test1
October 17, 2017, 06:03:33 PM *
Welcome, Guest. Please login or register.

Login with username, password and session length
News: You can attach photos hosted by the forum rather than using an external image hosting site, this means they will stay forever and not disappear after a year or two.
 
   Home   Help Search Login Register  
Pages: 1 2 [3] 4
  Print  
Author Topic: Converting Wacom GD-0912-R from Serial to USB  (Read 32068 times)
hermanvereycken
New Poster
*
Posts: 9


View Profile
« Reply #30 on: October 26, 2013, 01:51:41 PM »

Hi bernard!

i have made two hid_listen dumps, one for both problems.

1) unwanted pen pressure (not sure if it's 100%). in this test, i just hovered the pen over the tablet without touching it for the entire duration of the test

* pressure-point-log.txt (330.47 KB - downloaded 155 times.)


2) pen slips(offsetted) on the Y axis. in this test i was drawing 8's horizontally, i'm not sure if i was able to catch such an event

* y-axis-log.txt (496.52 KB - downloaded 147 times.)


hope it will provide some clues!

Herman

* pressure-point-log.txt (330.47 KB - downloaded 155 times.)
* y-axis-log.txt (496.52 KB - downloaded 147 times.)
Logged
bernard
Administrator
Hero Member
*****
Posts: 2584


pato mania


View Profile
« Reply #31 on: October 26, 2013, 04:23:37 PM »

The traces sometimes contains (?!) it seems some bytes are lost -- that might be due to the fact that the debug outputs too much and the buffers gets overloaded.  I see nothing else that sticks out. I could write a program to analyse this further, but I do not have that much time. We'll use the tools we already have, just need more time on your side.

The idea is to pinpoint something "wrong" in the data and discover a "pattern".

Let's try something else:  Put back the "normal" template (non-debug), then using my Wacom hid tracing tool, reproduce the issue and send me that.  The "best" is to be able to spot that precise area where it fails.  What you could do is to try to find that interactively:  you look at the data as you move your pen and when you the issue is occuring with your pen you attempt to look if you see the issue in the data (jump in x,y or pressure values). Then you find a way to send me your findings.

On another front, I might be able to modify the debug to specifically output the information we are seeking (can't output too much else buffers gets overloaded and data starts to be dropped (loosed).

The idea here is to find what exactly is going wrong and either adjust the parameters or add some special "filtering" to workaround the issue. I am sure it can be done.
Logged
hermanvereycken
New Poster
*
Posts: 9


View Profile
« Reply #32 on: October 27, 2013, 10:25:20 PM »

Hi bernard,

i'd be glad to help you out there. i have been able to trace the pressure point, below you will find some examples


* pressure-point-graph.png (12.44 KB. 655x465 - viewed 306 times.)


* pressure-point-log.png (23.83 KB. 935x167 - viewed 292 times.)

* pressure-point-log-2.txt (193.66 KB - downloaded 156 times.)


hope this gives you some more indications

* pressure-point-log-2.txt (193.66 KB - downloaded 156 times.)
Logged
hermanvereycken
New Poster
*
Posts: 9


View Profile
« Reply #33 on: October 27, 2013, 10:52:16 PM »

I've also been able to pinpoint the Y-axis glitch after some testing:


* y-axis-graph.png (32.53 KB. 483x894 - viewed 303 times.)


*first graph is a scatter plot x-to-y axis
**it's strange that we see on the x-axis a lot more glitches, however, these seem less noticeable in reality


* y-axis-log.png (43.23 KB. 1035x242 - viewed 298 times.)


* y-axis-log-2.txt (330.53 KB - downloaded 202 times.)


hope you see some clues in those extracts  Huh

* y-axis-log-2.txt (330.53 KB - downloaded 202 times.)
Logged
bernard
Administrator
Hero Member
*****
Posts: 2584


pato mania


View Profile
« Reply #34 on: October 28, 2013, 02:53:12 PM »

Good analysis!  It does not tell us the root cause, but it does helps a lot for pinpointing the data.  One thing I was looking if there was other data that was different, weird or out of whack in those cases.

