What if your camera were the size of a building?

Someone already turned an aircraft hangar into a pinhole camera. The physics of that 6 mm hole shows why a building-sized camera is barely sharper than a shoebox, and what you would have to add to beat it.

The Giant Camera, a walk-in camera obscura building shaped like a camera, on the coast at San Francisco
Photo: Dietmar Rabich, CC BY-SA 4.0, via Wikimedia Commons

If your camera were the size of a building, it would work, and it has been done: in 2006 six artists turned a 49 m-wide aircraft hangar in California into a pinhole camera and made a 34 × 9.8 m photograph with a 35-minute exposure. But a building-sized pinhole camera is not very sharp. Diffraction sets the best pinhole at about 6 mm for a 24 m-deep camera, so every point blurs to about a centimetre, and the whole giant print holds only about 2.6 million resolvable spots.

Key takeaways
  • It has been done: The Great Picture (2006) used a hangar as a pinhole camera and made a 34 × 9.8 m print.
  • The best pinhole is \( \sqrt{2.44 f\lambda} \): about 5.7 mm for a 24 m-deep camera, and the artists used just under 6 mm.
  • Each point blurs to about 11 mm, so the giant print holds only about 2.6 million resolvable spots.
  • Detail across the frame grows only with √(camera depth), while the f-number climbs to about f/4,000.
  • Put a 39 m mirror in the wall and the air, not the optic, limits sharpness unless adaptive optics correct it.

Sharpness grows only with the square root of size; to go further you need a lens or mirror, at which point your building is a telescope.

This is a tier 3 what-if, but it is the rare absurd premise that history has already tested. A room with a hole in one wall is the oldest camera there is, the camera obscura, described by scholars for more than a thousand years.

So the real question is what scale does to an image:

  • brightness;
  • exposure time;
  • and, above all, detail.

Let me solve it properly, with the hangar as the test case.

34 × 9.8 msize of The Great Picture, made in a hangar in 2006
~6 mmpinhole that formed it, 4.6 m above the floor
35 minexposure time on gelatin silver emulsion
3,200 MPRubin Observatory's LSST camera, first images June 2025

Has anyone built a camera the size of a building?

Yes. The Great Picture was made in 2006 inside a decommissioned F-18 hangar at the former El Toro Marine Corps Air Station in Irvine, California, and Guinness certified it as both the world's largest camera and the largest photograph.

The story

El Toro was being turned into the Orange County Great Park, and six photographers, Jerry Burchfield, Mark Chamberlain, Jacques Garnier, Rob Johnson, Douglas McCulloh and Clayton Spada, wanted one last record of it, made by the base itself. Their Legacy Project sealed hangar #115 against light, drilled a hole just under 6 mm wide in its metal door, 4.6 m above the ground, and hung a seamless muslin sheet coated with 80 litres of silver halide emulsion inside. The exposure took 35 minutes. Developing it was the hard part: about 80 volunteers worked for five hours in a vinyl tray the size of an Olympic pool, with chemicals pumped in and the print washed with fire hoses from two hydrants. The 111 × 32 ft negative print was unveiled in the hangar on 12 July 2006, and it went on to show in Pasadena, Beijing and, in 2014, the Smithsonian's National Air and Space Museum.

Wikipedia: The Great Picture (the Legacy Project, hangar #115 and exhibitions)

The numbers are worth listing, because they set the problem. The hangar was about 49 m wide, 14 m high and 24 m deep. The pinhole sat 4.6 m above the floor, and the emulsion-coated sheet hung about 24 m from the hole. Six artists and about 400 assistants built it. The bill of materials:

