On Tue, May 20, 2008 at 08:14:00AM +0100, Ian W. Wright wrote: > why does a low accel round corners more than a fast accel.?
By default, the emc trajectory planner tries to keep the greatest speed possible while touching each commanded line (or arc) in at least one spot. It does this by "blending", which means overlapping the deceleration part of one move with the acceleration part of the next move. By default (G64 without P-) the entire deceleration phase is going to be overlapped. When you specify a tolerance (G64 P-), only the deceleration phase that is also within P- of the specified end point is a candidate for being overlapped. The reason that the deviation is greater for a machine with a low acceleration is simple: The lower the acceleration, the greater the distance you have to travel before reaching a given speed. That distance is the distance from the programmed endpoint where emc will begin to blend away from the specified path. My own little machine has an inifile acceleration of 35 in/s^2 and the trajectory planner algorithm uses half of the available acceleration during blending, so it can decelerate from a cutting speed of 12 in/min in 11ms. In that time it travels only 1 mil (in metric: 900 in/s^2 inifile acceleration; F300 mm/min, 30 micrometer stopping distance). So even without setting a tolerance, there's essentially no deviation from the programmed path visible on the part. (stuff like backlash and screw irregularity almost certainly contribute more to error!) If your gcode will be running on a machine with different characteristics, then simply remember to program G64 P- with a tolerance value of your choice. Cutting speed will be lower at corners, but the cutter is guaranteed to move to within P- away from the programmed endpoints. Or program G61 so that emc will always move exactly to the programmed endpoints (but the cutting speed will drop to zero at those corners). Jeff ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2008. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ _______________________________________________ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
