« Bestsellers Up to the Minute | Main | Quote o' the Day »

Saturday, 24 September 2011


Feed You can follow this conversation by subscribing to the comment feed for this post.

One aspect of this discussion, print size, presents questions.

What percentage of photographers are making poster size prints, say 30in X 40in on a regular basis, at least one or two prints a month?

Up to what size do you normally print?

This might be a good subject for a poll.

Worst chart ever.

The thing I think is most obvious in the comparisons is that the J1 image is clearly less saturated than the DSLR images. Reducing saturation at high ISO is a fairly common trick to reduce the appearance of color noise. I suspect the J1 would look a lot less impressive in comparison if you ran the raw data for the three images through an identical image processing pipeline.

I think that, for many of us, it's the equivalent depth of field differences that scare many of us away from smaller sensors. It would take an 18mm f1 lens on CX format to equal the focal length and depth of field equivalent of a rather pedestrian 50mm f2.8 lens on FX.

"you settle on a camera and then, after exploring its capabilities, you use it within the areas where its competencies overlap with your tastes, expectations, and requirements."
Exactly! Now back to taking photographs

I don't know if this is a good comparison but I equate the Nikon DX sensor cameras to 35mm film cameras and the Nikon FX cameras to 220 film cameras.

Does this make any sense?

I've commented before about how I chose the Leica M8.2. Simply, it met my camera selection criteria better than alternatives. I also mentioned that, for me, it's all about the print.

But, after choosing the M8.2, Leica presented the M9 option. The key distinction was of course full frame versus cropped sensor. (There are other differences, too, including use of UV/IR filters, but I'll ignore those here.)

So, I tried to determine if I might benefit from moving to full frame. The simple answer is that, after looking at prints up to 16x20, I didn't see enough significant differences to warrant the change.

I don't like to complicate matters.

Wow! Thanks for the Imaging Resource site, Mike.

From pixel-peeping my Oly E-PL1 vis-a-vis my Canon (original) 5D, it seemed to me that the image quality was very close, maybe even good enough to make the Oly my basic travel camera.

IR doesn't list the older 5D in their Comparotomer, so I checked the Oly against the Nikon D700, which has the same 12 megapixel file size as the 5D and the E-PL1. I could tell little, if any, difference in the image quality and actually preferred the file from the E-PL1. So thanks very much for the confirmation that I'm on the right track as I assemble a small and light travel kit.

I agree in a very loose way. If you step up big, files become a lot easier to work with. You can get tons of quality from smaller sensors. It's just a lot harder. Like trying to get 6x6 results out of a 35mm camera.
The other variable, user processing software knowledge. The small sensor files need a lot more massage to squeeze out a high quality print. Ctein has the knowledge and the will to do that, a lot of other people would rather be spared and happy to use a bigger sensor that just spits out something a lot closer to a finished print.

Another parameter that gets little mention is the "quality" of the noise in high ISO, low light pictures. I have always been very taken by the look of the "mush" that my old Fuji F30 creates in these situations; very painterly in an impressionistic sense where the noise in my 5DII is, well, just noise (not an apples to apples comparison I know!). Looking at the low light samples on DPreview the "mush" had the same Fuji look about it and is not, to my eye at least, displeasing. As for the designs of the new Nikons, er, did I mention I've just found an old Kodak box camera in my mums loft?


If I may, I have a very specific format comparison question for you:

My E-pl1 makes fantastic 10x13 prints.

If I want to make 30x40 or 24x68 B&W prints at similar quality (say 300dpi), should I use 4x5 film? Or are the the film size advantages washed out by technical factors (dof, vibration, scan quality), and I would get less numerical dpi, but similar quality from a few (two or three) stitched frames from a recent FF or APS-C camera? By similar quality I mean of the "overlapping bars" type you described.

I would generally agree that, for instance, APS-C is not very far from "full frame" in most respects (some more DOF, slightly more noise, slightly softer...)

Yet, speaking of "holistic", above those image quality characteristics, to me "full frame" translates into a significantly larger/brighter viewfinder and access to higher quality wide angle lenses. (One of the most important lenses in my bag is the TS-E 24mm II, so APS-C simply wouldn't cut it, even if IQ was identical.)