JobMaterialAmount
Light-proofingBlack plastic sheeting2,200 m²
Light-proofingFoam gap filler4,900 litres
Light-proofingBlack tape2.4 km
Light-proofingBlack paint150 litres
"Film"Gelatin silver halide emulsion on muslin80 litres
DevelopingDeveloper / fixer2,300 / 4,500 litres
The hangar as a pinhole camerafrom the skyfrom the groundhole just under 6 mm, 4.6 m upground lands at the topsky lands at the bottomabout 24 m deep14 m high, 49 m wide, sealed against lightemulsion-coated muslin on the far wall
The hangar camera in side view, to scale for depth and height. Rays cross at the hole, so the 34 × 9.8 m print formed upside down on the far wall.
An 18th-century engraving showing light from an outdoor scene passing through a small hole into a dark room and forming an inverted image on the far wall
The camera obscura principle, from James Ayscough's 1755 account of the eye: a dark room, a hole, and an upside-down picture of the world on the wall. The Great Picture used exactly this, at hangar scale.Photo: James Ayscough, Public domain, via Wikimedia Commons
Building-scale cameraYearOpticImageWhat limits it
The Great Picture, El Toro hangar, California2006~6 mm pinhole, ~24 m to the screen34 × 9.8 m print, 35 min exposureDiffraction plus pinhole size: ~1 cm blur
Giant Camera, San Francisco (a walk-in camera obscura)Optics unchanged since 1946Rotating mirror and lensLive image on a viewing tableLens quality, daylight
Vera C. Rubin Observatory, ChileFirst images 20258.4 m mirror (6.5 m effective)3.2 gigapixels per shot, 9.6 square degreesAtmospheric seeing
Extremely Large Telescope, ChileUnder construction39 m mirror of 798 segments, 978 m²Instruments behind adaptive opticsDiffraction, if the atmosphere is corrected

How big should the hole be in a building-sized camera?

The best pinhole for a camera of depth \( f \) is about \( d = \sqrt{2.44\,f\lambda} \), which for a 24 m-deep hangar and green light is 5.7 mm, and the artists used just under 6 mm. Their hole was almost exactly the physics optimum.

Two effects fight:

  • Geometric blur. A big hole projects a big blob of light for every point in the scene.
  • Diffraction. A small hole spreads light by the bending of waves at an edge, which grows as the hole shrinks.

The pinhole formula of Petzval (1857), Lord Rayleigh (1891) and Young (1971) sits where the two balance.

Blur vs pinhole size, 24 m-deep cameraPinhole diameter (mm)Blur (mm)04812160102030best: 5.7 mm hole, ~11 mm blurhole sizediffractiontotal blur
Geometric blur (dashed) equals the hole width; diffraction blur (dotted) is 2.44λf/d. Their sum bottoms out at d = √(2.44λf), about 5.7 mm for f = 24 m and λ = 550 nm.
  1. Assumptions. Light of wavelength \( \lambda = 550 \) nm (the middle of the visible band); a screen at distance \( f \) behind a hole of diameter \( d \); a distant scene, so rays through the hole are nearly parallel.
  2. Geometric blur. A point in the scene lights a disc on the screen about as wide as the hole: \( b_{\text{geo}} \approx d \).
  3. Diffraction blur. A circular aperture spreads light into an Airy disc whose diameter, to the first dark ring, is
    \[ b_{\text{diff}} \approx 2.44\,\frac{\lambda f}{d} \]
  4. Minimise the total. Take \( b(d) = d + 2.44\lambda f/d \). Setting \( db/dd = 1 - 2.44\lambda f/d^2 = 0 \):
    \[ d_{\text{opt}} = \sqrt{2.44\,\lambda f}, \qquad b_{\min} = 2\,d_{\text{opt}} \]
  5. Plug in the hangar. \( d_{\text{opt}} = \sqrt{2.44 \times 550\times10^{-9}\ \text{m} \times 24\ \text{m}} = \sqrt{3.22\times10^{-5}\ \text{m}^2} \approx 5.7 \) mm. With the real 6 mm hole, \( b = 6 + 2.44 \times 550\times10^{-9} \times 24 / 0.006 \) mm \( \approx 6 + 5.4 = 11.4 \) mm.
  6. Count the detail. Across the 34 m print: \( 34{,}000/11.4 \approx 2{,}990 \) resolvable spots; down the 9.8 m height, about 860. Total \( \approx 2.6 \) million spots. Sampled at two pixels per spot (the Nyquist rule), that is roughly what a 10 MP sensor records.
The largest photograph ever made holds about as much detail as a decent phone picture. It is enormous, not sharp.

Walk up close and every edge is a centimetre-wide soft gradient, which is part of its beauty.

Why doesn't a bigger camera give a much sharper picture?

Because the number of resolvable spots across a pinhole image grows only with the square root of the camera's depth. Make the camera 100 times deeper and the image gets 100 times wider but only 10 times more detailed across, and, as the next section shows, 100 times dimmer.

