[argyllcms] Re: inter-instrument matching tables
- From: Roberto Michelena <colorsync@xxxxxxxxx>
- To: argyllcms@xxxxxxxxxxxxx
- Date: Fri, 18 Nov 2011 09:47:55 -0500
Graeme,
Of course there are several scenarios and one of them includes as you
mention the unavailability of some of the parts: either one of the
instruments, or the ink/paper combo. But in reality a model that
solves the inter-instrument matching conundrum on a present-time basis
(not trying to match across history) solves 90% if not more of the
problems which users commonly face.
Three scenarios come to mind in which I (and probably most users)
would love to be able to match instruments:
1) proof calibration and certification across sites: a company has "n"
proofing sites, all standardized on the same paper brand/model, and
the same inkset (say, has a variety of Epson 4880 and 3880); each site
has it's own spectrophotometer, some Eye-One because they're lower
volume, some DTP-70 because they're higher volume, some Isis because
they're the "always the latest".
Right now, on top of the repeatability of the prints themselves,
they'll have the fact that inter-instrument matching across that
variety might be off by up to 3dE.
2) proof calibration with one instrument and certification with
another: a particular site (which is settled on one printer they have
and one brand/model of proofing paper they use) has an Eye-One
permanently, but when doing a new calibration the "calibration guy"
comes in with his DTP-70 and does it in a snap. Unfortunately his
wonderful "0.5dE avg/2.9dE max" calibration, after he leaves,
certifies to just 1.2dE avg/4.5dE max" with the Eye-One because it's
not the same instrument used to calibrate. Furthermore, a couple weeks
later it may have drifted out of tolerance, while if measured with the
DTP-70 it'd still be quite within.
3) press calibration with a variety of instruments: customer wants to
place his press in as close to ISO or Gracol/G7 as possible; starts by
finding optimum densities for hitting the solids Lab values, and does
so with his press-side SpectroEye. Later he measures a G7 graybalance
chart to make the curves, but for the sake of speed (avoid cutting the
chart off) he swiftly does that with his Eye-One so he can output
plates asap for the second run. Finally he finishes the second run
with curves applied, and then for validation (or link profile if he
was too off) he reads the IT8/7.4 with his DTP-70.
But of course the solids he finely nailed within 2dE by trusting his
Spectro-Eye will be off tolerance or borderline when read with the
DTP-70, not to say the graybalance he built with the Eye-One. He might
be willing to (next time) avoid the Eye-One and just cut the
graybalance chart and use the DTP-70 for graybalance too, but for
press-side solids it can't be used, has to be the SpectroEye ; well,
might be the EyeOne but what a pain. By all means you can't take this
down to less than two instruments, unless you want to read the IT8/7.4
patch by patch with either the SpectroEye or the EyeOne.
I'm sure a lot more scenarios similar to this ones that make up the
daily lives of many users can be envisioned, in which inter-instrument
matching software will be a godsend, in spite of being limited to a
single ink/paper combination and to require availability of both
instruments at the same time and place at least once (or shipping
prints from one site to another in a well-protected and fast way).
For next week I'll try to get real data for posting from 4 instruments
reading the same chart:
- a medium-aged EyeOne rev C
- an old UVCut EyeOne rev A
- a DTP70
- same DTP70 with UV filter engaged
Now, the only chart I have that's formatted to be read by both EyeOne
and DTP-70 (and actually Isis too) is Idealliance's P2P25n ; hardly a
large profiling chart. I wonder if there's a way to make a version of
IT87.4 that'd be readable by many instruments too.
-- Roberto
On Thu, Nov 17, 2011 at 6:34 PM, Graeme Gill <graeme@xxxxxxxxxxxxx> wrote:
> Gerhard Fuernkranz wrote:
>> Spectral_readings and reference_readings are the two corresponding
>> measurement sets, acquired with the two instruments, from the same print
>> of the test chart (the rows of these matrices are the observations (i.e.
>> the patches), and the columns are the spectral bands; the columns of the
>> two measurement sets do not need to correspond 1:1, i.e. the wavelength
>> ranges and sampling intervals of the two instruments can be different).
>
> Hi Gerhard,
> very interesting, but I guess it doesn't cover some scenarios
> so well. For instance, say (hypothetically :-) a manufacturer were to change
> the behavior of one of their instruments between revisions - say they
> change their calibration reference and their U.V. light source. A
> customer that had a large number of historical records read by the
> old instrument, yet wants to compare them with readings made on
> the new instrument (because they don't have access to the old
> instruments - they are at some other location, or the old instruments
> are no longer serviceable), and who cannot re-read the old
> charts because it would take too long, or because they are no
> longer available or have aged, can't really make use of this
> approach. Ideally they want something that transforms the old
> readings into new readings independent of the paper or ink.
>
> I think an FWA compensation model could be applied, but the rest
> of the transform would have to be ink and paper independent.
>
> Cheers,
> Graeme.
>
>
Other related posts:
