Two small images distinguishable?

<p>The D810 (36MP) captures more detail than a D750 (24MP), but if you shot same subject with same lens, exposure, etc... and printed both images from 36MP and 24MP cameras at 8x10 or smaller, would you notice a significant difference? I know image from D810 will hold up better as you enlarge it but if you know you won't be printing poster size is there really a difference at smaller size images? The D810 users might say "definitely" and D750 users might say "not much" but I was hoping to hear from people who have actually compared the two side by side. Thanks.</p>

<p>ng ago I did a test with a D700 and a D800 and a 70-200 II on a tripod and printed the result on 24x30and with a normal view there was no difference, with your eye on the print the was some difference(because you wanted to see it???).</p>

<p>The usual rule of thumb is to assume 1 pixel = 1 print dot and a print resolution of 300 dpi. This means you need only 7.2MP for an 8x10, so even a D300 or D700 image would be downsampled for printing.</p>

<p>I've tested 16MP vs. 12MP vs. 6MP and found that in an 8 x 10 there is no difference if the image is shot the same.<br /><br />I also found that if you go bigger you STILL can't tell the difference at a normal viewing distance.<br /><br />So the higher res are unlikely to make much difference at 8 x 10 or smaller.<br /><br />But if you're gonna crop or do a LOT of image manipulation, conventional wisdom still tells me you want to start with all the resolution you can possibly get.</p>

At 10" x 8" there's almost no difference between

images shot with a D700 and a D800. Although on

subjects with very fine detail the D800 print will

still be slightly sharper on close scrutiny. However,

for the difference to be visible the print parameters

have to be set optimally. It's not true that one pixel

should be mapped to one print dot. A pixel

contains far more colour information than can be

shown in a single blob of ink, and it takes a cluster

of 4 or more ink dots to even come close. So

printing at 300dpi will not reveal the information

contained in even a modestly specified DSLR

image. The printer's maximum resolution should

always be used - 2400, 4800 or whatever dpi, and preferrably the image PPI should be an integer divisor of the print DPI. Best done by manual resizing rather than letting the RIP engine manage the interpolation.

<p>A lot depends on the quality of the image, the print and your technique. At scrapbook quality, you can't tell the difference between a Kodak Brownie and an Hasselblad.</p>

<p>Before scanners, we used to copy PR shots with a 4x5 process camera (a Speed Graphic on a rail) to reduce images for publication. The difference between commercial prints from 4x5 or 8x10 and those we made from 35mm or 2-1/4" film were pretty obvious in terms of sharpness and dynamic range.</p>

<p>The same is true for digital prints, or even images in PN reduced to 700 pixels wide. I think what happens is when downsizing, the sampling process tends to retain fine details which would otherwise be invisible if they were simply discarded, leaving a sense of texture and clarity. In order to be retained, those details actually have to be there. The subject matters too. Some of the best examples of this can be seen in the carefully crafted closeups Lou Meluso posts in these threads.</p>

<p>It's very hard to judge sharpness and detail in isolation. I've reported before that I have an old book on photography techniques that contained a lot of 35mm pictures which looked absolutely fine until there was a 6x6 shot on the same page. Few would argue that you can't make an A2-sized print from a 35mm image, but seeing one at Wildlife Photographer of the Year next to a number of high megapixel digital images made it look extremely soft. (The extinction resolution behaviour of film and digital is very different, of course.)<br />

<br />

Can you see the difference between pixel counts in small images? It depends how closely you look, and what you're looking at. Draw a black line one pixel wide on a white background on a 300dpi laser printer, and you'll likely see the pixels from some distance. Look closely enough and you can see the pixels at 600dpi. I wouldn't like to bet that the 746dpi of a D810 image at 10x8 wouldn't be visible. Have anything resembling the softer pixel-to-pixel transitions you're likely to see in the real world (and bear in mind that mapping the image to the resolution of the printer will introduce some softness and irregularity) and it's much less likely that you'll see either. You can argue the same for much smaller pixel counts too - improvements are incremental at every stage. I do have a 40" print of an image taken at 24mm with a D700 and, putting my nose in the right place for the perspective distortion to disappear, I can absolutely see the pixels. I'd see them from a D810 too, but less obviously. Billboard posters rarely look pixellated, but if you stand anywhere near one you'll see the resolution (even of the colour dither) is actually pretty poor. Look at the screen of a modern phone and you'll struggle to see pixels; stick a VR headset on your face and you'll absolutely see them.<br />

