OK it took me a while to understand that the merging with a blank photo
applies NEF compression in camera to the same original. I can see why
this is useful in his test because he does the layer overlay difference
comparison so each pixel needs to line up precisely. But... it doesn't
matter that much for my situation because I'm testing what can be
massaged out of the raw file and clearly seen without exaggeration. But
yes, it's possible I introduced a little camera shake in spite of the
tripod. I continue to maintain that his test is kind of meaningless
though if he's not forcing some highlight recovery adjustments. That's
the whole point is to be able to force more out of the file, not that it
would look any different at default conversion.

Alright, alright I tried it that way and while the file sizes are
slightly different I could not see any difference at all, so I'm
changing my mind & turning raw compression back on. Frankly I couldn't
figure out how he did the exaggerated overlay mask, normally I'd paste
one image over the other & set the top layer to mode 'difference' then I
might add curves or levels to exaggerate subtle differences but in this
case there was zero of anything at all to be seen. Pure black. I don't
know how he got what he got but it must be very very minor. Maybe he
used a different raw converter, maybe my dng conversion messed it up.
I'm using CS1 ACR. BTW I had to shoot the blank overlay twice in both
modes, using an uncompressed blank produced an identical sized file down
to the last KB. The blanks were shot with the lens cap at max 1/8000 second.

I'm going to try it my way again with mirror lockup, remote cable &
extra care not more the framing while changing the menu setting...

OK you win, there is no discernable difference. Using my overlay method
I could see a slight shift from punching the menus... my first test
apparently had camera shake from not using a cable release. So yeah, I'm
turning on raw compression. Maybe that other test showed some very minor
difference but if it's not perceptible under my somewhat extreme testing
conditions, it's not meaningful.

Let's see that comparison... I think no lens is immune to these effects
under extreme conditions and cheaper lenses would perform poorer.

The shadows can be greatly improved by pushing the highlights. With low
contrast scenes, huge improvements can be had. It depends on the photo
though.

On Jun 28, 1:54 am, achilleaslazari...@yahoo.co.uk wrote:

OK, I got curious so did a little experiment. I took this photo
http://www.pbase.com/al599/image/81277114
in uncompressed raw. Checking in raw magick (a raw converter), the red
and green channels are both blown somewhere (on the upper part), but
certainly most of the red and green raw values are above 215 (which
seems to be the value above which raw compression happens, at least in
the D70).

I then sent it to someone who has nikon capture, and he resaved it as
compressed raw (went down to 7mb) and sent it back, as well as the
converted version. Nothing was changed (except white balance, both set
to tungsten). I then overlaid the two files as layers (in 16-bit mode)
and used difference blending. Setting the whitepoint in photoshop at 2
(ie moved it from 255 to 2), I got this:
http://www.pbase.com/al599/image/81277117
(it's full size, click on "original" if it's not already).

Finally, I took the converted tiffs and sharpened them heavily (to
make slight differences more visible), and got this:
http://www.pbase.com/al599/image/81277115
(uncompressed) and
http://www.pbase.com/al599/image/81277116
(compressed).

I also checked with another converter (raw magick), same thing. Now I
did not do any highlight recovery, but if the data isn't there, I
don't see how using uncompressed or compressed raw can affect it
(after all, the only thing that can happen to non-existent entities is
for them to be brought to existence; this isn't going to happen, in
this case).

Some remarks:
a) Maybe ACR treats the raw files differently, and there is a bigger
difference if using ACR
b) Maybe nikon capture's conversion to compressed NEF uses a different
method than the camera.
c) maybe the highlight recovery algorithms of some converters change
depending on whether it's compressed or not
d) maybe what is described here
http://www.majid.info/mylos/weblog/2004/05/02-1.html
doesn't apply to the d200, and it does something else; this "else" may
be complicated enough that 1) it compresses the file by 50% with
almost no visible degradation, unless 2) we use the exposure
compensation slider set to -1, in which case the image softens.
e) other (or some combination of the above).