Quick Motion Problems

 ;TIME    TARGET    ACTUAL    ERROR  DRIVE TARSPD ACTSPD STATUS TARANA ACTANA ANASTS INTDRV  INTRATE
 
 17424     12766     12766        0  -4500   2692   2692  00068    300    -64  00044      0    0.000
 17448     12708     12708        0  -4500   2671   2671  00068    300    -14  00044      0    0.000
 17472     12648     12648        0  -4500   2637   2637  00068    300    -57  00044      0    0.000
 17496     12582     12582        0  -4500   2684   2684  00068    300    -68  00044      0    0.000
 17520     12516     12516        0  -4500   1617   2684  00068    300    -46  00044      0    0.000
 17544     12456     12456        0  -4500    324   2172  00068    300    -15  00044      0    0.000
 17568     12398     12398        0  -4500    969   1762  00068    300    -88  00044      0    0.000
 17592     12331     12331        0  -4500   2219   1301  00068    300    -70  00044      0    0.000
 17616     12519     12269      250  10000   3469    725  01070    300    295  00044     31    1.329
 17640     12756     12323      433  10000   4732    124  01070    300   -214  00044    149    5.035
 17664     12993     12353      640  10000   5986    431  01070    300    302  00045    332    7.789
 17688     13228     12403      825  10000   7198    892  01070    300    504  00044    577   10.459
 17712     13461     12478      983  10000   8384   1263  01070    300    485  00044    878   12.830
 17736     13691     12553     1138  10000   9587   1587  01170    300    518  00044   1229   15.011
 17760     13919     12617     1302  10000   9685   1835  01170    300    546  00044   1634   17.253
 17784     14143     12665     1478  10000   9549   1574  01370    300    591  00044   2000   15.632
 17808     14363     12694     1669  10000   9387   1425  01370    300    618  00044   2000    0.000
 17832
 [/quote]
 
  This does not look good.
 The time of interest is 17.616 seconds although the real problem occurred earlier.  Basically the problem is noise.  The system is running in open loop until a calculation decides it is time to ramp down.  The equaion that is used is velocity^2/(2*DecelRate) = RampDownDistance.   If the RampDownDistance is less than or equal to the distance left to go then the controller starts ramping down.  At this point the target position is set to the actual position and the target speed is set to the actual speed.  This is the weak part of the alogorithm.  If the actual speed is bogus because of a bad reading the controller can go into closed loop mode using a bad target velocity.  This is what is happening.  You can't see the bad reading in this plot because the data is captured only every 24 milliseconds on this plot.
 
 Velocity errors that are low don't trigger the transition to closed loop only the velocity errors that are too high do.  If you check out the equation you will see why.
 
 I will look into what could be done improve the noise immunity .  I hate noise.