1Ds Mark II, 16.7MP

[BJL]

[font color=\'#000000\']

The common estimate is that difraction reduces the MTF to zero at 1600/f and to 50% at 800/f. ... The Nyquist frequency is thus 69 lp/mm. But dSLRs show a limiting resolution more on the order of 2/3 the Nyquist frequency, or 46 lp/mm.

This MTF50 is a more refined approach than mine. For f/16, I get 50% MTF at 50lp/mm, and the 7 micron pixels also give 50% MTF at about 50lp/mm. The combined effects involve products of MTF factors, so at f/16 and 50lp/mm, the MTF is 50% of 50%, or 25%.

Thus, at f/16 you have lost a very noticable one stop of contrast relative to what the sensor alone is capable of at low aperture ratios. That suggests that visible degradation of fine detail will start somewhat before f/16, but I cannot tell from this how noticable it will be at f/7.

Even with that degradation, the resolution is probably comparable to about 11MP with no diffraction limitation, so still quite usable!

In fact, once processors get fast enough, I envision the combination of very small pixels and diffraction for low pass filtering, preventing moire and such without the need for the expensive added AA filter. For higher ISO settings, downsampling to a lower pixel count could then be used to restore about the same S/N ratio as using that lower number of photosites to start with.

This would be similar to over-sampling in digital audio.


P. S. I just checked Norm Koren's figures, and he is more optimistic about the 50% MTF threshold, giving about 65lp/mm for 7 micron pixel pitch. The aperture matchng that is about f/12, so it seems that at f/12 (rather than f/16), you have lost half the MTF that you get at very low apertures, out at 65lp/mm. I feel better about the idea that a small visible loss of resolution starts as one goes above f/7.

Interestingly, Norm Koren rates the Canon 24-70 at about 61lp/mm 50% MTF at f/8 (its optimal aperture?), so all the resolution numbers come together to suggest that 16MP is close to the useful limits of 35mm frame format.

Indeed, theoretical arguments like this some years ago lead to the suggestion that 16MP is about the useful limit for 24x36mm frame format (except by using good primes only, or for doing my "oversampling" trick).

Will Canon now settle into improving in other directions, like dynamic range?[/font]

[davidjl]

[font color=\'#000000\']"I'm assuming that the real world useful limit for the vast majority of available lenses is far less than 16 MP"

Well, sort of. The thing is, though, that even if the lens is the weak link in the system, a higher resolution sensor will still produce a better image than a lower resolution sensor.

The reason is that MTFs combine multiplicatively*.

Of course, this is very much a matter of diminishing returns with funky lenses or disoptimal f stops. But the higher res sensor will always look as good or better in a print of the same size.

So as long as the cost (i.e. the presumably higher noise assuming a comparison of sensors with equivalent technologies) isn't a problem, even when one has to use disoptimal lenses and f stops for practical reasons, you really do want the higher res sensor.

And I'd think that any Canon prime would work very well with the 16.7MP sensor at it's optimal f stop.

*: I first noticed this in the comparison review (on this site) of the 17-35 and the 16-35 zooms on the D30. Even the low-res D30 sensor saw improved image quality with the better lens.[/font]

[davidjl]

[font color=\'#000000\']"Most 35mm lenses perform optimally at around f11 and you start to visibly lose resolution at f16 or more due to diffraction."

Hmm. I tested all four (35 to 150mm) of my Mamiya 645 lenses (tripod, ugly but detail-rich cityscape, Provia 100F), and with an 80x microscope, I could maybe persuade myself that f/22 was worse than f/16. Maybe.

If someone had both a 6MP and an 8MP 1.6x camera and a prime lens, they could shoot a series from f/5.6 to f/22 with each lens and print them all at A4.

I'd be seriously surprised if f/16 with the 8MP camera didn't look better than any f stop with the 6MP camera.

