The exposure triangle explained in stops, not metaphors

Aperture, shutter and ISO are three ways of spending one currency, the stop. Count in stops and the triangle stops being a metaphor and becomes arithmetic you can check.

Close-up of the curved aperture blades inside a Canon EF-M 32mm f/1.4 lens
Photo: D-Kuru, CC BY-SA 4.0, via Wikimedia Commons

The exposure triangle is the trade between aperture, shutter speed and ISO, and it works because all three are counted in the same unit: the stop, a doubling or halving. Aperture and shutter speed decide how much light actually reaches the sensor; ISO does not add light, it only sets how much the captured signal is amplified and how the camera maps it to brightness. One stop off the shutter can be paid back with one stop of aperture or one stop of ISO, but only the first two change the noise you start with.

Key takeaways
  • One stop is a doubling or halving of light, for aperture, shutter and ISO alike.
  • EV = log₂(N²/t): equal EV means equal light per unit area on the sensor.
  • √2 ≈ 1.41 is the step between full-stop f-numbers, because light follows area.
  • ISO adds no light; it sets gain and the meter's target, so low light, not ISO, causes most noise.
  • EV 15 at ISO 100 is a typical sunlit scene, which is why Sunny 16 works.

The phrase was popularised by Bryan Peterson's book Understanding Exposure, and the top answer on Photography Stack Exchange's long-running thread about it makes the same complaint I have:

  • the three corners are not equal partners;
  • a triangle drawing hides the fourth variable, which is the light in the scene.

So this post skips the bucket-of-water metaphors and does the sums. Every claim below is a number you can check on your own camera.

The story

Late in the afternoon of 1 November 1941, Ansel Adams stopped on a highway shoulder at Hernandez, New Mexico, as the moon rose over the village and its white crosses. In his 1983 book Examples: The Making of 40 Photographs, he wrote that he could not find his Weston exposure meter. He remembered that the moon's luminance was about 250 candles per square foot, placed it on Zone VII of his Zone System, and worked out the exposure from that one number.

He then reversed the film holder to make a duplicate, but the sunlight slid off the crosses seconds too soon, so only one negative of Moonrise, Hernandez exists. His earlier 1943 account says he did read the scene with the Weston Master, so the two versions differ. Either way, the photograph came down to knowing the light in stops.

Wikipedia: Moonrise, Hernandez, New Mexico (Adams's 1943 and 1983 accounts)

What is the exposure triangle, in one equation?

The exposure triangle reduces to one formula: exposure value, EV = log2(N²/t), where N is the f-number and t is the shutter time in seconds. Two settings with the same EV put the same amount of light per unit area on the sensor. ISO then shifts which EV counts as "correct" for a given scene.

The base-2 logarithm is what turns light into stops. Each +1 EV is half the light. At ISO 100, f/16 at 1/125 s gives log2(256 × 125) = log2(32,000), which is 14.97, so EV 15.

Where Sunny 16 comes from

That single number is what the old Sunny 16 rule encodes: in full sun, set f/16 and a shutter time of 1/ISO. The rule works because a typical sunlit scene measures about EV 15 at ISO 100 in the ANSI exposure tables that Wikipedia's exposure value article reproduces.

EV 15typical full-sun scene at ISO 100 (ANSI PH2.7 tables)
2×light per stop, in either direction
√2 ≈ 1.41step between full-stop f-numbers
78 / HsatISO 12232 saturation-based speed formula

Why does each f-stop go up by 1.4 and not by 2?

Because the f-number is a diameter ratio and light depends on area. N is focal length divided by the entrance-pupil diameter, so the light gathered scales with 1/N². To halve the area you shrink the diameter by √2, which is why the full-stop series runs f/1.4, 2, 2.8, 4, 5.6, 8, 11, 16, 22.

Squaring the numbers shows the doubling directly: 2² = 4, 2.8² ≈ 8, 4² = 16, 5.6² ≈ 32. Going from f/2.8 to f/5.6 is not "twice as small", it is a quarter of the light, two stops.

