re:Digital vs Analog (and where to get a cheap scanner)

[P3D Michael Kersenbrock <michaelk@xxxxxxxxxxx>]

> 
> I don't know how to explain it any clearer.  Halftoning works because
> at some level, someone (e.g. a human applying a screen) or something
> (an algorithm) knows that you can approximate a grey level using only
> black/white "pixels" by trading off against resolution (you spread out
> your virtual, grey, pixel among some number of physical black/white only
> pixels, at the cost of a much larger "virtual" pixel).  In this
> arrangement, the real pixels that comprise a virtual pixel are related;
> a function related to the desired grey level that says which ones will
> be on and which will be off.

I agree with the purpose you had in saying what you did, however I'll
hold to the notion that dithering/halftoning "actions" are still what
they are even if produced by random-chance and that "knowledge" as such
can be used, but isn't required.  The density of silver crystals isn't
done by "knowledge" of one crystal with respect to the others around
it as to generate the proper density.


> 
> The grain in film has no such mechanism.  Each silver salt crystal acts
> on its own without regard for what its neighbor is doing.  The film
> CANNOT be binary as you suggest because there is no mechanism instrinsic
> in the film to distribute the virtual pixel information around to the
> physical grain structure.

I was using the term pixel as a way to describe the structure,  it was
one I was imposing upon it in order to more closely compare the structures
in order to more fully understand how to compare them honestly.  I 
didn't mean to imply that the structure was *created* that way.  I was
suggesting that we apply an artificial imaginary "screen" to the silver-image
and counting the size and quantity of silver crystals in each square and 
just delaring that the arbitrary square a pixel and the density of crystals
within the square as the pattern.  I wasn't talking about creation, I
was trying to speak about interpretation of the results.

> While it's true that a CCD imager is an analog device, the circuitry in
> a digital camera divides the continuous voltage changes from each pixel
> into a distinct series of steps (256 for an 8-bit pixel).  Any and all
> gradations between these steps that the analog CCD may be able to
> distinguish are lost, and the intensity vs. output graph becomes a
> stair-step, instead of a smooth curve.  So digital cameras are properly
> called digital, despite using an analog imaging element, because their
> output behaves in a digital way.

I kinda knew that, but thought it was funny anyway.  Sorry about that.  :-)


> >With digital, each pixel can be completely stark-black or completely
> >100% white at full resolution .... or some exact value inbetween.  I
> >don't believe that is true for each line in your resolution test as
> >it is usually measured (correct me if I'm wrong on that).
> 
> Not true.  Digital pixels do not change in intensity in a linear fashion,
> that's an analog function.  Digital pixels change in a step-wise fashion,
> with the amount of change being determined by the number of bits available
> to encode the range.  If your digital circuit encodes white as 1.0 volts
> and black as 0 volts, with 8 bits you have 254 intermediate values between
> these extremes.  If you want a grey that's somewhere between two of these
> values, you simply cannot represent it, without increasing the number of

Good point.  But doesn't it take an area of silver-image film to use
what I'm calling dithering to provide that continuous range of tone
or does the film use only density in the thickness-direction of the
film to create different gray tones, even for an infinitely small dot
in the film's area?


> 
> The line pairs per millimeter measure of resolution specification
> requires that you be able to *resolve* the lines as lines, which means
> that there is some minimum separation of intensity between them.  It
> certainly won't be 100%, as you say.  In reality, a digital system
> won't produce 100% separation either, since a monitor isn't usually
> capable of going from black to white within one dot time without some
> smear or ringing or other artifact.  Inks spread.  Etc.


Does this mean that film has low resolution if I look at
it through a lens that has vasoline spread on it, or if the enlarger
that made the print was out of focus (just kiddddding) ?

When talking about practical systems, then you make a good point about
the system's resolution, however my subject matter was trying
to understand the 3-Million pixel 35mm equivalence that I read about,
and in those terms, it's a discussion of the information content,
not about the resolution of the VACUUM TUBE technology we commonly
view images with.  Film has the clear lead in terms of all pieces
of the system, I'm mostly puzzled as to defining what numbers
are needed to define the point of equality because the measurements
used to define performance don't seem to be exactly the same (and
probably have the same inherent problems that "benchmarks" generally
have).

In conclusion, I suspect the 3-Million figure is probably wrong on
the low side, but I also suspect the 30-Million figure is probably
wrong on the high side, and I haven't the foggiest on how to decide
exactly what it is.

Nor is it particularly important. I just thought it interesting.  If
I were talking practicality, I'd bring in resolution of the weak
link in each system's chain, starting with lens resoluton of the
cameras through screen resolution for CRT viewing or Dye-sub 
printer resolution for digital-image printing (some of which look
awesome to me, I wish I had a high-end dye-sub printer!).

Thanks for the ideas, I'll stew on it for a year or so until the
subject comes back again for the next "round".  :-)

Mike K.

P.S.- I just got one of those $50 brand-new Primax color hand-scanners
      (www.onsale.com) for me to play with until I can get a really good 
      and proper flatbed.  I think most of the company's products are being
      dumped with the company's purchase by another.  Not quite a digital
      camera, but should scan a stereocard just fine.  They have two
      auctions per week, with Computer Geeks selling them.  At the moment
      $39 would put one on the list, but it doesn't end until Monday (then
      restarts).  Usually ends at about $50.  I haven't installed it yet,
      so it might be junk, but it certainly isn't expensive.  The unit appears
      to have been stock from Global (the distributor).  Still has a sticker.
      Claims to be up to 800 dpi, but manual suggests that the hardware resolution
      is 400 dpi max.  Still okay for a "toy".  :)


------------------------------