Diffraction doesn't seem to matter much - does it depend on the lens?

[Karim D. Ghantous]

Have a look here (cued up):

f/22 vs f/5.6 on a Nikkor micro lens. The lens is an SLR lens but the sensor on the camera is APS-C. However, this photographer is photographing negatives at close distances. Does that have an influence on how diffraction works?

The f/22 image is not as soft as I expected it to be. In fact it's so mild that the grain is still easily visible. Of course I wouldn't want to always shoot stopped down that much.

Maybe we should test our lenses for diffraction, rather than just accepting received wisdom? Perhaps if you want all that DOF, it's safe to do so, at least with some lenses.

 

[Stephen Pruitt]

Students of the great photographers like Adams, Strand, Bullock, Weston, etc., learned this long ago. Using their 8x10 view cameras, and without the ability to digitally manipulate anything, they modified their lenses and stopped them down all the way down to f/64 (hence the name of Group f/64). The result? Some of the greatest images the world has ever seen. Certainly diffraction is "a thing," but it really shouldn't be an issue at f/22 at all.

 

[Phil Holland]

Diffraction is dependent on both the lens and the medium/format. It comes in earlier in the higher resolution, more fine grained stocks as well as smaller pixel sizes. Very small pixel sizes for instance are prone to the worst of it. Slower film stocks to a degree as well depending on the stock itself.

I'm more on the side of do what it takes to get the shot and make the shot you want, but in modern times it's very easy to get a fairly soft image with some pixel sizes and designs. Film is a more interesting conversation due to the nuance, but higher speed films will show the effect, but will also maintain a tooth to the image due to grain size. Faster films, generally, go softer earlier in the stop range at an equivalent format size.

Interestingly, though absolutely not talked about at all, several camera companies have anti-diffraction technology now in mirrorless bodies. Somewhat similar to how we compensate for that drop in motion picture post production, but there are a few methods that work well. At the higher echelon of motion picture work, we basically match all material to keep it consistent no matter the T-Stop or f/stop within reason. Trickier when you take into account how some lenses are optimized for different working distances as well as varied performance across focus distances.

 

[Karim D. Ghantous]

Students of the great photographers like Adams, Strand, Bullock, Weston, etc., learned this long ago. Using their 8x10 view cameras, and without the ability to digitally manipulate anything, they modified their lenses and stopped them down all the way down to f/64 (hence the name of Group f/64). The result? Some of the greatest images the world has ever seen. Certainly diffraction is "a thing," but it really shouldn't be an issue at f/22 at all.

True that. Although f/64 on big sheet film is actually not that extreme if you think about it.

Trickier when you take into account how some lenses are optimized for different working distances as well as varied performance across focus distances.

So let's say that we have two lenses. They are identical in every way, except that one is a macro lens, and one is not. Is it possible that the macro lens could have less diffraction at any given distance?

Funny thing is that when I changed systems, I gave away all the adapters for the old system (not that they would be useful if I kept them). I think that in the future, I'd like to test different lenses and see how they differ. Vintage SLR lenses, medium format lenses, macro lenses, rangefinder lenses, etc.

 

[Tom Roper]

Rayleigh's criterion:

Aperture diffraction places a limit on the smallest spot of light that can be focused on an image plane. An Airy disc refers to the central spot of light surrounded by concentric rings that are produced when a point source of light is imaged through a circular aperture, such as a telescope or a camera lens. This phenomenon is named after Sir George Biddell Airy, who first described it in the 19th century. The size of the Airy disc and the surrounding rings depend on factors such as the aperture size and the wavelength of the light. In optical systems, the Airy disc serves as a fundamental limit to the resolution or ability to distinguish between closely spaced objects.

It's generally thought that when an airy disc overlays 2-3 pixels, the image becomes diffraction limited. There is a simple formula based on the wavelength of green light, and an airy disc size of 2 1/2 pixels, that you can enter your sensor pixel pitch (spacing), to give the smallest f-stop opening before the onset of diffraction limiting resolution.

pixel spacing (μm) x 1.863 = f-stop

Example: A full frame 36 x 24 mm sensor 8392 x 5594 = 4.29 μm pixel spacing

4.29 μm x 1.863 = f/8.0

As pixels get smaller, more become covered by the airy disc, and diffraction sets in sooner.

 

[Phil Holland]

As pixels get smaller, more become covered by the airy disc, and diffraction sets in sooner.

Correct, but unfortunately more variables at play. Low Pass Filters, different sensor designs, actual different sensor micro lens designs, and the spooky one; the individual lenses. It would be a much easier conversation if digital and/or film were a single surface/layer affair that was perfectly flat.

The math is a good guideline, but I've seen diffraction come into play "sooner" and it's dependent on all of the other factors. Usually around 1 f/stop variance.

 

[Karim D. Ghantous]

4.29 μm x 1.863 = f/8.0

As pixels get smaller, more become covered by the airy disc, and diffraction sets in sooner.

My problem is that theory and data don't always converge. Phil obviously shares the same sentiment. The guidelines are useful, but guidelines aren't data, and they don't convey understanding.

An example of this principle that actually matters: we still don't know if black holes exist or not. We actually don't. That's not a problem BTW. The problem is the assumption that we just know, when we actually don't know.

 

[Tom Roper]

My problem is that theory and data don't always converge. Phil obviously shares the same sentiment. The guidelines are useful, but guidelines aren't data, and they don't convey understanding.

You started with an observation that f/22 isn't as soft as you expected. Are you looking for support for that? If all parts of an imaging system are considered to be perfect, then the resolution of any imaging process will be limited by diffraction. That limitation is in addition to any others. No one is saying all parts of the system are perfect, quite the opposite. Nor is anyone stating an exact f/stop where diffraction is observable. How many sensor pixels an airy disc has to overlap before an image is perceived to be softened by diffraction is an open question. There is general agreement around 2-3 pixels. The simple formula I crafted is an algebraic reduction of Rayleigh's Criterion, to assume 2.5 pixels. The answer can be +/- 1 f/stop from that number generally. Strictly speaking, if the airy disc overlaps more than just 1 pixel, MTF50 resolution is being limited, but you might not notice.