FF Resolution Comparison

[bunnspecial]

Not too long ago, I picked up a Nikon D3s.

Although this is a somewhat dated camera, I love single digit Nikons(film and digital), and the D3s remains among the least noisy low light cameras available despite being ~8 years old.

In any case, after buying it I realized that I had Nikons with 12, 24, and 36mp sensors. With that in mind, I've tried a couple of times to set up a test to show how much of a real-world difference that makes.

For this test, the main cameras are the D3s, Nikon D600 at 24mp, and D800 at 36mp. All three of these cameras have anti-aliasing filters, which rob some sharpness from the sensor. Lower resolution cameras tend to be somewhat more aggressive in this department. For this reason, I'm also including "bonus" photos from a Kodak DCS 14/n. This is an otherwise terrible camera, but was the first Nikon mount full frame DSLR available, and also doesn't have an AA filter over its 14mp CMOS sensor.

In any case, for this test I used a Micro-Nikkor 105mm f/2.8D, and mounted each camera on my Manfrotto CF tripod with an Arca-Swiss B1 ball head. I had to do a little bit of moving around to keep the framing more or less similar due to the different size of each camera. All but the Kodak were leveled using the built-in electronic level(I left the Kodak on its own and just eyeballed it via one of the grid lines in the finder). I set each camera to ISO 200(the native ISO for the D3s, this was done to normalize everything, although the Kodak photos take a noticeable hit on noise) and pretty much let the Matrix meter work its magic. The only exception was the D600, which I know tends to overexpose and consequently used my standard for it. I manually focused on the stone and used the electronic rangefinder to confirm focus, although as you can see(unfortunately) I wasn't quite consistent in focus. The aperture was left at f/5.6-this is a very sharp lens, and beyond there diffraction starts showing up.

I know this is far from perfect-if the rain holds off tomorrow I'll try to reshoot with the focus nailed.

In any case, here's what I have-first images resized for the web(which show little difference).

D3s

Kodak

D600

D800

Now, a 100% crop. This started as an 800x800 pixel area on the D3s, and I did my best to get a 1:1 crop showing the same area on each subsequent camera(although obviously the images are much larger).

D3s

Kodak

D600

D800

I hope that this quick test might be somewhat useful to folks wonder if-for example-it's worth moving to a higher resolution camera.

 

[Darmok N Jalad]

Nice work. Would like to see a modern smartphone thrown into the mix.

 

[simonsi]

I'm confused, your 800x800 "100%" zooms cannot be 100% in the higher resolution cameras otherwise the image framing would have changed...as is you are showing little difference in normalising the different cameras to the same image size, frame and resolution, sure the higher-res camera's could go further (and the Kodak is already struggling), but that is known already surely???

 

[bunnspecial]

I'm confused, your 800x800 "100%" zooms cannot be 100% in the higher resolution cameras otherwise the image framing would have changed...as is you are showing little difference in normalising the different cameras to the same image size, frame and resolution, sure the higher-res camera's could go further (and the Kodak is already struggling), but that is known already surely???

Sorry, I'm not sure what you're asking/citing as a problem.

The D3s is the only crop that is 800x800, and on all the others I took a crop that covered the same AREA of the image and let the actual resolution fall where it may. If you enlarge the D800 image, you'll see that it's nearly 1600x1600 pixels, but shows the same area as the 800x800 crop from the D3s.

All the crops APPEAR as the same size in the normal view as that's how MR displays them, but if you click on them they will enlarge to their actual size.

If there is a better way to show this information, I'd love to hear how to do it-as I said I normalized the image area, not the number of pixels in the crops.

 

[simonsi]

Sorry, I'm not sure what you're asking/citing as a problem.

The D3s is the only crop that is 800x800, and on all the others I took a crop that covered the same AREA of the image and let the actual resolution fall where it may. If you enlarge the D800 image, you'll see that it's nearly 1600x1600 pixels, but shows the same area as the 800x800 crop from the D3s.

All the crops APPEAR as the same size in the normal view as that's how MR displays them, but if you click on them they will enlarge to their actual size.

If there is a better way to show this information, I'd love to hear how to do it-as I said I normalized the image area, not the number of pixels in the crops.

Ah that explains it, on a scroll down while at work I hadn't clicked through.

 

[kallisti]

So you are attempting to show the differences between MP sensor size on full frame. In other words, what is the real world difference between a 12 MP, 24 MP, and 36 MP sensor.

Since the focus is different between the shots, it's hard to really compare.