If I'm right, this would demonstrate that 16.7 MP is not too many MP, since 6MP is about the same pixel pitch as 16MP, and 8MP is about the same pixel pitch as 20MP full frame, on the simple logic that if there's room for improvement one hasn't gone too far.

(I have the impression that you are grasping at straws for a reason to object to the 1DsM2. But I really think that other than the price and weight (both of which are seriously objectionable, sigh), it's quite the right thing.)[/font]

[gwarrellow]

[font color=\'#000000\']EOS magazine link to the new EOS 1DS

http://www.eos-magazine.com/System_EOS1DsIIspec.html

They also provide the following info:
Price: £5999.99 (UK); 8699.99 EUR (Europe); both prices include VAT
Availability: November 2004[/font]

[Jack Flesher]

[font color=\'#000000\']David:  Your rough math matches mine, which is why I asked the question.  IMO there is a substantial difference between f7 and f16.  

If we take BJL's 1.22 at 7.2 microns and go the full f8.8 then take David's actual f17.4 (800/46) calc, we're still essentially two full stops apart...

So which is correct?[/font]

[BJL]

[font color=\'#000000\']

even if the lens is the weak link in the system, a higher resolution sensor will still produce a better image than a lower resolution sensor.

The reason is that MTFs combine multiplicatively.

Exactly, to balance my mild caution about diffraction limitation, increasing one factor in resolution (such as sensor resolution) will always increase overall resolution, even when it is already better than other factors.

As an example, suppose that with some combination of aperture ratio, pixel pitch and lens, you have diffraction, sensor resolution and lens resolution all giving MTF at 60lp/mm of 50% (one half, a one stop loss). These roughly fit the 1Ds mark II at f/12 with the Canon 24-70.

The combined MTF at 60lp/mm is then only 1/8 (12.5%), two stops worse. So clearly, improving one or two of those factors (say the sensor and lens) can greatly increase resolution at 60lp/mm, even if not all factors are improved (say diffraction due to staying at about f/12).[/font]

[davidjl]

[font color=\'#000000\']"it's very hard to "prove" ..."

Sorry: I seem to have misconstrued your earlier note. Rereading it, I really shouldn't have been so snappy.

I'm being enthused about the 1DsM2 on the theory that the sensor may appear in a camera less damaging to the travel budget/easier on the back.  In practice, I object to 24x36mm cameras that are substantially more awkward than my 645 or a 'Blad, and would probably pass on it whatever the price. So I'm not considering buying it at all, just talking. In the long run, I would like to replace 645 with 24x36 digital (scanning's a pain). But I suspect that's still a long way away. (There's still a rumor about a 12MP "2D", though...)[/font]

[russell a]

[font color=\'#000000\']Is it just me drinking "wishful thinking water" or does the view into the front of the iDs Mark II in the dpr Review appear as if the sensor isn't buried as deeply as the iDs?

http://www.dpreview.com/articles/canoneos1dsmkii/page2.asp[/font]

[russell a]

[font color=\'#000000\']Of course it's the same distance within the camera, I didn't make myself clear.  It just appears that the frame around the sensor is not as deep as the 1Ds or it may be a trick of the lighting.  When blowing off the sensor, I always hold my 1Ds upside down to limit crap floating down on it.  If they could address the CF card door access, it seems they could address sensor cleaning.[/font]

[mattlamb]

[font color=\'#000000\']All the math is very interesting and I will be thinking about it as it relates to Velvia when I am shooting with my 6x17 pano camera as I often stop way down ...

Re : 1Ds mark II diffraction, I would think that the worries are related to real world use of medium format backs as they have lower res lens (than 35mm L series)and  from memory the Kodak 22mb back is like a square 35mm ?
So if the 22mb back is better than the 11mb back all should be fine with the mark 2 [/font]

[BJL]

[font color=\'#000000\']

Medium format lenses are "lower res" then L series?

You obviously haven't used Zeiss or Schneider glass on a Hasselblad, Rollei or Contax MF camera.

