Why f/8 Creates the Same Blur on Every Lens

October 3rd, 2026

0 Comments

$What Diffraction Actually Is, and Why Photographers Worry About f/16$

A 24mm lens at f/8 has an opening 3.00 mm across. A 600mm lens at f/8 has one 75 mm across, 25 times wider, and it draws a diffraction pattern on the sky 25 times finer. Both land a blur spot on your sensor exactly 10.7 microns wide.

Every other limit in photography has a price attached. Aberrations yield to better glass, noise yields to a bigger sensor, flare yields to better coatings. Diffraction yields to nothing, because it is not a property of your lens at all.

Diffraction Is What Happens When You Delete Most of a Wave

Light leaving a distant point arrives at your lens as a flat wavefront, essentially infinite in extent. The Huygens-Fresnel principle says you can treat every point on that wavefront as a tiny source throwing off its own spherical wavelet, and the field farther along is the sum of all of them. In a complete, untouched wavefront, the wavelets heading sideways cancel each other exactly, and what survives is the wave continuing straight ahead. That cancellation is the only reason light appears to travel in straight lines at all.

An aperture destroys the arrangement. It removes every wavelet outside its rim, so the sideways contributions from the wavelets that remain have no partners left to cancel against, and energy shows up at angles no geometric ray ever went. The formal statement of this, the Fresnel-Kirchhoff integral, is a sum taken over the open part of the aperture only, which is why the size and shape of the hole set the entire pattern.

Diffraction is therefore not light bouncing off the metal of the blades, and it is not an aberration. Aberrations are errors in how the glass bends light, and a designer can drive them toward zero with better surfaces, exotic glass, and enough money. Coatings work on reflections at glass-air surfaces, which is a flare and transmission problem, not a pupil problem. Diffraction is a boundary condition set by the size of the opening and the wavelength, and the only way to eliminate it would be to make the aperture infinitely large.

Work out the sum for a circular hole and you get the Airy pattern: a bright central disc surrounded by faint rings, named for George Biddell Airy, who published the solution in 1835. The math produces a Bessel function of the first kind, the cylindrical cousin of a sine wave, a decaying oscillation with an infinite string of zeros. The first of those zeros sits at 3.8317, which is where the familiar constant 1.22 comes from, since 3.8317 divided by pi is 1.2197. Inside that first dark ring sits 83.8% of all the light in the spot. The first bright ring holds 7.2%, the second 2.8%, the third 1.5%.

A real Airy pattern, made by sending a laser through a pinhole and projecting the result onto a card. The bright central disc holds 83.8% of the light, and every ring outside it is energy the aperture threw sideways. Photo by Anaqreon, CC0. Source.

The diameter of the Airy disc measured to its first dark ring. d is that diameter at the sensor, lambda is the wavelength of the light, N is the f-number, f is the focal length, and D is the diameter of the aperture. The constant 2.44 is twice 1.22, which comes from the first zero of the Bessel function that solves the circular-aperture problem.

At 550 nm, the middle of the visible spectrum and where a Bayer sensor is most sensitive, that formula collapses to arithmetic you can do in your head. The Airy disc is 1.34 microns wide for every unit of f-number: 5.4 microns at f/4, 10.7 at f/8, 14.8 at f/11, 21.5 at f/16, and 29.5 at f/22.

Why the f-Number, and Not the Size of the Hole

The angle the light fans out into depends on the physical diameter of the opening and on nothing else. A wider hole diffracts less, which is why an astronomer buying a telescope cares about aperture in millimeters and never mentions focal ratio when discussing resolution.

You do not record an angle. You record a spot on silicon, and converting that angular scale into a distance at the image plane brings in the effective focal length. The diffraction angle goes as lambda over D, so the blur on the sensor goes as lambda f over D, which is lambda N. The exit pupil is a different thing and sits wherever the design puts it; in an image-space telecentric lens it is at infinity while the focal length stays finite. The diameter of the hole sits in the denominator of the diffraction angle and in the numerator of that lever arm. The two cancel exactly, and only the ratio of focal length to diameter survives.

How the f-number falls out of the physics. Theta-one is the angle from the center of the diffraction pattern to its first dark ring, lambda is the wavelength, and D is the diameter of the aperture. For the small angles involved here, the sine of the angle and the angle itself in radians are the same number. Multiplying that angle by the focal length f converts it into a radius r-one measured in millimeters at the sensor. The diameter D cancels, leaving only the f-number N.

Run four lenses at f/8. The 24mm has a 3.00 mm opening and reaches its first dark ring at 46.1 arcseconds. The 50mm reaches 22.1 arcseconds, the 200mm 5.5, the 600mm 1.8. Across a 25-fold range of physical apertures and diffraction angles, all four put a 10.73-micron disc on the sensor. Not approximately. Identically.