Those aspects, to name only two, likely have a more significant impact on my photography than the image quality differences.

I was wondering if you could explain that graph, as I find it more confusing than enlightening.

You mention that the x-axis is image quality (presumably as measured by the people polled). One guess at the length of the lines is that it represents the "spread" -- the amount of IQ assessed by different participants, or the difference in the averages for multiple prints, or ... ?

When I first looked at the graph, I assumed that each horizontal line was a different film format with the largest format on top and making its way down. But then you say

What I found, more or less, was that film formats (the "sensor size" of the film era) had a certain amount of overlap when mapped against enlargement size.
The phrase mapped against suggests that the vertical axis of your graph is the enlargement size rather than the discrete choice of film formats. What I would like to know is

1) What is on the vertical axis? Is it film/sensor size, or enlargement size, or something else?

2) What is the origin of the "fuzziness"? Does it come from the sample variation of responders, or is it from marginalizing over the variable mentioned in #1 that is not on the y-axis?

My best guess is that you have picked a fixed enlargement size for the final print, then queried the IQ for a set of prints and plotted format vs IQ with the horizontal lines given their length from the variation in responses. But it is not the only way I could read your text; it is not clear that this is really what you meant.

Mike, once again I have to agree with you. It just makes sense that larger sensors will produce better images. The D3/D700 is unbeatable at ISO 6400. And, don't forget the narrower depth of field associated with larger sensor size. I think that Ctein's analysis is correct from a theoretical standpoint, but in the real world, most pros are shooting FX and MF for a reason.

Wow! I know Canon is known for the "plastic skin" phenomenon, but looks like Nikon is just as bad!!!

(run and duck)

Most people choose their dSLR based on spec, which really is the wrong way to go, as most dSLR can produce images exceeding the ability of most photographers can produce. That includes ergonomic - you put an Olympus E-5 in the hands of a lifetime Nikon shooter and if that's all they have to use, then in no time they will manage the difference in control etc.

So what's the difference? To paraphrase Captain Reynolds of the good ship Serenity, it's love. The photographer must love photography, they must love what they do, where they are, and yes, what camera they are using.

So instead of spec, they should compare the love-factors :-)

Hey, you're on a roll. Even with the rudimentary graphics this made good sense. Of course the fuzzy boundary on the right is what we enthusiasts are always pushing. I shoot mostly landscapes with my 4/3 gear, which in some ways is silly (clearly larger sensors give more DR and resolution) but on the other hand it's still highly satisfying when I think I get an image just right.

Hi Mike


best wishes phil

"Careful photographers run their own tests."
- Fred Picker



Ctein knows his camera and adjusts his photography to fit the needs of the camera (lets it do what it does best). You Mike work from the opposite direction, you want versatility in the same machine. That's why you end up with a micro4/3 since these are very versatile camera's, not excellent in anything but great at a lot of things. Ctein knows full well (since he had a lot of marbles to begin with and hasen't lost to many of them on the way) that you can't compare a Canon G10-11-12...with a Hassy with Leaf Aptus snug behind it. But he also knows full well that in certain situations the G10 will get the better shot (try free climbing with a Hassy in you Rucksack). He also knows that given the right amount of light, the right amount of contrast and the right amount of photographic skills you can take great shots with a G10 that are hard to distinguish from a DSLR a A4 size. And he also knows that at A3 that becomes a little easier if the DLSR happens to be a full fram Nikon or Canon. At iso 800 things look decidedly different for the able Canon G10. Their the little camera loses his umpf and you wind up with a lot of artefacts in the .jpg. But if you go for .raw their are certain ways of overcoming the problems (external noise cancelation programs that work far better then the onboard computer of the Canon can provide, I know because Stein has written a book about restoration that I treat as my bible and there he devotes almost a whole chapter on noise cancelation). So it's like the Drake equation. Sensor size = skill x 1/iso used x depth of field needed x maximum image size on the wall x etc. ect. and that amounts to what Ctein so wisely states (IMHO):


But never the less, you both are right.

