1D3: BETTER THAN THIS?

RichA wrote:
Mark² lowest even number here wrote:
Annika1980 wrote:
Here's an actual sized crop from a photo I took today:
http://www.pbase.com/bret/image/79898815/original

This was shot with my 20D at ISO 800.
This is a straight conversion from the RAW file with no other post-
processing done in Photoshop. To me, that looks like outstanding
low- noise performance from an ISO 800 shot. I want to see some
samples from the 1DMKIII at high ISO settings.
Preferably, something similar with a smooth OOF background where
noise would be obvious.

Get to it, PW!

BTW, I'd hate you less if you e-mailed me some RAW files shot with
your new cammy.

I just posted four 100% crops...800...1600...3200...6400.

There are jpegs, straight from the camera. I didn't use RAW
conversion, simply because that can introduce variables.
Personally, I am very pleased with the high ISO performance. -See
what you think...

**There is a bit of texture to the felt...so I'll probably have to
post another set with a better subject.

Noise becomes obvious at 6400, but is amazingly restrained at all
levels (basically non-existent at 800).
See below:

**NOTE: Each "original" size file is about 0.8MB...
800:
http://www.pbase.com/markuson/image/79933906/original
1600:
http://www.pbase.com/markuson/image/79933903/original
3200:
http://www.pbase.com/markuson/image/79933904/original
6400:
http://www.pbase.com/markuson/image/79933905/original

There is noticeably progressive detail diminshment going up in ISO,
but the results are still astonishing.
6400 properly exposed (note dark area) is eminently usable.

I think the detail loss you see is due to high ISO noise reduction.
I can turn that off...so I'll have to try it again with zero NR.
The seal photos were shot with NR turned off... and the 2000 ISO shots were
excellent...though there was SOME visible.
--
Images (Plus Snaps & Grabs) by Mark² at:
www.pbase.com/markuson

On Jun 4, 8:44 pm, "Mark²" mjmorgan(lowest even number
wrote:
RichA wrote:
Mark² lowest even number here wrote:
Annika1980 wrote:
Here's an actual sized crop from a photo I took today:
http://www.pbase.com/bret/image/79898815/original

This was shot with my 20D at ISO 800.
This is a straight conversion from the RAW file with no other post-
processing done in Photoshop. To me, that looks like outstanding
low- noise performance from an ISO 800 shot. I want to see some
samples from the 1DMKIII at high ISO settings.
Preferably, something similar with a smooth OOF background where
noise would be obvious.

Get to it, PW!

BTW, I'd hate you less if you e-mailed me some RAW files shot with
your new cammy.

I just posted four 100% crops...800...1600...3200...6400.

There are jpegs, straight from the camera. I didn't use RAW
conversion, simply because that can introduce variables.
Personally, I am very pleased with the high ISO performance. -See
what you think...

**There is a bit of texture to the felt...so I'll probably have to
post another set with a better subject.

Noise becomes obvious at 6400, but is amazingly restrained at all
levels (basically non-existent at 800).
See below:

**NOTE: Each "original" size file is about 0.8MB...
800:
http://www.pbase.com/markuson/image/79933906/original
1600:
http://www.pbase.com/markuson/image/79933903/original
3200:
http://www.pbase.com/markuson/image/79933904/original
6400:
http://www.pbase.com/markuson/image/79933905/original

There is noticeably progressive detail diminshment going up in ISO,
but the results are still astonishing.
6400 properly exposed (note dark area) is eminently usable.

I think the detail loss you see is due to high ISO noise reduction.
I can turn that off...so I'll have to try it again with zero NR.
The seal photos were shot with NR turned off... and the 2000 ISO shots were
excellent...though there was SOME visible.

I think the form factor of the sensor is a good compromise. Most of
the image quality (maybe as much because it's new?) of the FF with
less of the edge issues you see with WA film lenses with the FF's.
If you can image at 6400, apply some NR in post-process and lose a
minimal amount of detail compared to 800, you are way ahead of most of
what is out there right now.
For Olympus, I'd settle for a dead clean 400 ISO and a reasonably
clean 800.

frederick wrote in news:1180939485.812421@ftpsrv1:

I haven't worked it out, but pixel density of the 1dIII has to be about
the same as for the 5d,

The 5D has the same pixel density as the 1DmkII. The mkIII has a slightly
higher pixel density.

--


John P Sheehy

"Mark²" mjmorgan(lowest even number wrote in
:

I think the detail loss you see is due to high ISO noise reduction.
I can turn that off...so I'll have to try it again with zero NR.
The seal photos were shot with NR turned off... and the 2000 ISO shots
were excellent...though there was SOME visible.

These JPEGs don't tell much of a story, IMO. They can all be explained by
Digic III. One ISO 6400 RAW would tell all I need to know about the only
meaningful variable, IMO; RAW read noise. There's a big fat unexposed
strip on the left side of every Canon RAW that tells exactly what the read
noise is, both statistically, and in character.

