accelerated aging, both environmental conditions can be accurately modeled. For example, the usual 300 sterilization cycles for qualification testing that can take up to two months can be done in about 3 to 4 days once a profile has been established. Normal surgical conditions can be tested within a few days to simulate the entire life of a handpiece. System qualification can be reduced to a few days, as long as a proper model has been derived during the development phase.

As can be seen from the previous summary, it is clear that we can use better and faster methods to substantially improve our product quality and reliability. Our current methods do not guarantee that a new product actual will be more reliable in the field than before. Reviewing the available Eyelite documents from 1996, a lot of the qualification, environmental and reliability testing that was performed was not all that much different than we do now. The NGL testing was more complete and thorough than the Eyelite. However, this still does not give us a lot of confidence that the NGL will do all that much better in the field, because our test methods in R&D and manufacturing are essentially the same low or no stress tests. For example, the NGL design margins in regard to temperature are very thin, since there is a risk that the system might overheat at a high ambient. If the laser engine runs hotter than the systems we have tested, the TEC's cannot keep up and the system will go into standby. During the NGL testing, we also saw a considerable amount of manufacturing defects from the new units we received. Part of it can be contributed to the learning curve of the assemblers of a new product, but many hidden defects were not caught using the traditional run-in methods.

The traditional temperature and vibration testing have been in use for more than 25 years in the aerospace and military industry, but this often did not always resulted in a reliable product. In fact, data analysis has shown that the actual field reliability had little correlation to the testing performed. The reason is again the same; the product was not appropriately stressed for its use environment.