Also there have been improvements in other aspects of sensor design/performance over the years that complicate a direct comparison.

One advantage of higher resolution sensors is that they give you larger files to work with. So it is easier to print larger without loss of image IQ or you can crop more aggressively without loss of image IQ.

To show this, you should crop to 100% on the highest resolution body (in your case the D800). Then crop the images from the other cameras to show the same field of view (which will result in crops at >100%). IQ differences with the smaller MP bodies would be expected to be more obvious in this case. They don't have the MP on the sensor side to resolve detail at this level of magnification.

The "real world" question is whether you need this level of detail for what you are shooting and how you plan to use your images. If you routinely need to heavily crop your images then the answer is yes. If you need to print large the answer is yes. One drawback of high MP sensors is that focus errors become more obvious.

Higher MP sensors aren't all about sharpness however. There are also issues of dynamic range, noise, leeway in how the files tolerate manipulation in post. Not strictly MP issues but also sensor generation issues. And also issues about how RAW files are managed by the camera. For example, the Nikon D850 produces largely gapless RAW files on histograms compared to the RAW files from the Sony A7R3 which have large gaps in their RAW histograms (data from DIGLLOYD.com).

I would suggest you shoot the comparison again with an emphasis on keeping the focus constant (which may be challenging as there may be AF differences between the bodies even with the same lens that aren't easy to correct--really need to focus with Live View to do this and some of your bodies don't have this ability).

I would also keep the ISO at base for each body rather than keeping it constant between bodies. Changes in ISO affect sharpness and base ISO for a given body will result in the sharpest images the body is capable of producing. If you are shooting on a tripod, then shutter speed is irrelevant and you can keep overall exposure constant between bodies.

Full scene images (scaled to web size) are useless for any comparison. Would only share 100% crops from the highest resolution body and then share similar field of view images from the other bodies (which will be cropped at >100%). That will show off (at least in one sense) what you gain with higher MP bodies.

 

[bunnspecial]

Thanks for the critique.

I'm going to give this another shot when I get a chance-I spent a while setting it up, and didn't realize the focus error until I actually sat down to put it together.

As I said, I did manually focus, but I think that for the next attempt I need to use magnified live view(although the Kodak doesn't offer that, but I think it's the lease relevant camera to this discussion/comparison for obvious reasons).

Also, I debated about whether to use base ISO or to shoot them all the same. I don't think going one stop above base ISO really affected the D600 and D800, but it DID make a big difference on the Kodak(which is 80).

The D600 and D800 are roughly the same age and same generation of sensor. The D3s is older, but the sensor actually holds up pretty well still in terms of dynamic range. The Kodak, of course, is just all around awful-if nothing else it's an interesting "last gasp" before an industry leader finally gave up on making high end cameras.

BTW, I gave the scaled to web photos to give an idea of how each camera renders the same scene-I know that they're not much use for actually comparing how detail is rendered, but at the same time they do give information that a 100% crop can't give.

One last thing-even working in the front yard, it's easy to forget how much work these kinds of tests are when you're trying to keep everything equal.

 

[kallisti]

Thanks for the critique.

I'm going to give this another shot when I get a chance-I spent a while setting it up, and didn't realize the focus error until I actually sat down to put it together.

As I said, I did manually focus, but I think that for the next attempt I need to use magnified live view(although the Kodak doesn't offer that, but I think it's the lease relevant camera to this discussion/comparison for obvious reasons).

Also, I debated about whether to use base ISO or to shoot them all the same. I don't think going one stop above base ISO really affected the D600 and D800, but it DID make a big difference on the Kodak(which is 80).

The D600 and D800 are roughly the same age and same generation of sensor. The D3s is older, but the sensor actually holds up pretty well still in terms of dynamic range. The Kodak, of course, is just all around awful-if nothing else it's an interesting "last gasp" before an industry leader finally gave up on making high end cameras.

BTW, I gave the scaled to web photos to give an idea of how each camera renders the same scene-I know that they're not much use for actually comparing how detail is rendered, but at the same time they do give information that a 100% crop can't give.

One last thing-even working in the front yard, it's easy to forget how much work these kinds of tests are when you're trying to keep everything equal.

Yes, they are quite a bit of work. Many variables all contributing to confuse the results. Some other suggestions (which you may be aware of):

Using magnified LV focus is critical. Optical view finder focus can be hit or miss for a given lens/body combo. It may also vary among bodies (which may be the explanation for the differences in focus in your original series).

