Olympus-OM
[Top] [All Lists]

Re: [OM] Set me straight, fine grain != good resolution?

Subject: Re: [OM] Set me straight, fine grain != good resolution?
From: "John A. Lind" <jlind@xxxxxxxxxxxx>
Date: Thu, 17 Jul 2003 21:41:34 -0500
At 04:40 PM 7/17/03, AG Schnozz wrote:

81 lp/mm?

81x36mm=2916 lp horizontal (round to 2900)
81x24mm=1944 lp vertical (round to 1900)

To extend this further and draw an analogy to pixels, these numbers would be doubled. With digital pixels it requires two rows to represent this; a black one next to a white one. As the Schnozz points out, you cannot work this in reverse starting with a CCD.

I hope AG will forgive me for amplifying on his math a bit to provide more about it and show that he did not dream this stuff up by plucking it off of a skyhook that drifted by.

With the Canon D1S, the CCD is 4064x2704 pixels. This gives a
maximum lp/mm of around 57 lp/mm or aound 2000 lp horizontal.

[Drum Roll]
[Horn Fanfare]

Wheel out Nyquist Theorem:
A theorem, developed by H. Nyquist, which states that an analog signal waveform may be uniquely reconstructed, without error, from samples taken at equal time intervals. The sampling rate MUST be equal to, or greater than, TWICE the highest frequency component in the analog signal. However, there is the problem of "line straddling" in which a detail lands on the boundary of two pixels. It ends up blurring reality. IOW, the Nyquist Theorem represents an _absolute_limit_ and NOT what is likely to be experienced in practical application. THAT requires a "de-rating" of the Nyquist Theorem limit.

However, that assumes perfect alignment of the pattern and the
array.  A more accurate resolution is more like 38 lp/mm which
is about 1350 lp horizontal.

[Longer Drum Roll]
[Louder Horn Fanfare]

Wheel out Kell Factor:
This is subjectively based on human perception about what truly reproduces reality in digitizing an analog signal. In this case, the analog signal is an image projected on a CCD. For still photographs (including digital cameras and scanners), 0.65 - 0.70 is generally the accepted Kell Factor value. For digital video, it's higher (0.85 - 0.9). What is lost or blurred in one frame is likely to be restored in the next one.

Velvia has almost 2900 lp across the width of the image.
Canon has almost 1350 lp across the width of the image.

Yuppers!  You got it right.

In all fairness, you would experience the gain in Velvia only if
kept in the analog realm.  The digitizing process instantly
negates most of the lp/mm gain in film.

You got that right!

Digital is coming...  It's only a matter of time before the
price and performance match and beat film.

I'm going to pretend I'm from MO; show me. I'll believe it when I see it. Just as the more classical Physics theories break down when things get small enough (giving rise to Heisenberg's Uncertainty Principle and things like Quantum Mechanics), I forsee digital advances hitting some very real limits with dimensions become significant at a molecular level. IOW, there's a very, very real wall in terms of how small something can be made; atoms and molecules do have FINITE dimensions.


AG (My OM with Velvia is still better than an $8000 DC) Schnozz

AND . . . so is my OM with Kodachrome 64!

-- John


< This message was delivered via the Olympus Mailing List >
< For questions, mailto:owner-olympus@xxxxxxxxxxxxxxx >
< Web Page: http://Zuiko.sls.bc.ca/swright/olympuslist.html >


<Prev in Thread] Current Thread [Next in Thread>
Sponsored by Tako
Impressum | Datenschutz