larger sensors

[John Camp]

If we are talking about a general purpose aspect ratio for all types of subjects and compositions, then the arguments in favour of 35mm are at least as strong as the arguments in favour of any other aspect ratio.
[a href=\"index.php?act=findpost&pid=93081\"][{POST_SNAPBACK}][/a]

This was what I was talking about. I don't think there is any escape in the near future from legacy lenses built for 35mm film. Given the image circle projected by those lenses, would there be other workable aspect ratios other than 2:3? I personally would prefer a "more square" ratio, like 3:4, or even 4:5, but not square. I'm not sure, however, if that is even possible given, say, Nikon 35mm legacy lenses -- as far as I know, the internal workings may be geared especially for the 35mm film aspect ratio, at least for larger sizes of sensor. I also think that if you surveyed most working pros, and actually gave them a chance to think about it, you'd find that if you could get 22mp in either a 2:3 or 3:4, that most would opt for the latter. But that's just what I think.

JC

[Ray]

I also think that if you surveyed most working pros, and actually gave them a chance to think about it, you'd find that if you could get 22mp in either a 2:3 or 3:4, that most would opt for the latter. But that's just what I think.
[a href=\"index.php?act=findpost&pid=93084\"][{POST_SNAPBACK}][/a]

That might be true and the evidence to support that is the prevalence of aspect ratios varying from square to 6x8cm in MF film cameras, with 6x9cm being less common. There's something to be said for the fact that a square gives you a greater capture area than any rectangle with the same diagonal.

On the other hand, pros tend to use (and can generally afford) the best tools for the job. If the client wants a high resolution panorama shot, the pro is (was) likely to use a 6x17cm panorama camera, In the absence of such wide aspect ratios in digital cameras, the best option might be to use, for all purposes, the camera with the biggest and highest resolving sensor, and crop to taste.

[bjanes]

Bt that's exactly what I've implied, Bill. At the end of the day, when the pixel count race is over and the minimum size photodiode for useful quality has been reached (and I believe Michael R has said this as around 5 microns), a FF 35mm sensor will hold more than double the number of pixels of a D2X size sensor and over 4x the number of pixels of an Olympus 4/3rds sensor.
[a href=\"index.php?act=findpost&pid=93075\"][{POST_SNAPBACK}][/a]

Yes, a 24 MP 35 mm full frame sensor could give excellent image quality at base ISO, but my argument was that 24 MP is more than can be made use of in hand held photography and many users would favor better high ISO performance and DR over the extra pixels. From what I understand, the Canon 1D M2 outsells the 1DsM2 by a large margin and most users of this type of camera are not cost constrained.

Bill

[Ray]

..but my argument was that 24 MP is more than can be made use of in hand held photography and many users would favor better high ISO performance and DR over the extra pixels. [a href=\"index.php?act=findpost&pid=93089\"][{POST_SNAPBACK}][/a]

I can't see it, Bill. It's certainly true that the bigger the enlargement, the faster the shutter speed required for a sharp print, but how much faster is an interesting question.

Supposing we start off from the 1/FL rule for an 8x12" print from 35mm. Let's say we're rather critical and demand 1/2FL. Let's assume also that from a purely resolution viewpoint, the 12mp 5D is equal to the best that 35mm film can produce.

A 24mp FF 35mm sensor will have 1.4x the resolution of the 5D. Does the rule then become 1/(2FL*1.4), ie. 1/2.8FL? Does that seem reasonable?

Consider a hand-held shot using a 100mm lens on a 24mp 35mm sensor. Without interpolation, at 360ppi, the 72mb file should produce a print approx. 11.5x17.25".

Our new rule gives us a shutter speed of 1/280th sec for good hand-held sharpness.

Let's suppose we are supercritical and demand nothing less than 1/3FL for a sharp 8x12" print and 1/(3FL*1.4) for an 11.5x17.25" print, because we're using the higher resolution of 360 ppi which is close to the limits of human perception. That gives us a shutter speed of 1/400th approx.

Let's further assume that the claimed 4 stop advantage of the latest Canon lenses with IS, such as the 70-200/4 IS, is a load of baloney, but 2 stops is very credible.

Our 1/400th then becomes 1/100th or back to 1/FL which seems to me a very usable shutter speed at ISO 100, and no problem at all at ISO 200-800.

Let's not create objections just for the sake of it   .

