1
Microcontrollers / Re: Spurious pin-change interrupts with AVR DB
« Last post by T3sl4co1l on Today at 10:44:31 pm »We need schematic and layout to know better, but fast bounce is typically not propagated along the PDN (power distribution network), differential mode, but common mode through wiring; including any boards that lack ground planes, to name one very basic error, but without layout provided, I cannot rule out the possibility.
Calling it an unqualified error is a bit over-the-top, but is simpler and more colloquial than saying "highly likely to be problematic". Again, without information, only broad sweeping answers are possible.
EMC is a complex and holistic topic. Rigorous, provable and comprehensive solutions are possible -- but the problem must be fully defined to every detail and nuance to be able to provide that certainty. In lieu of that, incomplete data lead to uncertain outcomes -- how much, depends on what data are missing, and as the picture gets less and less complete, the full range of possibilities, best case to worst case, quickly diverges; a likely path can still be explored through the problem space, but more and more assumptions must be made along the way: what components are used, how the layout and wiring is done, about what nodes are interfering and how, etc.
Note also, brushed motors arc internally, so create noise by themselves, without switching noise having to be generated by the circuit. Ceramic caps and ferrite beads on the terminals, and to the motor case or nearby chassis/enclosure (whatever might pass for relevant RF ground), are a typical addition. Needless to say, the switching source into the motor needs to be controlled, whether by local filtering, or switching slow enough. Note that filtering can be a small dampened LC filter to knock down the edges, or a full PWM filter (i.e. buck converter), depending on what you're doing with it (it might be better to use the full PWM filter if Fsw should be on the high side, motor control should be nimble, and the wires are long, preferring strong filtering).
There's also the semi-non-solution solution: if the noise is synchronous with the circuit (i.e. switching noise ultimately triggered by a timer), you might even blank the interrupt entirely, near the edge. This will require a cascade of timers, but might be reasonable depending on available resources.
Tim
Calling it an unqualified error is a bit over-the-top, but is simpler and more colloquial than saying "highly likely to be problematic". Again, without information, only broad sweeping answers are possible.
EMC is a complex and holistic topic. Rigorous, provable and comprehensive solutions are possible -- but the problem must be fully defined to every detail and nuance to be able to provide that certainty. In lieu of that, incomplete data lead to uncertain outcomes -- how much, depends on what data are missing, and as the picture gets less and less complete, the full range of possibilities, best case to worst case, quickly diverges; a likely path can still be explored through the problem space, but more and more assumptions must be made along the way: what components are used, how the layout and wiring is done, about what nodes are interfering and how, etc.
Note also, brushed motors arc internally, so create noise by themselves, without switching noise having to be generated by the circuit. Ceramic caps and ferrite beads on the terminals, and to the motor case or nearby chassis/enclosure (whatever might pass for relevant RF ground), are a typical addition. Needless to say, the switching source into the motor needs to be controlled, whether by local filtering, or switching slow enough. Note that filtering can be a small dampened LC filter to knock down the edges, or a full PWM filter (i.e. buck converter), depending on what you're doing with it (it might be better to use the full PWM filter if Fsw should be on the high side, motor control should be nimble, and the wires are long, preferring strong filtering).
There's also the semi-non-solution solution: if the noise is synchronous with the circuit (i.e. switching noise ultimately triggered by a timer), you might even blank the interrupt entirely, near the edge. This will require a cascade of timers, but might be reasonable depending on available resources.
Tim