The long lens really is resolving finer detail on the subject, by a factor of 25, and on a double star or a distant ridge line that is the number that matters. It does not translate into a smaller spot on the sensor, because the optics that bought the fine angular resolution magnified the scene by the same factor. The astronomer and the photographer are describing one piece of physics and quoting the term that matters for their own job.

There Is No Cliff, and There Never Was

The Airy disc is a useful mental picture and a poor measuring tool, because a photograph is not a field of isolated points. What you want to know is how much contrast survives at each level of fine detail, which is the modulation transfer function. For an aberration-free system, the MTF is the fractional overlap area between the pupil and a shifted copy of itself, and for a circle that overlap has an elementary closed form.

The exact diffraction MTF of a perfect lens with a circular aperture. Nu is spatial frequency, in line pairs per millimeter, meaning how fine the detail is. MTF is the fraction of the original contrast that survives at that level of detail, running from 1 when all of it survives to 0 when none of it does. Nu-c is the cutoff frequency, the finest detail the lens can transmit at all; lambda is the wavelength and N is the f-number. Above the cutoff the MTF is exactly zero.

The curve has a hard right-hand end, the cutoff frequency. At 550 nm, the finest detail a perfect lens can transmit is 227 line pairs per millimeter at f/8, 165 at f/11, 114 at f/16, and 83 at f/22. Past that frequency, the MTF is not small, it is zero, and no information exists to be recovered by any sharpening tool ever written.

Everywhere to the left of that end, nothing kicks in. The curve leaves 1.0 at zero frequency and starts down immediately, with an initial slope of 4 over pi, about 1.27 points of contrast for the first 1% of the way toward cutoff, then bends progressively toward zero. Three quarters of the way to cutoff the exact curve still holds 14%, where that opening slope carried on as a straight line would have run out at 4%. It is below 1 at every non-zero frequency at every aperture you own. Diffraction is not a threshold you cross at f/11. It is a tax you have been paying, in proportion, since f/1.4.

The curve also explains why nobody notices. A perfect lens at f/16 delivers 67% contrast at 30 line pairs per millimeter, roughly what a sharp 16 x 20 inch print demands at normal viewing distance, and 18% at 80 line pairs per millimeter, where pixel-level scrutiny lives. Both figures come from the same curve, which is why an f/16 landscape can look measurably degraded at 100% and print beautifully at exhibition size.

"Diffraction limited" is a compliment that photographers use as an insult. It means a lens has so little residual aberration left that diffraction is now the binding constraint, and the formal version, the Marechal criterion, sets the bar at a root-mean-square wavefront error under about one-fourteenth of a wavelength. A lens that is diffraction limited at f/4 is superb. "The camera's diffraction-limited aperture is f/7.1" is a statement about pixel pitch and means something entirely different.

What More Megapixels Actually Change

The 61 MP sensor in the Sony a7R V spreads 9,504 pixels across 35.7 mm, which is 3.76 microns per pixel and a Nyquist frequency of 133 line pairs per millimeter. The 24.5 MP sensor in the Nikon Z6III sits at 5.94 microns and 84 line pairs per millimeter. Feed the diffraction MTF those two frequencies and you get the honest version of the megapixel argument.

At its own Nyquist frequency, the a7R V retains 74% of contrast at f/2.8, 49% at f/5.6, 30% at f/8, and 10% at f/11. By f/13.7 the 550 nm cutoff falls below 133 line pairs per millimeter, which is the Nyquist frequency of an ideal monochrome grid at the a7R V's pixel pitch. The Z6III's matching crossing is f/21.6. Read those as ideal-grid landmarks rather than apertures past which the camera cannot alias, because a Bayer array does not sample green at every photosite and its green lattice is not that grid, and because shorter wavelengths cut off higher: at f/14 a 450 nm cutoff still sits at 159 line pairs per millimeter and does not drop below 133 until about f/16.7. The Z6III retains 54% at f/8, 38% at f/11, and 15% at f/16.

The Airy disc at 550 nm grows 1.34 microns for every unit of f-number, so it reaches the size of a pixel far earlier than the folklore about f/11 suggests. The pixel pitches here are computed from each sensor's real dimensions and pixel count: the disc matches the 1.22-micron pixels of an iPhone 17 Pro main camera at f/0.9, the 3.04-micron pixels of a 40 MP Fujifilm X-H2 at f/2.3, the 3.76-micron pixels shared by the 61 MP Sony a7R V and the 102 MP Fujifilm GFX100 II at f/2.8, the Nikon Z8 at f/3.2, and the 24 MP Nikon Z6III at f/4.4. The order is set by pixel pitch alone. Format and sensor size never enter the calculation, and at any given f-number the blur spot is identical on all of them.

