Is print resolution equivalent between APS and FF of the same MP?

Here's my question. I just ran some tests prints (12X18" portions of a 24X36" and 30X45" image files). I was pretty surprised to see the even at the larger size, there was no pixelization to be seen. I was impressed. My question is, for those of you that regularly print from both sensor, crop and FF, at base ISO, will the print details of either sensor size be equivalent, ie. indiscernible from one another? I am thinking of exploring landscape photography. I shoot with a 24 mp Fuji XT20 now. Would an equivalent 24mp FF camera, Sony, Canon or Nikon yield obviously more detail or would I need to move up the ladder to the 46-50mp sensor to make the noticeable jump in print quality. Remember, I'm thinking about base ISO only shot from a tripod.

 

Thanks,

Same resolution equals same resolution, bur resolution does not define everything in a print or image. Larger sensor cells tend to have lower noise. Smaller sensors require more resolution from the lens to achieve their full potential. A lens must have roughly 4x the resolution of the sensor in order to have a negligible effect on the results. In the absence of a tripod or image stabilization, camera shake reduces the effective resolution to the equivalent of 6 MP at the shutter speed defined as 1/F.

 

Under optimum conditions, you would need a print larger than 12"x18" to see any difference between two 24 MP prints.

9Sorry, Bob I fear you are starting your thoughts at the wrong end. "24MP, exposed by a perfect lens" are (or would be) 24MP, period.

Peep your pixels, judge your glass.

I have an old X-E1 with a pair of consumer zooms. - I believe in getting a fair share of resolution, good enough for 4K viewing the entire image but starting to fall apart, beyond that. I heard good things about Fuji primes but don't own any.

I could dig out some old, less than stellar lens, mount it on the Fuji and some FF body next. The result would be that if the lens delivers let's say 12 PMP on 24MP FF it will surely deliver less on APS. (Maybe a wee bit more than sensor size ratio wise expected, if it has a sweet spot in the image center but still the difference would be noticeable on the higher resolving APS camera.)

For measurements of the other brands: Dive into DxO's lens data base. They just don't cover Fuji and Leica. While some folks call them to be biased against some brands, I suppose they are a reasonable base for a pessimistic look at your shopping options.

 

If you have any lens you aren't happy with, while peeping it's pixels and the reviewing world claims it to be as good as Fuji lenses get at that focal length, it is either defective or maybe a reason to switch to a FF system offering something more promising. - Not all of them will necessarily do.

 

If you compare an X-trans 24MPsensor to a Canon 24MP FF and have perfect lenses for both, the Canon one will most likely noticeably suffer from it's AA filter; so prepare to buy a few extra MP in EOS land to get 24PMP out.

 

Myself I haven't tried high pixel density stuff like 24MP on APS. I am glad that my old Leica glass doesn't fall apart in front of 18MP FF...

 

I think: If FF was necessarily lightyears better in your use case, Fuji would offer such a system. Shoot what you have until you are ready to make a huge step, concerning upgrading options.

A lens must have roughly 4x the resolution of the sensor in order to have a negligible effect on the results.

 

- ???

Where does that figure come from Ed?

A 24 megapixel DX sensor has a theoretical (orthogonal) resolution of 125 lppmm. So you're saying you need a lens with a resolution of 500 lppmm to make the best of it?

 

I'm sure you know perfectly well that no such camera lens is available, anywhere. And even if it was, would only achieve that resolution with monochromatic light of short wavelength and at an impractically wide aperture.

 

Whereas in reality it's entirely possible to see resolution closely approaching the theoretical sensor limit. Even with a very affordable lens, and across the majority of the frame area.

 

So where does that crazy 4x figure come from?

 

The old (MTF of film x MTF of lens) formula no longer applies to digital sensors. Their MTF contrast is practically 100% right up to the Nyquist limit, although that limit varies with spatial-frequency angle.

 

So, yes, you do need a good lens to make the most of modern sensors, but no, you certainly don't need a lens that can't possibly exist!

To answer the OP's question:

 

There is a subtle difference in 'look', and a very noticeable difference in high ISO (>800) performance between APS and Full-frame. Their applications differ too. For example the choice of good superwide lenses is very limited for APS, while that format offers advantages for telephoto and macro use.

 

OTOH full-frame would be a better choice for portrait and fashion work because it's better able to control depth-of-field - at both ends of the spectrum - with smaller formats being more susceptible to diffraction and needing a wider aperture for a given 'bokeh'.

 

It's just a case of choosing the right horse for the course, or the right tool for the job. Simple as.

 

Both formats are technically capable of giving practically identical and near indistinguishable renderings of a resolution test chart, but that tells you nothing about their handling, or how suited it makes them to tackle different types of real 3D subject.

Edited November 15, 2018 by rodeo_joe|1

I shoot with a 24 mp Fuji XT20 now. Would an equivalent 24mp FF camera, Sony, Canon or Nikon yield obviously more detail or would I need to move up the ladder to the 46-50mp sensor to make the noticeable jump in print quality.

 

No. But as Rodeo says it depends on the ISO required to take the image. You gain a stop of noise improvement going from APS to FF, but with any half decent noise reduction program, I doubt you would notice it unless you are shooting above, say >3200 ISO. Whether any perceived difference is actually important is another matter and will depend on the subject matter.

Where does that figure come from Ed?

Assuming uncertainty (errors) follow a Gaussian distribution (random), contributions from the lens and sensor are additive by square law. Digital errors have discrete distribution, but to the first order of approximation, we can consider the uncertainty to be random on a visual basis.

 

If the resolution of the lens and sensor were equal, the net uncertainty would be the sum, or twice that of each component. Hence, the resolution would be half that of the sensor or lens taken separately.