--


John P Sheehy

John Sheehy wrote:
frederick wrote in news:1180939485.812421@ftpsrv1:


I haven't worked it out, but pixel density of the 1dIII has to be about
the same as for the 5d,


The 5D has the same pixel density as the 1DmkII. The mkIII has a slightly
higher pixel density.

Yes, but... (I forget the details...) but the m3 has some additional
tech to overcome that. Something about packing bigger pixels into the
space with less waste between them... someone will fill us in... also, I
believe more pixels does really give more detail & less noise... not per
pixel but per photo/print... probably at a small cost of dynamic
range... it all depends on how and what you count.

John Sheehy wrote in
:

"Mark²" mjmorgan(lowest even number wrote in
:

I think the detail loss you see is due to high ISO noise reduction.
I can turn that off...so I'll have to try it again with zero NR.
The seal photos were shot with NR turned off... and the 2000 ISO
shots were excellent...though there was SOME visible.

These JPEGs don't tell much of a story, IMO. They can all be
explained by Digic III. One ISO 6400 RAW would tell all I need to
know about the only meaningful variable, IMO; RAW read noise. There's
a big fat unexposed strip on the left side of every Canon RAW that
tells exactly what the read noise is, both statistically, and in
character.

Well, I got my hands on an ISO 3200 RAW from the 1Dmk3. Read noise is a
bit better than the 1Dmk2/20D/5D RAW noise (which are all about the same at
the pixel level). At 14 bits, the blackframe sigma is 24.3, or 6.08 in 12-
bit scale, as compared to the 9.2 or 9.4 in the other cameras, so it's
about 1/2 stop better. 6400 should be exatly double that, or 48.6 (12.15).
I got the RAW from a mk2/mk3 comparison, and the mk3, although having lower
statistical noise with generally low contrast, had more exceptional bright
pixels than the mk2 used in the comparison. It could be that the pixels
are not mapped out in that camera, though. The mk3 had less banding,
clearly.

--


John P Sheehy

On Jun 3, 11:23 pm, "Mark²" mjmorgan(lowest even number
wrote:
Mark² wrote:
frederick wrote:
John McWilliams wrote:
Mark² wrote:
Annika1980 wrote:
Here's an actual sized crop from a photo I took today:
http://www.pbase.com/bret/image/79898815/original

This was shot with my 20D at ISO 800.
This is a straight conversion from the RAW file with no other
post- processing done in Photoshop. To me, that looks like
outstanding low- noise performance from an ISO 800 shot. I want
to see some samples from the 1DMKIII at high ISO settings.
Preferably, something similar with a smooth OOF background where
noise would be obvious.

Get to it, PW!

BTW, I'd hate you less if you e-mailed me some RAW files shot with
your new cammy.

I just posted four 100% crops...800...1600...3200...6400.

There are jpegs, straight from the camera. I didn't use RAW
conversion, simply because that can introduce variables.
Personally, I am very pleased with the high ISO performance. -See
what you think... **There is a bit of texture to the felt...so I'll
probably have to
post another set with a better subject.

Noise becomes obvious at 6400, but is amazingly restrained at all
levels (basically non-existent at 800).
See below:

**NOTE: Each "original" size file is about 0.8MB...
800:
http://www.pbase.com/markuson/image/79933906/original
1600:
http://www.pbase.com/markuson/image/79933903/original
3200:
http://www.pbase.com/markuson/image/79933904/original
6400:
http://www.pbase.com/markuson/image/79933905/original

That's impressive. Your really slow ISO- 800, heh! shows just the
felt. The 6400 one, though, the speckles aren't just the felt. It's
noise, and maybe in five years we'll get noise free 6400, but what
you've got there now is fantastic.
==

It is impressive. But you can see from the loss of texture on the
surface of the ball and the threads, that some heavy NR is applied
in-camera to the jpgs.

That's likely true.
I'll have to shoot them again with all camera NR turned off and
compare.

That said... -For a bit of perspective, here's an image of the full frame
shot.
If this were printed at 8x12, I think you can imagine that it would be
extremely useable, and the noise would basically disappear:http://www.pbase.com/image/79936805/original

6400 ISO is entirely usable.

-Mark²

It takes balls to say that!

Paul Furman wrote in
. net:

Yes, but... (I forget the details...) but the m3 has some additional
tech to overcome that. Something about packing bigger pixels into the
space with less waste between them... someone will fill us in... also,
I believe more pixels does really give more detail & less noise... not
per pixel but per photo/print... probably at a small cost of dynamic
range... it all depends on how and what you count.

Well, what counts most is what you see.

I don't think that there is any loss of DR in the image, as long as overall
quantum efficiency isn't significantly undermined, and read noise
increases stay within very liberal limits. Having more noise per pixel and
more pixels can actually increase highlight capture. That's because larger
pixels or pixels that are binned will clip rather uniformly, while smaller,
noisier ones will dither in some extra highlight headroom. This can be
verified by binning RAW greyscale images 2x2 and comparing the results to
the original; the binned will have uniformly clipped, flat areas while the
original rolls in some extra headroom by having some pixels that do not
clip, when the clipping point is exactly 4x as high in the binned image.


--


John P Sheehy