[sane-devel] Epson gt-8000 transparency experiment

[Bart Buitinga]

I found a user level description of the Epson GT80FLU Transparency unit
(see http://www./xs4all.nl/~labrat/downloads/es600cua.pdf ) for some of
the GT series and it declares a lot of the difficulties:
1. This TPU is not just a hood with a lamp; it contains a carriage with
either a lamp or a sensor (and estimated cost of the tubes Epson put in
their earlier color scanners, I think a sensor) that is supposed to move
exactly synchronised with the scanners carriage. (So I'm not even sure
wether TPU mode, if faked, would actually "scan with lights off")
2. The TPU uses the narrow glass pane also used by the ADF unit for
calibrating the sensor. If calibration fails an error code blocks
further operation. Yet I've seen the scanner startup without this
calibration, but so far I can't reproduce it (my notes list earthed
7+8+4+15).

This, plus the latest paperclip results, (I found earthing pins 7+8
causes ADF detection, and a whole lot of ways to crash TWAIN32) pushes
me to give up on TPU mode.

I've also been experimenting with the parallel projection bundle, or
better, a slightly convergent bundle, but as said before: It takes a
very large lens. Perhaps one from an overhead projector would work.
Results for now have a strong sferic bias due to the enlargers condensor
lens (F 200 mm, 150 mm diameter) and suffer reflections of the scanner
tubes. Using a 58 mm wide 4,5/F240 projector lens just standing on the
scan surface, I got what you'll find at
http://www.xs4all.nl/~labrat/downloads/wlens.tif

But low tech solutions may become more feasible in the future anyway, so
I took some black paper and cut a mask for the carriage. Problem:
Calibration fails, and despite of plugging three paperclips on countless
combinations it just won't skip it.A white sheet underneath the black
mask, locked in in place with the scanners top (this one can easily be
opened) so that it is pulled away since it sticks when the carriage
moves) reenables the calibration, but requires to be put in place for
every single scan.
Using a thin white plastic bag for screen, a slide looks like what
you'll find at http://www.xs4all.nl/~labrat/downloads/wscreen.tif
A better mask could prevent the striping, which occurs because curls in
the paper edge influence the calibration in primary direction, and also
the leaking of tube light, which causes the flawed colors.

I think I'll just keep the GT-8000 for the countless things it's good
at, and look for a more primitive one for this project.

Thank you all for your attention and good luck with your project,

Bart Buitinga