[John Sheehy]

I can't see it, Bill. It's certainly true that the bigger the enlargement, the faster the shutter speed required for a sharp print, but how much faster is an interesting question.[a href=\"index.php?act=findpost&pid=93103\"][{POST_SNAPBACK}][/a]

One thing to consider is that any camera motion blur will be "thinner" with higher MP counts.  Suppose you were shooting a star; a point of light.  The 6MP camera moves by 3 pixels during the exposure; if it were 24MP, it would move 5 or 6 pixels during the exposure.  The 24MP case would result in a streak of affected pixels the same percentage of the image dimensions, possibly slightly shorter, but it would also be thinner, affecting a much smaller percentage of the pixels in the image.  This might make it easier to ignore, psychologically, or make it easier for a deconvolution to correct it (especially if the blur is straight, which is likely in subject blur, as well, such as an unsuccessful pan).

[John Sheehy]

Let's not create objections just for the sake of it   .
[a href=\"index.php?act=findpost&pid=93103\"][{POST_SNAPBACK}][/a]

IMO, time will tell that the idea that more and smaller pixels filling the same sensor space causes more image noise will join the Flat Earth in the Mythological Hall of Infamy.  More and smaller pixels only increases the noise of individual pixels, against their neighbors.  When each pixel and its neighbors become less significant to the big picture, their contribution to image noise diminishes.  When the pixels outresolve the lens, there is no need for antialiasing filters, there are no demosaicing artifacts, and you can resample to any smaller size without artifacts.  You can toss your bokeh-killing TCs into the trash, unless you want them as an optical viewfinder aid.

[Jonathan Wienke]

Consider the effect a square aspect ratio would have using your widest 35mm lens. It simply wouldn't be as wide in either the horizantal or vertical plane.

Ray, you're making the unfounded assumption that making a square sensor is done by chopping off the ends of a rectangular sensor,  i.e. making a 24x24mm sensor out of a 24x36mm sensor. If you made a 36x36mm sensor and put it behind the lens, you'd get just as wide of coverage as you would with the 24x36mm sensor, and have the luxury of cropping to either a vertical or horizontal format in post from one RAW, or not cropping at all.

[BernardLanguillier]

Ray, you're making the unfounded assumption that making a square sensor is done by chopping off the ends of a rectangular sensor,  i.e. making a 24x24mm sensor out of a 24x36mm sensor. If you made a 36x36mm sensor and put it behind the lens, you'd get just as wide of coverage as you would with the 24x36mm sensor, and have the luxury of cropping to either a vertical or horizontal format in post from one RAW, or not cropping at all.
[a href=\"index.php?act=findpost&pid=93131\"][{POST_SNAPBACK}][/a]

Yes, but a 36x36 mm sensor requires an image circle slightly larger than that of a 24x36 mm, since the diagonal is longer.

Regards,
Bernard

[Jonathan Wienke]

One thing to consider is that any camera motion blur will be "thinner" with higher MP counts.  Suppose you were shooting a star; a point of light.  The 6MP camera moves by 3 pixels during the exposure; if it were 24MP, it would move 5 or 6 pixels during the exposure.  The 24MP case would result in a streak of affected pixels the same percentage of the image dimensions, possibly slightly shorter, but it would also be thinner, affecting a much smaller percentage of the pixels in the image.

Whoa, there, buddy, you need your morning coffee or something. If you're keeping composition constant (which you're assuming, given your statement about the blur trail being longer), then the width of the blur trail is going to increase by the the exact same degree as the length, which is in direct proportion to the increase in sensor pixels. All you're doing is recording the same blur with more pixels, which tends to limit the effectiveness of putting more pixels on the subject.

Upping the MP count for a given composition makes motion blur and camera shake more noticeable and visually objectionable, not less, at least when viewing at 100% in PS.

[bjanes]

Supposing we start off from the 1/FL rule for an 8x12" print from 35mm. Let's say we're rather critical and demand 1/2FL. Let's assume also that from a purely resolution viewpoint, the 12mp 5D is equal to the best that 35mm film can produce.

Let's not create objections just for the sake of it   .
[{POST_SNAPBACK}][/a]

According to the Leica expert [a href=\"http://www.imx.nl/photosite/leica/technics/faq.html#Anchor-What-47857]Erwin Puts[/url] your rule of thumb has never been verified. Erwin states that a higher shutter speed is required for best results. And we are talking about an acceptably sharp image, not maximal resolution at 24 MP.

What we really need here are resolution figures measured in the field from your 24 MP camera with hand held shots from several expert photographers. Unfortunately, I do not believe that such data are available. Why does Michael use a tripod with his P45 back?