Can you make more live tests (no need to send me all the data, rather the results) and:

#1- When there is a "glitch" is it only for a "single" packet at a time? (for Y and Pressure glitches)
#2- Y: When there is a "glitch" what is the minimum y-delta?  What algorithm would work to safely ignore those while not "breaking" good cases? It is easy to remember what's in the past (although much more difficult to look into data that comes after).
#3- Pressure: same question as #2 but for the pressure?
Logged
Aerendraca
Administrator
Hero Member
*****
Posts: 1068


View Profile
« Reply #35 on: October 28, 2013, 09:45:23 PM »

I notice that you say that the x-axis has a lot more glitches than the y-axis, but that it is not as noticeable on the x-axis. In reality I believe that the y-axis is actually showing more glitches than the x-axis as the scatter chart offset is not the same on both axis. If you made it so that the scale on x was the same as the scale on y then you would get a better presentation of the scatter distribution, and I think you would see that the y-axis is significantly more glitchy than the x-axis, explaining why you don't notice it on the x. Lots of 'x' and 'y'x there but I hope you get the gist.

Actually on reflection it might not be as significant as I first thought, but still worth doing anyway.
 
« Last Edit: October 28, 2013, 09:58:34 PM by Aerendraca » Logged
bernard
Administrator
Hero Member
*****
Posts: 2584


pato mania


View Profile
« Reply #36 on: October 28, 2013, 10:59:14 PM »

One thing to check is if the glitch has to do with a specific "bit" being "flipped wrong". The only way to tell is to pinpoint it in the original serial port data.

Hopefully on the next round, we have enough information to help in pinpointing the problem in the serial data (i.e. when in debug mode).  We'll see.
Logged
hermanvereycken
New Poster
*
Posts: 9


View Profile
« Reply #37 on: October 29, 2013, 07:36:22 PM »

Hi Guys, can you just elaborate a bit on the bit flip thingie, are there some specific things that i can take into account and look for while doing the analysis? (besides the things mentioned by bernard)

cheers!
Logged
bernard
Administrator
Hero Member
*****
Posts: 2584


pato mania


View Profile
« Reply #38 on: October 29, 2013, 07:57:03 PM »

Hum, difficult to tell what to look for without writing a long post.  I'll try to summarize. The serial data contains values that is encoded as "bits" (binary), each bit is send one at a time. In Intuos boards, the pen data pressure is sent as binary over the air as well.  In binary, numbers are represented as series of 0 and 1. An 8 bit number range from 0 to 255.  If, for example you flip one bit, you can get a 1 where there shouldn't be (like 10000000 )  and this would yield the hexadecimal 80 or decimal 128.  So if the sudden "delta" difference is close to 128 then maybe that 8th bit was flipped to a 1.  The numbers to look for are powers of two:  32, 64, 128, 256, 512, 1024, etc.  Typically serial communication include advanced techniques to detect and even "correct" errors (bit flipping), so even if it occurs, it does not affect anything. In the serial data that Wacom sends, there is no such error detection or correction mechanism.  If the line is really "noisy", then it could yield bad data like that.  One reason for a line being noisy is that the grounds are not properly connected or the wires are very long and pickup various "radio" frequencies.  Only a tool like an oscilloscope can be used to see this condition normally.
Logged
SONB
New Poster
*
Posts: 17


View Profile
« Reply #39 on: October 30, 2013, 06:11:32 PM »

Hi guys! Yesterday I finished my first mod. This was my first experience with soldering as well, but I guess I did pretty well  Grin
I had to switch the tablet to 19200 to make it work. Now I'm in the same boat with hermanvereycken: Having random clicks and vertical jumps. I noticed an issue with the layout as well. It's like the tablet area is mapped with an offset. I mean when I touch the bottom edge with the pen, the cursor can't reach the Windows taskbar. When the cursor touches the top edge of the screen, there is still place left at the top of the tablet.
Logged
bernard
Administrator
Hero Member
*****
Posts: 2584


pato mania


View Profile
« Reply #40 on: October 30, 2013, 07:16:08 PM »

Oh okay, so two people having the same issue?  Hum, that actually makes me think it might be easier to fix. Maybe it is just the way the data is interpreted, not that something is broken inside a particular unit or not well connected.