<br />

I find print resolution to be largely irrelevant these days. If I want to look at a high resolution image, I'll open it in a computer. It'll be backlit, and the colours are likely better; a high resolution display can get rid of the worst of the obvious pixellation. If I want to look closely at a small part of the image, I can easily zoom in and scroll around - no need to rely on <a href="

optical zoom</a>. From that perspective, no resolution is "enough". Although I'll be quite interested when we start being able to capture holograms...</p>

<p>Back in early 2009 when Nikon first introduced the D3X, in an era when 24MP was a lot, I did an A/B comparison between the 12MP D700 and the 24MP D3X. It is a macro shot of a flower with a lot of details. I mounted the 200mm/f4 AF-D macro on a tripod and just swap bodies to get (almost) identical shots. I used the base ISO for both.</p>

<p>I printed both images to 8.5x11". At least to my eyes, there is no difference. Even 12MP is more than sufficient to capture all the fine details in the flower. I do have a D800E, but IMO 36MP is an overkill in most situations. Anything more is mostly marketing nonsense. Today, my primary FX bodies are 20 and 24MP. I only use the D800E occasionally. I also use the D700 occasionally for like product images for web display, where 12MP is an overkill.</p>

<P>

<IMG SRC="http://static.photo.net/attachments/bboard/00e/00e9Ls-565497584.jpg">

</P><div></div>

<p>I think it's hard to tell 36MP from 24MP.</p>

<p>But, I can certainly tell my current 24MP cameras from 6MP and from 12MP. </p>

<p>It's great to be able to print 20x30" from 24MP and feel little or no compromise. It's like medium format in a 35mm package. Our photographic equipment today is fabulous!</p>

<p>It seems most people think there's not much of a difference at sizes smaller than large prints. So then why buy a D810? I'm not saying you shouldn't buy one I'm just curious as to what everyone finds so appealing about all those megs (36MP). I'm counting on everyone here at Photo.net because I don't have a full frame camera and would like to get the best quality image possible within $2K - $3K. I've seen images from D810 and they look great but I can only compare it to my old crop sensor camera. The megapixel race does seem to be just a bit overrated but I've also heard/read the photos are (a lot?) sharper due to the lack of an "AA Filter" (whatever that is) on the sensor? I have good lenses and I like crisp sharp images but I'd rather not spend money on marketing hype and get more quality glass instead. It's always fun to buy new cameras but I get a lot more excited about new quality lenses than cameras. Over time cameras seem more disposable, lenses stand on their own long after you buy them. </p>

In terms of resolution, 36MP is definitely an incremental

improvement only over 24MP. Does it capture MUCH

more detail? No. Does it capture SOME more detail? Yes.

It's of the order of the difference between a D300 and a

D7000, or a D7000 and a D7100. For cropping or for

images I want to scroll around, I'll take the detail I can

get - there are times I've been on the limit of someone

being recognisable or text being legible.

 

The low-pass "AA" filter slightly blurs an image at the

sensor level, reducing the risk of aliasing effects and

especially false colours when looking at fine detail. (The

Bayer sensor in the camera, like the human eye, reads

different colours at different locations and heuristically

works out what the colour and brightness should be at

each pixel. If repeating light and dark patterns line up

with the sensor sites they look to the camera like colour.

The filter spreads out this detail so all the colour sites

pick up the signal, at the cost of blur.) You can

compensate by sharpening the image, but this can

introduce other artifacts. There are software ways to

remove false colours, although they're not perfect either.

In reality, false colour is rarely a problem, partly because

getting a lens to resolve perfect pixel detail is hard. Still,

the pixel level detail from a D810 is a lot better than, say,

the D700, which had a very strong AA filter.

 

I hesitated over upgrading from a D700 to a D800, and

what sold me was the dynamic range, not resolution.

The D810 does still have a small advantage there at

minimum ISO over the D750, though its high ISO

performance is a little worse. The handling of the D810

is also somewhat different - it's in the

D300/D700/D800/D500 lineage, whereas the D750 is

more like a D7x00 series. There's a big weight difference

too - the D750 feels like a toy in comparison. There are

other minor spec differences - fastest shutter speed

springs to mind.

 

If you want the best image quality from a Nikon, at least

until a successor shows up, the D810 is absolutely it -

for low ISO. Even if you never print big, for cropping or

scrolling around on-screen the extra detail is there. It's

not a huge difference, though. Have a look at

DPReview's scene comparison tools and select the two

cameras, and you can see the difference.

