[Ipe-discuss] Ipe on Tablet-PCs

[Gerhard Mesenich]

Otfried Cheong wrote:

> Thanks for the input!  Making Ipe usable on tablets was one of the
> design aspects when I rewrote Ipe 7 - which explains the freehand mode
> and the hooks in the code for the eraser functions - but so far I
> didn't get any feedback on that.


Dear Otfried,

that's good news! The lack of feedback is certainly due to the low
spread of the tablets. Good hardware is available, but all software I
have seen so far is still seriously lacking essential functions for
professional use. In my opinion Ipe comes closest in regard of the
needed graphics functionality.

Choosing the right hardware is quite difficult. Therefore let me give
some technical feedback based on my personal experience in this rather
long posting:


Tablet-PCs for academic and engineering use:

For handwriting and freehand drawing a good Tablet-PC is essential for
the needed haptic feedback. With a mouse handwriting is simply
impossible; the result will always look terrible. A pure Tablet like the
Wacom-Bamboo is a lot better and somewhat usable, but is also no
comparison to a real pen on a screen, which allows paper-like input and
precise control.

Handwriting enormously eases the input of complicated documents, which
are often a mix of text, formulas and sketches. After some training
handwriting looks good enough to even be used on quick internal
presentations. It can also be used for first layouts, over which a clean
publication drawing can then be drawn as an overlay. Basic drawings can
be made out of standard shapes, then small details and annotations be
added freehand. After some training this usually leads to a clean
appearance with little effort and time.


Pens:

A useful Tablet definitely requires an electromagnetic pen. The often
found touch-screen only PCs are only useful for 'multimedia' browsing,
they don't add any usefulness for drawing and annotation. Touch with a
pen combined has a tendency to interfere in unwanted ways with the
pen-input. The touch-pen combination only seems to make sense if mapped
to separated areas of the screen. The cheaper resistive type digitizers
don't seem to work very well and can not be recommended.

All good pens are basically of the 'Wacom'-type. These have an internal
chip, which draws its energy out of the electromagnetic field and
apparently generates a modulated and interpolated radio signal in a
ping-pong mode. The field is generated by a relatively coarse regular
wire-grid below the screen (spacing estimated with a search coil to be
1-2cm), which is consecutively exited horizontally and vertically by
short bursts; the signal between adjacent lines is then interpolated.

The center frequency of the bursts is either approx. 450 or 600 kHz
(measured it once, approx. values out of memory). Wacom pens use one of
these two frequencies for Tablet-PCs, the other for Tablets. Pens for
Tablet-PCs work on most (checked with HP2710p and Motion-Computing
LE1600, same pen working on both), but Pens for Graphic-Tablets and
Tablet-PCs are *not* interchangeable because of the different
frequencies alone (checked with a Wacom-Bamboo).


Touch screen:

Some of the newest models have a capacitive touch screen in addition to
the pen. My own Tablet-PCs don't have that and my experience with the
pen and touch combination on a Bamboo Tablet has not been convincing.
The intention there is to disable the touch whenever the pen tip is near
to the tablet. I finally disabled the touch because of a shaky switching
operation and the resulting frequent interference between the two functions.

On a Tablet-PC however it may be useful to map the Touch to a small
stripe on the left side only. This Touch-Strip would then serve as
modifier Toolbar for the left hand, whereas the rest of the screen would
the be touch-disabled. Interference would then be eliminated and most
haptic problems be solved. The touch strip would best be context sensitive.


Buttons:

Unfortunately the much needed modifier buttons are either missing or in
the wrong places.

All standard pens for Tablet-PCs have only one side-button, whereas pens
for Graphic tablets usually have two. This is a much needed enhancement,
the second button can be mapped as middle mouse button or some modifier
key. Wacom also sells replacements for Tablet-PCs with two side-buttons,
but I could not find any of the same size than the original. In this
case they don't fit into the 'pen-silo' and can easily get lost.
Replacement pens aren't cheap, usually they are sold for about 50 US$.