The scaling law

Here is the scaling law. Fix the field of view at angle \( \theta \). The image width is \( W = 2f\tan(\theta/2) \), and each spot is \( b_{\min} = 2\sqrt{2.44\lambda f} \) wide, so

\[ N_{\text{across}} = \frac{W}{b_{\min}} = \tan\!\left(\tfrac{\theta}{2}\right)\sqrt{\frac{f}{2.44\,\lambda}} \;\propto\; \sqrt{f} \]

The total number of spots, across times down, therefore grows in proportion to \( f \), the camera depth. With a 60° view:

CameraDepthOptimal holeResolvable spots
Shoebox30 cm0.63 mmabout 50,000
Hangar24 m5.7 mmabout 4 million
Skyscraper on its side300 mabout 20 mmabout 50 million

The chart shows the curve.

80× deeper camera, only ~9× more detail acrossShoebox, 30 cmblur spot 1.3 mm~50,000 spotsHangar, 24 mimage 80× wider, spots 9× widerblur spot 11 mm · ~4 million spotsNot to scale: the hangar frame would be 80 times wider than the shoebox frame.
Spots across the frame grow as √f: going from 0.3 m to 24 m multiplies the image width by 80 but the blur spot by about 9, so detail across rises only about 9-fold.
Resolvable spots in an optimal pinhole image vs camera depth (60° view, 3:2 frame)
See the data as a table
Million resolvable spots ( million)
0.1 m0.017
0.3 m0.05
1 m0.17
3 m0.5
10 m1.66
24 m3.97
100 m16.6
300 m49.7

Source: Computed from the optimal-pinhole relation d = √(2.44 f λ), λ = 550 nm (Young 1971, as summarised in the Pinhole camera article) · Each spot is 2d wide. Doubling the camera depth doubles the total spots and adds only 41% more spots across.

A 50 million-spot image is still less than a 200 MP phone sensor claims. That is the humbling result: with a pinhole, no building on Earth beats a good lens the size of your thumbnail.

The Vera C. Rubin Observatory building and dome on a snowy mountaintop in Chile
The Rubin Observatory on Cerro Pachón, Chile: a building-sized camera that works because it uses an 8.4 m mirror instead of a hole, and points up through thin, dry air.Photo: NOIRLab/NSF/AURA/C. Corco, CC BY 4.0, via Wikimedia Commons

How long would the exposure be?

A building-sized pinhole camera runs at around f/4,000, so even in sunshine it needs minutes to hours, depending on the emulsion. The Great Picture's 35 minutes is consistent with a slow photo-paper-like emulsion.

  1. Work out the f-number. \( N = f/d = 24{,}000\ \text{mm} / 6\ \text{mm} = 4{,}000 \). A normal lens might be f/8; this is 500 times narrower.

    Calculated in this post from the 24 m camera depth and just-under-6 mm pinhole in Wikipedia: The Great Picture; f/8 is only an example of a normal lens.

  2. Use the sunny 16 rule. In bright sun, a correct exposure at f/16 is about \( 1/\text{ISO} \) seconds. Exposure time scales with \( N^2 \):
    \[ t = \frac{1}{\text{ISO}}\left(\frac{N}{16}\right)^2 = \frac{1}{\text{ISO}}\left(\frac{4000}{16}\right)^2 = \frac{62{,}500}{\text{ISO}}\ \text{s} \]

    Calculated in this post from the sunny 16 rule (f/16 at a shutter speed of 1/ISO) and N = 4,000.

  3. Solve for the emulsion speed. With \( t = 35 \) min \( = 2{,}100 \) s, \( \text{ISO} \approx 62{,}500/2{,}100 \approx 30 \). If the effective screen distance was 17 m (the artists' best-coverage distance), \( N \approx 2{,}800 \) and \( \text{ISO} \approx 15 \).

    Calculated in this post from the 35-minute exposure and the 55 ft (about 17 m) best-coverage distance stated in Wikipedia: The Great Picture, and the sunny 16 rule.

  4. Sanity check. Photographic paper is a very slow emulsion for pinhole work, so an effective speed of 15 to 30 for a hand-coated emulsion is plausible. This is my estimate, not a figure from the project, and the light on the day is unknown.

    Calculated in this post from the effective speed of 15 to 30 solved above (Wikipedia: The Great Picture).

For comparison, a 30 cm shoebox at its optimal 0.63 mm hole runs at about f/470: under a minute on the same emulsion. Bigger pinhole cameras are slower, because \( N = f/d = \sqrt{f/(2.44\lambda)} \) also grows with \( \sqrt{f} \). Everything about a giant pinhole camera scales against you except the size of the print.

