Re: [CAD_CAM_EDM_DRO] Tuning EMC

mjf462001 wrote:

> I am now discovering the joys, and pitfalls of tuning a servo system.
> There are two problems I have not overcome.
>
> 1) using frequency mod, I step one axis away from zero in 0.100"
> increments and then back again, but it doesn't return to zero. there
> seems to be some backlash<?> Each axis has a different amount of
> error from 0.0009 to 0.005". If I keep stepping in the same direction
> it moves 0.100 like it should. could the encoder be loose?

All servo systems have some amount of deadband. If the DC gain is
very high, the static error should be quite small, no more than 1 or 2
encoder counts. Freqmod adds a whole bunch of complications. as
it simulates a servo system in software, and then you are driving a servo
system with the step pulses. Why don't you try stepmod?

How are you measuring this? With a dial indicator? That will cause
additional errors, as there is no such thing as no backlash. I'm
assuming you are using a rotary encoder on the motor, but that
you have no encoder feedback to the computer.

Yes, encoders can show subtle error conditions, especially when there
is dirt on the wheel. You might try making small moves while watching
with a dial indicator, and see if they are of uniform size. I mean moves
of perhaps .001, .005 or .010", and make a bunch of them. If some of
the moves are not uniform, that could mean the encoder is missing
some of the slots in the grating wheel. A loose encoder wheel or
coupling would usually cause violent movement when making small
moves. Any time the encoder is not following the motor, the system
goes "open loop", at least momentarily, and the servo loop accelerates
the motor, wildly trying to make the encoder start moving again. If this
is happening, you should be able to easily see and hear it, unless the
slipping/looseness is in very small amounts.

Ahh, you must be using Gecko drives, you should know that they
have a somewhat odd way of interpreting the direction signal. Basically,
the Gecko wants the direction signal to only change in coincidence
with the step pulse. If the particular step-pulse code you are using
doesn't do it this way, the Gecko drive may lose or gain one step
(encoder) position every time it changes direction.

> 2)in mdi I ask it to move .100" and the motors do not move nearly as
> much as a .100" step in manual mode. It seems I have overlooked a
> variable somewhere.

Oh, now this is REALLY odd. Make sure you have set either incremental
(G91) or absolute (G90) coordinates, as having the machine in the
wrong setting will add much confusion. How much is the difference?
If the acceleration is very high, you may have different velocities causing
the machine to stop on the near or far end of the mechanical and/or
deadband backlash, depending on how abrupt the stop was, and from
what speed.

Jon