Canon 1Ds replacement ... shooting the Canon 20D in Alaska

I just got back from 11 days in the backcountry, 9 days of smoke and rain, one
day of OK shooting on caribou and wolves, then yesterday was a wonderful day
with 4 hours shooting a grizzly at close range as it foraged nearby and 2 hours
shooting a wolf pack on a caribou kill at East Fork of Toklat river (Denali).
The kill was just a few yards from the bridge so we could shoot full-frame with
a 500 mm. Some days you stomp the grapes, some days you drink the wine.

One of the people with us for 5 days is a contract pro with Canon and had two
20D's with him, taking promo shots for Canon. He let us slip in a card and
shoot RAW+Jpeg (can only look at the jpegs until the RAW converters are
updated) and it was a pretty nice little camera from the images I saw,
definitely a step up from the 10D. I still prefer my 1D Mark II for faster
shooting and 1Ds for larger prints, but the 20D looks like a great camera.

He also mentioned that Canon will "soon" announce the 1Ds replacement, a 16
Megapixel model that will be pretty close to medium format in print quality.
From what he said it will be announced at the Sept photo show (Photokina?). I
wish he had one on this trip to play with

No guarantees, sometimes rumors don't pan out but this was from someone in a
position to know what's going on for real. So maybe hold off on the 1Ds
purchase for a few days to see what's announced.

Don't email me with questions, I'm back into the woods in two days to
photograph brown bears for a while. Life is good!

Bill

"Bill Hilton" wrote:

He also mentioned that Canon will "soon" announce the 1Ds
replacement, a 16 Megapixel model that will be pretty close
to medium format in print quality.

(IMHO, 16MP will be better than 645 for all practical purposes, no matter
how 645 is printed. Since 12 or 13.5MP would still be arguably not quite as
good as 645, I suppose there really is meaning to jumping to 16MP as soon as
possible.)

Hmm. I'd rather see 13.5MP (3000 x 4600) (or even 12MP (2830 x 4245)) and
lower noise than 3266 x 4900 (although 3300 x 4950 would be cute: 11x14 at
300 dpi).

I guess they're amused by the idea of 13x19s at better quality than 8x10s
from the 6MP camerasg.

The problem, though, is that 16MP requires 46 lp/mm at a fairly decent
contrast, and if you look at Canon's published MTF charts, they only go to
30 lp/mm, and lenses are often quite funky at the edges/corners even at 30
lp/mm. Especially wide angle lenses.

We'll probably see a lot of test shots taken with the 100/2.0 lens, one of
the very few lenses ever made that holds up out to the cornersg.

http://www.ocf.berkeley.edu/~ashon/photo/comparo6.htm

David J. Littleboy
Tokyo, Japan

"Bill Hilton" wrote:

He also mentioned that Canon will "soon" announce the 1Ds
replacement, a 16 Megapixel model that will be pretty close
to medium format in print quality.

(IMHO, 16MP will be better than 645 for all practical purposes, no matter
how 645 is printed. Since 12 or 13.5MP would still be arguably not quite as
good as 645, I suppose there really is meaning to jumping to 16MP as soon as
possible.)

Hmm. I'd rather see 13.5MP (3000 x 4600) (or even 12MP (2830 x 4245)) and
lower noise than 3266 x 4900 (although 3300 x 4950 would be cute: 11x14 at
300 dpi).

I guess they're amused by the idea of 13x19s at better quality than 8x10s
from the 6MP camerasg.

The problem, though, is that 16MP requires 46 lp/mm at a fairly decent
contrast, and if you look at Canon's published MTF charts, they only go to
30 lp/mm, and lenses are often quite funky at the edges/corners even at 30
lp/mm. Especially wide angle lenses.

We'll probably see a lot of test shots taken with the 100/2.0 lens, one of
the very few lenses ever made that holds up out to the cornersg.

http://www.ocf.berkeley.edu/~ashon/photo/comparo6.htm

David J. Littleboy
Tokyo, Japan

David J. Littleboy choreographed a chorus line of high-kicking electrons
to spell out:

We'll probably see a lot of test shots taken with the 100/2.0 lens,
one of the very few lenses ever made that holds up out to the
cornersg.

I've got one of those, get me a 1Ds II
As far as the lenses vs. pixels situation goes--well, one is a lot
easier to improve than the other.

--
__ (-o-) * A L L D O N E! B Y E B Y E!
(__ * _ _ _ _
__)|| | |(_)| \ "Mmm... unexplained bacon."

