Control electronics power - need for isolation?

In anything other than a trivial robot I *always* use two different
battery packs: one for the motors and the other for the electronics.
--
D. Jay Newman
http://enerd.ws/robots/

Whether or not you elecrically isolate your power system from your computer
system is a matter or preference and regulations for your application.

IMHO, if you are careful, and what you are creating is your own hobby, then
you can dispense with electrical isolation as it can get expensive.


Most moderate power DC->DC switchers are electrically isolated, but they do
something more important. They convert power with great efficiency. A
linear power supply is the worst, it burns power that you don't use.

Look at it this way:

Linear regulators use a semiconductor as a psudo-resistor, so 1 amp at 5
volts is five watts, but if you are driving your regulator with 12 volts,
you will still be drawing that 1 amp, so you'll be using 12 watts. You will
be burning 7 watts of power to produce 5.

A switcher uses a switched driver and an LC filter to product 5 volts. If
you drive it with 12 volts, your 5 volts 1A, may only be about 0.42A at
12V.

...

Thanks.  I've got half a rail of LT1171s kicking around as well as cores
and bobbins, so will knock a 24V->5V supply together for the logic.

I was considering a separate NiMH pack as an alternative, but am going
off the idea as this would require another charger AND a switching
regulator to keep the voltage stable against changes in NiMH pack voltage.

Cheers

M

--
Matthew Smith
Kadina Business Consultancy, South Australia
Work: <http://www.kbc.net.au>  Personal: <http://www.mss.cx>

Well, you could use two batteries and use two diodes to allow a charger to
charge both at the same time. That way, if the motors kill the drive
batteries, the computer can still operate and call for help.

(The only thing about using diode is the ~0.6 volt drop, you'ld need to put
the voltage sense on the other end of the power battery diode.)

Separate supplies provide for a number of simplicities, so it's a
trade-off that is fairly situation-specific. If you don't need to do it,
don't bother.

There are a number of devices on the typical robot that can cause power
supply glitches that affect electronics. Motors are one. Because motors
don't usually need regulation, it's often acceptable to use a large
battery for those, with maybe a battery-level monitor, but no regulator.
A smaller, separate set of batteries can be used for the electronics,
and this pack is regulated.

Back EMF from the motor can be coupled into the electronics, so
especially for larger motors, it is often desirable to use
opto-isolators between the control electronics and the motor bridge.
This is effective for both single- and dual-supply designs.

Noisy motors can also be "quieted" by applying disc caps across their
terminals. Some designers like to also put additional caps between each
terminal and the motor case, assuming a metal case on the motor, or
nearby to ground.

VERY noisy motors can induce RFI, which can interfere with electronics,
including audio and vision. I recently hacked a tracked-driven robot toy
I bought at Big Lots, and found the motors were so noisy that I'd have
to use RF shielding to prevent them from lousing up my camera signal,
and causing sputtering noises through the sound amplifier. It was easier
to use better motors.

There are additional methods that can be used to reduce noise from
motors, and these are well-documented in the lit, so I'll leave it at
that. (Sometimes even twising the wires to the motors helps a lot. Don't
overlook the little things.)

Some electronics pull lots of current when actived, and cause spikes in
the power supply. Adding decoupling caps doesn't always solve the
problem entirely. Consider separate little regulators for these. It
doesn't add much to the complexity of the design to create a small
handful of "local power regulators" for several of your sensors and
other electronics. Be watchful of some of the Sharp IR sensors, which
are known to cause noise in the power supply (affecting other
electronics, as well as its own signals). Especially bad are the
Polaroid sonar sensors, which pull from 1-2 amps each time they fire.

-- Gordon

Often times it is good to put an LC filter between the main power supply and
the motor aplifier as well as between the amplifier and the motor. The
power supply filter will round out the switching spikes of the drivers, and
the motor filter will smooth out the power to the motor.

