I finally carved out some time to really dig into these responses. There's a wealth of information here, and I hope others will benefit from this thread. I had no idea it'd grow the way it did.
In reading through Tim's response about snubbing and edge rates, it's really clear that the devil's in the details. I can't imagine that a lot of designers take the time to account for all of the nuanced relationships when producing cheap designs like the eBay circuit that I used to kick off this thread. I have always been a big fan of "best practices," so even knowing that all of this
exists is a huge benefit for my understanding and future designs.
Electro Fan, you had asked about the length of the power wire in this circuit. The supply line to the LED strip is about 6 feet long, and the strip itself measures about 5 feet. It's the parallel-chained style, something like five LEDs in series per chain.
I think this thread started with me being concerned about radiated EMI (antennas) due to the switching behavior. One thing I considered trying, but haven't bothered to yet, is to run shielded wire to the LED strip. My most recent testing seems to suggest that the PWM is causing more ripple than noise. I do pick up some "fuzziness" on my scope when the LED strip lights are on, but that's probably from the SMPSes I'm using to drive the other, non-dimmable strips on my bench. Turn out the lights and the noise goes away.
Tim, on a slightly off-topic note regarding the mechanical analogs: I'm actually an ME, but I disagree that we have it harder
. You electrical guys have to deal with things you can't see (electrons), RF magic, and other phenomena that behave in strange ways... it's not always intuitive to a grease monkey like myself.
Years ago I worked as a junior vibrations engineer doing predictive maintenance for heavy machinery (really BIG stuff that you cannot afford to have go down unexpectedly). Characterize the responses and you can predeict when the equipment is likely to fail. It's the same idea with resonant frequencies and filtering on the EE side, but for some reason, it's just way more intuitive for me to see that mechanical connection--for example, a peak on an FFT correlates to a specific gear in a transmission based on the number of teeth, rotational speed, characteristic curve for a "good" machine, etc. At the end of the day, I guess it's all relative, though.
More to come in a bit... been testing the all-LM393 PWM circuits posted a few replies up and hopefully a little closer to laying out a board. Scope creep...