 

Following the square law, a lens with twice the resolution would reduce the net resolution by 25%, etc. By convention, the contribution to uncertainty can be neglected if it is 5% or less of the net uncertainty. At 4x the resolution, the lens would contribute 1/16 of the overall uncertainty, or about 6%. That's close enough for government work.

 

A FF, 24 MP sensor would have 157 ppi, or 78 lppi. It is not unusual to determine the resolution of a premium FF lens at 400 lpi, in extremis, by the aerial method (MTF contrast in the weeds, so to speak). That would be about 5x the resolution of the sensor's resolution, reducing the net resolution by about 4%. The same lens on a 24 MP DX sensor would have only 3x the resolution, with an 11% effect. You would need a lens resolving over 600 lppi to have the same net resolution.

 

(I was out of graduate school 30 years before anyone thought of discrete Fourier transforms. It's never to late to learn, but I wouldn't be left much time to take pictures.)

For the most part, it's just the number of pixels and the size of the print. Whether it's APS-C or "full-frame" only matters as you get into arcane matters such as the quasi-mythical "fat Pixel" arguments.

 

:p

[quote="

If the resolution of the lens and sensor were equal, the net uncertainty would be the sum, or twice that of each component. Hence, the resolution would be half that of the sensor or lens taken separately.

 

 

Ed - I think that, overall, you are correct, but the sum of uncorrelated random uncertainties do not add linearly, but with the square root of the sum of squares. So, for your case, the total resolution would be 1/(square root of 2) = 0.707 of the sensor or lens taken separately, rather than 0.5.

You are correct - the rule is root sum of squares. I was thinking of sum-of-variances. It's been a very long time, even in doggie years. It actually works out better. The net uncertainty where one component is four times as accurate is (1^2 +(1/4)^2))^(1/2) = 1.03. In other words, at 4x, the lens contributes only 3% uncertainty, well under the arbitrary 5% (2 sigma) threshold. Edited November 15, 2018 by Ed_Ingold

You are correct - the rule is root sum of squares. I was thinking of sum-of-variances. It's been a very long time, even in doggie years. It actually works out better. The net uncertainty where one component is four times as accurate is (1^2 +(1/4)^2))^(1/2) = 1.03. In other words, at 4x, the lens contributes only 3% uncertainty, well under the arbitrary 5% (2 sigma) threshold.

 

- All well and good, even if the assumptions are uncertain, but I still want to know where you're going to get a real lens that defies diffraction and resolves 500 cpmm or lppmm?

 

Even 400 cycles/mm is pushing the boundaries of credibility.

 

"Whether it's APS-C or"full-frame" only matters as you get into arcane matters such as the quasi-mythical "fat Pixel" arguments."

 

- Not really mythical JDM. Photosites have to have interconnects between them, which eat up valuable light-gathering area, and the interconnects don't get any smaller on an APS sensor that on full-frame.

 

Hence the breakthrough of the rudely-named backside illuminated sensor.

Edited November 16, 2018 by rodeo_joe|1

(12X18" portions of a 24X36" and 30X45" image files

 

Frankly, I think there is a basic misunderstanding there. Your image file do not have dimensions measured in inch, centimeters, feet or any other standard that indicates distance. The image file has a lot of pixels, period. Its resolution refers only to how many pixels there are. The image itself hence does not have a print resolution. All you do in software is telling how many pixels per inch you want to use, and the dimensions you get are nothing but a calculation (=pixels/pixels per inch = inch). So, your image file has 24 million pixels, typically 6000 on the long end and 4000 on the short. The normal rule is using 300 pixels per inch for high quality prints, which yields 20"*13,3". And in fact the "rule" you always read that a high quality print requires 300 pixels per inch is not a rule, but rather a guideline.

 

If you get a different brand camera, or a full frame camera, with 24MP, you still get image files that are 6000 pixels on the long end, and 4000 on the short. So, in itself, that move will not give you more details or potential for larger prints (at the same ppi setting anyway). The sensor size has nothing to do with this: it's just about how much data you capture.

 

What is actually captured on those 24 million pixels is another story - and one where lenses come into play. You can cram 24MP in a phone camera, but being a small sensor with a cheap lens in front, it will not look as nice as something made with good optics. And sensor size can play a role here too, but it doesn't necessarily mean you can print larger if the sensor is larger.

 

(as for the 300 ppi - basically it is a good guideline for smaller prints, which will be viewed at short distances. The longer the viewing distance, though, the more you can drop the required pixels per inch. Billboards have prints with huge pixels and would look incredibly pixelated up close, but you never look at them up close..... so basically how large you physically can print based on your image resolution is not a fixed dimension)

All well and good, even if the assumptions are uncertain, but I still want to know where you're going to get a real lens that defies diffraction and resolves 500 cpmm or lppmm?

You may not find a lens which resolves 500 lpi (lppi was my mistake), at least for visible light. Diffraction is an easier nut to crack. At f/5.6, the Airy Disk for green light (520 microns) is 3.55 microns in diameter. This corresponds to 107 MP for an FX sensor, or 280 lpi. For a DX sensor, the resolution limit would fall to a mere 48 MP. The effective resolution is higher, because the Airy Disk consists of a series of rings, tapering from the center,

 

I don't have a specific reference for 400 lpi lenses, but is a reasonable figure, which I have seen in other publications. There are many lens tests which show diffraction limiting above f/5.6, whence the resolution begins to fall off. Like most real phenomena, diffraction limits are not a hard cutoff, so its effects can be seen as a peak and gradual decline past a certain point.

Edited November 16, 2018 by Ed_Ingold