David J. Littleboy choreographed a chorus line of high-kicking electrons
to spell out:

We'll probably see a lot of test shots taken with the 100/2.0 lens,
one of the very few lenses ever made that holds up out to the
cornersg.

I've got one of those, get me a 1Ds II
As far as the lenses vs. pixels situation goes--well, one is a lot
easier to improve than the other.

--
__ (-o-) * A L L D O N E! B Y E B Y E!
(__ * _ _ _ _
__)|| | |(_)| \ "Mmm... unexplained bacon."

"David J. Littleboy" wrote in message
...
SNIP
The problem, though, is that 16MP requires 46 lp/mm at a fairly
decent
contrast, and if you look at Canon's published MTF charts, they only
go to
30 lp/mm, and lenses are often quite funky at the edges/corners even
at 30
lp/mm. Especially wide angle lenses.

Not quite. The charts only show a line for 10 and 30 lp/mm. We have no
information how these lenses perform at higher spatial frequencies,
but we do know that the sensor itself poses a physical limitation with
its native sampling density, its AA-filter, its fill
factor/microlenses, and its physical size. The modulation of lenses
tapers off gradually, but usually exceeds the sensor by far.

The 1Ds for example has a sampling density of 8.8 micron which poses
an absolute upper limit of 56.8 cycles/mm, and that is before the AA
filter reduces the modulation. A 'full frame' 16MP sensor would
presumably have a sampling density of 7.3 to 7.4 micron, with an
inherent limiting resolution of 68.5 - 67.6 cy/mm dictated by the
Nyquist frequency.

In current practice (small sample) this works out to something like
this:
http://www.xs4all.nl/~bvdwolf/main/f...SFR_Graphs.png (I know
it's too crowded with lines;-))
It shows the measured MTFs of a few cameras with decent lenses,
adjusted for equal output size by using the same criterion as DPreview
does; Line Widths per Image Height but based on a fourier analysis
instead of on haphazard alignment of bi-tonal (square wave signal)
patterns with the sensor array. The triangles are at the Nyquist
limit. Any significant modulation (say 10%) beyond Nyquist will
result in visible aliasing (10% of a 100:1 subject contrast will be
visible to the human eye).

I'm well aware of the shortcomings of such a limited test, but it is
more useful than theory only. The test was performed with the help of
the Imatest application (www.imatest.com).

Bart

"Bill Hilton" wrote in message

SNIP...of the stuff that makes me GREEN with envy (Alaska)

He also mentioned that Canon will "soon" announce the 1Ds replacement, a
16
Megapixel model that will be pretty close to medium format in print
quality.
From what he said it will be announced at the Sept photo show
(Photokina?). I
wish he had one on this trip to play with

I suspect that we have Nikon to thank for this relatively quick upgrade (if
it's really coming this month). This most likely means that Nikon is about
to release their own full-frame show-stopper to compete with the 1Ds, so
Canon is compelled to push forward once again. This is getting VERY VERY
interesting.

If it's true, I'll be waiting to see what happens to prices.

"Bill Hilton" wrote in message

SNIP...of the stuff that makes me GREEN with envy (Alaska)

He also mentioned that Canon will "soon" announce the 1Ds replacement, a
16
Megapixel model that will be pretty close to medium format in print
quality.
From what he said it will be announced at the Sept photo show
(Photokina?). I
wish he had one on this trip to play with

I suspect that we have Nikon to thank for this relatively quick upgrade (if
it's really coming this month). This most likely means that Nikon is about
to release their own full-frame show-stopper to compete with the 1Ds, so
Canon is compelled to push forward once again. This is getting VERY VERY
interesting.

If it's true, I'll be waiting to see what happens to prices.

"Bill Hilton" wrote in message

SNIP...of the stuff that makes me GREEN with envy (Alaska)

He also mentioned that Canon will "soon" announce the 1Ds replacement, a
16
Megapixel model that will be pretty close to medium format in print
quality.
From what he said it will be announced at the Sept photo show
(Photokina?). I
wish he had one on this trip to play with

I suspect that we have Nikon to thank for this relatively quick upgrade (if
it's really coming this month). This most likely means that Nikon is about
to release their own full-frame show-stopper to compete with the 1Ds, so
Canon is compelled to push forward once again. This is getting VERY VERY
interesting.

If it's true, I'll be waiting to see what happens to prices.

