Bruce Cannon wrote:
> The thing is, in my idea of a super cheap 3v lith device there are as few
> parts as possible.  There is no brownout circuit and no regulator as the 3v
> is pretty flat and comes on without much of a ramp.  And I realize that
> mercury switches don't 'bounce', but they do 'jiggle'.  That is, it's likely
> that the power will cycle within the window of the startup timer.
>
> My questions are really "Will power cycling during the startup timer period
> just restart the PIC no problem, or can there occur a brownout-like
> condition?" and "Is a startup timer needed or useful in this scenario?"

This is why the recomendation to use an external "power latching"
system, using transistores. Once the mercury switch make the first
contact, it is done, power will be stable to the uC until it commands
differently, except if the mercury switch still contacted, when things
need to be done differently.

If you don't use a pulse discriminator to "wake up" the uC via Reset
connected to the mercury switch, it will holds the uC in reset state.
Using the discriminator pulse, the problem still the same, as many reset
pulses as many contacts the switch does, and if the switch still
contacted your program would never know, except if another pin port
check it out via a direct connection (resistor or something).

Remember that "cheap" doesn't mean "zero components". You can implement
a lot of components and build a nice circuit for less than 2 dollars,
that by the way, small quantities production speaking, 2 dollars is not
a fortune. Except if you are planning to produce thousands of units,
perhaps you are expending dozens of hours to save fifty cents in 20
units. Those hours of your time cost hundreds of times more than the
savings in the circuit.

This type of calculation is done all the time in the industry.  How much
we will expend in research to save how much in the final circuits?
Considering a professional costing $70/hour for the company, wasting 40
hours ($2800) to generate some savings in a circuit that will be
produced 100 thousand units, so the minimum cost reduction he =NEEDS= to
generate is 2800/100k = 2.8 cents per unit.