Bob Axtell cotse.net> writes: > I believe that a die shrinkage has occurred in the HC world AGAIN, and > some designs have had to be re-examined, because as dies shrink they become > more responsive to noise. I have tried to analyze the problem represented by the 'shrunk' dies as applied to pics and other chips. I am not an engineer and some of the following may be bull from the technical point of view. However, it is important enough to discuss it, imho. Die shrinks are usually accompanied by doping strength increases. To maintain the conductivity (esp. of outputs) the doping density is simultaneously increased. This causes un-important features (n and p etc wells vs. substrate for example) to become deeper and thus increases the capacitive coupling between the theoretically uncoupled wells on a die. Also the abrupt transitions in doping should be able to produce hotter injected carriers in the substrate and the better insulation should keep them longer once trapped. At the same time the vertical 'reverse' diodes in cmos devices become worse and worse, becoming smaller in area and longer in length. Also the speed of the 'wanted' features increases with lower area and higher doping concentrations. I believe that the resulting high sensitivity to esd is due to the sum of these factors, and that it affects all shrunk devices, including opamps, voltage regulators (oscillation-prone 780x clones for example) and even audio amplifiers like the lm386 and tda200x series which I have personally seen degrade from benign, easy to use chips, to oscillation-prone 'hair trigger' devices that require costly VHF grade pcb routing and strong decoupling and low pass filtering otherwise seen only in video circuits and above. Concluding, I would say that if there will be no breakthrough in technology, things will continue to become worse and worse in this direction. By breakthrough I would see perhaps a modified output cmos transistor structure that incorporates active turn-on when the attached pin exceeds the supply voltage in the relevant direction or other such measures. I cannot see that happening anytime soon, though. All of these things raise the implementation cost in ways that are sometimes unpredictable. I recently saw a simple design that involved a pic and had a total (and critical) parts cost of under $2 that had to have two costly Schottky diodes added (normal diodes would not do!) after the pcbs were designed and manufactured to prevent a current injection related malfunction that did not occur with any previous series of chips. That was a 10-15% parts cost increase, manual labor and ugliness of piggybacked parts not accounted for. There was not even a suffix change between the parts that needed the mod and those which did not. I feel that there is a limit beyond which I can see no point in ordering 'brand name' parts as opposed to using directly counterfeit c**p and designing accordingly from start to catch any 'unspecified behavior' that should not happen [tm]. I am not sure whether this limit has been reached yet. In any case I have no intention to whine against pics, which are usually good and rarely display undesirable 'improvements' (silicon erratas are not covered by this message - they represent a different chapter). As to the verosimility of counterfeit c**p, that it the one thing I am not afraid of. Unfortunately the largest improvements in counterfeit parts are seen in the exterior part marking. Only last year I went through three sets of power transistors for an amplifier, all being grossly counterfeit and of low quality, blowing up after 10 minutes to 6 hours. The fourth set was good because a) I massively overspecified it (this was a repair, not a new design!!) and b) I made sure I bought them from a reputable source (as opposed to the usual shops where the other three came from). Peter P. -- http://www.piclist.com PIC/SX FAQ & list archive View/change your membership options at http://mailman.mit.edu/mailman/listinfo/piclist