Bill

[Jonathan Wienke]

Yes, but a 36x36 mm sensor requires an image circle slightly larger than that of a 24x36 mm, since the diagonal is longer.

If you shoot with a 36x36mm sensor and crop to 2:3, the net outcome is exactly the same as shooting with a 24x36mm sensor. You're working with an image circle of 43.27mm either way.

If you crop 36x36mm to 4:5, your effective sensor size is 28.8x36mm, and your image circle is 46.10mm. The real-world penalty for the use of the extra 1.415mm of image circle outside the lens' design will vary from lens to lens, but isn't generally going to be catastrophic.

If you want a square composition, simply leave a enough room around the edges when composing to allow cropping away corner ugliness. Using the center 30.59mm of the sensor when shooting a square composition will use the same 43.27mm image circle as a 24x36mm sensor.

[John Sheehy]

Whoa, there, buddy, you need your morning coffee or something.

I was about 1/2 way through cup #1 when I posted that.

If you're keeping composition constant (which you're assuming, given your statement about the blur trail being longer), then the width of the blur trail is going to increase by the the exact same degree as the length, which is in direct proportion to the increase in sensor pixels. All you're doing is recording the same blur with more pixels, which tends to limit the effectiveness of putting more pixels on the subject.

3 pictures are worth 3000 words:

Motion blur and megapixels

Upping the MP count for a given composition makes motion blur and camera shake more noticeable and visually objectionable, not less, at least when viewing at 100% in PS.
[a href=\"index.php?act=findpost&pid=93134\"][{POST_SNAPBACK}][/a]

Of course at 100% the higher MP will look worse in every way.  I don't see the relevance of this to the entire image.  Pixels are only relevant insomuch as they make their proportional contribution to the image.

[BJL]

I once shot a D200 to 1Ds2 comparison, photographing a guy in a Paris café, across the room. Holding the cameras with elbows braced on a table. In this low light the shake and noise and focus problems cumulatively *wrecked* the D200 shots while the 1ds2 was still running strong.
[a href=\"index.php?act=findpost&pid=93007\"][{POST_SNAPBACK}][/a]

Edmund, I have a basic scientific question about your experiment: what lenses, aperture sizes, shutter speeds, focal lengths and ISO speeds did you use? If you used a larger aperture size with the 1Ds2 (as would be the case if you used equal aperture ratio and lenses covering the same FOV), all you are showing is the well known speed advantage of larger aperture sizes in low light situations. There is also of course the issue of comparing camera and sensors using different technologies and of very different costs. As retired(?) engineer and forum participant Howard Smith has often said, comparisons are best done with only one parameter varied, not many.

[Jonathan Wienke]

I was about 1/2 way through cup #1 when I posted that.
3 pictures are worth 3000 words:

Motion blur and megapixels
Of course at 100% the higher MP will look worse in every way.  I don't see the relevance of this to the entire image.  Pixels are only relevant insomuch as they make their proportional contribution to the image.

If you resample everything to the same pixel dimensions, or print unequal MP images to the same print size, the effect of MP on motion blur is irrelevant, as long as the motion blur has a greater negative effect on resolution than pixel count, i.e. motion blur is at least 1 pixel. Your sample images prove my point.

More megapixels don't make motion blur "thinner" or less noticeable; at best, they make no difference.

[bjanes]

Whoa, there, buddy, you need your morning coffee or something. If you're keeping composition constant (which you're assuming, given your statement about the blur trail being longer), then the width of the blur trail is going to increase by the the exact same degree as the length, which is in direct proportion to the increase in sensor pixels. All you're doing is recording the same blur with more pixels, which tends to limit the effectiveness of putting more pixels on the subject.

Upping the MP count for a given composition makes motion blur and camera shake more noticeable and visually objectionable, not less, at least when viewing at 100% in PS.
[{POST_SNAPBACK}][/a]

After a  brief google search, I was able to locate a scientific paper relating camera motion and effective spatial resolution: [a href=\"http://scien.stanford.edu/jfsite/Papers/ImageCapture/ICIS06_CameraShake.pdf]Stanford University[/url]

After studying the article, please refer to Figure 5 for a computer generated simulation of the 50% MTF in cycles/mm for sensors with the same die size but different pixel sizes. The analysis is for monochromatic light and a diffraction limited lens at f/2.8 and a SNR of 30dB.

In low light situations, the small pixel camera does worse, since it requires a longer exposure and has more camera shake than a camera with larger pixels. Under outdoor conditions, the MTF for the 7.4 um pixels approaches  100 cy/mm asymptotically, and the 3.5 um pixel size approaches 120 cy/mm. The resolution does not double as expected because of camera shake. The best that 1.7 um pixels can do is 200 cy/mm.

