Feature request: improved calibration technique

Hey there. I got a Nova 3, I like the product a lot, simple, good writing feeling. The first thing that does not feel right to me and that I can't solve is the accuracy of the pen. Despite calibrating it multiple times, I still get some minor offsets at different positions in the screen, some other locations are perfectly calibrated. It's a sub-millimeter offset, but big enough to be a bit inconvenient and disturbing. These offsets occur at different places after different calibrations.

I don't think this is a hardware issue though, but rather that it is possibly two-fold:

• A. calibration technique is imprecise
• B. the mathematical model of the coordinate transform might be too limiting

I can suggest this to solve A.: currently the model requires a 5-point calibration (top-left, top-right, bottom-left, bottom-right, center), and a single contact is made for each of these points with the pen. A single contact is very hard for a normal human being to do correctly because of hand tremor, and because we really can't anticipate the glass thickness before the contact. Hence we often fail at this (I'd expect a deviation as high as or even over 0.5mm-1mm depending on the person). I would suggest replacing the single contact by a short scribbling (dot shaped mini circles) motion on that point (generating 10s or 100s of coordinates per point) & then use a median filter on this set. This would take like 3 second per point for the user, but would give you a much more accurate estimate of the real center. [median is used instead of mean to better exclude outliers]

About B., I can't really tell without inside look. A better model might be needed (or not), or using the current one with 2 additional points (middle-left, middle right, total = 9 points). Anyway, I think A. should be tried first, to see if something is needed for 2.

Would be glad to help, I do have some knowledge in those fields. Just let me know. Keep up with the great work!

Cheers.

Raphael