The top button of the pen intended as eraser could probably best be
remapped as a pan and zoom button (one-click and two-click modifier,
gesture), which for Ipe seems the best use to me.

Another very useful enhancement for a right-handed person would be some
modifier keys at the left side of the screen. No Tablet-PC seems to have
that. Some few have some predefined keys in difficult to reach places.
In this case they mostly need double key presses (Fn + another) to reach
self defined functions. That is far from intuitive. With some effort
reprogramming can certainly be done (not directly supported), but then
other useful functions like screen rotation will have to be remapped
otherwise (i.e Motion Computing LE1600 slate). The LE1600 must also be
turned upside down, to have the keys on the left side. This works well
on a LE1600 and does not create any additional problems.

On the contrary the HP2710p does not have any useful self definable
buttons at all and can not be turned in the natural direction where the
connectors (DC,Lan,VGA) point to the rear with the originally supplied
screen driver. HP claims 'EMC-reasons/regulations' (WLAN-Antenna harmful
emissions?). In this setup all mayor connectors are pointing towards
your stomach and it looks exactly like the mess it sounds. Without any
recognizable problems this can be fixed however by installing a genuine
Intel screen driver against HPs intentions after clicking away a lot of
warnings. I have never seen anything that crazy before from an otherwise
respectable manufacturer.


Keyboards:

Another good solution for the input problem is a real small palm size
keyboard put on the left side of the Tablet-PC, which I currently use.
Normal size keyboards only get in the way during drawing, but they are
best for normal typing when the Tablet is used without the pen. For
universal use both are needed. They can both stay connected and be used
in parallel.

A problem here arises only, if different language layouts have to be
combined, which is the case with my mini-keyboard. It has an
international layout (not available in that size in German yet), whereas
the main one is German. WinXP then always needs a language switch first
(mapped to alt-ctrl) before using one or the other in order to get the
proper layout. Different language keyboards are not recognized by winXP
at the same time. This is quite a nuisance, but I don't switch the
keyboards very often.

At the desk I often use a dual-screen setup with the Tablet-PC and the
mini keyboard sitting on the left and a normal USB-Keyboard with a
1600x1200 Monitor on the right side. This setup greatly enhances the
usability and gives the best of both worlds during office use.


On-Screen-Keyboards (OSKs):

The Windows OSK works very well and even has character recognition which
after some training works very well even with normal handwriting. It
generates output just as a normal keyboard.

Linux does not have anything comparable yet. It has however several
OSKs, where you just select the keys with the pen key by key. For short
sequences like small annotations or filenames this works very well too,
which is sufficient for Ipe. In both cases the OSK of the operating
system should be hooked context sensitive whenever alphanumeric Input is
required.


Data rate:

Wacom Type Pens usually seem to be connected internally as RS232 devices
with a physical USB interface. The interface usually runs with 19200
baud, which results in a maximum data rate of about 50-100 points/s (2
coordinates/point). This is quite slow; on fast movements some lag can
be seen, but this does not seem to cause recognizable problems with
normal writing speed.

Taking Xournal as a reference for a good and proven algorithm, a very
densely handwritten A4 page with about 40-50 lines requires typically
200-400 kB there (normally written about 100kB), which can serve as a
rough estimate for the data load to be expected under optimized conditions.

Xournal stores its data in binary form (.xoj file). It runs well even on
slow machines (1Ghz single core LE1600). Xournals handwriting response
is always immediate even on multi page documents up to about 10 pages,
which are the longest ones I have.

On Ipe the data load is about 10 times higher under similar conditions
(tested with a few handwritten lines). This is caused by the the data
format: Ipe stores the pen coordinates as text file with very high
precision. Each coordinate takes 18 characters (including Linefeed),
where only 4 would be necessary in binary form with 16 bit precision.
This could be further reduced, if some kind of compressed differential
format would be used (but this may not be smart because of the
additional computational overhead). For data reduction a binary storage
with 14-16 bit precision of pen strokes should probably be considered.

