not enough area for continuous modeled stage position cal.
Added by Anchi Cheng almost 18 years ago
This is a question from mbycliaw.
The well correlated object moves out of the view during the measurement part of modeled stage position calibration. Can I reduce the interval to 1e-6 rather than the current recommendation of 5e-6?
Replies (1)
- Added by Anchi Cheng almost 18 years ago
The modeled stage position calibration fit the periodic deviation of stage movement due to the design of the goniometer.
Since we know from experience that the base period in x direction is about 60 um, we measure the amount of image movement for each 5 um mover at 25 starting position separated by 5 um so that the total distance covered is more than 120 um ( 2 period). The 5 um was determined so to get enough samples to fit 5 order harmonic series (For 60 um base period and 5 harmonic term, the sampling interval need to be =< 60/5/2 um, i.e. 6 um ).
As you see here, you do need to have objects to correlate in such a large distance if the measurements are done in one shot. However, since each movement is small, the measurements can be broken into segments so that this procedure does not need to be a continuous area.
For x model, for example, you can do 5 points at one place and then move mainly in y to another good area and do another 5 points ( so that x starting point is more or less continuous when you fit, not essential but convernient), and so on. As long as you are using the same label, all data points will be included in the fitting. If you sample very finely, you will have to take a lot more points total to cover at least one expected base period for that period to be fitted right.
What you suggested to reduce the interval is o.k. but would mean that you need to repeat the measurement at several other starting x positions since, in your case, 20 points only covers 1/3 (= 1um * 20 / 60 um) of the expected base period.