Digital RC Servo Jitter

It sounds like the various gains of the PID algorithm used are not tuned to
your system.

The classic tuning method is this:
(1) set all gains to zero.
(2) Increase the proportional gain until the motor starts to jitter.
(3) Increase the integral gain until the jitter stops.
(4) Increase the differential time until the response time is acceptable.

As far as I know for these RC servos, the controller gains are not
tunable. Typically, the microchips are built into the servo and
programmed for some single response (exceptions being some HiTec
programmable servos). Also, I'm pretty sure that standard RC servos
only do PD control.

I was just wondering if there might be some other way to address the
problem that doesn't involve controller gains, or if, perhaps, the
problem stems from something other than the gains programmed into the
servo.

-aaron

I would suspect the following:

Either your mechanism is not behaving/designed properly,

You are not providing the servos with a proper drive signal.

Mike

"blueeyedpop"


I would say the exact same thing.

When I was playing around with RC servos, I got the same jittery. But when I
was controlling the servo with the remote control and receiver, no jittery.
So my conclusion is that it must be possible to make it jittery free, it's
just a question of refining the PWM timing.

As for the 5900 steps, this is simply a number advertised by JR. They
also state it as a deadband of less than .001 ms.

I haven't really looked at the steadiness of my signal, however. I was
just assuming the signal was fine since I'm controlling the servo using
a 32 bit PWM signal from a Quanser board with Matlab and xPC.

Ground bounce and power supply stability could also be problems (as a
mechanical engineer I have near zero electronics knowledge). When I
drive it from the Quanser, though, I'm using a regulated 5 Volt DC
power supply. However, I've also been testing it using the 8 bit PWM on
a Mini-SSC II board. Power supply for this board is some crappy
regulated wall wart.

The stage itself is a compliant mechanism utilizing flexures rather
conventional mechanical components and its mass is around 100-200
grams. More significant is probably the spring force that needs to be
overcome for actuation, but at its max, it's only about .2N.

amhoov@gmail.com wrote:

Plus the best pulsewidth resolution you can get is about 4-usec.

amhoov@gmail.com wrote:

I'm not that familiar with digital servos, but the problem maybe with
your #'s ... 5900 steps in 120 deg. Roughly speaking, this corresponds
to 5900 steps for 1.5 msec change in pulsewidth [assuming 1 msec change
gives the usual 90 deg movement for these servos], and this corresponds
to 1500/5900 = 0.25 usec/step
resolution. I'll bet the servos were shipped to you with a setting more
like 1-2 usec step resolution and a deadband of 2-4 usec. This
would make them work, right out of the box, similar to std analog
servos.

Also, digital servos are much more prone to jitter on the incoming
signal than analog servos. A guy on another forum had a problem running
digital servos from the usual hobby-level servo controllers. The p/w
stability and resolution accuracy just wasn't good enuf. But this
should not be a problem with 32-bit hardware generated PWM.

Also, digital servos draw a lot more current than analog, so your power
leads may also be a problem. Inductance in long wires can lead to
ground bounce, which is the reference signal for the pulsewidth input.
You might try short power supply wires, and add bulk and bypass
filtering to the servo power buss. Also, check that your power supply
is very stable.


- dan michaels
www.oricomtech.com
=======================

On 24 Jun 2005 09:11:18 -0700, "amhoov@gmail.com"



Is a digital servo capable of resolving its current position into
5900 descrete steps in 120 deg of rotation? If so, how does it do
it? I assume it is not a typical pot that it uses.

A digital servo without its corresponding programmer is really just an
expensive analog servo with a little more torque. I am not familiar with
the JR digital servos, but others like those from Hitec ship with
"nominal" settings programmed in. If you want to take advantage of all
the features of the servo you need to program it.

If you think about it, .001ms deadband is about guaranteed to give you
jitter unless your signal is crystal-steady. This is seldom the case.

Finally, you don't mention the mass of the stage you are moving. These
are very small servos. The higher the mass, the more the overshoot, the
greater the jitter.

-- Gordon