> With the straight capacitive approach, if you do the math for the current
> you need, the required cap will be larger and more expensive than a
> transformer. Also, the value of "isolation" is primarily that of limiting
> the amount of current that would be available to sink into a hapless body
> that might contact it. Having one amp available kind of strains the
> concept. Too bad you have to have isolation; with the small amount of
> voltage to be dropped in your application, a simple 10W resistor for each
> supply would probably do it.

The problem with using any type of transformerless power supply in a device
is that ANY electrical connection (intentional or unintentional) between
that device and anything that's ground- or mains-relative (or between that
device and any other device with an intentional or unintentional connection,
etc...) poses a severe risk of blowing up one or both devices.

If you are designing an alarm clock which is fully enclosed in a plastic
case, the series-capacitor trick might be just the ticket if your displays
don't need much current.  If, however, you try to add a serial port to your
alarm clock, that port may have 120volts riding on its "ground" pin.  Even
if you put caps on both sides of your supply, tying ground of the device to
earn ground could cause twice as much current to go through the caps as is
suppose to.  Bad...

Personally, I'm just hoping someone will produce a small wall-dimple (smaller
than a wall-wart) with an unregulated 100KHz oscillator driving a small
transformer.  I think such a thing should be produceable cheaper than normal
wall warts (the cost of the electronics should be offset by the use of a
smaller transformer) but I have not seen this done; the only similar thing
I've heard about was one which produced a regulated output (and cost an arm
and a leg); I would think that an unregulated output could be much much chea-
per.