Re: [Gimp-print-devel] softweave confusion

[Robert L K. <rl...@al...>]

   From: sh...@al...
   Date: Sun, 13 Feb 2000 22:28:11 +0900

   Ok, I admit it, I really don't understand softweave.  I get the general
   principle, but the details clearly elude me.

   00000782  1b  i 12 01 01 af 00 30 00  (175, 48, 1) *0d 
   000008fb  1b  (  v 04 00 31 00 00 00 

   The 750 has 48 nozzles, so, we send 48 lines of data at once, one for each
   nozzle.  And then we move down 49 rows?  49?  Why 49?  Doesn't that skip a
   row?

OK.  This stuff is fairly complicated.  I'll explain very quickly, and
then point you to print-escp2.c which has it all explained in a long
block comment.

The first part is correct -- we send 48 lines of data at once, one for
each nozzle.  However, there are a couple of other things going on
here.  The most important is that the nozzles in the print head are
actually separated by more than one row spacing (1/720").  On the
Photo EX there are 32 nozzles spaced 8 rows apart.  On the 750 there
are 48 nozzles, there are 48 nozzles.  I assumed they were spaced 8
rows apart, but the stuff yesterday with the 740 convinced me that
they're 6 rows apart.  So there need to be 6 passes printed to fill in
the gaps between printed rows.

The obvious way to do that would be to advance by 1 row for each of
the 6 passes, then advance by (48 * 6 - 5) rows to start a new series.
The problem with that is that it creates a lot of banding (that's
explained in detail in the comment).  So the name of the game is to
ensure that each adjacent row is printed by DIFFERENT nozzles (some
nozzles are slightly bigger or smaller, or the spacing may not be
perfect, or something).

Even doing that carefully (the 720 Softweave mode) results in some
banding.  So what we do is use multiple passes to print each line, by
splitting up the dots and printing only a subset of them (1/2 in high
quality, 1/4 in highest quality) on each pass.  That way two or four
different nozzles are used to print each row, which further reduces
banding.

The 1440 modes are similar; 1440x720 uses two passes, 1440x720 highest
quality uses four passes.

It's pretty hairy stuff.  Read the comment very carefully.  The
function weave_parameters_by_row does all the cool stuff, but it's
almost impossible to understand (even for me, and I wrote the silly
thing).  It isn't perfect; there are some combinations that don't
work, although everything that I know of that's really important to us
(except maybe for the 440) works.

There's a program called weavetest.c that's actually a unit test for
it.  This program is absolutely essential to have any hope of
maintaining this stuff.

-- 
Robert Krawitz <rl...@al...>      http://www.tiac.net/users/rlk/

Tall Clubs International  --  http://www.tall.org/ or 1-888-IM-TALL-2
Member of the League for Programming Freedom -- mail lp...@uu...
Project lead for The Gimp Print --  http://gimp-print.sourceforge.net

"Linux doesn't dictate how I work, I dictate how Linux works."
--Eric Crampton