Years ago (nearly a decade) I needed to use a PIC in a product that,
when off, had to consume under 100 uA.

The "power control" processor was a PIC (A 'C54, as I recall) running at
32 KHz - so it pulled only a few 10s' of microamps - but as you guess,
about any inexpensive linear regulator at the time pulled far more than
that.

Using a standard Zener was suggested - but its problem is that at very
low currents (10's of microamps) a 5.6 volt zener may "regulate" to
anywhere between a volt and 4 volts, depending on the zener,
temperature, and phase of the moon.  (We tried it just for laughs.)

The workaround was to use 3  cheap red LEDs in series - yielding a bit
over 5 volts.  This voltage changed acceptably little (15-20% for the
LEDs that we used) over a range of a few 10's of microamps to 10's of
milliamps and it didn't change a lot with temperature.  The LEDs were
biased at about 30 uA, as I recall, and an emitter-follower was used
(plus a few caps, of course...) and the problem was solved.

This didn't provide particularly good regulation, but that wasn't
important in this application and the voltage was well within the 4-6
volt range of the of the PIC.  When the "main" supply came on (being
turned on by the PIC) it was diode-ORed and took over.  Several
thousands of these things were built, subject to extreme environments
(voltage, temperature, etc.) and nary a one ever had a problem with this
circuit.  As an unexpected side-effect, the 3 LEDs provided a valuable
diagnostic tool, being extremely dim (but visible in low light) when
"idle" and the flashing sequence (when the LEDs were given more current
as part of the start-up sequence) gave clues as to what might be going
on with main computer.

Nowadays there are certainly regulators with very low quiescent current,
but in a pinch, you could always resort to the "cheap LEDs + 2N3904" method.

CT

--
http://www.piclist.com hint: The list server can filter out subtopics
(like ads or off topics) for you. See http://www.piclist.com/#topics