Read that the wrong way and you conclude the high-resolution body is worse at small apertures. It is not, and the arithmetic says so plainly. The Airy disc on the sensor is 10.73 microns at f/8 whether the silicon underneath it is Micro Four Thirds, APS-C, full frame, or 44 x 33 medium format. The 102 MP Fujifilm GFX100 II has the same 3.76-micron pitch as the a7R V and behaves identically at the pixel level, on a sensor 1.7 times larger in area that needs less enlargement to reach the same print. Jim Kasson's simulations reach the same result from the other direction: at a fixed print size, finer pixels never make diffraction worse, and generally make it slightly better.

A 61 MP file at f/16 contains more real detail than a 24 MP file at f/16. What it loses is the advantage you paid for. Higher resolution does not create the degradation; it lets you see a degradation that was always in the file.

Why Every Lens Has a Best Aperture

Two things move in opposite directions when you turn the aperture ring. Residual aberration, the part of the blur that comes from the glass failing to bring every ray to the same point, falls as you stop down, because the narrower opening uses less of the lens and the outer zones where the errors are worst stop contributing. Diffraction rises in exact proportion to the f-number. Blur from independent causes adds in quadrature, so the total is the square root of the sum of the squares, and a quantity that falls against a quantity that rises has a minimum somewhere in between. That minimum is the lens's best aperture, and it is why one exists at all.

The position of that peak says something counterintuitive about optical quality. A superb lens is already close to its own build limit wide open, so there is little aberration left for stopping down to remove, and diffraction takes over early: it peaks wide, around f/2.8. A more modest design keeps improving further down the range before diffraction catches it, so it peaks narrow, around f/5.6. The better lens peaks at the wider aperture. What separates them is the height of the peak, not its location.

Three modeled lens designs plotted against the computed diffraction ceiling, each peaking at a different aperture and all three converging near the ceiling by f/16 — Three lens designs against the diffraction ceiling. Everything here is computed rather than measured: the dashed line is the exact diffraction MTF50, and each curve adds that ceiling in quadrature with a residual-aberration term that improves as the lens is stopped down and then flattens at the design's own build quality, plus a fixed sensor term. The excellent design peaks at f/2.8, the modest one at f/5.6, and by f/16 all three sit within a few percent of each other at 86 to 91 percent of the ceiling. The curves model the shape every published lens test shows; they are not measurements of any particular lens.

The right-hand side of that picture is the part worth internalizing. Once the diffraction term dominates, the aberration term stops mattering, and three designs that shared nothing at f/2 arrive at f/16 within a few percent of each other and of the ceiling. Published bench tests from the optics labs show the same shape on real glass, which is worth checking for whatever lens you own before you assume your copy is the exception. Past roughly f/11, better glass has almost nothing left to sell you, because you are no longer measuring the glass.

One caution about the number those tests usually report. MTF50 is the spatial frequency at which contrast has fallen by half, and it sits at roughly 0.4 of the cutoff, not at the finest detail the lens still delivers. It is tempting to take the ratio of two MTF50 figures, square it, and read off an equivalent megapixel count, and the move does not survive contact with what the number is. At f/16 a system keeps transmitting well past its MTF50 frequency, out to the cutoff. MTF50 measures what the aperture cost you in contrast, not how much of the sensor it took away.

That distinction also explains why two labs can test the same lens and disagree. MTF50 measures contrast, and sharpening raises contrast, so any measurement made on a processed file carries the processing in the result. The physics gives you a way to catch it: no unsharpened optical system can beat its own diffraction ceiling. When a published figure implies a lens has resolved past the limit its aperture allows, the sharpening is in the measurement, not in the glass.

Macro Is Where the Arithmetic Turns Brutal

The f-number engraved on the barrel is defined at infinity focus. Rack the lens closer and the image plane moves back, the cone of light reaching the sensor narrows, and the working f-number climbs. For an asymmetric design, the correction depends on the ratio between the exit and entrance pupil diameters, which telephoto constructions push below one and retrofocus wide angles push above it.

The effective f-number in close-up work and the Airy disc it produces. N is the f-number marked on the lens, m is the magnification (1 at life size, 5 at five times life size), and p is the pupil magnification, the diameter of the exit pupil divided by the diameter of the entrance pupil, which is 1 for a symmetric design, below 1 for telephoto constructions, and above 1 for retrofocus wide angles. N-eff is the f-number the optics are actually working at, lambda is the wavelength, and d is the Airy disc diameter at the sensor.

