On 04/06/2016 03:53 AM, Marshland Engineering wrote: > I finally have my mill table moving under CNC control. > > I'm using a 70 volt supply with DCM D6 drives in Torque mode, Mesa 7i48 cards, > Linear scales for position and 1.2 kw DC brush motors (my guess on the power > as they are 90 volt motors but no current rating) > > The only way I can get any control over the travel is setting P = 0.5 which of > course makes the position accuracy poor. Any more P and I get sever > oscillation. With the size of motor and the 1m table, there is quite a lot of > inertia in the system. > > One other problem you may be running into is backlash between the motor and the linear encoder. If there is more than about .001" of backlash, oscillation or hunting will be inevitable. The only hope is to put in a large amount of DEADBAND, but that has a number of other bad side effects. What happens is a small error between commanded and sensed position is seen, the PID decides to make the motor move to correct the error, and the motor starts to accelerate. But, due to backlash, the encoder sees no movement, so the motor continues to accelerate. Finally, the backlash is taken up, the encoder moves, but by this time the motor has picked up appreciable speed, and drives the encoder way past the small amount of error. Then the cycle repeats.
Look at velocity, or possibly position, with halscope. If the velocity shows a kind of square wave pattern, where there is zero velocity for a while and suddenly it jumps, that is a strong indication of backlash somewhere in the system. It could be the ballscrew end bearing, the ball nut, the mount for the ballnut, bowing of the ballscrew, or the link from the machine to the read head of the linear scale. It may be possible to put a long handle on the ballscrew in place of the pulley or coupling to get a feel for this. A large tap handle is great for this test. While you have LinuxCNC look at the encoder position, see how far you can move the tap handle before the DRO number changes. Ideally, you should not be able to even feel the slack before the DRO starts to move. Jon ------------------------------------------------------------------------------ _______________________________________________ Emc-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/emc-users
