[sane-devel] Re: [ANN] Plustek Backend update V0.45-1

[Jaeger, Gerhard]

Hi Gene,

On Saturday, 11. January 2003 22:08, Gene Heskett wrote:
> On Saturday 11 January 2003 10:05, Jaeger, Gerhard wrote:
> Announcing a 45-1 release.
>
> I unpacked it over the old 45 in sane-backends-1.0.9, did a make
> clean && make && make install, which apparently went well.
>
> Fireing up xsane from the icon, I had it do a preview, which ran as
> far as the end of the forward travel of the carriage, at which time
> it cleared Olivers logo from the preview window, and the whole
> thing went away about 1 second later.  The carriage went back home
> as it should have, but the lamp was left on.

Quite a strange behaviour! I recommend to check the .conf file each time
you make an update.

>
> Running it from a shell, it did the same thing, and reported a
> "segfault" to the shell window as it exited.

I remember this behaviour a long time ago, but after a complete
source code review and cleanup of my environment there are no
problems left here!

>
> The epson iscan-1.40 derived backend still works ok though.

It should do so;-)

>
> I'm going to do another make clean and get rid of the config.cache
> etc stuffs and try it again, as I've re-arranged the order in
> ld.so.conf a bit since that configure was done.
>
> Humm, that might have done something, in 2 restarts it hasn't
> crashed yet.

See above!

>
> However, the returned image seems to be quite dark in the 600 dpi
> mode, with the raw histogram only occupying about the left hand 25%
> of that window.  Also, all the britness and contrast sliders for
> both overall and rgb are pegged at the right border of the slider,
> and labeled 100 at that point.  The initial image doesn't look that
> bad in the preview window, but 1/2 second later when the autoadjust
> kicks in, the image drops to be dark enough to be obvious.  It can
> only be rescued by the options window britness slider being set for
> about +4 and a new scan done.
>
> Also, the white line artifact is (hooray!) gone when at 600 dpi.
> However, at 300dpi, its back and the output is too dark as before.
> Then at 1200 dpi, the inverse appears to be true, I'm on the third
> scan right now with the britness slider in the option window now at
> -15 and its still quite a bit too bright.  Not making too much
> progress, I'm trying -30 now.  Still too bright in the whites, but
> the darker colors are now all black, but the histograms haven't
> changed.  Now the contrast is set for -30 also, but its still too
> bright and contrasty. Final settings for a decent scan are
> brightness=0 and contrast = -65.  So thats a bit off.
>
> I also tried a 2400dpi scan (can we say slow?) at those settings and
> it looked pretty close to the 1200 dpi result except its beginnng
> to look like it was jpegged a wee bit too much.  Is this mode by
> interpolation?  Even the white line artifact is fuzzy.
>
> Then at 150dpi, the inverse is true, the usable setting is +7 and
> +50.  Ditto for the 75dpi setting.
>
> I don't recall this change dpi, change everything else as being this
> obvious before, and ISTR the main control windows sliders all were
> centered at 100.0 prior to this.  The motor you'll be glad to hear
> is running very quietly, only noticable at 150 and 300 dpi.
>
> This is odd indeed, I went back to do one more check at 600dpi, and
> had to virtually zero those sliders to get a decent scan again.  I
> think I've got ghosts or something...  Should this not be
> repeatable?

Well XSANE is a powerful tool, but there are per default too much things
that will be applied to the aquiered image...
For first tests, I recommend xscanimage or disabling all of the automatic
xsane stuff...

Anyway, I tweaked the driver using the information and register settings I
got from my EPSON 1260. When having a look at the 1250 and at least at
ISCAN, there seem to be some differences which we should pin down...

Cheers
  Gerhard