This is not always desirable. You often want the full voltage applied to
the motor armatures to retain torque. If you average out the voltage
reaching the motor, you also reduce current, and therefore reduce
torque. You don't want to simply raise and lower the voltage to the
motor if you care about retaining torque; for PWM you want to apply full
voltage and current for discrete periods of time.

-- Gordon

I'm not sure how to go about completely disagreing with you without
offending you, but I have to.

Torque is a measurement of force. A constant torque against a constant load
means a constant rotational velocity. Constant torque against a varying
load means a varying rotational velocity. The object of motion control is
to modulate the torque of the motor to control potentially changing
velocity against potentially varying loads.

So, to sum up:
If the rotational velocity is to remain constant as a load increases, the
torque must increase.
If the rotational velocity is to remain constant as the load decreases, the
torque must decrease.

Similarly
If the rotational velocity is to increase against a constant load, the
torque must increase.
If the rotational velocity is to decrease against a constant load, the
torque must decrease.

<snip>

All fine and dandy, but it neglects the actual comments I made (this is
becoming a common trait in your responses). You're trying to revise
Ohm's Law by insisting you can adequately drive motors under load simply
by varying their voltage. Reduce voltage and you reduce current. Reduce
current, and you...what happens to torque?

The fact that most PWM circuits do not attempt to "average out" the
motor voltage should give you a clue as to the design philosophies
behind them.

-- Gordon

First of all, Ohm's law has nothing to do with motors. Ohm's law dictates
the behavior of linear DC circuits with resistance. Motors are a reacive
load, and E=I*R does not work. If you apply 6 volts to a motor and measure
the current, and then apply 12 volts, you will not see twice the current,
the current will curve based on the reactance of the motor and the period
that the motor coils stay energized.

Second of all, You WANT to vary the torque. As the load increases you
increase the torque to maintain rotational velocity. As the load decreases
you reduce the torque to maintain rotational velocity.

Whether you know it or not, A PWM controller modulates the effective torque
of the motor. A wider PWM, more torque, and more narrow PWM, less torque.
That is is how they work.

If you are not varying the torque of the motor, how are you controlling
rotational velocity? Power to the wheels is a fairly simple equasion:

If your robot needs 8 ft/lbs of force to maintain a velocity of 8 RPM.

8ft/lbs = R * 8RPM.

If you increase the resistance on the wheels (going up an incline), say:

8ft/lbs = 1.5R * ?RPM

RPM has to equal 5.3. By maintainling a constant torque, you reduced speed.
Now to maintain speed:

?ft/lbs = 1.5R * 8RPM

The torque has to be 12ft/lbs to work.


Just because someone does something a certain way, or even that a majority
of people believe something is right is in no way proof that it is. Most
people believe that they are safe in a car during a lightning storm because
of the rubber tires. Anyone who knows anything about electricity knows it
is because of the metal cage that the car provides.

The truth is that it is easier and cheaper not to use an LC circuit,
however, you are forced to use low frequency PWM and get servo whine and
nice hot motors.

But you're not really espousing true PWM control. You're averaging out
the voltage from the PWM, and presenting THAT to the motors. Your
approach does not allow the motor to receive full voltage (and therefore
maximum current) during the ON time, however short or long that is. Let
the inductance of the coils and the mechanical momentum of the motor
average out current and torque. Speed, and torque, control are still
achieved.

While inductance may involve a complication in voltage/current
calculations for motors at varying speeds, coils are also resistive;
Ohm's Law works perfectly well in calculating maximum current through a
motor. You simply cannot escape the relationships between voltage and
current, nor between current and torque.

I'd really like to see some Web references to the design approach you're
recommending. Could you provide some links? Surely there must be
something out there.

-- Gordon

What?


Averaging out the power, yes. Voltage at the motor is a factor of back emf,
and velocity.



The motor's torque is a product of current.


Speed is a result of torque and load. You control speed by modulating torque
based on the load.


YIKES!!! Coils are NOT resistive!!!!! There may be parasitic resistance
because copper wire is not a perfect conductor, but resistance is *not* a
good part of the motor.


