Diffraction Limit: Why More Megapixels Don't Mean More Detail

May 10, 2026Technology

The Airy Disk — the physical boundary

Every sensor manufacturer promotes more megapixels — but physics sets a hard limit. Beyond aperture f/8, diffraction dominates, and additional megapixels only inflate the file without any real gain in detail. Gigapixel GmbH — the specialized large-format stock portal: real gigapixel captures from 100 MP, no AI upscaling.

Light is a wave. When light passes through the aperture opening of a lens, it is not just refracted in the geometric optics sense — it is diffracted. Every point in the scene is imaged on the sensor not as a sharp point but as a blurred circle: the Airy disk. This point spread function is the fundamental limit of every optical system.

The formula reads: d = 2.44 × λ × N

where λ is the wavelength of light and N is the f-number. At green light (500 nm) and aperture f/8, this yields an Airy diameter of d ≈ 9.76 µm.

In concrete terms: a single diffraction point at f/8 covers more than 4 pixels of a 50 MP sensor (a 2×2 pixel block). This means every point of light in the scene is smeared across at least 4 pixels — no matter how many more pixels the sensor offers.

As Wilson et al. state in the Leica Science Lab: "The spread of the diffraction-limited PSF is approximated by the diameter of the first null of the Airy disk". The Airy disk is not a theoretical construct — it is the measurable limit of every real optical system.

Diffraction limit per aperture — the numbers

The more the aperture is closed, the larger the Airy disk becomes — and the less usable resolution remains on the sensor:

• f/5.6 → Airy ≈ 6.8 µm → approx. 30–40 MP full frame usable
• f/8 → Airy ≈ 9.76 µm → approx. 30–50 MP full frame usable (Q1, Q3)
• f/11 → Airy ≈ 13.4 µm → effectively only 15–25 MP (diffraction dominates)
• f/16 → Airy ≈ 19.5 µm → effectively only 8–12 MP

The numbers are relentless: anyone shooting at f/16 has effectively only 8–12 MP of real image information, even with a 60 MP sensor. The rest is diffraction blur.

The rule applies across formats: APS-C reaches the limit earlier because its pixels are smaller. Medium format reaches it later because its pixels are larger. The physics of diffraction is universal — it affects every format equally; only the threshold shifts.

The Sweet Spot — f/8 as the "Golden Number"

Aperture f/8 holds a special position. Cabezos-Bernal et al. (2021) define f/8 as the optimal balance between lens aberrations and diffraction — and explicitly call it the "Golden Number."

The logic: at wide apertures (f/2.8, f/4), lens aberrations dominate — spherical errors, chromatic errors, coma. Stopping down slightly reduces these errors dramatically and improves overall sharpness. But beyond f/8, diffraction begins to dominate. Stopping down further degrades the image. f/8 is the point where aberrations are already minimized, yet diffraction has not yet taken over.

Q3 states: "f/8 is explicitly recommended for maximizing sharpness."

Other gigapixel systems, such as Kopf et al., cite f/11 as practical — not because f/11 is optically superior, but because deeper scenes require greater depth of field and the diffraction loss is accepted. The sweet spot depends on the subject, but the physics remain the same.

30–50 MP — the practical limit for full frame

At aperture f/8, full frame reaches its physical resolution limit at approximately 30–50 MP. More pixels on the same sensor then amount to "Empty Magnification" — enlargement without information gain.

Thomson et al. (2022) document in their MCAM study a resolution decline from 18 µm to 36 µm as pixel count increases, directly attributable to diffraction. Geometric lens errors grow non-linearly — commercial optics deliver less than 50 MP of genuine optical information in practice.

The consequence: a 100 MP full-frame single shot at f/11 yields diffraction blur. It provides no detail advantage over a 25 MP image at the same parameters. The file becomes four times larger; the image information remains the same.

How Gigapixel circumvents the limit — Stitching

If the full-frame single shot is limited to 30–50 MP, how do you achieve true gigapixels? The answer: by increasing the effective aperture area — through stitching.

Larger sensors (medium format) have larger entrance pupils and therefore fewer diffraction problems at the same f-number. But even the medium-format sensor has its limit. The decisive step is multi-row stitching: hundreds of individual shots at f/5.6 to f/8, each in the sweet spot, combining into a single image with genuine depth of detail (Q3, Q5).