This is also why the f-number, not the physical hole size, is what transfers between lenses: a 50 mm lens at f/4 and a 200 mm lens at f/4 put the same illuminance on the sensor even though the 200 mm lens has a 50 mm wide pupil and the 50 mm lens has a 12.5 mm one.

The diaphragm blades of a Hasselblad L3D-100c camera lens, closed down to a small opening
An iris diaphragm. Closing it by one stop cuts the opening's area in half, which takes a 1.41× change in diameter.Photo: D-Kuru, CC BY-SA 4.0, via Wikimedia Commons

How do aperture, shutter and ISO trade against each other?

One stop on every scale = 2× light Aperture Shutter (s) ISO f/2.8f/4f/5.6f/8f/11 1/10001/5001/2501/1251/60 1002004008001600 Left → right: less light per step for aperture,more light for shutter, more gain (not light) for ISO
The three scales in full stops. Moving one notch on one scale and one notch the opposite way on another leaves brightness unchanged; only the aperture and shutter rows change how much light the sensor actually collects.

One stop of any of the three can be swapped for one stop of another without changing the brightness of the final image. The table shows a run of equal exposures for a sunlit scene at EV 15 and ISO 100, and what each one costs you in side effects.

ApertureShutterISOEV (computed)Side effect you are buying
f/161/125 s10014.97Deep focus, some diffraction softening on small pixels
f/111/250 s10014.89Near-maximum sharpness on most lenses
f/81/500 s10014.97Freezes a walking person
f/5.61/1000 s10014.94Freezes most traffic
f/2.81/4000 s10014.94Shallow focus; many mechanical shutters top out here
f/161/1000 s80017.97Three stops less light captured, so more visible noise

The small differences in the EV column (14.89 versus 14.97) come from the rounded marked values: f/11 is really f/11.3 and f/5.6 is really f/5.66. Cameras use the exact values internally, so the exposures match.

The last row is the important one. At ISO 800 the camera meters the same scene as correctly exposed at EV 18, three stops less light than the first row.

The JPEG looks equally bright, but the sensor collected one-eighth of the photons, and that is where noise comes from.

For how much shutter speed you actually need to stop a moving subject, see my shutter speed chart for moving subjects.

Long exposure of a waterfall in Company Garden, Mussoorie, with the water rendered as smooth white streaks
Company Garden, Mussoorie, shot as a long exposure. Spending stops on time instead of aperture turns falling water into a soft blur.Photo: Navneet Sharma, CC BY-SA 4.0, via Wikimedia Commons

Is ISO really "sensitivity"?

No. On a digital camera the sensor's sensitivity is fixed by its silicon; ISO is an exposure index that tells the meter how much light to aim for, plus a gain setting that amplifies the signal after the photons are already counted. The ISO 12232 standard defines several ways to assign that number, and none of them change how many photons hit a pixel.

ISO 12232 dates from 1998 and was revised in 2006 and February 2019. The 2006 revision added two methods that most cameras now use, and the 2019 revision extended official definitions to speeds above ISO 10,000. The methods are:

  • Saturation-based speed: S = 78/Hsat, where Hsat is the exposure in lux-seconds that just clips the sensor. The constant 78 places an 18% grey at about 12.7% of saturation, leaving roughly half a stop of headroom for highlights.
  • Noise-based speed: the exposure that gives a signal-to-noise ratio of 40:1 ("excellent") or 10:1 ("acceptable"). Rarely used for consumer cameras.
  • Standard output sensitivity (SOS): S = 10/Hsos, the exposure that makes an 8-bit sRGB file read 118, which is 18% grey after gamma. It applies to JPEG output, not raw.
  • Recommended exposure index (REI): whatever the manufacturer judges gives good-looking images. Japanese makers must state whether a rating is SOS or REI.

That last option is why "ISO 800" on two cameras can need slightly different shutter speeds for the same scene, and why raw files from one body can look darker than another's at the same setting. ISO is a calibration choice, not a physical property.

