[sane-devel] EPSON 3490, calibration and raw data

[Krzysztof Halasa]

Krzysztof Halasa <khc at pm.waw.pl> writes:

> http://www.epson.co.uk/products/scanners/product_spec/Perfection3490Photo.htm
>
> Not nice, I wouldn't expect that from them.

I have changed:

#if 0
            pc[SET_WINDOW_P_GAMMA_NO] = 0x01;   /* downloaded gamma table */
#else
            pc[SET_WINDOW_P_GAMMA_NO] = 0x00;
#endif

and now, using 16-bit mode with "255" calibration (BTW the ppm_stat.c
program I sent to the list had a bug printing the same "different
pixel count" for G and B):

test1.ppm: different pixel values:
        R = 57545 (117-61776)
        G = 54193 (1002-62775)
        B = 54548 (2-63413)

It's 3192 x 19137 x 16bit (ca. 4 frames of 35 mm negative film at 3200
dpi) and it includes uncovered area (no film at all) and fully-covered
(metal film) one.

So unless the additional bits come from somewhere else it's really
a 16-bit (48-bit) scanner. I just have to figure out how to upload
16-bit gamma table to it (or set it to 1:1, the G and B ranges suggest
some correction applied).

Of course it doesn't automatically mean that all 16 bits from CCD
are usable, at least with a single-pass scanning (I can see some amount
of noise with 14-bit) - what I currently want is not a noiseless image
but a raw uncorrected data from CCD.

BTW:
calib-output-8b.ppm: different pixel values: (8-bit raw calibration data)
        R = 114 (66-234)
        G = 92 (71-229)
        B = 92 (75-233)
calib-output.ppm: different pixel values: (16-bit data)
        R = 117 (69-15104)
        G = 93 (71-14784)
        B = 93 (74-14976)

It looks like the black calibration (lower values, around 70) is the same
data in both 8-bit and 16-bit mode (i.e., in 8-bit mode the LSB is
present and 6-bit MSB is assumed to be zero).

I will experiment with this a bit longer - probably someone has an idea
how to set 16-bit gamma tables and/or to set "analog" gamma value (to
1.0) so the scanner doesn't modify the data?

Or maybe it's already transparent, i.e., the scanner does no gamma
but just brightness and contrast (already set to 128 = neutral)?
The peak values are a bit different from the 14-bit calibration
data * 4, the range should be something like 280 - 60500.
-- 
Krzysztof Halasa