Is the difference of DR on MFDB vs 35mm dslr discernible on print?

[BernardLanguillier]

Dynamic range should most definitely not be measured on a per pixel basis.  One perfectly valid way of increasing DR is to build a sensor with a great many noisy pixels, which then get averaged when making a print.  DXO properly corrects for pixel count in their DR measurements.

I guess that you mean that DR is affected by both pixel quality and enlargement ratio.

When performing a 12 images stitch with a d3x (3 rows of 4 images in vertical orientation), I am de facto simulating a sensor that is approximately 85x65 mm, and the enlargement ratio for a given print size will therefore be about 1.5 times smaller than that of - say - a P65+.

For a given print size, the per pixel DR of the P65+ would have to have to be 0.3 stop better to be able to deliver the same print DR achievable with a quick stitch. Spend a bit more time, do a 4 row of 5 images stitch and we are at 0.5 DR gap.

Cheers,
Bernard

[ErikKaffehr]

Hi,

I have that CD but I think it's not any longer available. It's also a couple of years P45 and D1s2, but I'm considering to make a couple of A2-s and have a look.

The "great shootout" is here: http://www.luminous-landscape.com/essays/back-testing.shtml

Best regards
Erik

Michael the websites "master" ;-) has done this very test. When you receive the cd it will be full of info and great raw files to compare... Enjoy!

[bjanes]

Dynamic range should most definitely not be measured on a per pixel basis.  One perfectly valid way of increasing DR is to build a sensor with a great many noisy pixels, which then get averaged when making a print.  DXO properly corrects for pixel count in their DR measurements.

I beg to differ. In the case you describe, you are trading dynamic range for resolution. Of course, the main rationale for an increased pixel count is to enable printing at a larger size and this is defeated by the lower resolution. No one buys a $30,000 MFDB to produce 8 x 10 inch prints--for 16 x 20 inch prints, you have to maintain per pixel performance. The standard definition of dynamic range is the full well capacity of the sensel divided by the read noise as shown here. This is on a per pixel basis. For a tradeoff between DR and resolution you should read about pixel binning. With 2 x 2 pixel binning in hardware 4 pixels can be combined into a superpixel and read with one read noise rather than 4. Thus far, such binning can be done only with CCDs and is usually restricted to monochrome images because of the complexity of the Bayer array with color. The latest Phase one backs have Sensor + technology, which enables such binning.

[Ray]

Dynamic range should most definitely not be measured on a per pixel basis.  One perfectly valid way of increasing DR is to build a sensor with a great many noisy pixels, which then get averaged when making a print.  DXO properly corrects for pixel count in their DR measurements.

DXOMark does both, and that's very informative. Pixel size is a significant factor in DR capability. It's interesting that the pixel size of the Canon 5D2 is about the same as that of the old 20D and 30D cropped formats, and consequently noise and DR is about the same, at the pixel level. (They are in fact very slightly better in the 5D2 because the 5D2 is newer technology, but nothing to shout about at the pixel level.)

Why do we need to know that? Because, for owners of a 5D2, the cropped format 20D and 30D have no advantage with respect to increased equivalent focal length. A 400mm lens on a 30D becomes effectively a 640mm lens in 35mm terms. An image from a 5D2 with 400mm lens, which is cropped to the same FoV as the 30D shot, also becomes effectively 640mm, and will be qualitatively the same in all respects as the 30D shot. Probably marginally better by a degree one would never notice on a print.

If you didn't already know this fact, and wondered about it, then DXO provide the answer. If you are an owner of a 5D2, you would have very good reasons to sell your old 20D or 30D.

Cropping images is a wide practice, isn't it? I frequently crop images. Don't you?

I believe the reason why this 35mm DSLR versus MFDB debate is so interesting is because of the huge price increase of the MFDB for what seems to be a relatively small image quality increase, in tandem with clearly worse performance in other areas such as frame rate, high ISO performance and general flexibility of use.

Futhermore, that image quality increase appears to be only applicable at base ISO on really large prints, which leads me to deduce that the only reasons anyone could justify the purchase of an MFDB system is for prodution of 30"x40" prints or larger.

Correct me if I'm, wrong. I once thought I was wrong, but then found I was mistaken.  

[Nick_T]

Futhermore, that image quality increase appears to be only applicable at base ISO on really large prints, which leads me to deduce that the only reasons anyone could justify the purchase of an MFDB system is for prodution of 30"x40" prints or larger.