Greetings, Ed

Suspect the only item that matters:
did you obtain the image you wanted?

If yes, stop complaing.
If not, look elsewhere.

Scratch the noise - it's all about the depth of field :)

" If you step up big, files become a lot easier to work with. You can get tons of quality from smaller sensors. It's just a lot harder."

I used to say, WRT to black-and-white film photography, that the bigger the camera, the harder it was in the field and the easier it was in the darkroom; the smaller the camera, the opposite.


We only tested film up to 4x5 and 20x24 prints. For the sizes you're talking about, I'd probably suggest a 24-MP FF camera or one of the medium-format options. It depends greatly on the subject matter of the pictures and the printing skill of the person making the prints.


"Yet, speaking of "holistic", above those image quality characteristics, to me "full frame" translates into a significantly larger/brighter viewfinder and access to higher quality wide angle lenses."

Yes, but one topic at a time. This post is about sensor sizes and image quality, not camera features and usability. Of course you're right, though.


Don't read so much into the "graph." It doesn't contain any information. I'm just trying to get across the concept of overlap. That is, when you compare 6x9 enlargements from 35mm and 645, there's not a great deal of quality advantage for the larger format. But by the time you get to 11x14, the quality of 35mm is beginning to fall off and the advantage of 645 is getting more obvious. At 16x20, the larger format kills the smaller one decisively. That's all.


"So instead of spec, they should compare the love-factors :-)"

See my reply to Charles... :-)


Museums and galleries tending to exhibit larger and larger prints of contemporary work may be a pretty good argument for some photographers to use larger sensors.

Outside of the big-landscape-as-decor market, most museums and galleries I get to couldn't give a toss about resolution ... and neither could the public. This isn't because their standards are low, just the works on the wall are exhibited, appreciated and sold on the basis of other, and arguably more important, qualities.

IR's comparometer presents jpegs only. So, in fact you're looking at the performance of the respective jpeg engine.

The 5DII e.g. doesn't come across well in these, because a) it's a bit long in the tooth (here: jpeg software algorithms) now and b) jpeg probably didn't figure that large in the development of the camera in the first place.

As another example, Panasonic's m4/3 still have that nasty yellowish tint, which makes many grab the Oly counterparts. If you use RAW, it doesn't matter.

I don't see how Ctein and you are really in disagreement. It's just that Ctein is talking from a practical standpoint and you from a qualitative one. The point is that unless you are shooting in low light or printing large, or printing something with lots of detail, it just does not matter. Your IR samples totally back this up.

Looking at the CX shots compared to FX, I think its pretty obvious that you could make pretty nice A4 prints at quite high ISO and as long as the subject matter did not require background separation or massive resolution it could work just as well from an artistic POV.

I bought your recommended book on Haas. I have to say I am quite blown away by it. But there are few pictures in there that could not have been taken with a Nikon J1. In fact I daresay if it is as responsive as claimed it could be a great PJ and street tool.

Sure if you are a technical photographer working in advertising then things are different, but for artistic or documentary photography concept beats resolution any day. I have seen far more shots that needed more DOF than less, but were constrained by film speed to use wider apertures.

Having looked through IR's samples I have to say that high ISO performance of the CX is quite impressive. Better in fact than most MFT cameras. It's only low ISO performance that is a little disappointing and that may well be down to the use of the standard zoom.

I can see a V1 replacing my P&S camera. I can also see it replacing my D90 (after all my D700 fits in the same bag). MFT could not do both jobs, whereas a V1 and D700 would make an interesting partnership with genuine separation of function and functionality. I think I may be talking myself into one but I will wait for the reviews.

When the original Canon 5d came out, it had image quality that was far above what the aps-c cameras of the day could offer. It was the first digital camera I was comfortable producing work with. The 5d MkII is even better in most every regard.

For a variety of reasons, I began shooting a Nikon D7000 this year. It's a smaller sensor, but the dynamic range and resolution really is quite good despite being a much smaller (and cheaper) package. I prefer files shot with this camera to my files from the original 5D. In fact, in many ways, it's the equal of many aspects of the current full frame cameras.

