Re: [CAD_CAM_EDM_DRO] Ball Bar Tests & accuracy

Greg Jackson wrote:

>For the most part, stepper versus servo will not be much of a comparison in
>accuracy. One of the big advantages of a servo is that, because of the
>feedback, you can sustain large dynamic errors and still not fault the
>machine.
>
><snip>
>The bottom line is that, while servos have a number of advantages, accuracy
>is not one of them. Steppers can provide a great deal of accuracy at low
>cost. With a typical 2000 line encoder providing 8000 counts per
>revolution, the 128 count fault level of a Gecko is 0.016 revolutions. A
>200 step motor will lose a step at 0.005 revolutions (1/200).
>

Gecko 320/340 drives are not true servos, in the sense that the position is
not relayed back to the CNC control.

With a true servo, such as with EMC, an encoder/DAC card and a real analog
servo amplifier, you have much more control and sensing of dynamic error.
You can set the error limits as tight as you feel necessary, and use the
sliding scale feature to make the error limit proportional to velocity.
You can also use the logging feature to record actual vs. commanded position
in real time and plot it.

On my Bridgeport conversion with a true servo system, I have an encoder
resolution
of 0.00005" (not a typo, 50 micro-inches) in X and Y, and 0.000025" in Z.
(1000 cycle/rev encoders on 5 and 10 TPI ballscrews.) It generally
idles within
+/- 1 count of null, and can run at least 30 IPM with error of about 3-4
counts.
Those are errors of .00005" at idle, and .0002" at reasonable cutting
speeds.
Error at 120 IPM is not important, as I don't ever cut at that speed.

(Yes, purists, these are errors at the ENCODER only, I know that. The
ballscrews
and the rest of the machine are not quite that accurate. I do have high
precision
grade ground ballscrews that probably cost $5000 each when new, so they ARE
very good.)

> In operation,
>a stepper will probably be well within 0.002 revolutions. In order to meet
>that sort of level, the Gecko servo would have to maintain an error of less
>than 16 counts. This can happen, but you cannot know that it has happened,
>all you know is that it never got greater than 128.
>
>

Right, the fixed error bound is a shortcoming. Note that the Rutex step to
servo drives have an error bound of 30,000 counts! As far as I can
tell, this
is burned into the PIC firmware in their drives.

One other comment: What you say about following error of full-step or
half-step
drivers is true, but many microstepping drivers regenerate a synthetic
(micro)
step pulse that is not the same as the microsteps coming out of the
computer.
This smooths the microsteps actually being fed to the motor, improving the
highest speeds before the motors begin losing steps, but it may cause errors
between the position the computer thinks the motors are at and the position
they really are. I'm talking about such devices as the Gecko 210 and 340
pulse multipliers, here. I've never gotten any info on how far the
multiplier
allows the input vs. output to be out of sync.

Jon