The paper also confirms that the rule of thumb of the reciprocal of focal length in millimeters as a guide for the exposure required for a sharp picture is a very rough approximation at best. There have been very few studies of hand held camera shake.

If anyone has additional data, please post.

Bill

[eronald]

John,

Or as a focus aid -
My 1Ds2 with the Canon 85 is more focus-limited than resolution limited.

Edmund

You can toss your bokeh-killing TCs into the trash, unless you want them as an optical viewfinder aid.
[a href=\"index.php?act=findpost&pid=93121\"][{POST_SNAPBACK}][/a]

[eronald]

BJL ( is that your name ?)

As for it being a scientific experiment, I never claimed that - I was speaking strictly as a photographer

However, simple geometry does indicate that when cropping a given sensor you will worsen the effects of camera shake, and equally the effects of noise as you thereby increase print magnification. The ratios are left for the mathematically inclined to work out - readers of this thread seem to be quite numerate.

Simple reasoning also indicates that neither factor will show up when tests are effected in good light where neither sensor noise nor shake are a factor ...

But your remark is interesting, maybe someone should concoct a "camera shaker" for more realistic but still scientific tests.

Edmund

Edmund, I have a basic scientific question about your experiment: what lenses, aperture sizes, shutter speeds, focal lengths and ISO speeds did you use? If you used a larger aperture size with the 1Ds2 (as would be the case if you used equal aperture ratio and lenses covering the same FOV), all you are showing is the well known speed advantage of larger aperture sizes in low light situations. There is also of course the issue of comparing camera and sensors using different technologies and of very different costs. As retired(?) engineer and forum participant Howard Smith has often said, comparisons are best done with only one parameter varied, not many.
[a href=\"index.php?act=findpost&pid=93157\"][{POST_SNAPBACK}][/a]

[Morgan_Moore]

I cant see how sensor size affects CShake if you have made a suitable change in focal length to get the same FOV

Of course you can see the shake better with a higher res chip

If you think about shooting from a train it will still move the same % of the total image in the exposure time irrelevant of chip size

I think you experiment was duff

PS an earlier postor mention a large chip and reultant lack of DOF as a disadvantage - only for those looking for a big DOF !

SMM

[Ray]

If you shoot with a 36x36mm sensor and crop to 2:3, the net outcome is exactly the same as shooting with a 24x36mm sensor. You're working with an image circle of 43.27mm either way.
[a href=\"index.php?act=findpost&pid=93138\"][{POST_SNAPBACK}][/a]

Jonathan,
I'm beginning to wonder if it was you who missed out on your morning coffee when writing that, or perhaps the hangover was too painful   . You seem to have forgotten the most well-known axiom of Pythagoras; the square on the hypotenuse equals the sum of the squares of the other 2 sides.

The diagonal of a 36x36mm sensor is almost 51mm. The consequences of putting such a sensor in a 35mm body would be more vignetting and degradation of the image in the corners with existing 35mm lenses, not to mention the problems of mirror clearance.

Happy New Year   .

[Ray]

According to the Leica expert Erwin Puts your rule of thumb has never been verified. Erwin states that a higher shutter speed is required for best results. And we are talking about an acceptably sharp image, not maximal resolution at 24 MP.
[a href=\"index.php?act=findpost&pid=93137\"][{POST_SNAPBACK}][/a]

Bill,
Of course a rule-of-thumb is a rule-of-thumb. I never thought the 1/FL rule to be more than a statistical average for an acceptably (reasonably) sharp image with the 35mm format. If you are a Parkinson's sufferer, then the rule doesn't apply. If you're shooting from a moving elephant's back or a rocking boat, then it also doesn't apply. If you are shooting with a cell phone or P&S camera, the rule doesn't apply and, if you are a novice who is completely oblivious to the requirement to hold a camera steady for a sharp photo,. then the rule also doesn't apply.

You should have noticed that I tripled the shutter speed in my example, making a new rule of 1/3FL. Are you seriously suggesting that a shutter speed of 1/300th sec would not be adequate for a sharp hand held shot with a 100mm (non-IS) lens attached to a 5D?

If you accept that it would be adequate, then do you agree that using a higher resolving sensor will require an increase in shutter speed in proportion to the increase in resolution which, in the case of a doubling of pixel count on the same size sensor, amounts to a 1.4x increase?