[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Digital Negatives with enough Tones for Pt/Pd
Wayde Allen wrote:
> Notice that it is "your" criterion of keeping the size of the divisions
> (sections) equivalent sized that depends on both bit depth and dynamic
Sorry you think so. So you think I am inventing criteria; well think
again. It is a fact that at a certain level, posterization of tones
will be detected in the print. This posterization is the direct result
of the total number of unique tones in the print. An easy demonstration
is to look at a photographic image with 2-bit and 4-bit data. This can
be done in Photoshop Idexed Color Mode.
> Having a condition that depends on two independent quantities
> doesn't make these quantities in any way dependent on each other.
It most certainly does when the "two independent quantities" are tied to
the "condition" as is the case.
I want to first get the information scanned from the negative
(generating enough data to produce a high quality Pt/Pd print, which
means recording enough tones), have the abitlity to manipulate that data
(the 16-bit GIMP looks like a good posibility for this), and produce a
negative (or, at present, a sandwhich of negatives to overcome the
defitiency of printers), and do this from my desk at home.
At this time, I can say for sure that I am confident of two things. A)
I am not yet sure how many tones and the distribution of tones that are
needed to print a Pt/Pd print with the same qualities as from
photographically exposed film. B) I am sure that a total of 256 tones
is not enough and the resulting posterization produces a print different
from that produced by photographically exposed film.
> The only reason to increase the dynamic range would be if
> what you are scanning exceeds the dynamic range of the scanner.
And this is what Case 3 is about below.
First, lets get your latest analogy straight. The dynamic range of the
object is the length of the object which is fixed and a constant. The
dynamic range of the instrument is variable as the ruler shortens or
lengthens in order to have its ends placed at the ends of the object
(This should remind one of the exposure adjustment tool in the
scanner.) The ruler can be thought of as being made of rubber. The
ruler is marked with a scale (bit depth). The "condition" imposed is
that the absolute accuracy of the measurement must be a minimum of 1/10
inch (Just pretend the shop manager required this.) The following
represent three possible scenarios.
1) When the ruler is in standard position, the divisions of its scale
are 1/10 inch. when the ruler is placed along side an object of the
same length, the object can be measured to 1/10 inch accuracy.
2) When the ruler is compressed as it is placed along side an object
half as long as the uncompressed ruler, the object can be measured with
an accuracy greater than 1/10 inch (to 1/20 inch as each division of the
ruler represents half its uncompressed value).
3) When the ruler is stretched as it is placed along side an object
twice as long as the unstretched ruler, the object can only be measured
with an accuracy less than 1/10 inch (1/5 inch as each division of the
ruler has been stretched to double).
In case 3 (a case similar to 8-bit scanning of negatives) the length of
the object cannot be measured to the required accuracy. Only half of
the object can be measured at the required accuracy. Now as it was
pointed out, two measurements (multiple scans) can be made to determine
that the length of the object is twice the length of the ruler with an
accuracy of 1/10 inch.
However, the boss says this must be done with only one scan. The
variable dynamic range of the ruler increases as the ruler is stretched,
but it loses accuracy. So what to do. Ah yes, we can assign a new
scale to the ruler (the boss has given us a budget to spend as
necessary). So we put this new scale of 1/20 inch divisions on the
ruler. Now any object up to twice the length of the ruler can be
measured with at least 1/10 inch accuracy.
Oh, but now we are given an object 4 times the length of the ruler. Not
to fear as the boss has deep pockets. We get a new scale of 1/40 inch
divisions and we're back in business.
What should be noticed here is that every time the length (dynamic
range) of the ruler (scanner) is increased so as to measure (record
data) an object (negative) of a given fixed length (negative's given,
fixed dynamic range) to the required accuracy (posterization level), we
MUST increase the scale divisions (bit depth). This very practical
purpose demonstrates a dependent relationship that bit depth increases
with dynamic range.
Jeffrey D. Mathias