When the motor is stalled and you have saturated the coil, yup, your nice
copper wire and carbon contacts will act like a resistor. All coils do.
That's why motors burn up, because rather than convert emf into rotaional
force (torque), people like to use them like resistors.



As I have said, torque is directly relational to current. Let me correct
this, torque is directly related to current on a non-stalled motor. Hell,
if the thing is stalled, it will act like a resistor, but it isn't supposed
to. Motors are supposed to produce rotational force, not heat.

A motor amplifier controls the power applied to the motor, NOT VOLTAGE.
Voltage is a product of back emf and motor speed. Motor speed is a product
of current (thus torque) and load.

This just caught my eye, and I didn't notice it before. I have *never*
talked about "voltage" to the motor. I have always talked about power.

The torque of a motor is proportional to the motor current, but the speed is
related to back emf, and thus the applied voltage.

Under a constant load and torque, there is a direct relationship between
voltage and current. Under varying load, the motor current must be adjusted
to keep the speed constant.

Okay, let's go with this. Here's an excerpt of one of your comments:
"and
the motor filter will smooth out the power to the motor."

You did use the term "power," but what does it mean? I've always
strictly considered "power" an output measurement, not an input, though
I relize the term is often colloquially used to indicate a voltage
source of some type. What electrical term are you meaning when you say
"power"?


Don't you mean that back emf is related to speed?


In robotics, the *typical* adjustment is to alter speed, independent of
load, in order to simply move more slowly, or to make realtime course
changes in response to odometer readings. We can simplify the discussion
by limiting ourselves to that aspect only for the time being.

What I'm talking about is "starving" the motor of the benefits of a full
inrush of current to its windings by averaging out its voltage BEFORE
that voltage reaches the motor terminals. The construction of the motor
itself amply averages out voltage and current (and therefore torque)
when given a pure PWM signal. Robotics is a high-torque application,
where speed variation may need to be 50% or more. Is this equally
achievable -- while maintaining efficiency -- using a servo control with
what is basically to a DAC voltage output?

Again, traditional PWM discussions are throughout the Internet. Please
provide some links that demonstrate your approach, either with a
discussion of the math, or with practical experiments. Thanks.

-- Gordon

Power is "power." The ability to do work. In electrical terms "watts."



Current to the motor is directly related to the torque the motor produces.
The speed that a motor spins at is directly related to torque and load.

Now, assume you are applying a constant current source to a motor under a
constant load. The motor will accelerate until the torque is in equlibrium
with the load and velocity:

tourque = resistence * velocity.

As long as the torque is greater than "resistance * velocity," velocity will
increase (acceleration). Once they are in equilibrium, there is no more
acceleration. Right?

Now, during the acceleration, the voltage measured at the motor will
increase proportionatly with the acceleration.



No, speed is a result of torque and load. We control torque in the presence
of variable load to maintain or adjust speed.


That makes no sense.  


Yes and no. A PWM signal, if it's frequency is too low will affect the
motors rotational velocity and make it too hard to control. If the
frequency is too high, the motors inductance will end up filtering the
signal.



A PWM amplifyer is not a DAC. A PWM system is more like a "power" source,
lile a current source that reacts to load. This is *not* a voltage source.

However, yes, a filtered PWM output will work better and more efficiently
than a raw PWM at the motor typically would.


I feel like someone trying to defend darwinism to creationsists. There are
*many* text books on the nature of electrical motors and control. I can not
cover all the theory in a few internet posts, and I can't spend a few days
looking for articles published on the internet.

I would suggest reading up on switch-mode power supplies for a solid
understanding of PWM amplifier theory. (Directly relatable to PWM motor
amplifiers.)

Howowitz and Hill, "The Art of Electronics," has a brief explanation of a
"step down" switching power supply. You'll notice that this resembles one
PWM corner of an H-Bridge. All the math still applies.

As for motors, they are harder because typical electronics books tend to
ignore the dynamics of them. A good treatment is in "Basic Robotics
Concepts" by John M. Holland, but you may have trouble finding it.