I've recently been experimenting with HDR and exposure blending using a 5D II and Photomatrix. You're right that it's easy to do - taking the shots is dead easy with a good tripod; working out the optimum settings for blending requires some effort.

The problem is that it's often impossible to avoid some subject movement between shots. One of my favourite subjects is Tokyo streetscapes, which as you can imagine are almost impossible to shoot without getting at least a few people in the frame, who have an annoying tendency to move around. Even if there are no people around, things like lanterns swaying in the breeze can be a problem.

A camera with an extra 4 stops or so of dynamic range would open up a lot of photographic opportunities. A medium format digital back might manage it, but I don't think I could justify the currently exorbitant cost. I'll probably buy one when the price becomes more reasonable, as I'm sure it eventually will.

That will be a difficult war to fight... since DR is anyhing but obvious as it manifests itself only in the shadows once correct exposure is applied (ETTR). DR today is mostly a romantic concept.

For DR to be understood, we will first need to make sure that the concepts of correct exposure in digital is understood. For example, all the photoraphers who believe that the device they are using as some sort of headroom towards the highlights has probably not put in place the required basis to understand what DR in the digital world means. Even the Fuji cameras are not exceptions to this rule. This will remain so as long as our sensors stay linear devices.

At the risk of looking again like a D3x fan, I will restate the fact that the D3x is significantly ahead of competition DRwise, and yet is assessed by many key reviewers as being very similar to its competitors.

The reason being that most raw converters apply a default curve to the raw data so as to generate a pleasing result. A pleasant result will determine the extend to which a user will want to keep using that particular raw conversion software and is therefore very important.

A pleasing result is one that has enough contrast, in this case global contrast since raw converters for DSLRs do not apply automatically local contrast enhancement, although their highlight/shadow recovery tools often do.

So the bottomline is that good or bad DR can only be assessed by looking at the shadows of a file after application of various methods to lift them.

Pretty often, tapping into the actual DR of a camera like the D3x can only be truly done by using some form of tone mapping, something that few photographers are willing to do on a regular basis.

So unless raw converters become smarter at making DR available, real DR we will remain difficult to sell beyond its romantic appeal. Look at how little recognition Nikon has gotten for their outstanding job with the D3x.

Cheers,
Bernard

I totally agree that improved DR in a DSLR is a very difficult concept to sell to anyone other than a professional photographer who understands the benefits.

The marketing departments of major camera manufacturers know this - they also know its far easier to market a camera as 'new' and 'improved' because it has 6 mega pixels instead of 5. Thats something every consumer understands on the basic level that '6' is better than '5' (hey this one has volume that goes to 11!).

Its the same in the display market - the marketing dept. of major display manufacturers knows that the purchaser understands that 'X' display has more pixels than 'Y' - but the punter most likely wont understand the benefit of a larger gamut; even though the larger gamut will give them a better more accurate picture. Hence, gamut is rarely if ever mentioned in a displays specifications.

Ultimately this all means that the marketing dept drives manufacturing to a large degree (duh....). Which means the MP wars are not over (or at least not over as far as the marketing dept. wants you to believe). The marketing depts' will continue to flog this dead horse for as long as they can sell cameras. And improvements in DR will come as part of natural progression - but they will be glossed over by the marketing depts.

I think you're probably right, unfortunately.

One factor that a lot of people perhaps overlook is that improvements sensor in resolution are becoming increasingly difficult to take advantage of in practice. Even a small amount of camera shake, focusing error, subject movement, or optical aberration can reduce a 21 megapixel photograph to less at 12 megapixels of information. With each improvement in resolution, the photographer has to take extra care (and perhaps use better lenses) to get the benefit from it.

Improved dynamic range, on the other hand, would not require any improvement in photographic technique to get improved results.

Considering that many MFDB owners boast about the highlight headroom of their sensors, I am really not sure that being a pro is a sufficient condition to understand this. The truth is that many of the top shooters have developped their skills in the film days and measure light with their back the way they used to measure with film, meaning that in essence they go for the 18% grey card or measure incident light on the subject.

The digital device they will like is the one providing them a safety margin in these conditions, meaning perceived highlight headroom, which means under-exposes relative to an ideal ETTR digital approach. The back histogram and raw conversion software (mostly proprietary with backs) is calibrated in such a way that this impression is maintained. Put it otherwise, the whole chain is calibrated to simulate negative film.

This is possible because of the good DR provided by backs, shadows remain reasonnably clean even with limited under-exposure that it not perceived anyway since the raw conversion software compensates for this when applying the camera curve.

DSLR like the D3x are typically calibrated in such a way as to behave like slide film, meaning that the histogram shows something that is extremely close overall to the actual level of clipping in the raw file. It results in totally clean shadows, but often un-satisfied users since blown highlights show up more often (and with less pleasing transitions on some of the DSLR due to a lower real bit depth). In a way DSLR manufacturers are over-optimistic about the actual abilities of their photographers.

Do pros needs to care or be aware of these aspects? Probably not, they are paid for their shooting, artistic and Mgt abilities, not for their understanding of the technology.

Cheers,
Bernard

Tonality, Tonality, Tonality.

Oh and Noise, DR, etc. Did I mention diffraction?

I own both crop and FF BTW....

I agree with you on those ones.



Diffraction is actually the same on crop and FF. That's because diffraction scales with *DOF*, not sensor size. For example, f/22 on APS-C has more diffraction than f/22 on FF, *but*, it also has more DOF. You only need f/14 on APS-C to get the same DOF as f/22 on FF. Then diffraction becomes the same.

Thanks; this needs to be said so many times to people stuck on comparing only at equal aperture ratio and equal ISO speed.

