I moved the second big filter between the the tcon board and the lvds controller, but I also rotated it 90º from the previous position, which may have broken whatever magnetic entanglement was going on in the original position.
The screen has CCFL lights, and - as could be expected for interfering screen - jitter is worse near them, becoming quickly unusable in the last 10 mm nearing the edge.
Oddly enough, it is not as bad near the lower lamp as it is near the upper lamp, where I suppose the tcon board is not fully shielded.
The main effect seems to be a "distortion" on the y axis reading of the position of the pen (where the general Jitter is, instead, mainly an x axis issue: looking at the driver test page, aside the worrying tool serial number 0x00000001, both x axis tilt and position jitter is more than twice that on the y axis).
The pen appears to "leave" the cursor behind, but I am not able to tell whether it is the lamp, or the metallic case still present distorting the electromagnetic field in that zone, the source of the issue.
Combined with the tablet not reaching the left and right borders, I'd say that it is better to keep the drawing area inside of 2 cm from each border, which suits me well.
As a long time paper dog, the border area is sacred for me (it's used to hold the paper on watercolours, or editorial annotations on comics...) and I never go there on paper, even when drawing sketches.
Also, the unusable space is mainly occupied by the Windows status bar (on the right), the Photoshop status bar (bottom), the Photoshop toolbox (on the left - I don't really need to reach it, as I use hotkeys for tool selection, but it is a bit difficult to get what is your current tool without it) and other seldom used PS options (top).
________________________________________________________________________________________________
On a side note, I moved the UBiQ where I used to keep the CabinetiQ (Simply, I can't concentrate much on a sofá).
So, I finally found why C..Q was so randomly, with his jitter.
The earth dispersal is less than ideal, and noise just goes around from one appliance to the other (no idea how much would cost fix it... more than change the screen to a less jittery one, for sure).
In the new place, some random clicks are back and jitter is a bit worse, and gets a lot worse when the other computers - and monitors - in the room are active. Or if the hard drives work hard.
Also, there is some more jitter near the vertical edges, than what was there in my flat. I suppose it is a matter of the ground not dispersing the noise generated by the ccfl in the metal casing.
Finally, I'd say that the main problem is still not what I call "random" jitter (which could be mitigated a bit more, maybe) , but rather what I call "geographical" jitter.
This is shot of a line made with a ruler, going back and forth five times.

Senza titolo-14.jpg (185.32 KB. 1680x1050 - viewed 503 times.)
The random jitter is the reason why the line is not just 3 pixels wide, but the overall wavy shape is a proof of the presence of a "fixed" decorrelation between the movement on x and y axis, that depends on the position/movement of the pen (by the way, also my beloved Huion H610 has the same issue, and that without placing a screen over it).
_________________________________________________________________
On a side note, I have an half Idea of building a couple of capacitive "wheels" and some touch buttons, to place under the glass.
I bought an Arduino clone that should arrive next week, and it seems simple enough. Maybe.
I've seen a library that would do the trick, using one digital out as " source" and measuring the time that the various pins need to react, and I think two wheels would "eat" up 7 digitals I.O. , leaving 7 inputs for buttons.
Not so easy... the tricky part is designing the wheel plates, taking into account that the glass is 3mm thick, and then the software so that it ignore false touches.
On the other hand, the rotary encoders in my current, repurposed mouse wheels have a 15º step (24 positions in a complete turn), and I am not sure that I can manage to be able to do the same (the issue, mainly, being the glass thickness getting in the way of the single sensor spatial sensibility... the examples I have seen around, of pcb with drawn wheels, have something like 5 sectors).