heyneman
11-15-2010, 04:43 AM
I'm starting this thread to discuss the possibility of implementing an improved calibration procedure, after it came up in the Xournal thread (http://www.mobileread.com/forums/showthread.php?t=87593)
I haven't been very happy with the calibration of my DR800 (with the standard pen), especially around the edges. To be fair I haven't had it very long, but coming from a Lenovo tablet which (I believe) uses the same era Wacom technology, I was somewhat disappointed with the performance.
First, is this something other people are unsatisfied with? Does using a nicer penabled pen help significantly?
Assuming that this is an improvement worth pursuing, I poked around in the gtktscal code tonight and I'm pretty sure I understand what they're trying to do: build a linear mapping from raw x/y to calibrated x/y. This first time through the math looks a little funny (there's an extra averaging step which I don't think should be there). But, hinky looking math aside, they try to come up with numbers (a-f) so that x = a + b*x_raw + c*y_raw, and y = d + e*x_raw + f*y_raw.
If anyone knows what source code handles general touch screen events, it should be an easy matter to change out the calibration math to something more effective, and then correspondingly change the calculation of the tap position in the general touch screen code.
Doing it this way would let all applications benefit from improved calculation. If for some reason we can't get into the general touch screen code (which doesn't seem like it should be a problem), it could be worked in on an application-by-application basis :(
I haven't been very happy with the calibration of my DR800 (with the standard pen), especially around the edges. To be fair I haven't had it very long, but coming from a Lenovo tablet which (I believe) uses the same era Wacom technology, I was somewhat disappointed with the performance.
First, is this something other people are unsatisfied with? Does using a nicer penabled pen help significantly?
Assuming that this is an improvement worth pursuing, I poked around in the gtktscal code tonight and I'm pretty sure I understand what they're trying to do: build a linear mapping from raw x/y to calibrated x/y. This first time through the math looks a little funny (there's an extra averaging step which I don't think should be there). But, hinky looking math aside, they try to come up with numbers (a-f) so that x = a + b*x_raw + c*y_raw, and y = d + e*x_raw + f*y_raw.
If anyone knows what source code handles general touch screen events, it should be an easy matter to change out the calibration math to something more effective, and then correspondingly change the calculation of the tap position in the general touch screen code.
Doing it this way would let all applications benefit from improved calculation. If for some reason we can't get into the general touch screen code (which doesn't seem like it should be a problem), it could be worked in on an application-by-application basis :(
