21MP 1Ds Mark III Announced

On Aug 24, 4:36 am, frederick wrote:
Rita Ä Berkowitz wrote:
acl wrote:

Lol - I think so. The big surprise is that it's only $500 more than
the 1DIII, and the famed 17-35 also just got gazumped by new AF-s
f2.8 14-24.

That 14-24 looks great! I hope now prices of used 17-35mm will come
down to earth...

They will briefly drop in price when the market gets momentarily flooded
with them. Once Canon users realize that the new 12-24 and 24-70 are G
lenses and won't work on EOS bodies the price of the 17-35/2.8 are going to
soar into Noct range. I know I'll be buying several more 17-35s when they
get nice and low.

According to Ken, prices for the new lenses a
AF-S NIKKOR 14-24mm f/2.8G ED, $1,799.95
AF-S NIKKOR 24-70mm f/2.8G ED, $1,699.95
AF-S NIKKOR 400mm f/2.8G ED VR, $ $8,799.95
AF-S NIKKOR 500mm f/4G ED VR, $7,899.95
AF-S NIKKOR 600mm f/4G ED VR, $9,499.95

You are IMO an optimist on the 1DIII retaining value, and also the
17-35. Nikon users replacing 17-35 with 14-24 will outnumber Canon
users wanting to use the lens on 5d etc by far.

The 16-9.net tests show the Nikon 17-35 to be no better than Canon's
own 17-40 anyway. I had considered using the Nikon on my 1Ds2, but my
own tests supported the 16-9 results.

I also tested the Canon 16-35 Mk1 and Mk2 and still ended up staying
with the 17-40.

On Aug 23, 5:55 pm, Rita Ä Berkowitz ritaberk2O04 @aol.com wrote:


Though the Nikkors are much better in build and optical quality the Canon
equivalents are a bargain for the casual shooter.

Rita's ass is talking again!

"Doug McDonald" wrote:
David J. Littleboy wrote:

The D300, on the other hand, is nuts. Too many pixels in too small a
space. If they had used the D3 sensor and the same FF/DX switching with,
say 3 fps and 6 fps, they'd have a 5D killer, forcing Canon to get off
their butts and release a 16MP 5DII.

The 5D-II should be 1t least 20 megapixels

8*1.6^2 = 20.5

I think not. That's the 1DsIII pixel count.

The 5D had the same pixel size as the 1DII, so I'd expect the 5D to have the
same pixel size as the 1DIII.

The 1DII - 5D - 1Ds system was 8 - 12 - 16.


So 1DIII - 5DII - 1DsII should be 10 - 16 - 21.

David J. Littleboy
Tokyo, Japan

Rita Ä Berkowitz wrote:
frederick wrote:

That 14-24 looks great! I hope now prices of used 17-35mm will come
down to earth...

They will briefly drop in price when the market gets momentarily
flooded with them. Once Canon users realize that the new 12-24 and
24-70 are G lenses and won't work on EOS bodies the price of the
17-35/2.8 are going to soar into Noct range. I know I'll be buying
several more 17-35s when they get nice and low.


According to Ken, prices for the new lenses a
AF-S NIKKOR 14-24mm f/2.8G ED, $1,799.95
AF-S NIKKOR 24-70mm f/2.8G ED, $1,699.95
AF-S NIKKOR 400mm f/2.8G ED VR, $ $8,799.95
AF-S NIKKOR 500mm f/4G ED VR, $7,899.95
AF-S NIKKOR 600mm f/4G ED VR, $9,499.95

You are IMO an optimist on the 1DIII retaining value, and also the
17-35. Nikon users replacing 17-35 with 14-24 will outnumber Canon
users wanting to use the lens on 5d etc by far.

The 1D Mk III will hold its value for a while. I'm not going to defy fate
by keeping it more than 18-months. The new 14-24/2.8 is going to be a
winner with the people that stay with the D200/300 that want to go wide.
Plus, you are going to see a lot of Nikon shooters grabbing up the 17-35.
If the new 14-24 were a D lens the Canon shooters would go crazy over it
and
make the 17-35/2.8 worthless.

I disagree.
Already some 1DIII users are complaining that Canon has
dropped them deeply in the brown sticky stuff. Some because
the announced 40d with higher burst rate and features then
the 30d will do for them what the 30d wouldn't have done -
when they think they just spent over $3k that they didn't
need to, some complain because they wanted a D3 equivalent
from Canon. Your 17-35 on a 1d ain't like a 17-35 on a 5d
with the crop factor. The 1d is a sports cam, and I'm sure
a very fine one it is too.
A d200-300 user isn't likely to really want the 14-24. For
starters it's got integral hood and no filter thread, and
isn't as wide as the 12-24 that costs half as much, which
isn't as wide as the Sigma 10-20 which costs half as much
again. If the f2.8 really matters and the pockets are deep
enough to finance it, then surely a d3 is within reach.
Thom Hogan's comments:
"D300 is pegged at an initial production of 60k units a
month, D3 is pegged at an initial production of 8k units a
month"
The D300 is only one of four (or is it 5 or 6?) current
Nikon DX sensor cameras. The full frame market is expected
to be small - if you an call anticipated 100,000 units per
year of a pro camera small.

