Speed control

[Nathanael Henderson <jude@pconline.com>]

    I know we have a few electrical engineers and other sorts here, so I
thought I'd toss this idea out and see if it floats.   The trouble with
controlling speed on subs (and other large electrical engines) as I
understand it is that there's no good way to control the ammount of
voltage going to the motor.  Variable resistors aren't practical I'm told,
so speed is often controlled by switching banks of batteries in and out
of series (ie. you might have 4 12V batteries, and open and close switches
to provide 12, 24, 36, or 48 volts (giving four speeds.)   The psubs web
site mentions 'pulse width modulation' but the term doesn't really mean
anything to me, but I've taken a guess:   Switch the power to the motor on
and off very rapidly, with the longer the intervals of power being 'on'
the more power the motor will receive (and generate.)    For instance, one
power setting might provide power to the motor for one tenth of a second
twice a second, another setting might provide power to the motor for a
tenth of a second 5 times a second, etc.   The problem is how on earth to
do this?   A good relay seems like the first step.  (A relay being a
switch that opens and closes a higher power circit when you give it a
little electricity, such as a 12V signal opening a relay switch that might
control a 240V power supply.)    With the relay doing the actual switching
of the power supply, we can use a much smaller 'signal' source of
electricity to open and close the relay.    Generating the signal is a
little more complicated, but I think a fairly cheap microprocessor and a
variable resistor could do the job nicely.   Using a microprocessor with
basic analog/digital conversion capabilities (which can be had for a few
dollars each) you could hook the variable resistor (any 'volume' type
control) to the analog input of the processor (so it would be able to tell
how high the 'volume' was turned up) and the processor could then generate
the apropriate sequences of on/off signals and send them to the relay.
Getting an apropriately programmed processor would be a pain in the arse
(it'd cost a few hundred bucks to get the basic tools to program them and
you'd have to learn the programming language) but the end parts cost
should be reasonably low, the circuit fairly simple, and the control over
engine power quite good (completely variable--just turn the knob.)
    So, the question to Them That Knows Better is:  Is this a good idea, a
bad idea, or something else?  Is there an inherent flaw in the idea?