The 1.36 gigapixel example from Cabezos-Bernal et al. (2021) demonstrates this: 90 individual shots (10×9 grid), each in the optimal diffraction range, yield a composite image in which every pixel carries genuine optical information.

Brady et al. (2012) summarize the physics: "Diffraction-limited resolution is proportional to the aperture area". Stitching increases the effective aperture area without opening the aperture. The diffraction limit is not violated — it is circumvented.

The result: real pixels everywhere, no diffraction blur, no AI upscaling.

Viewing distance and print resolution — what the eye actually sees

The full physiology of the eye with all figures on foveal resolution (94 ppd, 60 arcseconds practical) can be found in the article Physiological Limit of the Human Eye.

Here the focus is: What does diffraction mean for printing practice?

For stretch ceilings and aluminum composite at 1–3 m viewing distance, 70–120 ppi are sufficient. The Gigapixel rule: 600 ppi is a museum standard, not relevant for any practical application. What matters is the viewing distance.

Printer output works at 1,440 dpi — this is the internal machine resolution of direct printing (12 colors, Lindner Group / REPRO ONLINE), not equivalent to the ppi the eye perceives in the print. The swissQprint Karibu 2 achieves 1,350 dpi addressable and 2,540 dpi visual.

The decisive formula: diffraction-limited resolution × viewing distance = required source file resolution. For diffraction-limited captures (f/11), a lower source file resolution suffices compared to diffraction-free captures (f/5.6) — making format choice directly relevant (→ File Formats Compared).

For printing on large surfaces this means: anyone bringing source data to 20 m² needs to know how many of their pixels carry genuine diffraction information. Only these determine true print quality (→ Print Methods Compared).

Real vs. AI upscaling — what physics decides

Real gigapixels mean: every pixel carries physical information captured within the diffraction limit by the sensor. AI upscaling, by contrast, generates textures that never reached the sensor — and produces semantic hallucinations at extreme magnification.

Moser et al. (2024) demonstrate in their Chain-of-Zoom study: real pixel crops preserve the "recursive fractal complexity of nature". AI upscaling cannot invent this complexity — it interpolates based on training data that captures natural structures only statistically.

At the pattern level, the difference is clearly distinguishable: diffraction patterns (Airy disks) exist in the original, not in the AI upscale. Anyone printing an image at 20 m² and verifying the original data at the pattern level can distinguish real pixels from generated ones.

Conclusion: For maximum informational accuracy — heritage conservation and large-format printing — real stitching is irreplaceable. AI upscaling has its place in aesthetics, not in documentation.

Practice Table — Resolution reduction by aperture and sensor

The table shows: anyone shooting at f/16 on APS-C has effectively 5–8 MP — no matter what the sensor manufacturer promises.

Sources

- Q1: Wikipedia (2023): Diffraction-limited system. Airy diameter d = 2.44λN; at f/8 → 9.76 µm.

- Q2: Wilson, M. et al. (2023): Leica Science Lab — Microscopic Resolution. Sparrow criterion, NA dependence.

- Q3: Cabezos-Bernal et al. (2021): Documenting Paintings with Gigapixel Photography. f/8 as "Golden Number", 1.36 GP via 90 shots.

- Q4: Thomson, E. E. et al. (2022): MCAM — Gigapixel imaging, eLife. USAF test targets, resolution decline 18→36 µm.

- Q5: Brady, D. J. et al. (2012): Multiscale gigapixel photography, Nature. Diffraction-limited pixels proportional to aperture area.

- Q6: Ashraf, Chapiro & Mantiuk (2025): Resolution Limit of the Eye. 94 ppd foveal, 60 arcseconds practical.

- Q7: Lindner Group (2026) / REPRO ONLINE (2024): Print Standards. 1,200–2,400 dpi, 1,440 dpi direct print, 12 colors.

- Q8: swissQprint (2026): Karibu 2 Specs. 1,350 dpi addressable, 2,540 dpi visual.

- Q9: Moser, B. B. et al. (2024): Chain of Zoom (CoZ). Real vs. AI upscaling, semantic errors at extreme magnification.

Gigapixel GmbH — the specialized large-format stock portal: real gigapixel captures from 100 MP, no AI upscaling. Print licenses from €90.