User Tools

Site Tools


slow_speed_considerations

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revision Previous revision
Next revision
Previous revision
slow_speed_considerations [2016/09/07 15:26]
asiadmin ↷ Page moved from documentation:slow_speed_considerations to slow_speed_considerations
slow_speed_considerations [2016/10/04 15:13] (current)
Line 23: Line 23:
 ===== Optical Tracking Tests ===== ===== Optical Tracking Tests =====
  
-{{ :​documentation:​slow_1.jpg?​direct&​300 |}}+{{ slow_1.jpg?​direct&​300 |}}
  
 The next three plots show actual stage motion as determined by looking at the position of an object optically and plotting its location. ​ The first plot shows a stage with 1.59 mm pitch lead screws and rotary encoders traveling at its slowest controlled speed, about 0.1 μm/​sec. ​ There are distinct glitches in the motion every 10 seconds or so, corresponding to motion of about 1.25 μm. The next three plots show actual stage motion as determined by looking at the position of an object optically and plotting its location. ​ The first plot shows a stage with 1.59 mm pitch lead screws and rotary encoders traveling at its slowest controlled speed, about 0.1 μm/​sec. ​ There are distinct glitches in the motion every 10 seconds or so, corresponding to motion of about 1.25 μm.
  
-{{ :​documentation:​slow_2.jpg?​direct&​300 |}}+{{ slow_2.jpg?​direct&​300 |}}
  
 The second and third plots show motion where the stage was commanded to go about 1 μm/sec and 5 μm/sec respectively. ​ Both plots also show the glitches, still about 1.25 μm apart, now occurring more frequently because of the faster speeds. ​ Hence, it appears that the glitches are due to some fixed mechanical disturbance to the system. The second and third plots show motion where the stage was commanded to go about 1 μm/sec and 5 μm/sec respectively. ​ Both plots also show the glitches, still about 1.25 μm apart, now occurring more frequently because of the faster speeds. ​ Hence, it appears that the glitches are due to some fixed mechanical disturbance to the system.
Line 37: Line 37:
 D =  \frac{1.59}{141} ​ =  11.3 μm/rev \end{equation} D =  \frac{1.59}{141} ​ =  11.3 μm/rev \end{equation}
  