The photodo tests at suggest that MF lenses have some what lower resolution than good 35mm format ones in the technical sense of line pairs per millmeter (specifically, MTF at 40lp/mm) and perhaps in particular at the corners of the larger MF frame. However this does not directly relate to final print resolution, because
a) if you use the full MF frame, it has more "millimeters" than 35mm, adding more resolvable line pairs across the total image, and
 if you give up some of that extra frame size by using MF lenses with a smaller sensor, the worst, corner, performance is taken out of the picture.

So I can easily believe that MF lenses outdo 35mm in terms of "line pairs per picture height" over their full intended frame size, and it is possible (though less clear) that they can hold their own in lp/mm comparisons restricted to resolution within the 35mm image circle.

The latter is close to the practical reality that most digital MF users see; for example Michael's 16MP Kodak back uses only a central 36x36mm part of the 42x56mm 645 frame, and he often crops to an even smaller rectangle.


Is there a source of MTF curves as a function of distance from center for some medium format lenses?[/font]

[Gary Ferguson]

[font color=\'#000000\']A few more details are emerging. E-TTL II, 4 fps and an 11 shot buffer in RAW, a wireless facility as per Nikon, and it ships in November. No change to the 1Ds body style (at least looking at the front photograph), guess that's in part why it gets the evolutionary title of "Mark II" rather than a revolutionary "1Dt" or "2Ds".

Looks like Canon are still the masters of camera marketing, although the buffer size increase isn't quite the hike I was hoping for. The 1Ds manages 3fps with an 8 shot buffer in RAW, so with the 1Ds Mk II I'd still occasionally be taking an unplanned cigarrette break waiting for normal service to be resumed![/font]

[jd1566]

[font color=\'#000000\']I dunno if it's a dud..  Although waiting a few days will certainly clear the air of any mistery, I think it's real. Look at the embossed Canon signature on the prism head..
Also the numbers are not wild but within what people are expecting...

Chasseur look like they were looking at the next generation model!

Still it means we all have to go shopping for extra memory cards by the dozen!

:-)[/font]

[Christian]

[font color=\'#000000\']

The flyer everyone is reacting to is a bogus, according to native Japanese speakers and people fluent in Japanese.

They claim the advert even admits it is bogus.

It looks pretty real now...

http://www.dpreview.com/articles/canoneos1dsmkii/[/font]

[rokkitan]

[font color=\'#000000\']A more detailed preview at Rob Galbraith: http://www.robgalbraith.com/bins/multi_pag...cid=7-6459-7231[/font]

[BJL]

[font color=\'#000000\']

One warning about the 7 micron pixel pitch of the EOS-1Ds Mark II; at apertures smaller than about f/7, diffraction will start to reduce resolution noticably. Exteme high resolution and great depth of field are in conflict when pixel counts gets this high.

Care to share the math on that?

My math is very rough, but it does sem to match experiment in previous cases. For example, Norm Koren puts the resolution of a film like Provia at equivalent to 8 to 10 micron pixel pitch, and around f/8 to f/10 is indeed the rough limit beyond which difraction starts to be a noticable limit on resolution.

I definitely recommend experiments though: has anyone measured th f-stop beyond which diffraction limitation sets in on various DSLR's when using very sharp prime lens?


Onto the rough math:

Diffraction spot diameter (out to the first null of the diffraction pattern, since you asked) is about
2.44 * (wavelength of light) * (aperture ratio)
(the Airy disk diameter: see Norm Koren's site, or any optics text.)

The wavelengths of visible light is around 0.5 microns (slightly more for green in fact), so
diffraction spot diameter is about 1.22 * (aperture ratio).

Thus if the aperture ratio is roughly equal to pixel pitch in microns (like my example of 7 microns, f/7),
diffraction spot diameter is about 1.22 * (pixel pitch in microns).

