> SMPS are even worse, because wire-losses are almost
> unpredictable. 

JT> They're scary. The cheapest supplies are designed by geniuses.
JT> Since the highest single component cost is the transformer,
JT> they make that as cheap as possible, and design around it. They
JT> could do away with all the protection, spike and noise
JT> suppression and quite a lot of other problems if they used a
JT> good transformer. You'd end up with a text-book example power
JT> supply that way- but it would cost a mint- all due to the
JT> transformer.

 It's not so much cost... they're on the edge of what's possible
anyway. The original EMI TV set had the ultimate transformer. It used
thin flexible PC boards etched to provide a low-loss interwinding
shield, and when I stripped it all out to see what it actually did...
sweet FA. I've tried winding them on toroids, enclosed pot cores, you
already use stranded winding wire to cut HF losses... it's just that
they're utter bastards. MOSFETs work best, because the recovery diode
is inbuilt, and the FET switches "slowly". It's as fast as a
transistor, but it's nicer (if that makes any sense).

 The basic problem with EMI radiation is that the switch *has* to
operate outside the range fo the transformer response. If you make the
transformer wideband, it switches faster and stresses the diodes more!
That's what all the spike protection is - to soak up the transients
and slow the fucker down.

 I shudder to think that the house of the future will have dozens of
these bloody things running lights and bloody-nearly everythign else!
They already have what they call a "power factor" correction spec, and
that frightens me too... power engineers talking about power factor
for an SMPS! 

Regards,
Bob
  

--- BQWK Alpha 0.5