Moose thus howled:
> The GX7 scores worse than either of the Olys. I have all three, I've
> processed a lot of low light images. The GX7 has
> slightly lower noise, not higher, at least up to ISO 6400. To me more
> importantly, the GX7 noise responds better to NR,
> AND, when noise is reduced the GX7 has FAR fewer 'wormy' artifacts
> left/revealed than the Olys at the same ISO. At the
> end of the day, the GX7 is, as a practical, visible matter 1 to 2 stops
> better than the Olys in high ISO noise. This is
> a much greater difference, and in the opposite direction, than DXOMark gets.
This is EXACTLY the kind of information that is hard to quantify, but
is where the rubber meets the road. Sometimes, it isn't a matter of
total noise elimination, but a matter of whether the noise that is
there improves the image, takes away from the image, or is easily
neutralized with no ill effects. The "wormy artifacts" may show up as
remarkably low scores on a quantifiable measurement, but are anything
but desirable. I would pick the otherwise "noisy" image over a "wormy"
image any day. Worms are NOT good. Not even cooked in bacon.
> DXOMark says the Olys actually under perform rated ISO, which would make the
> differences even greater. I've done no
> serious comparisons, but when I was taking ISO test shots of a real life
> subject, the exposures were the same and the
> resulting images looked the same in exposure visually and on the histogram.
> (Color, contrast, saturation, etc. are
> different, via ACR.)
This gets back around to one of my original complaints about DXOMark
and the ISO methodology they use. In essence, they redefined what
exposure ISO means and made it completely irrelevant to real-life
usage. When they first launched this method, I, along with a number of
other people, addressed it as flawed, but was out-shouted by the
landscape photographers who were so intent on ETTR as the one and only
method of exposure determination. That exposure method is bogus for
the vast majority of photographic applications, as well as flash
lighting. DXOMark continued on with their screwed up redefinition of
ISO and as a result, we have no idea what any of this really means.
Then the camera manufacturers got into the game by seeing how DXOMark
measures things, so they altered the processing engine slightly to
tweak how the sensor responds to the extreme thresholds. One tiny bit
of dithering noise changes everything.
A case in point on this: The lowly E-1, has a MASSIVE effective
dynamic range, because of how the distribution of noise occurs across
the range from the CCD as well as the inclusion of additional
dithering noise. The same era Canons (10D) had extremely low noise, in
comparison and had a longer real dynamic range, but a lesser effective
dynamic range. You can lift the shadows of an E-1 image, but can't do
that with the 10D image because of the lack of dithering noise results
in visible posterization. There are more modern comparisons today, but
those two example cameras really showed the difference in a huge way.
While, the E-1 lost out on ultimate discussion of high-ISO noise as
well as resolution, it did punch way above its weight when it came to
the visual quality of the color, contrast and overall tone of the
image. All stuff for which there is no measurement.
> If their measurements and ratings are clearly inaccurate for cameras I have,
> How can I put any credence in their ratings
> of other cameras? I accept that much larger, heavier, more expensive, FF
> cameras have better low ISO performance than my
> µ4/3 gear. But from the DXOMark data I have no idea how big the difference
> may be.
>
> I understand our culture's obsession with measurement, rankings and winners.
> But when the rankings do not reflect actual
> performance differences, I can't see how they are of use. What the H are they
> measuring? Why?
I don't think you can, unless your photography subjects and exposure
metering style match their exact methods and subjects. They don't
differentiate which colors are where the breaks occur, either. Nor is
there any real-world aspect to this. Granted, they've tried to include
something meaningful with the color depth thingy, but even that is
questionable. However, it is a data point. Just as DPReview's tests
are a data point, and as Zone-10's tests are a data point. And I look
to the Moose Tests as another critical data point.
--
Ken Norton
ken@xxxxxxxxxxx
http://www.zone-10.com
--
_________________________________________________________________
Options: http://lists.thomasclausen.net/mailman/listinfo/olympus
Archives: http://lists.thomasclausen.net/mailman/private/olympus/
Themed Olympus Photo Exhibition: http://www.tope.nl/
|