About the Offset: Are you sure that the intuos2 is correctly mapped to the whole screen in the configuration? 

There *is* actually a permanent vertical offset because of the way the "top button strip" is handled. But normally the provided templates were supposed to "hide" this.  This is controlled in WaxBee configuration's Y coordinate mapping. Are you getting a "menu" on the screen when you move your pen at the top-most area of the tablet?  Or is there some other weird configuration of the Intuos2 that could have an effect on this (look for disabling/enabling the menu strip maybe?).
Logged
SONB
New Poster
*
Posts: 17


View Profile
« Reply #41 on: October 30, 2013, 11:06:13 PM »

A small update:
Sorry, there is nothing wrong with the top area of the tablet. Those 1200 points in the template, reserved for the menu strip, perfectly match the offset on the tablet. The menu itself doesn't work. Tried copy, paste, cut etc. and switching between modes (pen, mouse), nothing worked. For example, I opened a text file with some text and every time I touched a button in the menu, the window just lost its focus. Like on a simple click on the desktop. And there is no visual feedback, no numbers appear at the top, nothing. I never used the menu, so it's not a big deal for me.

The problem is at the bottom area of the tablet. Like I said, I can't reach the taskbar with my pen. It stops close above its top edge.
Then I noticed that there is a template converting a 9x12 to 12x18 and flashed it. And it's mapping perfectly. Well, almost perfectly. There is a tiny offset on the X-axis. But now I can reach the taskbar. I guess I should flash the initial template and play with the Y-numbers.

I don't know, if it's important, but I'm using Windows 8.1 Pro (64bit) with the latest Wacom legacy driver for Windows 7. Might be the reason, why the menu strip doesn't work.
Logged
bernard
Administrator
Hero Member
*****
Posts: 2584


pato mania


View Profile
« Reply #42 on: October 31, 2013, 03:01:51 AM »

Make sure you do not have the "keep proportions" option ON or other special options like that in the Wacom Driver. 

Yes, you can always adjust the "mapping" by playing with the numbers -- either within the Wacom Driver settings itself (the "counts") or within the WaxBee configuration before flashing it.
Logged
SONB
New Poster
*
Posts: 17


View Profile
« Reply #43 on: November 09, 2013, 03:40:39 AM »

I thought I'd share photos of my mod. Managed to hide the tweensy inside the tablet. The reset button is the cap of a stabilo pen  Grin


* mod_01.jpg (399.07 KB. 1500x1125 - viewed 330 times.)



* mod_02.jpg (302.64 KB. 1500x1125 - viewed 330 times.)



* mod_03.jpg (267.96 KB. 1125x1500 - viewed 302 times.)



* mod_04.jpg (453.1 KB. 1500x1125 - viewed 337 times.)



* mod_05.jpg (282.07 KB. 1125x1500 - viewed 302 times.)



* mod_06.jpg (229.06 KB. 1125x1500 - viewed 304 times.)



* mod_07.jpg (187.69 KB. 1500x1125 - viewed 306 times.)


Logged
bernard
Administrator
Hero Member
*****
Posts: 2584


pato mania


View Profile
« Reply #44 on: November 09, 2013, 04:28:09 AM »

Wow! Really a perfect fit! The cable connector inside like this is really great!  Have you put some isolation between the teensy and the main board?  they seem to overlap and be quite close...

And this is one of the nicest reset button I have ever seen!
« Last Edit: November 09, 2013, 04:31:15 AM by bernard » Logged
Pages: 1 2 [3] 4
  Print  
 
Jump to:  


Powered by MySQL Powered by PHP Powered by SMF 1.1.21 | SMF © 2015, Simple Machines Valid XHTML 1.0! Valid CSS!