With the current text format a maximum of about 2-4 MB of data for a
densely handwritten page can be expected for Ipe. Handwriting is clearly
possible with Ipe, but the storage algorithm should probably be improved.


Calibration:

Due to the interpolating nature of the pen calibration is always
required. The calibration procedure is part of the pen driver and is
done by touching 4 spots about 1 cm diagonally away from the edges of
the screen. This leads to a typical accuracy of about +- 1 mm for the
tip of the pen inside the rectangle defined by these calibration spots.
This is good enough once you get used to looking at the cursor spot
instead of looking at the tip.

Close to the edges of the screen calibration often gets much worse (up
to +-3 mm). The active field of the digitizer also has very little
overlap to the screen edges, so that sometimes it is not possible to
reach the Autohide-Taskbars anymore. Due to the nonlinear nature of this
fault sometimes the Autohide Toolbar can only be reached in certain
places, not over the full length of the edge. On my HP2710p for example
the 'sweet spot' is the lower left corner under normal calibration,
touching the lower edge in any spot further away than about 3 cm to the
right does not bring up the toolbar anymore.

Deliberate miscalibration to eliminate the dead zone can overcome this
problem, but this throws off the calibration accordingly for the main
part of the screen. I have seen this insufficient pen range on several
Tablet-PCs. The pen-drivers still leave a lot to be desired.

It should be considered to have a separate offset calibration table with
a denser calibration grid, especially close to the screen corners, to
get better precision out of the pen in Ipe until better drivers are
available.


Pen-Cursor offset bug:

There is also a cursor offset bug showing with Tablet-PC-Pens only; not
with Graphic tablets or with the mouse. With Ipe the bug shows as a
stable offset between the cursor and the strokes (Lines, all shapes) of
about 5 pixels to the right and about 3 pixels upwards, only during Pen
use. A similar bug seems to be in some other programs too (i.e Foxit).

This nuisance is probably caused by a different definition of the cursor
reference point by the Wacom driver or some unaccounted calibration
correction. Bug or Wacom-feature?, culprit unknown. Whatever the cause
may be, this hopefully can be fixed easily just by correcting this
offset in Ipe.


Background-grid and Screen Orientation:

The best orientation for most work seems to be Landscape mode for the
screen whereas the 'paper' is in portrait mode. With Xournal I usually
use a 2mm graded A4 Paper displayed in full width as background, so that
approx. 1/3rd of the page is shown on the screen. Such paper can be made
with the free 'Graph Paper Printer':

http://www.brothersoft.com/graph-paper-printer-download-298633.html

Something similar should be implemented in Ipe with proper margins and
mayor and minor graduations with a precise snap (probably already there,
haven't looked seriously yet). Such setup seems to be most practical to
me and leaves enough space for Toolbars on top and side. It also makes a
pleasant appearance.

Suggestion for a good standard A4 background 'Paper' with margins:

grid-size 19x28cm, centered on 21x29.7cm A4-size
2mm minor, 10mm mayor gridlines, thin light gray, grid changeable to
other tastes

This layout has proper margins, makes good use of A4-paper and
grid-boxes are regular and are not 'cut-off' or hanging in the air.
Something similar should be set up for other common sizes too. Important
is the precise binding of the 'paper' grid to the program grid, so that
the final drawing does not show unpleasant offsets. The best 1:1 zoom
factor for a 1280x800 standard Tablet-PC would be exactly 5 or 6 pix/mm;
this gives a very nice fit for the grid and eliminates all grid
artifacts due to interpolations.


