Here is a slightly edited mail exchange with someone who asked about principles of ESC operation. I'm putting this ASCII file up because several others asked me the same question. I hope I didn't take too many liberties with the truth. Please send corrections and complaints to behr@math.niu.edu.
>We desperately need information on how ESC works.  We would greatly
>appreciate any information you can send us.  The E-mail address is:

Well, I'm no expert, but here's some info:

The modern remote control transmitters/receivers for model aircraft,
cars etc. use what's called pulse width modulation. The radio sends out
signals which look like pulses in which the strength of the signal
goes up, stays up for a while, then goes down. On a diagram it looks
like this:
  +---+       +-----+
  |   |       |     |      Notice that I made the second pulse wider
  |   |       |     |      than the first one; this is the crucial
--+   +-------+     +---   data - it tells the receiver what information
                          the transmitter is sending, and that's why
the method is called "pulse width modulation". It's actually a little
more complicated than this, because the pulses are not transmitted
as they are by the radio but they are encoded in one more level of
complexity (how the sine wave sent by the radio actually represents
the signal level -- AM or FM), but don't worry about it right now.

When you move the joystick or a wheel on the transmitter back and forth,
the pulses grow narrower or wider. So to detect what you are doing at
your controls, the receiver (in a car, boat or plane) has to measure
how wide the pulse is and compare it with the "normal" width; if it
sees the pulse 10% narrower than normal, it tells a "servo" (a little
special motor) to move the rudder a bit to the left; if the pulse is
20% wider than normal, the servo turns quite a bit to the right, and
so on. This lets the receiver see not only that you moved the stick or
a knob, but how far and in what direction. It's called "proportional"
control. This method is also quite resistant to radio interference and
noise, e.g. from power lines and such.

But it's also quite difficult to make such equipment: the radio wave
which carries the signal swings back and forth about 70 million times
each second, and the pulses last (depending on their width) only about
1/1000 to 1/500th of a second. That's like a very fast camera shutter.

One more problem is that you want to send information about several
things at once: rudder (or steering wheels) position, engine speed,
maybe ailerons. In radio control equipment there are usually two or more
channels -- in one cycle two or more pulses are transmitted: channel 1
pulse, channel 2 pulse, channel 3, etc., then it gets repeated. The
information about the state of each channel is sent in "frames":
repeating patterns of pulses, usually 50 frames per second or so.
Channel 1, then channel 2, etc., then a brief pause, and then another
frame. The receiver must untangle all this and send the right signals
to its output connectors 

With an ESC, there is no servo which moves things: you move the
throttle stick or press a trigger on the transmitter, and you want an
electric motor that drives the car or a plane's propeller to go faster
or to slow down. The ESC must look at the pulses, decide how their width
compares with the normal width, and depending on that send more or less
power to the motor.

These days the first part (comparing the widths) is done quite easily
with simple chips called operational amplifiers or comparators, and a
few resistors and capacitors. These parts cost maybe $2 at Radio Shack,
and will put out a voltage that depends on the width of the pulse -- say
0V if the received pulse has "normal" width, -0.5V if it's a bit shorter,
and +2V if it is quite longer than standard. This part is not a very big
deal.

Trouble is that most modern solid state circuits can't handle high
currents (and even small motors in cars or airplanes can require several
amps when they start up or go at full power, more than a garage opener!)
To make this low-power signal control a power-guzzler like a motor,
you'd need a power amplifier, like the one in your stereo only larger.

So people invented wonderful gadgets, "solid state switches", which
are relatively small but can handle high currents. One of them is a
"power MOSFET", which is really a special type of a transistor.
You apply a small voltage to one pin, and the other two pins start
conducting (like a wall switch), and can carry a very large current
with little resistance; you take the voltage away, and the two power
pins stop conducting.

When choosing MOSFETs, pay attention to the highest "source to drain"
voltage (normal R/C equipment uses up to 20V, but because of inductive
spikes and such you want at least 60V); maximal continuous current (as
much as you can get! 10A is the least for a rather small motor); gate
"on" voltage (only "logic level" FETs will turn fully on at the 5V or
so at which a typical ESC operates - with other types you must play
various voltage doubling tricks); and the "on" resistance: the lower
the better, and you can decrease it by wiring several FETs in
parallel.

But this is still "on and off", not proportional. You want the motor
to run slower or faster, not to hiccup on and off. So most ESC's take
this small voltage (remember, the one that comes from the comparator),
and chop it up into pieces -- they apply the control voltage to the
MOSFET on and off but very fast: a thousand times a second or more.
Depending on how high the control voltage is, they keep the MOSFET
turned on for a shorter or longer period of time. The motor keeps
turning (no motor can stop and start 1000 times a second), but it
gets less or more power, so it turns more slowly or faster. This is
exactly the way the dimmer in your living room works -- it sends
power in short spurts to the bulb, but you don't see it blink, only
get dimmer or brighter:

  +-+       +-+       +-+        +-----+   +-----+   +-----+
  | |       | |       | |        |     |   |     |   |     |
  | |       | |       | |        |     |   |     |   |     |
--+ +-------+ +-------+ +--    --+     +---+     +---+     +--
 not much power coming in      lots of power, motor turns fast

For various rather obscure reasons this method of pulse driving
power equipment is very good and efficient compared with the old
fashioned "huge variable resistor which sinks loads of energy"; it
makes motors and friends start with higher torque and eat up less
power than they normally would.

There are a few complications (making the motor run backwards and
forward, or making it _brake_ the car, etc.) These involve arrangements
such as a braking FET which swallows (shorts) energy generated by the
motor which is being turned by airflow or inertia - and can charge the
battery while it's doing so!; and an H-bridge which uses several FETs
to send current flowing _backwards or forward_ to the motor, making it
go forward or backwards.

But the main idea is simple: a) measure the width of a pulse from the
receiver, b) translate this into a small "analog" voltage signal,
c) turn the power on and off so fast that the motor will run steadily
but slower or faster depending on the pulse width.

In newer digital speed controls the second analog step is not needed --
a microcomputer measures the input pulse and figures out what signals
to send to the FETs.

OK, good luck with the project; class dismissed!