Use this to check what you just read: set the depth of your own room or building and see the hole, the blur and the exposure it needs.

Design your own building-sized pinhole camera

What if you put a lens in the wall instead?

With a lens, detail is no longer set by the hole but by the lens's own diffraction limit, \( \theta = 1.22\lambda/D \), so a building-sized lens could in principle resolve detail thousands of times finer. In practice the air between you and the subject limits it to what a 5 to 20 cm lens can already do, unless you add adaptive optics.

  1. Diffraction limit of a huge aperture. For a 39 m mirror, the size of the ELT:
    \[ \theta = 1.22\,\frac{\lambda}{D} = 1.22 \times \frac{550\times10^{-9}}{39} \approx 1.7\times10^{-8}\ \text{rad} \approx 3.5\ \text{milliarcseconds} \]
    At 10 km that would separate points 0.17 mm apart.

    Calculated in this post from the 39 m mirror in ESO: ELT facts, green light of 550 nm, and 10 km as an example distance.

  2. Now add the atmosphere. Turbulent air bends light randomly. Its scale is the Fried parameter \( r_0 \), which at visible wavelengths runs from about 5 cm at sea level to 20 cm at the best sites. Any aperture larger than \( r_0 \) is limited by the air, not the optic.
  3. The real limit. Good seeing is about 1 arcsecond, \( 4.8\times10^{-6} \) rad. At 10 km that is a 4.8 cm blur. Near the ground, where warm air rises off roads and roofs, it is worse.

    Calculated in this post from good seeing of about 1 arcsecond (ATNF/CSIRO) and 10 km as an example distance: 4.8 × 10⁻⁶ rad × 10,000 m.

So a building-sized lens pointed across a city would collect enormous amounts of light (a 39 m mirror gathers 978 m²) but see no finer than a lens the size of a dinner plate.

Why the giant cameras we actually build use mirrors

That is why the giant cameras humans actually build follow three rules:

  • they point up, through the least air;
  • they sit on dry mountains, above most turbulence and water vapour;
  • they use adaptive optics, deformable mirrors that cancel the turbulence hundreds of times a second.

The square-cube law adds a second reason they use mirrors: a solid glass lens's mass grows with the cube of its diameter while its supporting edge grows only with the diameter, so beyond about a metre lenses sag.

The largest refractor ever used for astronomical research, at Yerkes Observatory, has a 1.02 m (40-inch) lens; a 39 m one would need to be segmented mirrors, which is exactly what the ELT is.

LSST: From Science Drivers to Reference Design and Anticipated Data ProductsŽeljko Ivezić et al. · The Astrophysical Journal · 2019 (arXiv 0805.2366)

This reference paper describes the camera and telescope of what is now the Vera C. Rubin Observatory: an 8.4 m primary mirror (6.5 m effective aperture), a 9.6 square-degree field of view and a 3.2-gigapixel camera. Each pointing is a pair of 15-second exposures, the system covers about 10,000 square degrees in three nights, and over ten years it images about 18,000 square degrees roughly 800 times, reaching a single-visit depth of r ≈ 24.5 and a stacked depth of about 27.5. The planned database holds about 32 trillion observations of 40 billion objects. The limit: it is a design document; real performance is being measured now that operations have started.

Read the paper (free)
ELT facts and figuresEuropean Southern Observatory · official project data · 2025

ESO's project page lists the Extremely Large Telescope's 39 m primary mirror, its 978 m² collecting area and its 798 hexagonal segments, with technical first light planned for 2029. In April 2025 ESO reported the dome had reached its highest point and the project was more than 60% complete. The limit: schedules for projects of this size move; treat dates as plans, not facts.

Read the facts (free)
Panorama of the Extremely Large Telescope construction site on a desert mountaintop, with the circular dome structure rising
The Extremely Large Telescope under construction in the Atacama Desert in 2022. Its 39 m mirror is made of 798 segments, because a single lens or mirror that size would sag under its own weight.Photo: G. Hüdepohl (atacamaphoto.com)/ESO, CC BY 4.0, via Wikimedia Commons

So what would a building-sized camera be good for?

A building-sized camera is good for two things: making an image people can walk into, and collecting light. It is bad at detail, unless it is also a telescope with adaptive optics.

