Airy disc size and modern DSLR cameras

Hopefully someone more knowledgeable than myself can shed some light on this. I came across the following issue while doing some research that relates to my astrophotography equipment (specifically to optimize a camera to a given set of optics) and the implications for normal photography really shocked me.

If these numbers are correct then most of the DSLRs made in the past six years lose resolution beyond about f/6.3 - f/8. I find this very ironic since stopping down is normally done to increase your depth of field.

Using the wavelength of 555nm - yellow-green light, the size of the airy disc at various f-ratios is:

Byte size: ZERO | Content warning: NOT AN IMAGE

HTTP response: 404 | MIME changed to 'text/html' | Byte size: ZERO

I am not sure how you came up with f/3, but here is the visual representation of pixel size against airy disk size at different apertures for selected cameras (not the newer ones, admittedly), as well as diffraction limit calculator.

http://www.cambridgein​colour.com …ffraction-photography.htm

That utility shows the Airy disc becoming larger than the pixel size of a Rebel XT beyond f/5.6, so my table is a bit more conservative. But it doesn't matter, f/5.6 as a limiting f/ratio for getting the fullest resolution that camera can offer is just really surprising. If the Airy disc is the final word on resolving power and NOT the camera chip's resolution, then what practical use is an 18mp camera over an 8mp camera? It means you could upgrade to a much higher resolution camera but in practice you wouldn't gain any resolution unless you respect an incredibly low f/ratio every time you shoot.

I have to believe I'm missing something here.

Hermeto wrote in post #12690738
I am not sure how you came up with f/3, but here is the visual representation of pixel size against airy disk size at different apertures for selected cameras (not the newer ones, admittedly), as well as diffraction limit calculator.

http://www.cambridgein​colour.com …ffraction-photography.htm

With the conclusions here:

http://www.oldham-optical.co.uk/Airy%20D​isk.htm

could you suppose a full frame camera would be better suited, if only marginally?

imhotep wrote in post #12690836
That utility shows the Airy disc becoming larger than the pixel size of a Rebel XT beyond f/5.6, so my table is a bit more conservative. But it doesn't matter, f/5.6 as a limiting f/ratio for getting the fullest resolution that camera can offer is just really surprising. If the Airy disc is the final word on resolving power and NOT the camera chip's resolution, then what practical use is an 18mp camera over an 8mp camera? It means you could upgrade to a much higher resolution camera but in practice you wouldn't gain any resolution unless you respect an incredibly low f/ratio every time you shoot.

I have to believe I'm missing something here.

Well, if we accept that photography is the art of compromise, we also have to accept that so is the art of designing camera sensors.

No, you are not missing anything, you are quite correct.
We have to understand that, within the limits of current technology, each and every sensor has it`s optimum resolution.
Pushing over that limit, brings nothing but problems..

Let`s hope that some day, camera manufacturers` marketing departments will understand that too.

Logicus wrote in post #12690873
With the conclusions here:

http://www.oldham-optical.co.uk/Airy%20D​isk.htm

could you suppose a full frame camera would be better suited, if only marginally?

Nope.

The diffraction limit for full frame is the same when measured against depth of field as it is for a crop camera. Yes, the aperture you can use before you hit the diffraction limit is narrower on full frame than for crop, but the full frame camera needs an additional 1 1/3 stops narrower aperture to get the same depth of field as the crop camera, and as it happens, with identical resolution, the full frame camera hits the diffraction limit 1 1/3 stops later than the crop camera does.

kcbrown wrote in post #12691594
Nope.

The diffraction limit for full frame is the same when measured against depth of field as it is for a crop camera. Yes, the aperture you can use before you hit the diffraction limit is narrower on full frame than for crop, but the full frame camera needs an additional 1 1/3 stops narrower aperture to get the same depth of field as the crop camera, and as it happens, with identical resolution, the full frame camera hits the diffraction limit 1 1/3 stops later than the crop camera does.

Gotcha... you know, I was recently noticing that the dof seemed almost "exaggerated" when moving lenses over to the 5dmkii- especially on my 10-24mm which is an asp-c lens, but works down to 13mm, which is pretty wide, full frame. It's an f/3.5-4.5 lens, but the shots look more like f/2 or so when actually at f/3.5. Makes sense.

Logicus wrote in post #12691648
Gotcha... you know, I was recently noticing that the dof seemed almost "exaggerated" when moving lenses over to the 5dmkii- especially on my 10-24mm which is an asp-c lens, but works down to 13mm, which is pretty wide, full frame. It's an f/3.5-4.5 lens, but the shots look more like f/2 or so when actually at f/3.5. Makes sense.

You've just discovered the primary advantage full frame has over crop: additional shallowness of depth of field.

Even in terms of noise, the only reason full frame does any better than crop is that you're using a shallower depth of field to achieve it. Equalize the depth of field, and you suddenly equalize the noise. That assumes you keep the shutter speed the same for both, of course, which is reasonable (if you can decrease it on one camera, you can decrease it on the other).

Full frame also is able to retain detail a bit better, but that's only because it's not as demanding of the glass. But sufficiently good glass on a crop camera and it'll be able to retain detail every bit as well as a full frame camera of the same resolution.


I'm interested in full frame, but only for its extra depth of field latitude. I won't sacrifice anything to get it, either, which is why I haven't yet bought a Canon full frame camera -- none has the combination of autofocus capability, responsiveness, and price that the 7D has. The 7D does everything and does it really well. Canon will have to be willing to potentially eat into some of its 1Ds series sales in order to meet my requirements (and that of many others here, frankly). As much as we're anticipating the 5Dmk3, I'm skeptical that Canon will put into it what it needs to. Guess we'll see. If they don't this time around, with plenty of time and market pressure, then I'll assume they'll never do so, and will just stick with my 7D or, if I fancy full frame enough, jump to Nikon.