 

But - you're right to wonder about lenses. They're not

always an investment (they don't go obsolete as fast as

DSLRs, but they do get updated - a 500 f/4 AI-P is a lot

less valuable since the latest range of big telezooms

came out) but you can certainly continue to use them on

new cameras (within reason). How good are your "good

lenses"? I end up stopping down many lenses on my

D810 that I'd use wide open on a D700 because you can

see the aberrations - but it's also not true that you won't

see detail improvements except with something like an

Otus. Still, if you were happy using an AF-D 50mm wide

open, you probably now won't be. Software can help

make lenses better, too - shockingly so for a 14-24, in my

experience.

 

Lenses will last across many camera upgrades - but a

better camera will make all your lenses look better. Pick

your poison. Welcome to Nikon Acquisition Syndrome.

(And probably go and try the cameras in a store - If you

hate the handling of one, that'll probably decide you.)

<p>The main benefit I gain from the D810 and it's 36MP sensor is that when shooting images (with 55mmf/3.5 micro Nikkor) for online mail order plant sales I can place my tripod setup back a considerable distance from the subject to gain both better DOF and also to allow better light (diffused,natural light) onto the subject ( I used to be so close that I blocked some light) Then crop tighter and be able to retain as much detail as I require for my online image files which can be 2048 pix wide sometimes.</p>

<p>Admittedly I'm comparing the D810 to my previous camera the D700 with it's 12MP sensor.</p>

<p>Matthew: Do you actually need the extra field of view? It sounds like you're using the pixel density, and could achieve the same with a teleconverter or a longer (maybe not even macro, since you're further away) lens. Even the DX crop of a D810 has more pixels than a D700. I'm not discounting that the D810 can do this, it just doesn't seem like the only solution!</p>

<blockquote>

<p>Then crop tighter</p>

</blockquote>

<p>The DOF is the same in the end, so you are not actually gaining anything. The DOF from going close and filling the frame will be the same as when you took it from further back and cropped to the same size.</p>

<p>I did some calculations and graphing a few months back. If you shoot at 1/60 second, using a 50 mm lens without image stabilization, you will get 6 MP resolution no matter which sensor or lens quality is involved. Camera shake is the great equalizer.</p>

<p>I would not normally attempt hand-holding at any speed slower than 1/200s except in a dire situation where I have no choice. Normally I typically hand hold at 1/500s or faster; if I plan to crop I may go to 1/1000s. If the intention is to get movement blur I may choose a slower shutter speed but I am really not good at this technique of using slow shutter speeds for movement effects.</p>

<p>At 8x10 inches, printed on my Epson inkjet, I can see the difference between 12MP and 24MP at apertures that I typically use, but the difference between 36MP and 24MP is more subtle and I don't really regard it an important difference. The gradation of sharpness into blur is smooth and beautiful with the 36MP, and you can more easily see where exactly the focus has fallen. The D810 has some features, however, such as electronic front curtain shutter (EFCS) which can, on tripod, be used to improve sharpness of macro shots and telephoto images in some circumstances where significant blur might occur without. Mostly I appreciate the D810 for its quietness, relatively low shake, its remarkable dynamic range at ISO 64, and general niceness of the camera and its refined feature set. It offers a split image live view which allows the image to be studied in two zoomed-in windows (they must be along the same lines of the long axis of the frame), and this is great for adjusting tilt in PC-E Nikkors when the tilt direction is suitable. For landscape type subjects, architecture etc. it is a very nice camera. However, for most circumstances the D750 is very capable and could be used in its place, and in some situations (very high ISO i.e. 6400-12800) it yields a better result. </p>

<blockquote>The DOF from going close and filling the frame will be the same as when you took it from further back and cropped to the same size.</blockquote>

 

<p>At the risk of making an idiot of myself, I don't believe that's true. Rather than replicating the maths myself, Ansel Adams claims that depth of field increases with the square of the subject distance. Depth of field decreases proportionately to image magnification (I think?), so as you crop to match you'll get half the depth of field at twice the distance. Intuitively, the same lens at the same aperture is projecting a narrower cone of confusion at a more distant subject. Of course, since perspective has changed as well, that's not the only difference - which is why this isn't the type of "equivalence" I normally think about. If you "crop" by using a longer lens with the same f-stop, <i>then</i> I believe it balances out - because the larger physical aperture gives you the narrower depth of field.</p>

 

<blockquote>I did some calculations and graphing a few months back. If you shoot at 1/60 second, using a 50 mm lens without image stabilization, you will get 6 MP resolution no matter which sensor or lens quality is involved. Camera shake is the great equalizer.</blockquote>