I don't do big prints for a living. I take photos for a living.

I currently use an Ixpress 384 (multi-shot 16MP), a Nikon D3, an H3D31, and a 5D2.

In a typical week of shooting I'll use all four.

Nick-T

[Ray]

I don't do big prints for a living. I take photos for a living.

I currently use an Ixpress 384 (multi-shot 16MP), a Nikon D3, an H3D31, and a 5D2.

In a typical week of shooting I'll use all four.

Nick-T

Then please tell us in what circumstances you find the H3D 31 an advantage. I understand that 31mp on a larger sensor is likely to be better than 21mp on a smaller sensor, for large prints.

[BJL]

In the case you describe, you are trading dynamic range for resolution. Of course, the main rationale for an increased pixel count is to enable printing at a larger size and this is defeated by the lower resolution.

As has been said a number of times before:

The same photographer with the same camera might on different occasions prefer different trade-offs between resolution, shutter speed, noise levels, DR and such. So it can make perfect sense to get a higher pixel count camera because you sometimes want the extra resolution, but also on other occasions to trade away the extra resolution for the sake of better low light handling, increased DR or whatever.

We did that a lot with film cameras, by choosing different film types for different situations, or with push or pull processing.

So it makes sense to compare what, for example, a camera with 24MP sensor can do with images processed down to "12MP worth of resolution" against what a camera with 12MP sensor can do ... if 12MP is sometimes enough, but sometimes not.

And to repeat another point: film has billions of photosites each with pathetically low "per pixel DR" ... but no-one measure the DR or S/N ratio of film at the "per pixel" level

Also, no-one compares film cameras in different formats exclusively at equal degree of enlargement: larger formats often give better DR (fine tonal gradations etc.) by allowing the use of a lower degree of enlargement.

[Jack Flesher]

I shoot both a P45+ and a Canon 1Ds3 with regularity.  The easiest area where you can see the MF DR advantage is in any image with clouds.  With a typical exposure balanced for a subject lit by the outdoor light, on the Canon file the best you can do is get clouds that have mostly blown white areas except for the more obviously darker gray patches; IOW Canon clouds usually appear kind of DR posterized.  However with a similar shot taken with the P45+, you get clouds that look a lot more like real clouds, with smoother and broader transitions from the darker grays into the full whites -- and very often,even  the full whites are not clipped in all three channels as they are in the Canon file. And yes, it DOES show up in the final print regardless of size.  Not scientific, but empirical and I think you'll find most MF shooters share that experience...

Cheers,

[Ray]

And to repeat another point: film has billions of photosites each with pathetically low "per pixel DR" ... but no-one measure the DR or S/N ratio of film at the "per pixel" level

That would be very difficult and not at all useful, BJL. Those film 'pixels' vary in size and uniformity and tend to form in clusters and clumps, or dye clouds in the case of color film. However, the characteristics of film types per unit area were well defined using other terminology such as absolute resolution at a specific MTF, independent of any particular lens. For example T-max 100 B&W film would have a specification of 50 lp/mm at 100% MTF and 100 lp/mm at a still impressive 65% MTF.

One might deduce broadly from the DXO sensor tests that using a Canon 1Ds3, compared with a P65+, would be like using 35mm film compared with a similar but slightly slower film type on 645 format. However, that slower film type on the 645 format would also have the characteristics of rather poor 'push-processing' potential, whereas the slightly higher ISO film for the the 35mm format would have excellent push processing potential.

[BernardLanguillier]

I shoot both a P45+ and a Canon 1Ds3 with regularity.  The easiest area where you can see the MF DR advantage is in any image with clouds.  With a typical exposure balanced for a subject lit by the outdoor light, on the Canon file the best you can do is get clouds that have mostly blown white areas except for the more obviously darker gray patches; IOW Canon clouds usually appear kind of DR posterized.  However with a similar shot taken with the P45+, you get clouds that look a lot more like real clouds, with smoother and broader transitions from the darker grays into the full whites -- and very often,even  the full whites are not clipped in all three channels as they are in the Canon file. And yes, it DOES show up in the final print regardless of size.  Not scientific, but empirical and I think you'll find most MF shooters share that experience...

Jack,

What happens when you expose with your 1ds3 so as to avoid blown clouds?

Cheers,
Bernard

[rainer_v]

Jack,

What happens when you expose with your 1ds3 so as to avoid blown clouds?

