[LRflex] Re: More contoversy!

David,
it bothers me a little that the 20D was set at 3200, this is more or=20
less a "pushed" setting - for those who don't know the 20D, the default=20
ISO eq. range is
set to a default of 100 - 1600,you have to set 3200 in the menu - the=20
typical "stone-wall effect" is very prominent in the darker parts of the =

Canon shot, but is also visible in the DMR shot just to the upper right=20
of the squirrels tail.
What I would like to see (or maybe I'll do it myself) is the difference=20
between the stop-shift method from 400 to 1600 ISO and true 1600 with=20
the Canon - just to see if there are any significant differences in the=20
s/n ratio at these settings.Maybe someone with a DMR could run a similar =

test too.
A credible  test for noise levels can't easily be made  between  DMR=20
and  20D, mainly due to the different  sensor  parameters, what we=20
really need are two versions of the same view with identical shooting=20
parameters and lenses, then subtract one image from the other to be able =

to see the noise part of the image as a "difference image", =20
(destructive interference method), I ran some tests on some shots some=20
time ago to test the effects of noise reduction, sharpening and=20
de-spotting (unfortunately now deleted, but the sharpening tests were on =

Xavier's shots of an old camera to compare NeatImage and a couple of=20
other sharpeners), the method was easy enough to run, even in Photoshop=20
Elements 3.0.
It may even be worth considering shooting without any lens on the camera =

at all, just to get  predominantly noise images for comparison purposes.
There may even be a spectral component in there somewhere, it seems to=20
me that the green wavelengths are quite strongly represented  in shadow=20
areas of the 20D shot, but less so in the case of the DMR
cheers
Douglas.
IIRC, Peter Werner did some testing with DMR against Canon quite some=20
time ago, with a yellow lily or iris flower, which were reasonably=20
comparable.


Philippe Amard wrote:

>Very fine pixes indeed David, thanks for sharing
>Only I have to now join, was it Jerry? who early on this thread said one=
=20
>should modify only ONE variable
>Until you manage to do this, and whatever your -proven - talents as a=20
>photographer and a presenter, there will always remain a flaw in the=20
>conclusions we may draw.
>
>For my part, i can't see the edge of the DMR over film, yes film, on the=
=20
>last two pixes.
>And if I had to choose one of the  last three posted, I'd say the=20
>Canon's best.
>Sorry about this.
>Your , was it grandson? in the anorak the other day, was much more=20
>telling than these two.
>
>Waiting for "analytical" comparisons.
>Enjoy the weekend.
>Mine will be gray again I'm afraid
>Phileicang=E9nieux
>
>
>David Young wrote:
>
> =20
>
>>Just when you thought it was safe to go back into the computer/camera r=
oom!
>>
>>OK.  We've talked before about the "work around" that Ted seems to hate=
=20
>>so much, for the high ISO noise problem with the DM-R, as opposed to th=
e=20
>>simple cranking up the ISO with the Canon 20D.
>>
>>I now offer you three photographs.
>>
>>The first two were taken within minutes of each other, from the same=20
>>vantage point, same DM-R (I can't afford two!) and with the same lens..=
=2E=20
>>the Telyt 400 f6.8 and 2x converter.   On the first shot (a squirrel)=20
>>the ISO was set to 400, with the exposure compensation set to -3 stops.=
 =20
>>The second shot (a woodpecker) was also at ISO 400, but with the=20
>>exposure compensation set to -2 stops.   Both shots were developed with=
=20
>>identical parameters, other than the exposure compensation, which was +=
3=20
>>and +2 stops, respectively... bringing them back to "normal" exposure. =
=20
>>No more. No less.  Thus, we have two shots with **effective** ISO speed=
s=20
>>of 3200 and 1600, respectively.
>>
>>As a "control", I have found a similar style shot (of a Rufous-naped=20
>>Wren), taken with the 80~200/f4 Vario-Elmar and 2x converter, on the=20
>>Canon 20D at ISO 3200... with no exposure compensation.
>>
>>This 'control' shot was "developed" with Silkypix, with identical=20
>>parameters as the two above.=20
>>
>>All three were converted from RAW to .TIF files with Silkypix, and then=
=20
>>to .jpg files in PS7.  All three had the unsharp mask (131%, 0.2 pixel =

>>Radius and 0 Threshold) applied.  No other image manipulation of any=20
>>kind (other than to add copyright notices) has been done.  No noise=20
>>reduction software of any kind has been used on any of these shots.
>>
>>The large files are all in excess of 1mb, and may cause long download=20
>>times for dial-up users.  (Sorry!)  And a screen resolution of 1024x768=
=20
>>will make life easier.
>>
>>Noise is most visible in the background areas of a photo where large,=20
>>out of focus areas exist.  All three shots meet this requirement, thoug=
h=20
>>the background colours do change a bit.  I'm sorry, Jerry, but this is =

>>not as scientific as either you or I would like, but it's the best I=20
>>could do, from photos at hand.  I think it still tells a valid story!
>>
>>I make no claims about superior noise levels. Certainly, the DM-R at IS=
O=20
>>1600 is not a usable camera.  However, to my eye, both the shots at the=
=20
>>*effective* ISOs of 1600 and 3200 are noticeably superior to the Canon =

>>at ISO 3200.   Moreover, I think all three (which could all be further =

>>improved with Neat Image or Noise Ninja) have less noise than film at=20
>>the same ISO!
>>
>>All of this fallderall can be seen at:=20
>>http://www3.telus.net/~telyt/Noise.htm
>>
>>Once again... Nomex suit on... Kevlar vest in place.
>>
>>You take a look.  You tell me what you see.
>>
>>Cheers!
>>
>>=20
>>
>>   =20
>>
>
>
>------
>Unsubscribe or change to/from Digest Mode at:
>    http://www3.telus.net/~telyt/lrflex.htm
>Archives are at:
>    http://www.freelists.org/archives/leicareflex/
>
> =20
>

------
Unsubscribe or change to/from Digest Mode at:
    http://www3.telus.net/~telyt/lrflex.htm
Archives are at:
    http://www.freelists.org/archives/leicareflex/

Other related posts: