...

Note: I found no documention regarding how a lightpen for Vectrex
      exactly works. I found a dissassembled ArtMaster Rom
      file at a certain ftp site (look at links...).
      Thanks for that again Fred!
      I tried some reverse engeniering,... and it actually works now
      for the three lightpen using images I found so far:

      Art Master
      Melody Master
      Animaction

      I don't know if I got it completly right. If something does not
      work... you may try contacting me...

How I think the vectrex lightpen works:
(forgive me if something in here doesn't sound very professional,
 cause I really am a technical illiterate, I just describe what I found
 out looking at the Art Master Rom...
 and ... I never programmed a 6809 processor)

Artmaster uses sort of two different ways to locate the lightpen:
a. check known positions if something is there
b. search the hole screen for a lightpen

a. To put it simple: The last lightpen position is known, Vectrex thinks
   the lightpen must still be somewhere near, so it draws something like
   a spider net arround the last position.
                                   /-----\
                /---\             /       \
    /-\        /     \            I       I
    I*I        I  *  I            I   *   I         ...
    \-/        \     /            I       I
                \---/             \       /
                                   \-----/

   If it is found somewhere on the WEB Vectrex is happy :-)!
   If not it stops searching and is unhappy :-(!
   You will see that very often using Artmasters Sketch or Connect
   Menu (everytime you release the button!).

   Emulating this was fairly simple, since the only thing to be added
   was mouse position information.
   The lightpen will generate an interrupt on the PIA line CA1.
   So I only changed the analog section of the emulator, if
   the position of the vector ray was near the mouse position
   then...
   t6522IFLG|=E6522IFLG_CA1;
   change the PIA interrupt register so that it thinks an interrupt had
   occured. Actually I didn't like to put it in the Analog section,
   since it is a PIA interrupt, and all other PIA interrupts are triggered
   in the PIA section of the code... but than again it is triggered sort
   of by the vector beam too...

b. The method used in animation modus is somewhere different, for
   you can put 'new' lines anywhere on the screen, so 'webbing' would
   be sort of unefficent.

   ArtMaster scans the whole screen for the lightpen position.
   It draws lines from bottom to top (search_screen_for_scanlines (0721)).
   (0x7a lines by the way (0721)) These lines are drawn every
   0x0200 (0766) vectrex coordinates (y-axis that is).
   It starts at position 14944 (decimal)(0726) and goes up (screen upwards)
   to -17888.

   A line is drawn all the way, after it is finnished there is a check if
   a lightpen pick occured, if so then further testing regarding the
   horizontal position is done (find_point_of_intersection (076d)).
   If it doesn't find a pick at all it leaves the routine
   (being unhappy again...).

   The horizontal position is somewhat trickier to get. ArtMaster
   calculates the position. It redraws the scanline (0770...) where
   it found the lightpen pick, starts a counter (register b=0x7f (0798))
   draws the line and while decreasing the counter waits for an
   interrupt to occur. (lightpen interrupt on PIA CA1, 0776...)
   If an interrupt occurs, the position is calculated and stored
   (process_ISR (07B7)).
   Finished.

   So I had to implement a new interrupt, the vectrex emulator
   did not react on CA1 interrupts (this was done in five minutes, since
   it was virtually nothing!).
   After me having optimized the processor to multi-ticks (that's what
   I call it) calculation of the horizontal position was somewhat
   difficult. I actually had to reimplement the old single-tick cpu
   again to calculate the position correctly, furthermore an analog
   tick HAD to occur every cycle, otherwise the emulator could miss
   the position it was looking for.
   The beam position had to be integrated every tick and could not
   be calculated per ticks since last integration.
   All this makes lightpen emulation sort of inefficient :-(.
   BUT I only use this sort of testing and integrating when BOTH
   mousebutton is pressed and an CIA interrupt is enabled... so
   most of the time emulation goes smooth still.

   While we are at it... I cheated somewhat here!
   ArtMaster tries to disable interrupts with following code
   (disable_interrupts (07aa)):

   clra
   sta   0x0e

   Which doesn't really clear anything, since the interrupt vector
   to be cleared must be specified, it should be something like:

   lda   #0x02
   sta   0x0e

   This incorrectness caused a slowdown once the interrupt was enabled
   the first time and the mouse button was pressed...
   The workarround to that was, that EVERYTIME an interrupt is cleared
   bit two is set!
   (emu6522.c function: UBYTE f6522Addr0EWrite(UWORD tAddress,UBYTE tData))
   This shouldn't really concern any other vectrex rom, since that
   interrupt is only used by the lightpen (I think).
   But I really hated doing that, since emulation
   perfection is somewhat broken...
   The other way would have been to change the ROM, but that would
   have been a none trivial task, since there is not so much
   space for new code in there...

   The width for a mouse position to be recognized is 0x100 vectrex
   coordinates, I derived that value of the ArtMaster
   search_screen_for_lightpen routine, which scans in intervalls of
   0x200. Sometimes with the emulator lightpens are not recognized,
   this relates directly to the above mentioned 0x100. It could
   probably be a bit wider range, but a miss happens fairly seldom,
   so why bother, I guess it happened with a real lightpen too.
   To take a wider range than 0x100 would have made
   for example connect and scetch somewhat inaccurate, so I left it at
   0x100...

...