Re: [CAD_CAM_EDM_DRO] reprap build started

On Sun, Oct 9, 2011 at 6:59 PM, Kevin Martin <kpmartin@...> wrote:

> I agree that there is no point using servos for filament feed (nor, for that matter, for table motion).
> Your comment about filament slippage instantly gave me an image of a second wheel measuring filament motion for the servo feedback. Now it has occurred to me that the pinch roller (which pinches the filament against the drive shaft) could be given markings that an optical encoder could read, thus giving the servo feedback.
>
> But back to reality. Servos probably aren't worth the trouble.
>
> It should be pointed out that, unlike on the axis drives, the occasional missed step on the filament feed is generally not a problem because each step represents so little volume of plastic, and it is not necessary to move the same number of steps in reverse and return to the starting point.

It is true that a single missed step or two is not an issue, but it is
rarely just one or two missed steps. Missed steps on the X or Y is
trivial to detect, but missed step issues are extremely difficult to
detect on the extruder, since it is hard to know how much filament
should have been extruded. What you end up with are weak, parts, poor
bonding to the build surface, and other issues that may not be easy to
immediately identify the cause, so having feedback to alert the user
that your extruder is not working properly would be a huge benefit.
And of course feedback on the extruder could also pause the print and
alert the user in the event you run out of filament or your filament
jams or gets tangled, so all in all it is much more useful than
feedback on the X & Y (our axes are light, so even with NEMA 17
motors, missed steps are not a major problem).

>
> Before going around replacing axis steppers with servos, a better approach would be to put the acceleration smarts into the gcode interpreter/stepper drivers, and make it possible to adjust these tuning parameters without having to build and load new firmware. Properly-tuned steppers could go a long way in curing (perceived or real) stepper performance problems.

Putting the smarts in the drivers would mean all custom drivers...
Certainly possible, but you would need to make a pretty compelling
argument to justify doubling or tripling the driver cost (the most
widely used reprap stepper driver costs $13). The most widely used
firmwares do require reflashing the firmwares to tweak the
acceleration (and just about any other) settings, but runtime config
is becoming available. I use SJFW, which is completely configured at
runtime. It is a bit more bleeding edge than the others, but so I
don't recommend it broadly yet, but it is the way most of the
firmwares will likely go in the future (although the most popular
firmware, Sprinter, may not since the author is dead set against
runtime configuration).