Digital BW, The Print

The thing is it doesn't matter which space you work in if you use colour
management to manage the translation. (I'm talking the greyscale case here.
There are many other issues when we move to a colour world.)  In this case
it doesn't matter if you work in GG2.2, Gray Lab or DG20.  We, generally,
tend to work visually and not by the numbers (and any by-the-number work,
say by graphic artists, is always governed by a prescribed colour space).
(A good display profile is hence critical.  There's no sense working
visually if what you're seeing isn't at least somewhat accurate.)

The choice of workspace only becomes an issue when you aren't using colour
management.  Then, in a "Same as Source" work environment, you send
unconverted numbers to the printer.  If the printer has been tweaked for a
particular workspace, not using that workspace is then therefore
detrimental.  QTR Create ICC made QTR Gray Lab obsolete because it allows us
to use colour management to translate between our chosen workspace and the
print space.  So it's best to set "general practice" around widely in-use
workspaces, such as Adobe RGB/GG2.2.  This is given even more weight if
Epson has tweaked Adv B&W to be more in-tune with Adobe RGB/GG2.2.

The other thing which I am the first to admit I need to understand a little
more is the following.  People talk about Lab being closer to human visual
response.  (This language used to come up repeatedly when people talked
about how best to linearize a RIP and it always used to baffle me.)  CIELab
is a perceptually uniform space meaning the distance between two points also
reflects the relative perceived change in colour between the two points.  In
this way it was designed as an improvement on CIE XYZ.  (People are
generally familiar with the xy chromaticity chart which maps out the range
of human vision.  The issue with the xy chromaticity chart is that it
doesn't factor in the non-linearity of the eye and so from a chart or plot
perspective the distances between two points are distorted.) But CIE Lab
doesn't model human vision any better than XYZ.  It just uses co-ordinates
that make it more perceptually uniform in plot.

Various workspaces cover greater or lesser parts of the xy chromaticity
chart.  (Prophoto RGB, for example, is far broader than Adobe RGB.)  But
that doesn't make one workspace any less "closer to human visual response"
except in so far that it covers less of the human visual response's field of
view.  (Prophoto is still an RGB space.)  When we come to greyscale, all of
this collapses to no difference because the greyscale is common to all the
workspaces.  (I am sure there are some really technical anomalies here but
in general terms I think this is correct.)  QTR Gray Lab does not model
"human visual response" any better than CIE XYZ (it's just a perceptually
uniform representation of the same thing) and Adobe RGB or even GG2.2
contain all the greyscale information that is in the human visual response
system.  The notion that Lab (and in our case the focus is on L*) more
closely models the human visual response is, I think, a bit of a misnomer.
It's advantage is that the steps in the numbers coincide with steps in
perceived change - that's all.  From a practical point of view, this is
really only useful when we look at a step wedge.  Then the L* numbers align,
inversely, with the % K numbers.  That's fine but if using it throws other
things out of whack (such as any tuning in a non-colour managed printer
driver such as Adv B&W) then gaining that uniformity is actually
detrimental.

[For colour work I use ProPhoto RGB because there are many printable colours
that we can see which are outside the Adobe RGB space.  The monitor is
actually the big bottleneck at the moment. We need more capable monitors and
higher bit depth graphics cards.  As monitors and printers become more
capable, watch the steady movement away from Adobe RGB to a broader space.]


> From: Ernst Dinkla <E.Dinkla@...>

> If there's one mailing list and group of people that could
> promote a perceptual greyscale curve for use in B&W
> photography it should be this list. Roy gave us that space and
> it is closer to human visual response than any of the gamma
> (old photography attempt to perceptual) and dotgain
> (completely different concept) choices.
> 
> I know that Gamma 2.2 works and isn't in contradiction with
> the rest of the workflow but why not promote a conversion to
> the QTR space as it is available ?
> 
>                     --
>            Ernst Dinkla