PWM generation details

PWM is a technique that allows smooth changes in motor output, using
simple on/off switches.  This is achieved by switching faster than the
mechanical system can respond, such that the mechanics smooth out the
switching to average the switch state.

Sensor feedback (such as from shaft encoders) is required for accurate
movement; but for simple tasks it is often sufficient to make a few
measurements such as "75% PWM is about X[m/s]" and base your control
on that.

Later,
Daniel

The point of PWM is not that you can use on/off switches.  The most
common "switches" used in motor drives are transistors and are quite
capable of operating in linear mode to achieve speed control.
However, doing that is inefficient because it's essentially the same
as using a big variable resistor.

Further, low frequency PWM is not a great idea.  Driving motors with
pulsed current makes them whine annoyingly and may cause damage.  It
is also less efficient because the pulsed nature of the current
results in greater resistive losses in the windings.  Ideally you
operate PWM drives at a sufficiently high frequency that the current
is smoothed by inductance.  In this mode, the motor driver is
essentially a variable dc-dc down-coverter.  The "inductance" here can
actually be the windings of the motor but in that case you cannot use
noise suppression capacitors.

You'll never be able to get precise speed control without feedback,
but in the case of treads, even precise speed control doesn't
translate to precise dead reckoning.

-chris.

Also, you can use one bit per motor rather than enough bits to get the
desired resolution.


I think you misread his post as suggesting that you sit there toggling
a physical switch back and forth with your index finger.  He pretty
clearly meant something that could be switched "fast enough" to work
well, though he focussed on the mechanical system's response speed
rather than induction.

    Generally PWM rates are in the audio range, from hundreds of hertz to
hundreds of kilohertz.  The trend is towards higher frequencies, outside
the human hearing range.  See the Atmel ATA6831, which is a dual PWM
driver for automotive applications designed to run at up to 25KHz.

    This is for brush-type DC motors.  There's also polyphase PWM, where
several phases are driven separately and in sync with motor rotation.
Brushless variable-speed motors are driven that way.  This requires
more elaborate drive electronics but is more efficient.

    These things are actually quite difficult to design well.  They
have to deal with brush noise, inductive kickback, short circuit and
overcurrent protection, and thermal protection.  The ones designed
for RC motor speed control are usually quite good at current capacity,
efficiency, and protection, but dumb - no feedback or computer interface.
The ones designed for hobbyist robotics tend to be smarter, but bigger and
not as good on protection circuitry.  The ones for industrial motor control
are all over the place; some are very robust, and some are underdesigned,
and they're all much bigger and more expensive than the R/C units of
comparable current carrying capacity.

    It's still harder than it should be to get small, fully-protected
motor controllers with encoder feedback at low cost.

                    John Nagle

No.  This is not a matter of "fast enough".  Inertial and inductive
averaging of pwm driver output are different processes, use different
switching rules, are described by different math, and he pretty
clearly explicitly named the former.  A small dc motor driven with a
50% duty cycle at 1khz will behave quite differently from the same
motor driven at 20+khz, and for most robotics applications you want
the latter behavior.  The low frequency pwm response won't even be
linear.

-chris.

 For the first time in history , we have everything

 we need , everything is far faster than needed .

 Yet some people still use PLL , P.I.D. loops ..

 It's job security . They can take it from the

 electronics "books" , all of which are obsolete .

  We use faster ARM mcu's , with adaptive

 circuits , nixing the slow an old Proportional

 integ , dir' loops .

  instead of loops , we measure the motor

 and slam it with a square wave and the result

 is predictable movement , thus no need to

 put it in a control loop . less circuitry .

  Added to that motor are linear motors and

 brakes that decrease it's inertia .  mcu's easily

 handle the brakes and linear motors .

   But it dont work if we "study" engineering

 from text books . this is Luddite , job security

 for college professors .

  In the future , there will be no income for

 textbook engineers , since the public will be

 paying "directly" , for what they need .

 No more govt control of what is sold .


 Make one tiny mistake , and you're branded

 a fool .

 It amases me , how long our industrial

 robots did NOT evolve !

 Luddites claimed they used the most modern

 engineering ! Cant be improved !


 Linux , WXP , the PC , cars and trucks

 stepper motors , PIC mcu's , Cell Phones ,

 Campbels chicken soup , taxi's , doctors

 hospitals ...all obsolete , will not be used

 past 2030 !

    There will be NO textbooks to "learn"
 from .  No colleges , no professors .
  No engineer will "teach" ..

  It's the future ..