[sane-devel] Adding S1300i support

[Peter D'Hoye]

H Ben,

> I'm trying to add support for the S1300i to the epjitsu backend.

So am I (although not active currently)

> 1) There are quite a few commands used by the S1300i Windows driver
> which aren't mentioned in epjitsu.c. All the unknown command sequences
> I've seen are collected together in unknown-commands.log:
>
> - b0/b2/b3/b4: these seem to set a byte (b0/b2 could be a boolean)
> - b5/b6: these seem to get a u16 (b5 always gives 0x19, 0x00 and b6
> always gives 0x10, 0x1e)
> - d8: probably sets a boolean (I've seen both 0 and 1)
> - e1: another byte/boolean
> - 24: receives 256 (!) bytes (no idea what for, but it only appears in
> init.log, not any of the scan logs)
>
> Have you come across any of these before? If not, is there any Fujitsu
> documentation? Or has the whole driver been reverse engineered?
>
> 2) As assumed in epjitsu-cmd.h, most of the magic data I've extracted
> is the same for a given resolution. However it seems that the c3/c4
> cal headers are different for every combination of settings I've tried
> (compare e.g. 300 at color-300-simplex-stop.pcap.h with
> 300 at gray-300-simplex-stop.pcap.h). Could this actually be the case
> with the S300 too? There are commented-out values for these for the
> S300 so possibly there was some confusion as to their correct values.
>
> 3) Do you have any tips for deducing the integer values in model_res?
> Do you just work backwards from what the Windows driver transferred?
> For example, I've got the scanner as far as the first c6 coarse cal
> but it's under reading at the moment. The Windows driver reads 48584
> bytes at 300dpi but epjitsu only attempts 24576 (line_stride 24576,
> height 1) if I copy the S300 settings.

I assume you started your work off patch 21?

If you search back a bit in the mailinglists (mid March) you can find
some USB dumps of me too... I have quite a few notes but they are in
analog form (read: physical notebook). I got kinda stuck between doing
the work based on s300 (which fails on coarsecal) or s1100 (which
fails on coarsecal2).

Allan Noah is the reference/author of the fujitsu backend, he told me
everything he knows is in the sources but the comments are sometimes a
bit thin.

I'll dive into this again and I hope we can crack this one together :)


Peter