It would be nice to have a "format comparison myth-buster" web page hosted by someone credible, with this sort information.
Next item: why comparing cameras in different formats with lenses of different minimum f-stops at equal ISO speed and shutter speed and thus equal f-stop is not very illuminating ... (It's partly a DOF thing, again.)

However, a lot of landscape shots will be taken with everything at or close to infinity, so there will be no need to stop down for DOF.

And in that situation, the apertures are large enough that diffraction is not a problem. With SLR formats, diffraction only comes into play when one stops well down from wide open for the sake of DOF control. The practical rule of thumb seems to be that diffraction only has a significant effect on resolution when the aperture ratio is about twice or more the pixel spacing, so f/8 and beyond with the smallest current DSLR photosites (The 18MP EF-S 7D and 12MP 4/3). If you are worried about diffraction hurting the fine detail in your landscapes and your lens fails you as soon as you open up a bit beyond f/8, then a decent lens would probably be the first upgrade to consider!

I shoot landscapes and nature almost exclusively, and I practically never focus at infinity. Usually there's some sort of foreground or at least middle ground that needs to be in focus.

When I shot landscapes I used a 70-200 more than anything wider, I have only 2 pictures shot with my 17-40L. Now with stitching and urban landscapes I'm using a 100mm lens and f22 isn't enough. I don't care about the calculations or the physics or the tirades, on a crop camera I would have more problems with diffraction as I would still have to stop down.

pom, this should be very easy to understand without any fancy calculations: equal f-stop and equal FOV with a smaller format gives more DOF, so when getting enough DOF is the issue, you still have to stop down, but not to as high an f-stop. (If I may risk one calculation, about f/14 on EF-S would give as much DOF over the same FOV as f/22 in 35mm, f/35 in 645 ...)

So please stop comparing diffraction effects at equal, high f-stop: a smaller format never has to be used that way.

What? I wonder if you ever shot a landscape.

Yes, I have.

Yes, the compromise between DoF and diffraction is better with DX than it is with FX.

Cheers,
Bernard

I don't know what "tirades" you're talking about, but you don't need physics or calculations to know that f/14 on DX has the exact same DOF and diffraction as f/22 on FX.



No. You would not still have to stop down, because, again, f/14 on APS-C gives you the exact same DOF as f/22 on FF.



You are mistaken. The compromise is the exact same with every format size.

Anyone like to prove that you get more DOF with crop for a given FOV to the extent of a stop and a half? Or that for a given DOF the diffraction will always be the same?

For me the answer is yes, but your mileage may wary. Basicly, crop camera is a subset of a full frame. It usually offers you tighter pixels, faster operations and less expensive price. But if we are talking about image quality in general: yes, bigger sensor area with practically same quality of lenses and technology you will naturally get better image quality.

An exaggregated example of image caption area impact: If I'm looking www.shorpy.com I can usually spot 4x5 scans from 35mm even from 400pix thumbnails. More accurate colors, more accurate details, better everything.

A visual demonstration requires two cameras with different format sizes and different focal lengths, but the same sensor MTF (total number of pixels), the same OLPF MTF, and the same lens MTF (relative to L/PH, not lp/mm) at the f-number scaled by crop factor. Furthermore, when the MTF of each lens is plotted over defocussing in image space at a given spatial frequency (relative to picture height) at the scaled f-number, the curve must be the same for both.

Personally, I don't have any two camera/lens systems that meet those specifications. All the lenses that I own have sufficiently different OLPF and aberrations that it can skew the DOF results. If there is someone with a D300, D3, 400mm f/2.8, and 600mm f/4, my guess is they would meet the qualifications.

Fortunately, there are experts that have already measured all that stuff. They created DOF calculators. That way people like you and I can calculate DOF without getting results that are skewed by lens aberrations.

But since you do not believe in DOF calculators, there is nothing I, personally, can do to prove it to you.

With the slight correction that the DOF change is a factor 1.5x or 1.6x, only slightly over one stop, both are easily proven. But to do so online require using physics and some calculations, which you seem to disdain.

So either do some simple experiments (using crops on the same camera with a 1.6x change in if you wish), or read some basic optical text books.
For DOF, the experiment could be as simple as:
200mm, f/22 vs 125mm, f/14, same lens; crop to same FOV; compare equal sized prints.

But to recap some elementary and uncontroversial optical facts:
1. reducing focal length by factor 1.6 (to fit the desired FOV onto a 1.6x smaller sensor, say) with equal f-stop reduces the circle of confusion size (disc into which a point is blurred by OOF effects) by a factor of 1.6 squared (square of focal length: see any DOF formula).
2. It also reduces image size by the same factor, so enlarging to equal size enlarges the circle of confusion size on the print at each point to 1/1.6 as big as with the longer focal length: "1.6 times more DOF."
3. reducing f-stop by the same factor of 1.6 increases CoC size by factor 1.6 (standard DOF formulas again), and thus in combination with the above, it gives equal sized CoC on equal sized prints: equal DOF and OOF effects when focal length and f-stop are adjusted in the same proportion.
4. The diffraction spot size is determined only by f-stop and varies inversely with f-stop, and so in step 3, it becomes 1.6 times smaller at the focal plane when the f-stop is reduced with the shorter focal length. With the extra factor of 1.6 in enlargement to get an equal sized print, one gets equal sized diffraction spots everywhere.

Conclusion: adjusting focal length, aperture ratio and degree of enlargement in the same proportion given an equal sized final displayed image with equal sized circles of confusion at each point and equal sized diffraction spots: the same DOF/OOF/diffraction trade-offs.