We couldn't reproduce the problem in the lab. Two days later, the customer had more data. He contended that the problem happened only outdoors. Oh great, we thought—the old “works-in-that-spot-but-doesn't-in-that-other-spot” failure. Given the limited amount of observation, we wondered whether the outdoors connection might be mere coincidence. The customer speculated that the unit had a temperature problem—that the units were overheating in warm weather or in direct sunlight. So, we asked him to test the units by running them for extended periods in a hot car. He reproduced the problem, but not consistently.

　　W

e began thinking it possible that direct sunlight was overheating the liquid-Crystal material and rendering it dysfunctional. We tried some outdoor testing ourselves but couldn't get the LCD to fail by leaving it out in the sun. The LCD was a complete OEM module on whose glass its maker placed the controller Chip to run its pixel array, providing designers with an easy mICroprocessor-controllable display product. We didn't expect to have to deal with inadequacies of the LCD physics, because we are eleCTRonics designers.

　　A little bit of extra education provided just the insight we needed, along with some finally consistent observations from the customer. Using several units, he managed to get the same behavior from each one. If he powered them up in the sun, the units faiLED to show anything on the LCD. If he powered them up while shading the LCD screen, they booted fine! Our educated theory? Somehow, sunlight reached some part of the controller chip, and a photoelectric effect of intense light hitting the chip-on-glass assembly was causing the SILicon to malfunction, but only during its power-up cycle. It sounded far-fetched, but we consistently confirmed the behavior. Now, we needed a solution. And our customer had already shipped some of his first production units and was dreading the returns.

　　Fortunately, the LCD module had a SOFtware-reset command. We tried using it on software start-up. OCCasionally, it worked, but occasionally wasn't good enough. We then found that a brute-force loop of some 50 or so reset commands to the LCD-controller chips eventually brought around any that were failing to start up properly in the sunlight. Luckily, we had designed the product's firmware for an easy fiELD update, so we solved the problem for production and for the already-delivered units. It seemed pretty kludgy to us, but it always worked, and, in real time, it moved too fast for anyone to notice it.

　　Remember: Your engineering education began with fundamental science for good reasons. It helps you understand real-world problEMS, because chips and software aren't always pure electronics-theory Tinkertoys that snap together and just work. And make sure beta-testing happens and that the users are good observers, with at least a little sense of scientific method.

　　英文原文地址： http://www.edn.com/article/CA6491161.html

上一页  [1] [2]