Viewing observations:

The standard screen on all mainstream Tablet-PCs is now 12" - 1280x800,
which is sufficient. The best working position of the screen is lying
slightly inclined on the table in landscape mode (my personal preference
is 10-15 deg., you have to put something under the back of the tablet,
which also improves venting a lot). In such position the eye is closer
to the screen than with a typical monitor setup. My personal viewing
distance is approx. 50-60cm to the desktop monitor, but only 30cm to the
Tablet. For the same viewing resolution angle the tablet screen
therefore only needs half the pixel pitch than a desktop screen.
Therefore 12" screens work well and could even use a better resolution
than available now in that size. With the pen even small icons can
easily be reached. Ipes icons are just fine, they could even be a little
smaller for my taste. Personally I also prefer old style separators
between each icon, to get additional guidance. Oversize icons are only
needed for touch with a finger, if a separate Touch-Strip at the side
gets implemented.


The typical power consumption of a 12" display with LED backlight is 4W
typically. Total power consumption of my HP2710p is about 8W under light
load (1.8Ghz dual core, normally throttled to 900MHz by power
management), the older LE1600 (1Ghz single core, CFL-Backlight) takes
about 10W under same conditions. Ubuntu Linux and winXP have about the
same power consumption in a proper setup. It pays to make a clean
installation and throw out the junk, this can greatly enhance battery
life. Windows Vista for instance is said to be a notorious battery eater
(I did not test Vista longer than about 15min, it was the first thing
thrown out). On all tested setups Ipe runs easily and usually puts only
light load on the machine.

Screen brightness is always sufficient indoors and always insufficient
outdoors in bright sunlight, even with the most expensive displays. The
reason is simply technical: A 12" screen has an area of approx. 430
cm**2, direct sunlight has a power of approx. 43W on that area and not
much less even with a cloudy sky. To be able to see the picture on a
passive screen, the light from the back must be stronger than the light
from the front, which is almost impossible to achieve with the little
power the backlighting can possibly have; even if direct sunlight is
shaded away and taking the power reduction be the incoming angle into
account. The best way to cope with viewing deficiencies is finding a
good working place. That way even inexpensive screens give good results.


Choices:

Good selection guide:

http://forum.tabletpcreview.com/what-tablet-pc-should-i-buy/30006-exhaustive-tablet-pc-faq-all-you-ever-wanted-know-about-tablet-pcs.html

Currently the best deal seems to be the HP-Touchsmart TM2, which has
everything needed. Comparing this to my much more expensive HP2710p
'Business Tablet' it has the following advantages: Much better battery
life up to 9 h under light load, a standard 2.5" SATA HD, which can be
replaced easily with a SSD, which are a lot less expensive in 2.5" than
1.8" that most "professional" Tablet-PCs have. The TM2 also has
capacitive Touch in addition to the Pen. Computing power is comparable,
the only disadvantage seems to be a 'mirror'-type screen which also has
a smaller viewing angle. The glare can be overcome by choosing a good
working place and with an anti-glare screen protector.

Right now the HP2730p can be found on sale for about 800 Eur, which is
about half the common price a few weeks ago. These have been replaced
recently by an almost identical 2740p, which has now touch in addition
to the pen.

http://geizhals.at/deutschland/a451798.html

There are also some Motion Computing LE1600 for less than 300 Eur in
almost as new condition on Ebay here in Germany, which are very
professional and sturdy build with an excellent docking station. It has
only a 1 GHz single core and only a 30 GB 1.8" IDE HD (quite expensive
to replace), screen Resolution is only 1024x768; but with a clean and
lean setup these run very well and are excellent machines for Ipe.

http://cgi.ebay.de/MOTION-TABLET-PC-LE1600-WLAN-BLUETOOTH-FLEXDOCK-XP-/270625323710?pt=DE_Technik_Computer_Peripherieger%C3%A4te_Notebooks


It pays to shop around, all the above are very good choices.


> One problem is that I do not have a functioning tablet right now, and
> the tablet-specific Qt-functions do nothing without a tablet. But
> I'll try to look at it next time I work on Ipe (which is probably not
> soon, as the semester is just starting).


Summary:

There is no way to do Tablet-PC developement without one. Get a
Tablet-PC asap, you will not regret it :-).


Thanks for Ipe,
Best regards from Germany,
Gerhard