-{{ :​documentation:​slow_3.jpg?​direct&​300 |}}+{{ slow_3.jpg?​direct&​300 |}}
  
 With the spacing of the glitches at 1.25 μm, we are seeing about 9 glitches per turn.  This turns out to be exactly the number of teeth on the primary motor gear in the anti-backlash gearhead on the motor. ​ The anti-backlash gearhead consists of two identical parallel gear trains, locked in preloaded torsional opposition at the motor and output shaft. ​ Perhaps the output shaft jumps ahead as a gear tooth is released and then abruptly halts as the next tooth catches it.  At the very slowest speed, the servo loop has plenty of time to ensure that the motor is turning quite uniformly, even in the presence of torque variations caused by the glitches. ​ Indeed, data taken from the motor encoders show very short glitches of 5-6 encoder counts, whereas the displacements seen in the optical measurements would be closer to 20-30 encoder counts lasting much longer – consistent with this explanation. ​ It is clear that regulating strongly on the rotary encoder so that the motor will turn at a uniform rate will not significantly reduce the speed variations in the output shaft. ​ It would be more effective to regulate on stage linear encoders and attempt to modify the motor speed to compensate for the glitches. ​ Because of their sudden time varying nature, this approach will only yield marginal improvement,​ especially at intermediate speeds between 1 μm/s and 20 μm/​s. ​ Experiments would need to be done to see if aggressive servo loop tuning could improve the response enough to make a significant difference using the linear encoders. With the spacing of the glitches at 1.25 μm, we are seeing about 9 glitches per turn.  This turns out to be exactly the number of teeth on the primary motor gear in the anti-backlash gearhead on the motor. ​ The anti-backlash gearhead consists of two identical parallel gear trains, locked in preloaded torsional opposition at the motor and output shaft. ​ Perhaps the output shaft jumps ahead as a gear tooth is released and then abruptly halts as the next tooth catches it.  At the very slowest speed, the servo loop has plenty of time to ensure that the motor is turning quite uniformly, even in the presence of torque variations caused by the glitches. ​ Indeed, data taken from the motor encoders show very short glitches of 5-6 encoder counts, whereas the displacements seen in the optical measurements would be closer to 20-30 encoder counts lasting much longer – consistent with this explanation. ​ It is clear that regulating strongly on the rotary encoder so that the motor will turn at a uniform rate will not significantly reduce the speed variations in the output shaft. ​ It would be more effective to regulate on stage linear encoders and attempt to modify the motor speed to compensate for the glitches. ​ Because of their sudden time varying nature, this approach will only yield marginal improvement,​ especially at intermediate speeds between 1 μm/s and 20 μm/​s. ​ Experiments would need to be done to see if aggressive servo loop tuning could improve the response enough to make a significant difference using the linear encoders.
Line 46: Line 46:
 The problems of uneven slow-speed operation reared its head again during slow speed scanning experiments using the controller’s SCAN mode functions. ​ Again the problem was traced to the periodic variations in speed caused by the motor gear teeth on the anti-backlash gearhead. ​ The problems were reported in both rotary and linear encoder mode.  Since the anti-backlash gear preload is suspected to be the origin of the uneven movement, several experiments were done using bare motors, i.e., motors not installed in stages. ​ SCAN firmware was used to initiate several low-speed scans, and the encoder divide-by-N signal pulses were recorded on an oscilloscope. ​ Data was taken first using a standard off-the-shelf stage motor, then with a motor where the preload on the anti-backlash gear had been removed by pulling the gearhead and replacing it with no torsion on its gear trains. ​ Typical data recordings for the two motor configurations are shown below. The problems of uneven slow-speed operation reared its head again during slow speed scanning experiments using the controller’s SCAN mode functions. ​ Again the problem was traced to the periodic variations in speed caused by the motor gear teeth on the anti-backlash gearhead. ​ The problems were reported in both rotary and linear encoder mode.  Since the anti-backlash gear preload is suspected to be the origin of the uneven movement, several experiments were done using bare motors, i.e., motors not installed in stages. ​ SCAN firmware was used to initiate several low-speed scans, and the encoder divide-by-N signal pulses were recorded on an oscilloscope. ​ Data was taken first using a standard off-the-shelf stage motor, then with a motor where the preload on the anti-backlash gear had been removed by pulling the gearhead and replacing it with no torsion on its gear trains. ​ Typical data recordings for the two motor configurations are shown below.
    
-[{{ :​documentation:​slow_4.jpg?​direct&​400 |Click to Enlarge}}]+[{{ slow_4.jpg?​direct&​400 |Click to Enlarge}}]
  
-[{{ :​documentation:​slow_5.jpg?​direct&​400 |Click to Enlarge}}]+[{{ slow_5.jpg?​direct&​400 |Click to Enlarge}}]
    
 Removing the anti-backlash preload significantly smoothes out the motion. ​ The table below summarizes the results of several runs using various scanning speeds and encoder counts per output pulse. Removing the anti-backlash preload significantly smoothes out the motion. ​ The table below summarizes the results of several runs using various scanning speeds and encoder counts per output pulse.
Line 77: Line 77:
 Based upon the above considerations,​ a stage equipped with linear encoders and a standard spur gearhead motor without the zero-backlash feature was set up and tested. ​ A representative typical scan is shown below for the stage traveling at 10 μm/sec with encoder pulses every 240 nm.  This stage was equipped with 1.59 mm pitch lead screws. ​ None of the very obvious speed fluctuations seen with the zero-backlash gearhead are present. ​ The speed is still not perfectly uniform, undoubtedly because of small mechanical imperfections in the motor and bearing systems. Based upon the above considerations,​ a stage equipped with linear encoders and a standard spur gearhead motor without the zero-backlash feature was set up and tested. ​ A representative typical scan is shown below for the stage traveling at 10 μm/sec with encoder pulses every 240 nm.  This stage was equipped with 1.59 mm pitch lead screws. ​ None of the very obvious speed fluctuations seen with the zero-backlash gearhead are present. ​ The speed is still not perfectly uniform, undoubtedly because of small mechanical imperfections in the motor and bearing systems.
   ​   ​
-[{{ :​documentation:​slow_6.jpg?​direct&​400 |Click to Enlarge}}]+[{{ slow_6.jpg?​direct&​400 |Click to Enlarge}}]
  
 Quantitative measurements of the velocity uniformity were made for various commanded scanning speeds. ​ The results of the tests are summarized below. Quantitative measurements of the velocity uniformity were made for various commanded scanning speeds. ​ The results of the tests are summarized below.
slow_speed_considerations.txt · Last modified: 2016/10/04 15:13 (external edit)