That second use is where the money goes. Rubin's 3.2-gigapixel camera released its first images on 23 June 2025, including a view of the Trifid and Lagoon nebulae built from 678 exposures taken in just over seven hours. The focal plane is more than 60 cm across and holds 189 individual sensors.

For the everyday camera in your hand, the lessons are the same physics at small scale: diffraction caps sharpness when you stop down too far, which I cover in depth of field explained and the exposure triangle. The scaling argument is also why megapixels matter less than you think.

Verdict: possible, but huge is not sharp

A building-sized camera already exists: a 6 mm pinhole in a 24 m-deep hangar, 35 minutes, a 34 m print. Physics then caps it.

The best pinhole is \( \sqrt{2.44\lambda f} \approx 5.7 \) mm, the blur is about 11 mm, and the whole print holds about 2.6 million resolvable spots. Detail grows only with \( \sqrt{f} \) across, and exposure time grows with \( f \).

Put a 39 m mirror in the wall and the diffraction limit drops to 3.5 milliarcseconds, but across a city the air (\( r_0 \) of 5 to 20 cm) blurs it back to the level of a dinner-plate lens. The version that works is the one astronomers build: segmented mirrors, a dry mountain and adaptive optics.

Calculated in this post from the 39 m mirror in ESO: ELT facts and the 5 to 20 cm Fried parameter range in Wikipedia: Astronomical seeing.

What if next?

The opposite approach, one fixed small lens for a whole year, is in what if you could only shoot one focal length. More from the studio at Kala.

Questions people ask

What is the largest camera ever made?

Guinness certified the hangar used for The Great Picture (2006) as the world's largest camera; it was about 49 m wide, 14 m high and 24 m deep. The largest digital camera is the LSST Camera at the Rubin Observatory, with 3.2 gigapixels across 189 sensors.

Why is a pinhole camera image blurry?

Because a hole cannot focus. A big hole lets each point spread into a disc as wide as the hole, and a small hole spreads light by diffraction. The best compromise, \( d = \sqrt{2.44 f\lambda} \), still leaves a blur about twice that size.

Does a bigger lens always give a sharper photo?

Only up to the atmosphere. Beyond an aperture of about 5 to 20 cm in visible light, turbulence sets the blur at around 1 arcsecond in good conditions, so larger lenses gather more light but not more detail unless adaptive optics correct the air.

Can you turn a room into a camera obscura?

Yes. Black out a room facing a bright view and make a hole a few millimetres wide in the window covering; an upside-down image of the outside appears on the opposite wall. For a 4 m room the optimal hole from the formula is about 2.3 mm, and give your eyes several minutes in the dark before you judge the image.

Calculated in this post from the optimal-pinhole relation d = √(2.44 f λ) (Wikipedia: Pinhole camera) with these example inputs: a 4 m room and 550 nm light.

Sources

  1. Wikipedia: The Great Picture (source for the hangar size, 6 mm pinhole, 35-minute exposure and materials; its references include the Guinness certification)
  2. PR Newswire: Guinness certifies world's largest photograph and camera (source for the Guinness records)
  3. Wikipedia: Pinhole camera (source for the Petzval, Rayleigh and Young optimal-pinhole formulas)
  4. Young: Pinhole optics, Applied Optics (1971) (origin of d = √(2.44 fλ))
  5. Ivezić et al.: LSST: From Science Drivers to Reference Design, ApJ (2019) (source for the 8.4 m mirror, 3.2 GP camera and survey figures)
  6. Rubin Observatory: First images from the 3,200-megapixel camera (2025) (source for the first-light date, 678 exposures, 189 sensors)
  7. ESO: ELT facts (source for 39 m, 978 m², 798 segments and planned first light)
  8. ESO announcement ann25002 (2025) (source for the dome milestone)
  9. Wikipedia: Astronomical seeing (source for r₀ of 5–20 cm and 1 arcsecond seeing)
  10. Wikipedia: Yerkes Observatory (source for the 40-inch refractor)
  11. Wikipedia: Camera Obscura, San Francisco (source for the Giant Camera)
camera obscurapinhole cameraThe Great PicturediffractionRubin ObservatoryExtremely Large Telescopeastronomical seeinglargest photograph
Sapta

Sapta (Saptarshi Sarkar) is a mechanical engineer turned marketer in Bengaluru and a solo touring rider. He writes the seven Saptaverse blogs. Work with him or read more of Artblog.

Keep reading in Artblog

From the other blogs