Archived posting to the Leica Users Group, 1998/04/20

[Author Prev] [Author Next] [Thread Prev] [Thread Next] [Author Index] [Topic Index] [Home] [Search]

Subject: Re: [Leica] an objective evaluation of leica M lenses??
From: Erwin Puts <imxputs@knoware.nl>
Date: Mon, 20 Apr 1998 16:40:06 +0200

>The evaluation of the Noctilux made by the french review "Chasseur
>d'images" is objective, i.e made with the MTF analysis.Leica Camera does
>the same and publishes some of the results (not for the Noctilux, of course
>!It's too bad!)

First of all: Leica's policy is this: they do not publish MTF graphs of
older lenses in general and not because of bad results. You will not find
published MTF graphs for the Summicron-M 50 and the Summilux 75, themselves
star performers in the current Leica line-up.

The MTF measurements of Leica are totally different from the ones of CdI
and for that matter: both differ from Zeiss. The reports by CdI cannot be
objective. What happens is this: MTF measurements are made at several
spatial frequencies and the contrast figures in exact numbers are graphed
over the image field. So we can say that at an image point 6 mm from center
the 10lp/mm are recorded with 90% contrast transfer. At 40lp/mm the same
point reaches 40% in the tangential direction and 35% in the sagittal
direction. This information is objective (but no longer interpretable by
most persons).
Now what does CdI do: they produce a weighted average of these figures (the
weightings are unknown to me, but maybe Dominique knows). And this weighted
result is attached to a word, like fair or very good. Here CdI assumes a
linearity that does not exist.
I do know of a Swedish  magazine that also uses MTF figures (from
Hasselblad). They put the figures (very accurately) into a spreadsheet and
manipulate the numbers to get a condensed figure or word. Their weighting
and numberjuggling  favors the performance at f/8 in the center. Hardly a
noteworthy exercise.
As far as the Noct goes: the results at full aperture for the 40lp/mm must
be approached with some background knowledge (read my report in
PhotoTechniques soon). The quality at the 10lp/mm however offsets the
results at 40lp/mm and the Noct adds its own special characteristics
(excellent rendition of fine textural detail, which at f/1.0 is much more
important than the ultimate in resolution).
The defects of the CdI method are quite clear: They use one set of
calculations to generate one overall qualification. This calculation
however is only one of many possible weightings and without the actual
weighting and explanation of their choices, we have the famous case that it
is easy with statistics to prove anything you want. CdI themselves note in
their explanatory article that wide apertures strongly influence negatively
the overall score because of their weighting in the overall calculation.
They use the spatial frequencies of 10, 20 and 50 lp/mm and even pick up
100 lp/mm. Now 100lp/mm is so far removed from realism that using these
results would be quite dangerous. But even 50lp/mm is a bit on the wild
side. They also neglect the 5lp/mm that most commentators regard as very
important for overall sharpness impression.

Erwin