Cheers,
Bernard

[Jim2]

Jack,

What happens when you expose with your 1ds3 so as to avoid blown clouds?

Cheers,
Bernard

I would assume he meant that all things being equal, the MF can still have enough DR to capture the details in the clouds while still having proper exposure for the shades. If the Canon was exposed to avoid the blown clouds, perhaps the details in the darker areas will get lost because he'd have to underexpose the whole scene?

[BernardLanguillier]

I would assume he meant that all things being equal, the MF can still have enough DR to capture the details in the clouds while still having proper exposure for the shades. If the Canon was exposed to avoid the blown clouds, perhaps the details in the darker areas will get lost because he'd have to underexpose the whole scene?

Maybe, but I haven't had such problems with my DSLRs since the d2x days, so I am surprised. I didn't mean the question to be cynical, I understand the value of shooting a camera without having to think about the highlights, but in the end it is just a matter of how the histogram is displayed and of real ISO value. Highlight headroom simply doesn't exist with digital, but some cameras are calibrated in such a way as to make you think that there is headroom. I would personnally prefer to know exactly when highlight blow.

Assuming that ETTR is a given, the shadows of my files are extremely clean and C1 Pro does a perfect job at filling them when needed. The difficult part can be to map this amazing DR in a pleasing way, but that is a different matter all together.

One recent photograph shot in these conditions (for what it is worth, I don't really like this image):



This was in fact a bit under-exposed (about 1/2 stop), but the shadows are still totally clean and full of details.

Cheers,
Bernard

[Ray]

I would assume he meant that all things being equal, the MF can still have enough DR to capture the details in the clouds while still having proper exposure for the shades. If the Canon was exposed to avoid the blown clouds, perhaps the details in the darker areas will get lost because he'd have to underexpose the whole scene?

I assume that also, but the D3X is a different kettle of fish   .  At any print size it's got about 2/3rds of a stop greater DR than the P65+. But that is not to say greater resolution in the midtones. (In fact, less resolution in the mid-tones, obviously).  Without the evidence of test results in front of my eyes, I would say that at any print size the D3X image would show greater detail in the deepest shadows with equal detail in the highlights. Or greater detail in the highlights with equal detail in the shadows, depending on ETTR accuracy.

[Jim2]

[deleted - probably irrelevant]

[ErikKaffehr]

Hi,

Thanks for info! From a technical viewpoint I cannot understand this but there are many hidden parameters in the process. Exposure bias in raw converters, different kind of histograms (like JPEG based histograms).

Best regards
Erik

I shoot both a P45+ and a Canon 1Ds3 with regularity.  The easiest area where you can see the MF DR advantage is in any image with clouds.  With a typical exposure balanced for a subject lit by the outdoor light, on the Canon file the best you can do is get clouds that have mostly blown white areas except for the more obviously darker gray patches; IOW Canon clouds usually appear kind of DR posterized.  However with a similar shot taken with the P45+, you get clouds that look a lot more like real clouds, with smoother and broader transitions from the darker grays into the full whites -- and very often,even  the full whites are not clipped in all three channels as they are in the Canon file. And yes, it DOES show up in the final print regardless of size.  Not scientific, but empirical and I think you'll find most MF shooters share that experience...

Cheers,

[Doug Peterson]

I assume that also, but the D3X is a different kettle of fish   .  At any print size it's got about 2/3rds of a stop greater DR than the P65+. But that is not to say greater resolution in the midtones. (In fact, less resolution in the mid-tones, obviously).  Without the evidence of test results in front of my eyes, I would say that at any print size the D3X image would show greater detail in the deepest shadows with equal detail in the highlights. Or greater detail in the highlights with equal detail in the shadows, depending on ETTR accuracy.

Can't imagine where you get that from, but it sounds like you need to shoot a P65+ and a D3X for a weekend so you can see how things come out when the rubber hits the road.

Doug

Doug Peterson
__________________
Head of Technical Services, Capture Integration
Phase One, Leaf, Canon, Apple, Profoto, Eizo & More
National: 877.217.9870  |  Cell: 740.707.2183
Newsletter: Read Latest or Sign Up
RSS Feed: Subscribe

[bjanes]

I shoot both a P45+ and a Canon 1Ds3 with regularity.  The easiest area where you can see the MF DR advantage is in any image with clouds.  With a typical exposure balanced for a subject lit by the outdoor light, on the Canon file the best you can do is get clouds that have mostly blown white areas except for the more obviously darker gray patches; IOW Canon clouds usually appear kind of DR posterized.  However with a similar shot taken with the P45+, you get clouds that look a lot more like real clouds, with smoother and broader transitions from the darker grays into the full whites -- and very often,even  the full whites are not clipped in all three channels as they are in the Canon file. And yes, it DOES show up in the final print regardless of size.