For something like this, it is critical that you use good tripod technique. Pressing the shutter on a camera mounted on a tripod can introduce shake that will influence results. Use a wired or wireless remote release. Or have a long delay from shutter press to shutter release.

Turn off VR on your 105mm micro lens. For tripod work, it can introduce blur if left on.

Possible you know this, but important to keep all of this in mind when doing tests like this.

 

[bunnspecial]

For something like this, it is critical that you use good tripod technique. Pressing the shutter on a camera mounted on a tripod can introduce shake that will influence results. Use a wired or wireless remote release. Or have a long delay from shutter press to shutter release.

Turn off VR on your 105mm micro lens. For tripod work, it can introduce blur if left on.

Possible you know this, but important to keep all of this in mind when doing tests like this.

I didn't mention this in my write-up, but I used the self timer for all photos. The next time I set this up, I'll most likely use a cable release. I have a couple of the ones that work with the D800 and D3s, but I couldn't locate one when I set up for this. I have the little wireless remote that works with the D600, and I always manage to forget that the Kodak can use a standard Compur-type threaded release.

I'm using the older "D" version of the 105mm f/2.8 Micro, which does NOT have VR. I actually sold the VR version because I found the D version sharper in actual macro use, but that's neither here nor there. In any case, VR doesn't factor into this one, but it's an easy one to overlook when doing tripod tests.

 

[kallisti]

Something else to consider.

Yes the Nikon 105mm micro is sharp. On film and lower MP digital cameras. But optically it wasn't made for high MP digital cameras. Per DXOMARK testing (which isn't perfect and has it's own issues), the VR version has a maximal resolution of 21 MP. So you won't necessarily see "better" sharpness on a sensor above 21 MP. The "D" version of the lens hasn't been tested, so possible it is sharper than the VR version. Also possible the site's testing doesn't reflect real world conditions. But it was reported on a D810 (which doesn't have an AA filter).

https://www.dxomark.com/Lenses/Nikon/AF-S-Nikkor-VR-105mm-f-2.8G-ED-mounted-on-Nikon-D810__963

Can quibble about the accuracy of DXOMARK testing. But based on their data, you may not see the full effects of a higher MP body when using this lens (because the sensor may out-resolve the lens--i.e. the sensor is capable of resolving more detail than the lens can deliver).

My suspicion is that if you can normalize focus between the bodies you are testing and then look at files from the D800 at 100% compared to the lower MP bodies you mentioned, you'll see noticeable differences in IQ.

Using a "sharper" lens that is able to resolve closer to 50 MP (like the Zeiss 135mm f/2) you *will* see a very obvious difference as the lens can out-resolve the sensor even with a D850.

But the important question is whether this matters in "real world" use and not just pixel peeping. If you print big or crop heavily, it can. But printing at smaller sizes you lose some theoretical sharpness because of the limits of printing resolution and it becomes something of a moot point (though there are other lens characteristics that still matter in the final print).

 

[steveash]

I know this is very much personal opinion but resolution apart, the older the camera, the nicer, more natural looking the image. Compared to the Kodak, the D800 is saturated with too much contrast. Were these jpegs or raw files? Any processing?

 

[bunnspecial]

I know this is very much personal opinion but resolution apart, the older the camera, the nicer, more natural looking the image. Compared to the Kodak, the D800 is saturated with too much contrast. Were these jpegs or raw files? Any processing?

The Nikon files were shot as RAW, but processed in Capture NX2 with no PP and using the settings as specified in camera. All are set the same in-camera: AWB, "Vivid" color, "normal" sharpening, and a few other settings I forget.

The Kodak files are SOOC JPEGs, primarily since the RAW files are clunky to work with and the camera is even clunkier when writing them.

Aside from that, the Exif will report processing in CS6, but that was mostly for resizing/cropping. The "full frame" image you see was resized to 1500 pixels on the long dimension for all photos. Crops were taken from the full resolution image in CS6 in the manner already described(800x800 for the D3s, all others with a 1:1 aspect ratio of the same area of the image).

The late Kodak DSLRs were very polarizing cameras. To someone who likes over the top color, they're not that appealing. A lot of portrait and studio shooters LOVED the colors they could get from them, though.

 

[tizeye]

Nice work and understand your rational for keeping the visual crop size. What threw me was the Kodak as I missed it on the initial reading that it was a Kodak camera. I thought that with your statement of liking Nikon single digit cameras, film and digital, that it was Kodak film that you scanned to digital. Now that would have been an interesting comparison.