There's a certain look to the 5D Mk II files that I find myself missing, although I can't exactly quantify that. The files appear crisper with a more pleasing color rendition. Something like that.

I think I'd be much happier if Nikon made a 35mm-ish DX prime. There's just so many more wide-ish lens choices for a full frame camera, even if they tend to be expensive.

john camp's note reminds me of why I prefer my personal portraits to be shot in black and white, from a very long way away...
Point might be, though, to what purpose we need the resolving power of an electron microscope mated to a photo sensor? What we gonna DO with it, really?

Are we now The Online Peepologist? :-)

Hendrik, that's an interesting comment because the newspaper journalists (and similar) I know here usually shoot in JPEG, either with Canons or with Nikons. Yes, they do tweak the photos but they usually cannot afford the time to develop RAW. So it's JPEG, tweak if needed, send.

I wish I could get background separation. But when I most need it, it's not an option. I've done about all I can -- I'm shooting a full-frame D700. Trouble is, at in-room distances and focal lengths that get people interacting, f/2.8 doesn't blur the background anywhere enough. Yeah, I know, I'm using slow zooms (well, fast zooms, to most people; slow lenses, which happen to be zooms, to me). Next time I have $6k sitting around I'll be sure to buy the 24-35-50-85/1.4 kit.

But in a lot of cases, I can't focus accurately enough to actually use f/1.4. Not when people are moving around a lot; certainly not in sports action.

So, for me, it's all about adequate shutter speed and low noise; background separation is a nice concept, but I can't afford to make that the top priority, or I'd never get a picture.

I agree with the depth of field comment. None of the small sensors, even APS-C in most cases, really gets me the bokeh I want. I've always been more of a landscape shooter where shallow dof didn't matter much, I usually shot f11,f16 etc. Small sensors are great in these cases. As I've gotten older, I am getting more into people pictures. Ability to work in low light and getting the nice out of focus areas trumps the resolution issues. I'd like these comparison sites to have a bokeh test.

I would like to echo what Hugh Crawfored noted but in a slightly different context. The other day, just for grins I decided to try some of my old M8 files with software which didn't exist when I owned that camera. I found I could do more with the files, but shockingly I discovered the limits of those files rather quickly. Nailing the exposure with the Leica was much more imperative than my current setup.

I should say that I agree 100% with Mike that there's no such thing as a neutral test. Neutral isn't especially useful either, unless it uses a workflow/software package similar to what the users do. dcraw may be reasonably even-handed, but the results still aren't near what people will actually get. It's also not realistic to try and tune output for every camera.

That said, JPEG noise reduction and sharpening settings do vary widely among manufacturers, more widely than most RAW converters in my opinion. I suspect that ACR default conversions for all their faults are a lot closer to what most folks in the pixel-peeping realm can expect.

Mike wrote:
Digital is similar, except there are at least three more complicating factors obscuring a clear result: ISO, image size (pixel count), and sensor generation.

I would add two more factors:

1) Sensor manufacturer.
2) Firmware interpretation of sensor data.

Currently, there is a lot of difference (at higher ISO) between a Sony APS-C sensor and the equivalent from Samsung, exemplifying point 1 above. As for point 2, it should be well known that Nikon and Pentax extract higher quality images (again, especially at high ISO) from Sony sensors that does Sony itself.

I've looked at some image comparisons between the Nikon mirrorless and a couple μ4/3 cameras, and the Nikon comes out on top at higher ISO despite the smaller sensor.

To paraphrase Einstein, not only is the sensor size issue more complicated than we imagined, it's more complicated than we could ever imagine.

John Camp: You are correct, we don't need increased resolution. We need better shaving materials!

I think the most interesting aspect of the new Nikon system is how big the lens are given the very small sensor size.

To me this says that it is hard to economically make very small, high quality, lens.

Point and shoots do it but they are dealing with a single lens so the engineering problem is less and also they can spread the upfront costs over a much larger number of lenses.

If this is true, then making the sensor small does not help - the size of the overall system is determined by the economics of lens production.