I doubt very much that the problem you are seeing in your cloud images with your Canon has to do with dynamic range problems. What are the raw pixel values in the brightest and darkest areas of these clouds?

Not scientific, but empirical and I think you'll find most MF shooters share that experience...
Cheers,

As long as this discussion remains in the realm of subjective impressions with not even photographic examples demonstrating the alleged differences, much less any quantitative evaluation, this discussion is going nowhere.

This statement is just as valid today as it was over a hundred years ago:

"In physical science the first essential step in the direction of learning any subject is to find principles of numerical reckoning and practicable methods for measuring some quality connected with it. I often say that when you can measure what you are speaking about, and express it in numbers, you know something about it; but when you cannot measure it, when you cannot express it in numbers, your knowledge is of a meagre and unsatisfactory kind; it may be the beginning of knowledge, but you have scarcely in your thoughts advanced to the state of Science, whatever the matter may be."  Lord Kelvin (Sir William Thomson) [PLA, vol. 1, "Electrical Units of Measurement", 1883-05-03]

According to Sir William, your knowledge is meagre and unsatisfactory. Why don't you post examples or much better, raw files? Then we could do some proper scientific analysis.

[Doug Peterson]

According to Sir William, your knowledge is meagre and unsatisfactory. Why don't you post examples or much better, raw files? Then we could do some proper scientific analysis.

Just out of curiosity, have you shot both a modern digital back and a modern dSLR?

I see so little argument on this topic out of anyone who has.

Doug

Doug Peterson
__________________
Head of Technical Services, Capture Integration
Phase One, Leaf, Canon, Apple, Profoto, Eizo & More
National: 877.217.9870  |  Cell: 740.707.2183
Newsletter: Read Latest or Sign Up
RSS Feed: Subscribe

[Jack Flesher]

Maybe, but I haven't had such problems with my DSLRs since the d2x days, so I am surprised. I didn't mean the question to be cynical, I understand the value of shooting a camera without having to think about the highlights, but in the end it is just a matter of how the histogram is displayed and of real ISO value. Highlight headroom simply doesn't exist with digital, but some cameras are calibrated in such a way as to make you think that there is headroom. I would personnally prefer to know exactly when highlight blow.

Assuming that ETTR is a given, the shadows of my files are extremely clean and C1 Pro does a perfect job at filling them when needed. The difficult part can be to map this amazing DR in a pleasing way, but that is a different matter all together.

One recent photograph shot in these conditions (for what it is worth, I don't really like this image):



This was in fact a bit under-exposed (about 1/2 stop), but the shadows are still totally clean and full of details.

Cheers,
Bernard

Here you go Bernard.  In your image above, clouds in bright sunlight, it is my opinion that your clouds are seriously clipped -- sorry, but they do not look like real clouds the way the grays so abruptly transition to your blown whites. Maybe if you had underexposed another full stop, you'd have captured the clouds better, but then your shadows would suck -- and that tradeoff is the entire point of this discussion.  (As an aside, your sky has the typical crappy Canon cyan-blue too. Granted, thats a different argument, but one I thought I'd mention.  By contrast, most MF backs get clear skies the correct color blue.)

Here is a shot with in similar bright sunlight I took earlier this year in Moab (actualy this is a stretch of the Petrified Forest) with my P45+.  Again, in my humble opinion only, these clouds look a lot more like real clouds with similar exposure values for our main foreground subjects and nice, clean detailed shadows. (And the sky is the proper blue.):



As to what happens if I underexpose the scene with my 1Ds3 to keep the clouds from clipping, that is a stupid question. The answer is the shadows block up to the point they look like crap and are unrecoverable.

Doug is right -- there is an awful lot of sanctimonious argument on this question from armchair "experts" that have NEVER even picked up an MF camera...  So buddyjanes, here is my post, now let's see yours from your DSLR that proves me so "wrong" -- or is this another one of YOUR typical meager and unsatisfactory armchair arguments?

Cheers,