Do you use apertures narrower than f/8-f/11?

[Daniel Browning]

Jon already answered this one excellently, but since I already wrote up a response I may as well post it.

While it is true that the medium for capture doesn't make any difference on diffraction, the type of digital sensor does.

I kindly disagree. The type does not make a difference: diffraction is the same for all of them. This can be seen in the actual prints, but I'll try again to explain why.

With the increasing resolution of digital sensors, which I presume is a matter of squeezing those pixels tighter together, isn't this a disadvantage for diffraction?

You are correct that increasing megapixels for a given sensor size involves making the pixels smaller. However, smaller pixels are not a disadvantage for diffraction.

• More pixels always result the same or more resolution.
• Sometimes it's a lot more.
• Sometimes it's slightly more.
• Sometimes it's so infinitesimally more that it might as well be the same.
• Never, under any circumstances, is it less resolution.

The increase in linear resolution is directly proportional to the decrease in pixel pitch. Even after diffraction causes the MTF to drop below the Raleigh criterion for luma, additional gains are possible in order to oversample chroma infomration and make it possible to remove the OLPF.

For example, going from the 5D1 to the 5D2 the linear resolution increased 28% (sqrt(21.1/12.8)). Translated to line pairs per millimeter (the standard measurement of resolution), at 2 pixels per line pair (Nyquist limit), the 5D1 has is 60 lp/mm whereas the 5D2 is 78 lp/mm.

With both cameras at f/8, the 5D2 gives the full expected 28% increase. As the aperture closes down, the resolution increase gets less and less, but it's never goes negative: just diminishing returns.

For example, the 5D has about a 13 Mp sensor, vs the Mk 2's 21. Diffraction kicks in above f/13 on the 5D, but around 10 on the Mk 2.

It doesn't "kick in" because it was always there. It's just that the large pixels of the 5D1 were too coarse and blurry to make out the fine detail that was squashed by diffraction. If the blurriness of large pixels is improved by switching to the 5D2, it becomes possible to see the diffraction blur that was always there. You can always resize 21 MP back down to the same spatial resolution as the 5D1 if you want to hide the diffraction.

So, in this case, with full frame sensors on both, one is clearly different (worse) than the other.

Yes, the 5D classic is worse.

Diffraction IS a undesirable thing

Agreed. Gravity can also be a bother sometimes. [] They're constant, like death and taxes.

A solution seems to be, then, to make a larger sensor, but how practical is that?

A larger sensor does not help at all because depth of field scales with diffraction. If you increase sensor size and keep f/number and field of view the same, then diffraction improves... but depth of field gets thinner. So you stop down more. Which worsens diffraction back to the exact same spot you were when you started. A smaller sensor doesn't help either. You can use a wider f-number, but the smaller pixel size cancels out the advantage. Another way to see it is that diffraction scales with the aperture diameter (AKA focal length divided by f-number).

The reason why so many people get the wrong impression about diffraction and pixel size was explained aptly by RDKirk. Some users perform very flawed comparisons. Chief among these flaws is comparing two completely different magnifications through the use of 100% crop. The spatial frequency must be scaled before any analysis is completed. This can be done, for example, by looking at actual prints, or simply resizing images before creating crops.

Kind regards,
--Daniel

[Alan]

Jon, Daniel, thanks. Lots of info to digest.


I got lost in all the verbage about seeing the difference in the print. If I'm using the same lens, at the same f/stop, carefully ensuring proper steadiness of the camera (tripod, mirror lockup, etc.), then exactly what will the prints show, in terms of difference?


If there is a difference, and the diffraction is the same for both cameras, then what does that difference translate into? Resolution?


Thanks for the helpful insight.

[Daniel Browning]

...what will the prints show, in terms of difference?

28% more resolution, in the case of the 5D1 -> 5D2 at f/8, no aberrations, and sufficiently large print sizes (or magnification).


0% more resolution at f/64.


Somewhere in between at other f-numbers, by an amount that varies with post processing (since demosaic, diffraction deconvolution, and USM sharpening can recover some resolution).

[Jon Ruyle]

0% more resolution at f/64.


Somewhere in between at other f-numbers, by an amount that varies with post processing (since demosaic, diffraction deconvolution, and USM sharpening can recover some resolution).

It isn't clear to me that deconvolution can't recover more resolution on a higher resolution sensor, even at very high f numbers such as f/64. In fact, I'd be very surprised if it couldn't. (I've never actually *tried* it, though [:S])

[Colin]

Yes, the 5D classic is worse.

Until I have a 5D mkII, can we just say that the 5D classic isn't just quite as awesome? Daniel, you've sold me. I want more pixels. I really do. I just don't want to pay for them right now []

Agreed. Gravity can also be a bother sometimes. [img]/emoticons/emotion-5.gif[/img] They're constant, like death and taxes.

No Kidding. My new torpedo-sized target arrows, between 20 and 30 yards, drop more than a foot. Makes diffraction limitations seem like a relatively minor thing []