Interpreting spectrophotometry data

Margaret Morris <mmorris [at] cca__qc__ca> writes

>We recently purchased a spectrophotometer and have just begun using
>it to determine colour change in our photographs collections. With
>the help of Roy Perkinson's article Statistics Without Anesthesia we
>have set up a spreadsheet in Microsoft Excel which incorporates the
>standard deviation calculation and the student 't' test, and which
>tells us the delta E result before and after readings. We now have a
>mechanism for determining if a statistically significant change has
>occurred. My question is how important is that statistically
>significant change?

Delta E (dE) is a valuable measure, but it needs interpretation. A
dE of 2 is just perceptible in the middle L value, when compared
side by side by an experienced observer.  A dE of 5 is difficult for
an experienced human to differentiate in the very light and darker L
values.

If you look in the Wilhelm book, now online at
<URL:http://www.wilhelm-research.com/book_toc.html>, you will find
the dark storage life of most photographic materials.  Light
exposure lifetime is different from dark storage, but in many cases
dark storage trends (not values) are valid.  Older data (formerly
found) on the Wilhelm web site will give light exposure data for
many common digital and photographic prints.  The basic premises is
that a 1.0D patch in the material will change 20-30%, at the end of
the predicted lifetime.

Assume that a 10 year exposure of a 1.0D patch (neutral) resulted in
a 25% change in L.  The L value (Luminosity in Lab color space)
would change from L=34 to L=51 over the 10 year exposure.  If you
calculate dE between Lab 34,0,0 (start) and 51,0,0 (end), it would
be 17.  This is a noticeable change.  It might be humanly
perceptible from memory.

If density of the measure area started at 2.0D, an average dark in
modern photographic print, and it changed 25% to 1.5D with exposure,
the change would be 12 dE.  This would be a change that an
experienced human could perceived when compared side-by-side, but
little else.  Try it in Photoshop: 2.0D is RGB 31, 31, 31; 1.5D is
RGB 84, 84, 84.

Statistically significant means that at a particular confidence
level, commonly 95% confidence, two values will always be the
different in 95 out of 100 experiments, and the same in 5 out of 100
experiments.  You can calculate a rough statistical significance by
making 9 separate measurements of the same 1.0D patch, starting with
the instrument "off" each time.  If the values vary a mean of 1%,
that would be quite good. If one will use different equipment to
make the before and after measurements, the variation will be much
larger; see Wilhelm Book's, Table 7.3, pp 260.

Wilhelm has a chapter (7) on making long term color measurements.
Data from 54 days of exhibition showed almost no change, only 10% of
the prints had statistically significant changes.  The three albumen
prints darkened with time, something that Paul and I observed in our
albumen studies.

Fuji Crystal Archive is predicted to have a life on display of 75
years (20% change in one dye) at 1.92 Mlux-hrs exposure per year, or
144 Mlux-hrs exposure in its life.  Wilhelm actually recommends a
lower percent change for collectors and museums: 9-13%.  So, assume
a 72 Mlux-hrs exposure life for Fuji Crystal Archive photographic
paper in a museum, the best photographic paper available today.

For 300 days of exhibition at 45 lux, 12 hrs/day, over 300 days, the
radiation dose would be 0.16 Mlux-hrs.  That would be less than 0.2%
of its lifetime, and a 0.0002D change (0.002% of 10%).  I'm sure
that you could not find an instrument to measure that change.

If the material had an exposure life of 8 Mlux-hrs, a tenth of Fuji
Crystal Archive's exposure life, a 0.16 Mlux-hr exposure would be a
0.002D change, maybe some research grade equipment could make this
measurement.  If the color material had the predicted exposure life
of 0.8 Mlux-hr, the loss of color may just be perceptible using the
same machine.  Eight Mlux-hrs is about the accepted exposure life
for colored works on paper, a tenth of that would have to be a
extremely sensitive color artifact.  B&W prints are assumed to have
a life of 100-200 years on display at 1.92 Mlux-hrs exposure per
year.  Yellowing of the Dmin (whites) is the probable mode of
failure in historic B&W prints; experience shows that yellowing is
acceptable for older materials.

Tim Vitale
Conservator and
Digital Imaging
1500 Park Ave
Suite 132
Emeryville, CA 94608
510-594-8277


                                  ***
                  Conservation DistList Instance 17:57
                  Distributed: Tuesday, March 2, 2004
                       Message Id: cdl-17-57-001
                                  ***