 

<p>6MP seems a bit low, but I believe you. However, post-processing can restore motion blur to some extent. Heavier lenses, VR, leaning on things and (of course) tripods help. I certainly feel I'm getting close to the sensor resolution, if not perfect pixel sharpness, hand-held on a D810 without jumping through excessive hoops.</p>

<blockquote>Mostly I appreciate the D810 for its quietness, relatively low shake, its remarkable dynamic range at ISO 64, and general niceness of the camera and its refined feature set. It offers a split image live view which allows the image to be studied in two zoomed-in windows (they must be along the same lines of the long axis of the frame),</blockquote>

<p>Agreed with Ilkka on the selling points of the D810 (the quiet mode is shockingly effective compared with the D800). Ilkka, interested that you're using the split live view for the reason I requested it of Nikon! (I've barely used it myself, partly because it wasn't quite what I wanted and partly because my tilt-shift lenses are a pain to use.)</p>

<blockquote>

<p>I certainly feel I'm getting close to the sensor resolution, if not perfect pixel sharpness, hand-held on a D810 without jumping through excessive hoops.</p>

</blockquote>

<p> <br>

Pixel peeping is an exercise in humility. It's very difficult to find an edge or object fine enough in nature to project less than one pixel width on the sensor. Printed text comes close, but not from an inkjet or laser printer. Cobwebs, maybe. Stars should project as a singularity, but the motion of the earth will produce tracks in 5 seconds at the pixel level. The width is subject to various aberrations in the lens, not to mention scattering. Taking extreme care, I find that faint stars have a strong central track 1 pixel wide, with a faint pixel track on either side probably due to diffraction.<br>

<br>

That is with an heavy tripod, cable (flexible) release, IS off, and a really great lens (Batis 25/2). If I leave IS on, I almost always see a doubling of the finest details, spanning 2-3 pixels.<br>

<br>

Camera shake is about 2 degrees/second, on the average. For a 50 mm lens, with a FOV of about 45 degrees, at 1/50 second, that amounts to an uncertainty of 8 pixels on a 42 MP sensor, the equivalent of 5.2 MP. The same uncertainty would encompass a few less pixels on a 36 MP sensor, but the end result is the same.</p>

<blockquote>Pixel peeping is an exercise in humility.</blockquote>

 

<p>Absolutely, but post-processing can do wonders (though not magic). You generally don't want a line resolved to a single pixel unless it's absolutely centred on that pixel, because you'll get aliasing as the alignment drifts. Of course, the amount of detail you <i>can</i> resolve still scales with sensor resolution even in this case. To an extent, it's not single items that are one pixel wide that are as important as preserving a sharp edge that doesn't drift across multiple pixels. And yes, VR will mostly turn a very soft image into an only-slightly soft one; it can also turn a perfectly sharp image into a slightly soft one. Shutter speed is still your friend.</p>

 

<blockquote>It's very difficult to find an edge or object fine enough in nature to project less than one pixel width on the sensor.</blockquote>

 

<p>Huh? Every leaf of grass will project to less than the pixel area unless you're close to it. Of course, "project" is a relative term - optical aberrations spread out pixel-level detail to a greater or lesser extent. You want this with stars, or they'd be resolving different colours as they hit different parts of the Bayer sensor. Many lenses can resolve information at 36MP, but how much contrast they have at that resolution is another matter. I need to do some more star shots (when the sky eventually clears).</p>

<p>Andrew: try it and see. It won't work. The depth of field is only apparently greater because the subject is smaller in the image or further away: as the object is smaller the difference in sharpness is not so apparent. When cropped, the lack of focus becomes magnified, bringing it to equivalence to the fully-framed image.</p>

<p>Oops, sorry Robin, I missed you'd posted this.<br />

<br />

I believe that as you move away, the depth of field increases linearly because the magnification of the subject is reduced, then increases by another linear factor because the lens's aperture looks smaller to the subject (reducing the angle of the cone of confusion). In total, then, the depth of field changes according to the square of distance. If you crop to remove the reduction in magnification by being further away, which removes that linear factor (or, put it another way, reduces the depth of field by a linear factor), you've still got the depth of field increased by the subject's reduced view of the aperture - so the depth of field is still increased. That is, unless you "crop" by using a longer lens, which (at the same f-stop) increases the aperture from the subject's perspective.<br />

<br />

That's maths, though, and it wouldn't be the first time I had conclusively proved some gibberish to be true. I'll attempt to persuade myself with an experiment at some point.</p>