Avoid reading too much into superstition. Parallel capacitors of graded values are a direct example of superstition in engineering. There is some merit to the idea, but it's almost never executed correctly, and we can see from the types of associated components that it's utterly irrelevant to apply here.
Despite electronics being an eminently scientific field, the fact remains that most engineers will never have the time (or interest, frankly) to test every element of every design. Some elements are calculated from reasonable design equations and related assumptions (and, occasionally, those assumptions are remembered correctly, and their applicability confirmed..), but many are also just left to straight up tradition -- or superstition.
Consider: What bypass capacitor is default -- why 0.1uF? What pullup resistor is default -- why 10k? Probably because someone once said so; or one got the idea that, because these show up most commonly, there's something special about them. And so the meme reproduces, unchecked by critical analysis.
(As it turns out, bypasses and pullups are usually of such value that, anything in a wide range will do. It would be nice to use maximum-value pull-ups to reduce current draw, or minimum-value bypasses to reduce component size/cost, but even these constraints are often so minor that they can be ignored, and default values chosen instead.)
And by "interest", there, I mean not just the engineer's care to go and test something, even when the means and opportunity to do so are available (which often are not, due to budget or time constraints of a project); but even just the sheer idea that there is something here that *can* be tested, let alone *how* to test it.
Case in point: to characterize bypass capacitors, we would ideally perform an impedance measurement with an RF bridge (typically the bridge comes as part of a vector network analyzer (VNA) instrument). Install the passive components on the board (leaving nearby chips off), wire a transmission line into the respective chip(s)'s power pins (this must be done carefully; for example a coax cable must be installed with shield widely bonded to ground plane first; assuming a ground plane design was used on the PCB in the first place -- and the whole exercise is a whole hell of a lot less representative, let alone meaningful, if not!), perform whatever open/short/terminated calibration is necessary given the test conditions (short can be done on-board by shorting the coax, but open might be done before soldering the cable down, or by lifting the signal connection off the board momentarily), and then run the test. A variety of tests and testing methods are possible just for the passive networks we use for power distribution, signal filtering and etc., let alone more complicated systems like device-level EMC, or software testing (beta testing, automated fuzzing, code analysis, etc...). (Software is more-or-less a subset of electronics these days, but offers a depth and breadth of complexity far different from ordinary electronics. Hence computer science (CS) being its own thing.)
And who would ever think to test a bypass cap (or array thereof)? You almost never see it discussed. Even though having reasonably clean, low-impedance voltage supplies is the underpinning of all modern electronics; so, you'd think there would be some interest there, or pressure to verify it. But nah. In fact, assumptions underpin a disturbing amount of engineering; and, despite that, real work can get done, so, I mean -- it works. When it does, anyway.
Tim
Home
Help
Search
Login
Register
