[Openicc] Drop size calibration

[Hal V. Engel]

On Monday 28 January 2008 04:37:45 Robert Krawitz wrote:
>    Date: Mon, 28 Jan 2008 10:52:25 +0100
>    From: Gerhard Fuernkranz <nospam456 at gmx.de>
>
>    Graeme Gill wrote:
>    > Robert Krawitz wrote:
>    >> I'm experimenting with another approach to drop size calibration.
>    >> This uses the new segmented dither algorithm to print stripes using
>    >> different drop sizes.
>    >
>    > I guess I'm a bit puzzled as to why you need such a calibration.
>    > By definition a larger drop is going to result in higher density
>    > than a smaller drop. So as long as the dither/screen progresses
>    > through the dots in order, the result should be monotonic, even
>    > if it's not very linear. As long as the input precision to
>    > the screen/dither is high enough, and the resulting raw screen
>    > transfer curve is smooth enough, the calibration will linearize the
>    > result.
>
>    The calibration algorithm just must be prepared of course, that the
>    theoretical raw transfer curve of a typical multi-level screen
>    which dithers between two adjacent drop sizes, may have
>    discontinuities of the 1st derivatives at the boundaries between
>    the drop sizes, i.e. the theoretical raw transfer curve is
>    segmented, where each segment is continuous. But I think
>    linearizing such curves should still be no problem, if the channel
>    response is measured with sufficiently small steps (if necessary
>    iteratively with two linearization passes). In practice one likely
>    does not even notice these discontinuities as sharp corners in the
>    measured raw response curve, due to noise, so in practice I'd
>    assume that the raw curve is smoother than theoretically anyway.
>
>    So I'm not so sure, whether it is really necessary to tune the
>    dither algorithm just on order to give a smoother raw response (if
>    this tunig does not improve other issues too), but it is likely
>    more important to keep the algorithm's behaviour monotonic as long
>    it is granted that drop size 1 < drop size 2 < drop size 3 ... The
>    dither algorithm should be designed in a way, so that even wrong
>    assumption about the absolute sizes of the drops should never
>    result in a non-monotonicity.
>
> This tuning is on a per-printer, per-drop size basis; it doesn't
> involve actually modifying the dither algorithm.
>
> The other thing is that I suspect that not every combination of
> printer, paper, resolution, etc. will wind up getting properly
> linearized, so I want to have reasonably good estimates of the drop
> size out of the box.  I could always simply set sizes of 0.25, 0.5,
> 1.0 for everything and be done with it, but then a lot of resolutions
> on a lot of printers will look rather ugly until someone gets around
> to doing a lot of work.  This method is a lot faster than what I was
> doing before (when I had to recompile the driver and print a whole
> bunch of test strips and then try to estimate matches to make any
> changes at all to the drop sizes).

I suspect that drop size calibration has a lot in common with calibrating the 
light/dark ink transitions in CcMmYKk type printers.  A few weeks ago I 
printed out some targets with C, M, Y, K and CMY ramps and measured these 
with my DTP20 UV.  I did this with my Espon R2400 using the default settings 
in GutenPrint.  I then imported this data into a spread sheet and plotted the 
response curves and I also plotted the deviation of the L* response curves of 
the C, M and Y channels from a linear L* distribution.  The plots of the L* 
channel curves showed for the most part that these curves appeared to be 
smooth.   It was the plot of how much the curves deviated from a linear L* 
response that proved to be the interesting one .  It is posted here:

http://lprof.sourceforge.net/images/Channel-Error.jpg

In this case error = % deviation from L* being linear.  The left side of the 
curve is the light end of the scale.  Above the 0% line is darker and below 
it is lighter that a linear L* response.

Notice how the cyan curve has a significant bump and the magenta curve starts 
out above (darker than) the L* line and then shifts below it (gets lighter).  
Both of these appear to happen in about the same location in the curve.  I 
suspect that this area corresponds to the the Cc and Mm transition points but 
it could also be an ink drop size transition point or a combination of both.  
The area where these anomalies occur can be seen in the CMY ramps as a 
noticeable green shift.  The reason I suspect that this is the light/dark ink 
transition point for these two inks is that the yellow curve, which only uses 
one ink, has a very smooth curve compared to the cyan and magenta curves.  In 
later tests I also plotted the same thing for the K channel and I could see 
the same sort of anomalies in the curves but in two locations rather than 
one.  Since the K channel on this printer has three shades of ink this seems 
to support my conclusion that these were light/dark ink transitions.  

FYI these where 40 step ramps.

Hal