With PWM, you turn the motor on/off very rapidly,
and use duty cycle to control the speed. It is really
the only way to go for this project. Since it is
the only way to go, we really can't talk about saving
power or reducing average current, as that implies
a comparison to something else.
The concern is not exhausting the batteries to the
point that the microprocessor (mp) won't run. The
motor needs hundreds of times the current that the
mp needs. When the batteries drop low enough that
they can't run the motor, they still would be able
to run the mp.
Therefore, exhausting the batteries is of no concern
when looking at keeping the mp running. What is of
concern is the temporary voltage drop created when
you first turn on a motor that is not moving. You
get a very large surge current, far in excess of the
motor's steady state draw. That causes the voltage
to drop briefly, until the motor gets going.
Your LDO voltage regulator needs at least 5.6 volts in
to provide 5 volts out. If you use a 7805, you need at
least 7.5 volts in. So the idea is to keep at least
7.5 volts available to the mp when the voltage from
the batteries drops due to the surge when the motor
is first turned on. The job of the capacitors at the
input of the regulator is to do that.
Your LDO regulator is capable of 100 mA, so we must
assume you are keeping the mp from drawing more than
that. Using 100 ma as worst case, and figuring your
capacitors at 80% of their value, they will keep the
regulator satisfied for ~ 300 mS with the 7805 regulator,
and longer with the LDO LM2931.
Whether 300 mS is sufficient I can't say. I don't know
how long it will take the motor to get up to speed, nor
how low it will drag the voltage at startup. Among other
things, the duration of the start up time (getting the
motor up to speed) will depend on the load on the propeller
and the condition of the batteries. Weak batteries will
have a higher internal resistance, and therefore a
greater voltage drop.
A separate battery for the mp avoids the problem, but may
not be the best solution. It could be that a separate
battery is not needed - the existing circuit may do the job
nicely. Do you have any specs on the motor operation at
startup?
Ed