That reminds me of something... I made a small GPS logger based on a
Logomatic v2 board. One of the things that board has is an SD card
socket, and a warning that we are supposed to manually request the
data to be flushed to the card before removing power, to avoid this
corruption problem. I needed a reliable system that I could turn off
power from entirely, using a magnet and reed switch, so I had to
figure out a way to stop the device safely. I did it with an LED, an
aerogel capacitor, and not a lot else. The idea is that the Li-ion
cell keeps the cap charged, and when power is removed the (carefully
chosen) LED arranged for a voltage drop that immediately put a logic
line below thrshold, to logic 0, forcing data flush. It had just
enough time to do this before that cap discharged.
So it was always
safe after that. I came up with a basic rul that for any system
needing that method, the capacitance should be rated at one farad per
amp drawn by the load. (And the cap rated for the voltage used too). I
documeted the doing on Sparkfun Electronics forum using the same name
I do here.
The point of that is, that some systems might use a similar method.
Power fading can be managed so a reservoir persists long enough to
finish a cache flush to disk or card before critical power loss
occurs. The question is, is this actually done? I like my method
because it doesn't depend on an order given. It physically makes sure
there IS always enough power to do the deed. This is true nio matter
how the main source fails. Basically, like a UPS writ very very
small...
The main problem with this idea is that it may not be easy to apply
in many situations. You'd need a way to send a 'logic low' signal to
force a card driver to flush the cache and finnish writing,
immediately on receipt of signal, and I've no idea if that's a
standard thing. Even if it were, it might need a few parts changed or
moved, using SMT soldering in most cases. At least the Logomatic
makers thought about this, but even there I had to do some work of my
own.