"David J. Littleboy" wrote in message
...

"Rita Ä Berkowitz" ritaberk2O04 @aol.com wrote in message
...
David J. Littleboy wrote:

He's currently correct when it comes to the dSLR body. I'll give
Nikon a year to totally trash the Mk III.

ROFL. They're still nowhere close to even the 1DsII. And that's been
more than 3 years. And won't be until they go FF.

I never mentioned the 1Ds. You're correct that Nikon won't be competing
on
the FF field. Nikon will make a decent 1D Mk III killer.

ROFL again. They remain nowhere close to the 1DII. The D2x is a joke for
sports and PJ work.


ROFL, then every camera used for sports before digital must be a joke, too!
And yet they got good photos anyway!

--
www.mattclara.com

"frederick" wrote in message
news:1187852244.906171@ftpsrv1...
David J. Littleboy wrote:
"frederick" wrote:

snip

The D3 is out of my price range, but the D300 looks like the camera I've
been waiting for. I'm glad I skipped the D200, stuck with my D70 and
spent what I would have done on lenses instead.

Don't chuck the D70 just yet: you may find yourself better off using D70
images straight than noise reducing and downsampling D300 images in low
light.

If Nikon made the D300 with worse noise than the D200, then IMO that's a
mistake if the option was to stay at 10mp and improve noise performance.
But my guess is that it's going to be good (but no - not a 5d). Take a
D2xs sensor and add a few years of development, and it darned well ought
to be improved.
And no - the D70 has been excellent, but it's time to move on. At about
30,000 clicks, Even if I trashed it now, I've saved more than 5x what it
would have cost me in film and processing, and the results have generally
been better, sometimes much better than I ever got from 35mm. I shoot raw,
and use an R1800 for printing. Viewing my old Cibachrome collection shows
me very clearly how much things have moved on in a relatively short space
of time.

The D70's great for IR photography, whereas the D200 and, I assume, the D300
are not. Keep it if you can afford to.
People worry that adding the extra pixels to the sensor will make the D300
noisier, but if DPReview is right, the D300 uses a CMOS instead of CCD, and
thus may alleviate that problem to some degree.

--
www.mattclara.com

acl wrote:
On Aug 23, 7:31 am, "Roger N. Clark (change username to rnclark)"
wrote:
Ilya Zakharevich wrote:
[A complimentary Cc of this posting was NOT [per weedlist] sent to
acl
], who wrote in article .com:
What someone who knows how to parse DNG files can do, is write a program
that does this within DNG files, for any camera with DNG.
For me, C is 2 lines a day, so I won't do it .
No C needed - if dcraw can MERGE a header from one DNG file with a
TIFF of "very raw" data.
a) Convert DNG to TIFF;
b1) use -fx of ImageMagick to do the transformation; or
b2) convert to .txt via ImageMagick, use any scripted tool to edit
.txt, convert back to TIFF;
c) Merge the modified TIFF into the DNG file.
(I have no idea whether 'c' is available).
Hope this helps,
Ilya
If you convert to tif, you are not seeing the raw data,
but the results of the interpolation algorithm to interpolate
the RGB values. Converters tend to average adjacent pixels
in the interpolation and I typically see and 1.6x lower
noise in tif images from true single pixel raw data.

You need to extract one color pixel at a time from the raw file
and not interpolate it to a tif image.

Here's the manpage for dcraw,
http://www.pochtar.com/jb/dcraw.man.htm
dcraw -D -T should give just the raw data in a tiff file (and it does).

Oops. Sorry, I was on the road. I haven't used dcraw in a while.
Do you then separate the red, green and blue pixels in your analysis?
Each would result in different signal levels therough their
respective color filters.

Roger

Ilya Zakharevich wrote:
[A complimentary Cc of this posting was sent to
Roger N. Clark (change username to rnclark)
], who wrote in article :
I see no reason why any explanation is needed. According to
http://www.clarkvision.com/imagedeta...mance.summary/,
table 4, 1D mII has read noise about 17 electrons at iso100. Add to
this electron noise, and you get noise about 20 electrons.

With full well close to 64K, this gives noise of 20 units.
log2(20)4; There is practically no need to have more than 16-4=12
bits for storage. (Correct appropriately for mkIII; I do not know
relevant numbers.)

The reason the noise appears high in the low iso camera data
is not limitations of the sensor, but limitations of the A/D
converter. At 14-bit converter will help about 1 bit, but
is still limited by the A/D.

Do not think so. With full well of 64K, 12-bit converter has
quantization noise (IIRC) about 6electrons. With 17e noise of
amplifier, the total noise is 18e. So quantization noise is not
measurable when (honest) 12bits are delivered.