At life size with a symmetric lens, the multiplier is exactly 2. A marked f/8 is a working f/16 and a 21.5-micron Airy disc, which is 5.7 pixels across on a 3.76-micron sensor. A marked f/16 is a working f/32 and 11.4 pixels. To hold the disc to two pixels at 1:1 on that body, you would have to shoot at a marked f/2.8, and no macro lens delivers usable depth of field there. Push to five times life size, the range where the discontinued Canon MP-E 65mm operated, and on that same symmetric assumption a marked f/16 becomes a working f/96 with an Airy disc 129 microns wide, about 34 pixels.

A honey bee photographed head-on at high magnification; the head is about 4 mm across. The photographer built the frame from a stack of 18 exposures at a marked f/5.6 on a bellows rather than stopping down, which is the answer the arithmetic recommends. Photo by Gilles San Martin, CC BY-SA 2.0. Source.

This is why the minimum apertures on macro and macro tilt-shift lenses look absurd and are not. The Canon TS-E 90mm f/2.8L Macro stops to f/45, and Canon's two non-macro tilt-shifts stop at f/22. The macro lenses got the narrow settings because magnification eats most of what they offer. Venus Optics goes the other way on its 25mm 2.5-5X and stops at f/16, and that lens is the best argument in the article for why the pupil term is not a footnote. The symmetric shortcut would call a marked f/16 a working f/96 at five times life size. Enrico Savazzi measured the lens's pupil ratio at 1.79 at that magnification, and running the real formula gives about f/61, a difference of roughly one and a third stops, with an Airy disc near 81 microns rather than 129. Still punishing, and not the number the shortcut produces.

The real answer in macro is not an aperture at all. Shoot at the lens's sharpest setting and buy depth of field with frames. The same arithmetic holds in landscape: at f/22 on a 42 MP full frame body, the diffraction blur circle is 6.5 pixels wide at the plane of perfect focus, while the worst-focused point in a properly spaced f/8 stack carries about 2 pixels of defocus plus 2.4 pixels of diffraction. The stack's worst point beats the single frame's best point. Elia Locardi's landscape course covers the capture and blending side of that workflow.

Sunstars Are the Same Physics, Aimed at Something You Want

A straight edge diffracts light into a fan running perpendicular to itself, the one-dimensional relative of the Airy pattern. A polygonal iris is a ring of straight edges, so a bright point in the frame throws a streak along the normal of every blade. That is the entire mechanism behind a sunstar, and the reason the effect gets stronger as you stop down is that the pupil becomes more sharply polygonal and the edges become a larger fraction of its perimeter.

The blade-count rule is exact for a regular polygon iris. Every edge throws its streak in both directions, so n blades always generate 2n streak directions. The only question is whether any of them land on top of each other. With an even number of blades, opposite blades are parallel and share a normal, so the directions collapse in pairs and you see n rays. With an odd number, no blade is parallel to any other, every direction is distinct, and you see 2n. I checked this by building regular polygon pupil masks and taking the squared magnitude of their two-dimensional Fourier transform, which is exactly the diffraction pattern they produce: five blades gave 10 rays, six gave 6, seven gave 14, nine gave 18, thirteen gave 26, and every recovered ray angle sat on a blade normal.

A sunstar from a stopped-down lens, with the rays countable against the sky. Every ray runs perpendicular to one straight blade in the iris, and the effect strengthens as you stop down because the pupil becomes a sharper polygon. Photo by The wub, CC BY-SA 4.0. Source.

Rounded blades, chosen to keep out-of-focus highlights smooth, wash the star out at wide and middle apertures, because there is very little straight edge left to diffract from. Close them far enough and even a rounded iris folds into a polygon, which is why lenses with rounded blades still throw stars at f/8 and beyond. Cosina's M-mount Voigtlander lenses use 10 or 12 straight blades and are the enthusiast reference for tight, well-defined rays. Among autofocus lenses, the nine-bladed Nikon Z 14-24mm f/2.8 S gives the 18 points the odd-blade rule predicts. More blades does not mean a better star; the energy in the streaks is fixed, so more rays generally means each one is fainter.

Most lenses need about f/11 before the star is pronounced. That is the same aperture the internet spends its time warning you away from, and on a 61 MP body it costs you 90% of your contrast at the pixel level while leaving 77% of it at print-relevant detail. Next time you stop to f/16 for a sunstar over a ridge, count the rays. A nine-bladed lens will give you eighteen, and every one of them is built from light your aperture threw sideways, the same light that is quietly spreading every other point in the frame to 21.5 microns while you watch.

Lead image by D-Kuru/Wikimedia Commons, CC BY-SA 4.0. Source.

Education