So diffraction is smearing the image of each point of the subject out over a region a bit bigger than a pixel. Bayer sensor resolution is not quite as fine as pixel pitch due to interpolation and such, but it seems that with aperture ratios significantly higher than this, diffraction will reduce resolution.[/font]

[davidjl]

[font color=\'#000000\']"Care to share the math on that?"

Here's my shortcut: The common estimate is that difraction reduces the MTF to zero at 1600/f and to 50% at 800/f. 3300 pixels over 24mm is 138 pixels per mm. The Nyquist frequency is thus 69 lp/mm. But dSLRs show a limiting resolution more on the order of 2/3 the Nyquist frequency, or 46 lp/mm. In other words, if you can provide decent contrast (MTF) at 46 lp/mm, you'll have as nice images as you could ask for. That means 800/f = 46, or f = 16.

So my call is that you _may_ see some degradation due to diffraction moving from f/11 to f/16, but f/11 will be fine, f/16 probably fine as well, and only by f/22 will you begin to be noticeably unhappy.[/font]

[BJL]

[font color=\'#000000\']Why do we have to do any maths on anything? All that matters is, is it better than the 1Ds?
Sample images look promising. ...[/font]

[font color=\'#000000\']Of course in the end, the RIGHT sample images can answer all relevant questions (and my math IS very rough, so I will not comment further until I do a more refined calculation).

The samples I would want to compare are a sequence, varying f-stop from about f/5.6 up to f/16 or beyond. Otherwise, the possibility is there that they can make the images look good, but only by avoiding certain DOF challenges.

Actually, I would be happy to see such a test done on any good DSLR, with a lens good enough not to be to much of a factor.[/font]

[didger]

[font color=\'#000000\']

even if the lens is the weak link in the system, a higher resolution sensor will still produce a better image than a lower resolution sensor.

Yeah, I'm sure this is true.  The big question is how much better an image, and what are the aperture limitations.  In the end, whether the new camera is a good investment is a complex formula of how much money you can spare and how big the improvement will be for the lenses you'll be using and what kind of shooting you do.  Those of us that only do landscape shooting will have to be much more careful in making a decision, since we have to use small apertures for greatest possible DOF and since most landscape shooters also use wide angle lenses a lot and these lenses are the weakest (unless you can the a rare 18 or 21mm Zeiss distagon).

And I'd think that any Canon prime would work very well with the 16.7MP sensor at it's optimal f stop

I don't know if this is necessarily true.  Most 35mm lenses perform optimally at around f11 and you start to visibly lose resolution at f16 or more due to diffraction.  The diffraction effect cuts in at lower f stops than f11 with a sensor with smaller pixels.  I don't know how real life important this is, but as you say, there's the issue of diminishing returns for less than optimal lenses and apertures.

Having said all that, if I had very large financial resources, I'd go for the new camera in any case, especially since all my lenses are world class and I'm sure I'd see some genuine improvement with 16.7 MP.  However, at this point I don't want to spend my travel money and end up with an improved camera and no money to go out and shoot.  11 MP isn't all that bad.  There's still plenty of folks that do great landscape photography with slides and make good money at it too.[/font]

[didger]

[font color=\'#000000\']

I really shouldn't have been so snappy.

Hey, I'm enough of a forum veteran (including some pretty rude ones) that your comment didn't even register on my "snapometer".

I'm also not thrilled about the weight of the 1ds, but for what I want to do (digital for sure) it's still pretty much the only game in town in spite of the weight and a few other warts.  I can't think of anything that's even remotely comparable quality that would allow me to take a camera and 4 lenses and tripod and batteries and a solar panel out for about two weeks of backpacking.  I'll be most interested to see if a 12MP 2d or 3d ever appears and how much it'll weigh and cost.  Nikon's new 12MP is pretty attractive, especially since they're supporting the format with lenses particularly optimized for it.  However, the price and weight are not all that attractive.  For now I'll keep on keeping on with what I have.[/font]