Microsemi became Microchips and changed the numbers on all their technical documents, I'm too busy tonight to find the correct paper.
?
Basically when a diode is reverse biased and goes from conducting to 'turned off' there is a short time when electrons in the PN junction snap to the cathode. This can produce an intense, extremely short, burst of EMI.
Placing a low value non inductive resistor and capacitor across each diode will mitigate or eliminate the EMI.
Think of the diode as a switch, when the switch opens under load there can be a short arc. My first experience with diode commutation noise that sounded like my mother's sewing machine. I solved that problem by adding a capacitor across the brushes, each brush to the case, bypassing the AC mains with? 0.1uF 1000V disk cap.
[Note all of this violate current regulations and sane design guidelines. Do not use DC capacitors in a 120/240V AC circuit, they will fail shorted. BTDT.]
?
Ideally one would use a storage scope and measure the size and width of the commutation spike and pick resistor and capacitor values to achieve the lowest noise.
Each situation will be different because every transformer (for different projects) will be difference, the current drawn will be different (for different projects.)
The real nightmare comes with a variable DC power supply because the best value for one current won't be the same for higher, or lower current, so you have to comprimise.
I've been known to place diodes in a separate shielded box with capacitor feedthroughs and the parallel RCs. Of course this also requires each winding be bypassed by a suitable capacitor, a suitable commercial EMI mains filter, add an inductor to the ground leg (or buy an EMI filter with one already there.)
?
Another issue is local AM (or HF transmissions such as a ham or CBer) signal might reach the diode and produce God only knows what mix products. This can be a nightmare to solve. I passed on designing a replacement power supply for a bank of tape cartridge machines because the field strength was 10V per meter at 630kHz. I had serious doubts I could keep the RF out, the OEM damn sure couldn't.
?
Serious doubts should be taken to mean "no way in hell." I knew my limits.
?
10V per meter is an extreme RF environment, I was happy I didn't work there, while RF isn't ionizing radiation, there are studies that suggest exposure to intense RF fields for significant spans of time can cause cancer. Even intense 60Hz fields are problematic.
?
?
10V per meter will cause 6 (or is it 8) foot florescent lights used in commercial settings to glow in your hand. And you get a small zap every time you touched anything metal.
?
I'm pretty sure the station violated OSHA regs in 1998. I decided "Not my circus, not my monkeys."
?
The uA723 allows tighter regulation then you can achieve with a LM78/79X, if you obey all the magical restrictions. [And there are many if you want to do it right.]
?
As to "how much current can a coil in a filter pass."
I'm not sure of any way except a smoke test, start out with a light bulb 1/10 the desired load, measure the coil temperature, increase the wattage with 'lager' light bulbs until the coil gets warm.
?
I'd use a non contact laser pointer thermometer, paint a black dot on the coil! Black radiates heat allowing a more accurate reading.
[I learned this the hard way while trying to read the temp from a shiny heatsink. I could boil spit but the IR thermometer read room temp. I had a flash of smarts, painted a black spot and wow instant accurate heat reading. I should have known before starting the test but I was in a hurry with my brand new spiffy laser IR thermometer. Lesson learned, think it through before you start any project! And yea I felt pretty damn silly.]
For "for hire" projects I use new parts, I charge out the ying yang, don't get much business today which is more then fine with me. I do support customers I liked, and I generally liked people who treated me like a person with respect.
