Hmm. We may wish to review this analysis, being careful to distinguish lines
and line pairs. As noted below, there are some ambiguities that make it
unclear what the conclusion should be.
I haven't read the original article, and so have no opinion on its merits.
At 1:15 PM +0000 1/27/03, olympus-digest wrote:
>Date: Mon, 27 Jan 2003 17:42:35 +0800
>From: Albert <olympus@xxxxxxxxxxx>
>Subject: Re: [OM] Medium Format vs. Digital
>
>http://www.luminous-landscape.com/reviews/shootout.shtml
>
>This is fishy to me... Only because the mathematics of it doesn't work...
>
>Ok, let me assume that the Pentax's lens resolves at 70 lines/mm. This
>is a fairly conservative number, and so I've picked it. Most Zeiss
>lenses etc.. will resolve 80+ lines, but we'll go with 70 lines.
To be precise, it's line pairs (aka, "cycles") per millimeter. The literature
is often sloppy about the distinction, causing much confusion. When someone
says that the resolution of something photographic is X lines per millimeter,
what they almost always mean is X line *pairs* per millimeter.
MTFs (Modulation Transfer Functions) are usually stated as contrast versus
spatial frequency in cycles per millimeter. In this case, the cycles are
sinusoidal, while the film-test cycles are square waves.
I guess I don't follow Pentax lenses very closely, but 70 line pairs per
millimeter is very good for a 35mm prime lens, while 35 lp/mm would be pretty
bad, so perhaps a more representative value is between these extremes. Can we
point to some test data on the lens in question?
>6x7 yields an effective area of 56mm x 69.5mm = 3892mm^2
>35mm yields an effective area of 24mm x 36mm = 864mm^2
>
>That means there is 4.50 [times] more surface area on the 6x7 than the 35mm.
That sounds about right.
>Ok, let's do some math here...
>
>Canon 1Ds = 11.1 megapixels. OK.
>
>If we take a pixel as a crossroad of 2 lines, we would then be able to
>effectively calculate how much resolution a lens in 35mm format would
>have to yield, to take advantage of the 11.1 megapixel sensor..
>
>sq-rt (11,100,000 / 864) = 113 lines / mm
Yes, but the "lines" of "70 lines/mm" are in fact line pairs or cycles, so the
crossroad of two of these cycles contains four pixels.
>I currently don't know of any lenses in 35mm format world that resolves
>this much. Also, as velvia has a 125 line/mm resolution limit, we know
>that the Pentax lens is not using up the full resolution capabilities of
>the film.
>
>Therefore, we must conclude since the film-size is 4.5 times bigger,
>that between the digital CMOS, and the lens on the Canon, since the film
>of choice is not the limiting factor, then that means the Canon lens is
>4.5 times sharper then the Pentax lens? Or the combination of the CMOS
>+ Canon lens is sharper?? No way. Not possible, because the Canon lens
>cannot resolve that much!!
>
>This was the mathematical reasoning behind the 4/3's system! Something
>is very fishy with this test; as there is really no way that an 11.1MP
>camera will beat a decent MF camera (and at 6x7cm no less!!) NO WAY.
>The 35mm lens is the limiting agent here. There's no doubt about it.
>The CMOS can be made finer then film; but the lens in front of it won't
>resolve to it.
>
>The film is not the limiting reagent, the lens is... The math becomes
>more rediculous with the Kodak 14 Megapixel camera... A lens would have
>to resolve to 127 lines/mm to equal "squeeze" out 14 megapixels worth of
>quality... Someone nominate a lens that will do this that is currently
>in the Nikon mount line...
I think that cameras and film have coevolved, so the best-case resolution of
expensive cameras and expensive film will be within a factor of two of one
another. It doesn't make economic sense to make one greatly better than the
other, as the added performance would be wasted, at added cost.
Joe Gwinn
< 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 >
|