Shrinking the sensor reduces ISO and other aspects of quality and gives you little benefit in terms of system size.

Having a large sensor means you can have a single system that appeals to both high end and casual users - having a small sensor does not allow that. Many casual users will prefer to use the same basic system as professionals and other high end users even if they buy cheaper versions of the cameras and lenses.

I personally think APS-C sensors will end up being the dominant size for interchangeable lens systems because of all this.

Dear Miserere,

The quote you're thinking of was:

It is my supposition that the Universe in not only queerer than we imagine, is queerer than we can imagine. --

John B. S. Haldane


Dear DF, Marcin et.al.,

While "many" photographers are greatly concerned with achieving shallow depth of field, there are many, many, many, many photographers in total. I only devoted one sentence to that segment, because you really aren't a significant percentage of the market (no matter how many folks you know personally who feel the same way). You're like the photographers who do B&W almost exclusively, or the ones who really need clean ISOs at 6400 and above or fast lenses at 500mm and longer. Each of these subsets of photographers does good and important photography, but you're each maybe 1-2% of the market. Your needs will never enter into design considerations except in an ancillary way or for specialty niche cameras. All you can let it do is dictate your personal buying choices, but trying the rally the manufacturers or the visio populi to your side is not going to bear fruit.

pax / Ctein
-- Ctein's Online Gallery http://ctein.com
-- Digital Restorations http://photo-repair.com

Somehow this post AND the comments have generated a lot more heat than light.

It just seems like too many variable have been glommed together to go anywhere beyond "more" pixels is better. I can report that my D300 has better IQ than my D70 but that's 12 Mpix versus 6 Mpix. Back in the day (when was that, last week maybe) larger sensors meant more pixels, and more pixels implied better IQ. Now, who knows?

Michael and Mike,
Concerning jpeg's and the comparometer...a lot of us (well, some of us at least) are concerned only with the subject matter, composition and telling the story. You know, like street shooting stuff, or award-winning photo essays, etc. I simply do not have to, nor do I want to spend any time at all in front of a computer when I shoot jpegs. Spending any time in front of my computer, for me, takes away from the joy of photography. I know it adds to the joy for others. I remember reading about the award-winning, Pulitzer Prize winning photograph from a decade or so ago, which was shot by a pro photojournalist, with a C5050 compact Olympus. I'm just sayin'.

What always mattered most to me in sensor size was depth of field (or lack of). You just cant emulate a fast 50/1.4 on smaller formats than what is now called "full frame" or "FX".

Ill take a low resolution full frame camera over a higher resolution crop sensor anytime.

Controlling DOF is one of the most important aspects of photography (unless all you do are landscapes). Something that is practically taken away from the artist when he has to use small sensor cameras.

To John Camp: Lucas Davenport doesn't have these problems. Suck it up, buttercup.

In the history of photography, "controlling depth of field" most commonly has meant "trying to get enough depth of field to make your picture work". That's why view cameras were used for product photography, the tilts let you manage DOF to get the whole product sharp.

Yeah, sometimes you also want to lessen DOF, but that's very rare overall. And, in my experience, when I want it most, I can't do it; f/4 or even f/2.8 isn't wide enough to blur the backgrounds as much as they need when the issue is an overly-busy background, in most indoor shots I've done.

Dear DDB,

I've run into a similar situation. In those (uncommon) cases where I want to throw the background out of focus to avoid distraction, either I can't get it enough out of focus or else the usable depth of field becomes so razor thin than I get too much unsharpness in the subject of interest (which, annoyingly and incomprehensibly, seems hardly ever to lie in a perfect plane parallel to the camera back).

What I'd really like is a lens whose light-cone (really, a light hourglass neck) stays unusually narrow for a short range around the point of best focus and then balloons rapidly. So I'd get better than expected overall sharpness within the DoF and substantially worse outside of that.

I think there are some telecentric industrial lenses that pull off that trick, but I've not seen anything for real-world use. Possibly, it's impossible to design over a wide range of operating distances? I dunno.

pax / Ctein

The comments to this entry are closed.



Blog powered by Typepad
Member since 06/2007