Try actually looking up the specs of A/D converters that run
in the 50 to 100+ megahertz range. There is a delicate balance
between speed, power, and precision (meaning in a low power
camera, it is a challenge). E.g., see:
http://www.analog.com/IST/SelectionT...n_table_id=124
12-bit A/Ds get in the 63-70dB SNR range, while 14-bit converters
rise to 68-78 dB SNR. 12-bit A/Ds in the current Canon cameras are
delivering about 70 dB. From Canon's statements, the 14-bit
converters are delivering about 12.6 bits (~76 dB).
(Ideally, 12 bits = 72dB; 14 bits 84 dB.)


Look up some A/D specs. As the number of bits increases, the noise
at the bottom increases too and you don't gain linearly for fast
A/Ds needed in DSLRs.

Have no idea what you talking about... I think you mix up
quantization noise with the amplifier noise.

See above.

It looks like we really need 16-bit converters.

??? What is needed is, e.g., two 11-bit converters working in
parallel, one with variable gain (governed by the ISO setting),
another at maximal possible gain. Collect both numbers. There is no
need for many "in one basket" bits...

That strategy might work if you could accurately match
the two gains so the intensity versus signal does not have
a kink in it. No easy feat.
Fast A/D 16-bit converters deliver 78-82 dB SNR, and would be
a simpler solution and would likely require less total
power than than 2 12-bit A/Ds plus compute power to combine
the two signals.

The sensor read noise of the 1DII is 4 electrons
and the full well about 80,000 electrons.

Anyway, for art photography, there is little reason to require read
noise below 3e. And 3e must be accessible today with the scheme
above; the price is a (few?) mm^2 of the die... Have no idea why they
did not implement it yet...

Probably because of above. 14-bits is cheaper and easier.
Scientific chips use 16-bit converters for the above SNR
reasons. The fast 16-bit converters would deliver about
14-bits of clean performance, which is what we need for these
sensors.

Roger

On Aug 24, 5:05 pm, "Roger N. Clark (change username to rnclark)"
wrote:
acl wrote:
On Aug 23, 7:31 am, "Roger N. Clark (change username to rnclark)"
wrote:
Ilya Zakharevich wrote:
[A complimentary Cc of this posting was NOT [per weedlist] sent to
acl
], who wrote in article .com:
What someone who knows how to parse DNG files can do, is write a program
that does this within DNG files, for any camera with DNG.
For me, C is 2 lines a day, so I won't do it .
No C needed - if dcraw can MERGE a header from one DNG file with a
TIFF of "very raw" data.
a) Convert DNG to TIFF;
b1) use -fx of ImageMagick to do the transformation; or
b2) convert to .txt via ImageMagick, use any scripted tool to edit
.txt, convert back to TIFF;
c) Merge the modified TIFF into the DNG file.
(I have no idea whether 'c' is available).
Hope this helps,
Ilya
If you convert to tif, you are not seeing the raw data,
but the results of the interpolation algorithm to interpolate
the RGB values. Converters tend to average adjacent pixels
in the interpolation and I typically see and 1.6x lower
noise in tif images from true single pixel raw data.

You need to extract one color pixel at a time from the raw file
and not interpolate it to a tif image.

Here's the manpage for dcraw,
http://www.pochtar.com/jb/dcraw.man.htm
dcraw -D -T should give just the raw data in a tiff file (and it does).

Oops. Sorry, I was on the road. I haven't used dcraw in a while.
Do you then separate the red, green and blue pixels in your analysis?
Each would result in different signal levels therough their
respective color filters.

Roger


I tried both separating them and not (for blackframes the CFA
shouldn't matter!). Each channel seems to have different whitepoints
in my D200, for example. Anyway, no difference whether I split or not.

Unfortunately, these days I am finishing up writing a few things, so
this is going slowly (not at all at the moment, in fact...).

[A complimentary Cc of this posting was sent to
Roger N. Clark (change username to rnclark)
], who wrote in article :
Do not think so. With full well of 64K, 12-bit converter has
quantization noise (IIRC) about 6electrons. With 17e noise of
amplifier, the total noise is 18e. So quantization noise is not
measurable when (honest) 12bits are delivered.

Try actually looking up the specs of A/D converters that run
in the 50 to 100+ megahertz range. There is a delicate balance
between speed, power, and precision (meaning in a low power
camera, it is a challenge). E.g., see:
http://www.analog.com/IST/SelectionT...n_table_id=124
12-bit A/Ds get in the 63-70dB SNR range, while 14-bit converters
rise to 68-78 dB SNR. 12-bit A/Ds in the current Canon cameras are
delivering about 70 dB.

IMO, using dB with video data is a NONONO. Raw video data measure
"intensity". Raw audio data measure "square root of intensity".
("Intensity" is what is additive when [unrelated] sources are
combined.)

So the normal convention for audio data is that 10dB is 10x increment
in intensity; so 2x increment in raw data is 6db change. This is
convenient and leads to very little confusion.

Using same rules, 2x increment in raw data of video should be a 3dB
change. So 12-bit would be 36dB, not 72dB. Since some jerks use a
misplaced analogy with audio signal processing, and assign 6dB to 2x
change, dB in video context became nothing but a source of confusion.

IMO, we should use bits, and avoid using dB whenever possible.
(Although it is hard to do when quoting, as you did before.)

Hope this helps,
Ilya