speed control with modified servos?

The farther the control pulsewidth gets away from the "center" [or
null
speed] position, positive or negative, the faster the servo turns, up
to
its particular max speed. Should be easy to demonstrate.


had within +/-200 usec or so of the center point, but still
controllable.

<snip>

Misconception, but a common one. As Dan notes there is a speed
differential about center (in this case, the neutral position, or 1500)
us. This is inherent in the servo position algorithm to ramp down the
speed as the servo reaches its desired destination -- cuts down
overshoot.

The problem is, if the servo exhibits speed changes with only +/- 200 us
(I've found it a bit more limited, but it's servo brand/model specific),
and the average servo has a deadband of about 8-10 us, you can see the
speed control isn't very granular.

If you don't like the limited speed resolution, you can always gut out
the electronics and use your own, and then you have a very nice
gearmotor that's easy to mount and attach wheels to, and that probably
cost less than most surplus gearmotors.

-- Gordon

Yes, that's a good point.  An even easier modification than just
converting it for continuous rotation, I would imagine -- just rip out
the whole control board and get two leads onto the motor itself.  Thanks
for pointing this out.

Best,
- Joe

You still need to convert it mechanically. The output gear on most
servos has a stop on it that needs to be removed. You usually also need
to decouple the pot from the output gear, and this may involve removing
the pot, drilling out the bottom of the output gear, or removing the
slotted clip placed in the bottom of the gear (GWS servos are like
this).

-- Gordon

Thanks again.  It's heartening that at least some servo manufacturers
have noticed that robotics hobbyists are using them this way so often.  
Though this seems to be mostly the new kids on the block, like Bioloid;
the old R/C servo companies seem to be surprisingly slow to adapt.  
(Though at least GWS makes the hack easier, which is something.)

Best,
- Joe

    Take a look at this circuit.  You can build a real speed controller
for the thing, using the brush noise as an encoder.

                John Nagle

What circuit?  Did you mean to include a URL or something?

Thanks,
- Joe

What I meant to say was that [from my recollection], a servo adjusted
for
zero speed at 1.5-msec will hit max speed forward and backwards at
about
1.7 and 1.3 msec, give or take a bit. Also, over the range of 1.5 up
to 1.7-msec
the speed will be graded, although possibly not a linear relationship.
It doesn't
get much faster for pulses > 1.7-msec or so. But helps to check
specific servos.

    If you want a speed controller that understands R/C pulses,
get one.  Every R/C car has one.  An R/C servo is the wrong tool
for the job.

                John Nagle

You missed the point of the post. Joe wants to use a modified RC servo
and still be able to have reasonable control of speed, rather than just
off, full CW, and full CCW. It's a common request.

-- Gordon

The SSC-32 Servo Controller i have been using has a rudimentary speed
control operation (http://www.lynxmotion.com/Product.aspx?productID95 ).
You can direct the servo to reach a predefined position X in T amount of
time. I think underneath it is done by splitting the total X position in
small segments and pulse the servo to reach these segments every few ms
creating a 'slower' impression. The fastest then would be the servos default
capability.

regards

Konstantinos Dermitzakis

Almost any servo controller has that, but it won't work with a servo
hacked for continuous rotation, which is what's under discussion here.

Best,
- Joe

In fact, I don't want that; I'm generally of the opinion that R/C pulses
are a very crude way for devices to communicate.  A serial protocol
makes a lot more sense.  But I do see other advantages to using servos
as gearboxes -- and they certainly are popular, so I'm trying to be sure
I understand all the ramifications of using them that way.

Best,
- Joe

To get a wider range of the speed ramp in a servo, you might want
to look at the below page.  

http://www.seattlerobotics.org/encoder/200009/S3003C.html

I have one of Gordon's servo tanks, this one has modified Parallax
servos. I measured linear tread speed [inches/sec] versus servo
pulsewidth [usec] as follows.

pulse speed
1500  stopped
1525  0.9 ips
1550  1.9
1575  2.2
1600  2.6
1650  2.8
1700  2.8
1750  2.8
... etc