Regarding the OP's original question, I think the issue is slightly more complex.

The radius of the first null of an Airy disk is given by d = 1.22 * wavelength * focal ratio (see Wikipedia or Goodman if you want a rigorous explanation) so for a f/8 system I get 5.42 microns. The reference is evidently quoting double this which gives the diameter of the disc to the first null. It's more common to use the fwhm as the resolution limit as this corresponds to the view through an eyepiece when trying to resolve binary stars at the limit of a scope's performance.

The complicating factors are that the camera has a Bayer colour filter array and an anti-aliasing filter in front of that. These conspire to degrade the camera's true resolution to something less than 1/4 of the number of quoted pixels, so the camera will have resolution limit that's roughly 2 pixels wide. In fact, the resolution limit is somewhat better for green light than it is for blue or red. Slightly related to this, you might find this article interesting - particularly the bit about a 2003 vintage, 6 megapixel mono camera producing a sharp 60" print...

Unfortunately, the perfomance of AA filters is kept opaque to the general public. (As others commented earlier - why let reality get in the way of marketing.) Since you have a scope - presumably on a polar aligned mount - you might be able to determine the camera's true resolution by examining on axis starfields. With a 1000mm focal length, a 6" or larger scope should resolve 0.8 arcsec to fwhm (notionally at least) and 4.7 micron pixels subtend 1 arcsec.

Under good seeing (say at the WSP) you should only be limited by the camera's resolution. It would be interesting to see what the combination really achieves.

Your calculations are out the correct 'Diffraction Limited Aperture' (DLA) for the cameras you listed are as follows:

T2i - f/6.8
T3 - f/8.4
XS - f/9.3
XT - f/10.4

As a matter of interest the 7D is f/6.8 (Same as T2i, T3i), and I have tested the DLA and you will see it start to kick in and degrade the image on a 100% crop at f/8 however with relation to depth of field you are using a 1.6x crop sensor and will therefore get 1.6x the depth of field in comparison to a full frame 35mm sensor (5D/II,1DS etc) so f/6.8 on a crop sensor is almost the same as f/11 on a full frame with regards to depth of field.

Obviously depth of field is greatly dependant on the focal length of the lens (The DOF will increase as the focal length gets shorter) and the focusing distance (The DOF will increase the further away you focus) and it is important you learn your hyperfocal distances. As an example if you are using a 10-22mm lens on a 7D at a focal length of 11mm at f/6.4 and your focus point at 3.5ft (Approx. 1M) then you depth of field will be 1.71ft (Approx. 0.5M) to infinity.

For a lot of photographers DLA and DOF won't make a lot of difference with regards to the higher MP sensors that now exist which is probably why Canon seems to be on some kind of MP crusade, where it hurts is if your trying to take landscapes without a very wide angle lenses (I'm guessing this effects macro photographers as well, but I don't shoot macro so I'm guessing). I do shoot landscapes sometimes and that is why I shoot 40D's, personally my favourate crop sensor was the 12MP in the XSi/450D which was a good comprimise between MP and sensor density.

Not even going to the 5D MKII would help in regards to the DOF/DLA as it is technically worse than the 7D. Guess if you want a high depth of field without DLA and you don't want to spend the equivelent of a very good car you are left with a choice between an older low resolution DSLR or a medium format film camera.

WayneCornish wrote in post #12692137
Not even going to the 5D MKII would help in regards to the DOF/DLA as it is technically worse than the 7D. Guess if you want a high depth of field without DLA and you don't want to spend the equivelent of a very good car you are left with a choice between an older low resolution DSLR or a medium format film camera.

The nature of digital probably introduces some complexity here, but I would expect that at least to a first degree approximation, the higher resolution sensors don't really limit you in diffraction terms any more than a lower resolution sensor would. You can always downsize the larger resolution image to match that of the smaller resolution image, and by doing so you'd probably wind up dispensing with at least a large portion of the effects of being diffraction limited. That probably depends greatly upon the resizing method chosen.

More interestingly, the greater the resolution ratio of the high resolution sensor to the low resolution sensor, the less of a difference it'll make in terms of diffraction effects when targeting the same final output resolution (e.g., when printing to the same target size using the same printer).

So for us high/drunk and possibly foreign artists...

In layman terms a higher f-stop results in worse IQ?

Bang Bang Boy wrote in post #12692468
So for us high/drunk and possibly foreign artists...

In layman terms a higher f-stop results in worse IQ?

In layman's terms, if you stop your lens down, the quality will reach a peak and then diminish. Where that peak is depends on the lens and how large the pixels on your sensor are. The more pixels on your sensor, the smaller they'll be. The smaller your sensor, the smaller the pixels will be. The smaller the pixels, the less you can stop your lens down before the quality starts to diminish.

Full frame can be stopped down more than crop without losing quality because its pixels are larger. But that doesn't mean much, as it happens, because you have to stop it down more than you would a crop camera in order to achieve a given depth of field. Bottom line is that it all works out the same in the end if you think of it in terms of depth of field as opposed to aperture size.

kcbrown wrote in post #12692603
In layman's terms, if you stop your lens down, the quality will reach a peak and then diminish. Where that peak is depends on the lens and how large the pixels on your sensor are. The more pixels on your sensor, the smaller they'll be. The smaller your sensor, the smaller the pixels will be. The smaller the pixels, the less you can stop your lens down before the quality starts to diminish.

However, there's one additional point to bear in mind. Diffraction will not make the image on a sensor with more, smaller, pixels worse than the image on a sensor of the same size with fewer, larger pixels. The image will, in theory, be better or the same.

Thanks, had a rough night yesterday. This is really bad news for landscape photographers.