On Thu, Feb 3, 2011 at 12:12 PM, Kirk Wallace <[email protected]>wrote:
> On Thu, 2011-02-03 at 15:16 +0200, Slavko Kocjancic wrote: > > That's not problem of friction at all!. > > > > The worst you have is 'sticky' axis. > > That means you need a lot more power to start to move axis and when > already > > moving the required power is a way smaller. > > And if that ratio is big then tuning PID is near imposible. You can have > low > > static error or low (or none) overshots but not the both!. > > So in this point of view the mechanics is bad nad electronic's can't do > the > > job. > > The ratio can be measured with spring scalle. > > Just make arm on screw and then pull that arm with scale. Slowly pull > until > > axis move then stop pulling and you have reading just at the point before > > start and when stoped. (you need some support for pulling as if you pull > > with bare hand then the 'stop' can't be measured as hand is to 'springy') > > And if ratio is biger than 1:2 then you have serrious problem. > > One can see these forces if you use an oscilloscope to look at the > current going to the axis motor. If there is stiction the current will > quickly spike then fall at the start of a move command. If there is a > delay then a spike, there is flex or backlash. > > I would look at the maximum motor RPM, decide on an maximum joint rate, > then set up a pulley ratio to match, so that the motor has the maximum > mechanical advantage. > > Guys! Please! This is _NOT_ sticktion! When I took the motor off this rotary table (dealing with bad resolver conversion), I took channel lock type pliers and tried to turn the worm screw by hand. I experienced a very typical, regular FRICTION, that is, resistance did not depend on whether the move was just starting or a move in progress, how fast I was turningt, etc. It only depended on the direction of where it turned. I am experiencing a friction problem and not a sticktion problem. see below > Another thing that comes to mind. > > C=pi*D=3.14*6"=18.85" 2min/360deg=2/21600=0.00009259 > 0.00009259*18.85=0.001745" > > So if the table is accurate to 2min, at 3" from the center, the linear > accuracy is close to 0.0017". I would tend to hope for better accuracy. > It would be interesting to put an indicator on a post at a 3" radius and > see what a 10 degree command produces. > > Yes, and it means that the servo motor shaft's position is approximately 4 degrees off from where the motor shaft ought to be. It is pretty substantial. It seems to me, that EMC has compensation for, say, weight of the Z axis (FF1), or constant force acting in one direction, spring effect (FF0), inertia (FF2), but no compensation for "friction", which is a constant force that is opposed to the movement. Being a computer programmer, I would like to consider taking a stab at it and adding a "FFF" compensation, that is, compensation against a constant force that is opposed to the direction of movement. Am I on the right track at all? i ------------------------------------------------------------------------------ Special Offer-- Download ArcSight Logger for FREE (a $49 USD value)! Finally, a world-class log management solution at an even better price-free! Download using promo code Free_Logger_4_Dev2Dev. Offer expires February 28th, so secure your free ArcSight Logger TODAY! http://p.sf.net/sfu/arcsight-sfd2d _______________________________________________ Emc-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/emc-users