"Bart van der Wolf" wrote:
"David J. Littleboy" wrote in message
...
SNIP
The problem, though, is that 16MP requires 46 lp/mm at a fairly
decent
contrast, and if you look at Canon's published MTF charts, they only
go to
30 lp/mm, and lenses are often quite funky at the edges/corners even
at 30
lp/mm. Especially wide angle lenses.

Not quite. The charts only show a line for 10 and 30 lp/mm. We have no
information how these lenses perform at higher spatial frequencies,

Yes, but it's a good guess things are a lot worse at 45 lp/mm than at 30
lp/mm. Especially for normal and wider lenses. (Medium telephoto primes will
be fine all the way out to the corners, I suspect, even at 45 lp/mm. My
jaundiced view of photographic technology is due in part to my intest in
subject-free images of large spaces. Voyeurs, birdwatchers, and insect
molesters have more fun.)

but we do know that the sensor itself poses a physical limitation with
its native sampling density, its AA-filter, its fill
factor/microlenses, and its physical size. The modulation of lenses
tapers off gradually, but usually exceeds the sensor by far.

By now we know that digital systems cough up decent contrast at 2/3 the
Nyquist frequency, and to do that, the lens has to cough up decent contrast
at 2/3 the Nyquist frequency. And Canon wide angle lenses clearly won't do
that at the corners at 45 lp/mm. That's a reasonable inference from those
graphs.

The 1Ds for example has a sampling density of 8.8 micron which poses
an absolute upper limit of 56.8 cycles/mm, and that is before the AA
filter reduces the modulation. A 'full frame' 16MP sensor would
presumably have a sampling density of 7.3 to 7.4 micron, with an
inherent limiting resolution of 68.5 - 67.6 cy/mm dictated by the
Nyquist frequency.

Unlike you, I'm interested in _practical_ imaging. And I've looked closely
at the charts, and you're right: if you look at 800% on the screen, you can
see that, for example, the 1Ds really does resolve the charts well above 2/3
of the Nyquist frequency. But for making images of real scenes that people
will actually look at, stuff that you can only see at 800% really isn't
relevant.

So I take 2/3 Nyquist frequency, where, at 100% on the screen one can barely
count the lines in the chart, as a point where I need decent contrast.
Agonizing about what happens above that seems seriously pointless.

It seems to me that this "practical imaging cutoff frequency" is quite a bit
lower for film. Theory should, after all, reflect practical reality. And
film images look really grody at high magnifications.

By the way, I'm finding that with my eyes, 200 dpi looks pretty good at 10
or 12 inches, but doesn't hold up at 4 inches (whereas 330 dpi from 1Ds
images dog). But 200 dpi is a 17x enlargement from the 300D sensor.

My experience with film is that 17x enlargements are nowhere near acceptable
at just about any distance.

Of course, 200 dpi from the 1Ds sensor is more like a 14x magnification.
Film looks pretty poor at 14x, nowhere near as good as digital at 200 dpi.

So your data (graph below) that sharpened Provia "has a better MTF" than the
1Ds doesn't jibe with subjective reality.

In current practice (small sample) this works out to something like
this:
http://www.xs4all.nl/~bvdwolf/main/f...SFR_Graphs.png (I know
it's too crowded with lines;-))

Actualy, not enough lines. It would be interesting to see no sharpening vs.
aggressive sharpening for the 10D.

Interesting. Especially that the 20D without sharpening is the same as the
10D, but that sharpening really helps. I wonder if they're using the same AA
filter but that the extra pixels allow sharpening to rescue more detail.

But it fails to capture the perceptual experience that scans are really
really ugly. Resolution isn't worth doodly squat if you can't make pictures
with it, and the detail out at the tail of the MTF curve dies a hideous
death in the noise. And I'm not convinced NeatImage helps half as much as
you think it does.

It shows the measured MTFs of a few cameras with decent lenses,
adjusted for equal output size by using the same criterion as DPreview
does; Line Widths per Image Height but based on a fourier analysis
instead of on haphazard alignment of bi-tonal (square wave signal)
patterns with the sensor array. The triangles are at the Nyquist
limit. Any significant modulation (say 10%) beyond Nyquist will
result in visible aliasing (10% of a 100:1 subject contrast will be
visible to the human eye).

I'm well aware of the shortcomings of such a limited test, but it is
more useful than theory only. The test was performed with the help of
the Imatest application (www.imatest.com).

Yes. A neat idea (if I understand correctly): basically a single edge
transition gives you enough information to calculated the whole MTF curve.

David J. Littleboy
Tokyo, Japan