Help with calibration with waxbee

[bumhee34]

Found little bug in waxbee template.

If you look at the template PEN to Graphire3, the slave protocol is set to wacom adb.

If I modify to ISDV4 protocol, it works. But still jitter like inking..

[bernard]

Thanks for pointing out the bug about the Penenabled-Graphire3 template, indeed using the ADB protocol is totally wrong!

For your "jitter", we can investigate to see what is the root cause of the issue and if there is something that can be done about it.

Emulating a USB penenabled would be doable, but it is not implemented. Takes time to do these things and beyond having a 1:1 match with the original hardware there isn't much concrete benefit to it other than benefiting from a screen calibration utility (which is a nice feature actually).

Like I explained earlier, the next step is to analyze the "jitter" closely.  For that, map a tiny area of your board to a huge area on the screen (like the whole screen).  Close enough to see the resolution of the tablet -- if you draw on the screen with just 1 pixel dots (no lines), you would be supposed to see a spaced "imaginary grid" of dots. Like one dot per grid intersection. Then once we have that setup, we clear the canvas and draw a single diagonal line following a ruler:  The dots that are picked should follow a "regular stairs" pattern. There shouldn't be dots appearing outside of that stair pattern, if there are then it is jitter if not, then you are seeing the resolution of the tablet.  There is also a possibility that the scaling logic introduces issues with "aliasing". (before getting on the screen, there are 2 scaling that are happening:  coordinates of the original tablet scaled to the emulated USB tablet, then it is re-scaled again to the screen coordinates.

One way to reduce the chance of aliasing is to use an "emulated" tablet with the highest resolution. Here I am thinking the Intuos2 12x18 -- there is well enough precision in there o avoid any effect from the double re-scaling.

[bumhee34]

Cause I am a newbie, I have one stupid question.

"aliasing" algorithms are different from OS? I am using Win 8 64bit now. Not sure whether it is related.

And I will try your suggestion later.

[bernard]

I cannot answer your question because I do not understand it.  I think I'll just say: Forget about the word "aliasing" -- I wasn't sure how to describe the type of "number rounding-errors" that can arise when performing 2 "scaling" operations in a row. (under certain conditions).

[bumhee34]

Using tablet setting, I can remap small portion of sensor board to the whole screen.

In this way, I do not need to modify waxbee template, am i right?

So, now I need to test with the small area of sensor board.

[bumhee34]

And this is the capture from the drawing with a ruler.

If I understood wrong again, I would feel shy, haha.

Anyway, is this correct way to check?

Anyway, I can see jitter-like drawing cleary.

[bernard]

I think you got the point, just that we need to draw dots -- not lines -- I mean -- if you draw with your mouse and move very fast you should see many pixels that are not touching each other. Because of the lines we cannot analyse the data. 

Whatever software you use to draw the dots does not matter.

[bernard]

Ok, I tried to find a free software that would do this but no luck. They all "connect the dots"!

So I wrote one in javascript. ->  http://jsfiddle.net/SAEfU/   

Hit "Run" and draw in the black rectangle on the right. (tried in Google Chrome)

[bumhee34]

Here, I tried to draw rectangle.

[bernard]

I mean, with the special mapping described above?

[bumhee34]

Yes, I track the cursor and try to draw rectangle cause the sensor board area is remapped into small area.

Is not it what I should do?

[bernard]

yeah, but it seems it is not "powerful" enough to see some space between the pixels.  Try to map a much smaller area on your tablet (like 1 cm or even less) -- put tape or something to remember where it is physically on the tablet.  What you can do is to modify the "counts" (section number 3. in the device driver window) instead of playing with the little square (for which there is not much precision).   Make the count difference in each axis to be around 500. For example,  lefT:10000 and right:10500  and top:10000 bottom:10500.   Hopefully 500 is significantly less than your screen resolution. (like at least half less).

[bumhee34]

What you can do is to modify the "counts" (section number 3. in the device driver window) instead of playing with the little square (for which there is not much precision).   Make the count difference in each axis to be around 500. For example,  lefT:10000 and right:10500  and top:10000 bottom:10500.   Hopefully 500 is significantly less than your screen resolution. (like at least half less).

I made around 2.5cm square because it is minimum requirement in the tablet setting.
Anyway, I do not get it. Modify "counts"? I don't know what device driver window you mean.
Sorry for my poor understanding.

[bernard]

Counts is the english label of the item 3 in the dialog woth the 4 numbers - I mean your screenshot in this thread. I cannot reproduce the labels because it uses an alphabet I cannot know how to reproduce here.

[bumhee34]

Ok, but as the setting says I cannot make it 500 cause it is too small.

Anyway, I set like as follows, as small as possible i can achieve.