Noise-Optimal Capture for High Dynamic Range PhotographySamuel W. Hasinoff, Frédo Durand, William T. Freeman · CVPR 2010 (MIT CSAIL)

The authors model sensor noise as photon shot noise plus additive noise that enters both before and after the ISO amplifier, and test it on cameras including the Canon 1D Mark III.

Measured in electrons, the additive (read) noise falls as ISO rises, because the post-amplifier noise is divided by a larger gain. So at a fixed shutter time and aperture, a higher ISO gives a cleaner image of the dark parts of a scene, even though it saturates sooner. Their optimised capture sequences gained up to 19 dB of worst-case SNR in the darkest regions over standard ISO 100 bracketing, and more than 10 dB for most DSLRs they modelled. The limit: the advantage depends on each camera's noise profile, and newer "ISO-invariant" sensors show far less of it.

Read the paper (free)

Does high ISO cause noise, or does low light?

Low light causes most of it. Photon arrival is random, so a pixel that collects n photons carries shot noise of about √n, and its signal-to-noise ratio is √n. Four times the light doubles the SNR. Raising ISO is what you do when you could not collect more light, so the two travel together and get blamed on each other.

The Hasinoff paper above separates them cleanly:

What you hold fixedRaise ISO and…Effect on shadow SNR
Shutter time and aperturethe signal is lifted above read-out noiseSlightly better
Final brightness (auto mode)the camera shortens the shutter, so fewer electronsWorse

The second case is what most people experience, which is why "high ISO is noisy" is a useful rule and a wrong explanation.

This also explains the "expose to the right" habit for raw files: give the sensor as much light as the highlights allow, then pull brightness down in editing. I go through that trade in RAW, dynamic range and ETTR.

Exposure value of common scenes at ISO 100
See the data as a table
EV at ISO 100 (EV)
Snow or light sand, full sun16
Typical scene, full sun15
Hazy sun, soft shadows14
Cloudy bright, no shadows13
Heavy overcast12
Open shade, clear sun12
Neon and bright signs9
Bright street scenes8
Night street scenes, window displays7
Offices and work areas7
Home interiors5
Night vehicle traffic5
Full moonlight-3

Source: Wikipedia: Exposure value, Table 2 (from ANSI PH2.7-1973 and PH2.7-1986) · Where the table gives a range (signs 9-10, night streets 7-8, offices 7-8, homes 5-7, full moonlight -3 to -2) the lower value is shown. Each step is one stop, a factor of 2 in light.

Why is distance absent from the exposure triangle?

Because, for anything that is not a point light, distance cancels out. Walk twice as far from a sunlit wall and your lens collects a quarter of the light from it, but the wall's image on the sensor also shrinks to a quarter of the area, so the light per unit area, which is what exposure measures, stays the same.

This is the answer the Photography Stack Exchange thread on the question settles on, and it is why a handheld meter reading taken at the subject matches a reading from across the street, provided the light falling on the subject is the same. The image-plane illuminance depends on the scene's luminance and on N, nothing else.

Distance does matter in two places:

  • Flash, where the light source itself is close and obeys the inverse-square law;
  • tiny bright objects such as stars, whose image stays a point however far away they are.
Why distance cancels out of exposure Subject distance d Light collected: L Image area: A L / A = exposure distance 2d Light collected: L/4 Image area: A/4 (L/4) / (A/4) = same
Doubling the distance quarters the light from the subject and quarters the area of its image, so exposure per unit area does not change. Flash is the exception, because the light source moves too.

How do phones break the triangle?

Phones keep aperture fixed and replace one long exposure with a burst of short ones. A main camera usually has a single fixed f-number, so the triangle collapses to two corners, and the phone adds a fourth: the number of frames it merges. Merging N equal frames averages away random noise and improves SNR by about √N, so 8 frames buy roughly 2.8× the SNR, or 1.5 stops.

Burst photography for high dynamic range and low-light imaging on mobile camerasSamuel W. Hasinoff, Dillon Sharlet, Ryan Geiss, Andrew Adams, Jonathan T. Barron, Florian Kainz, Jiawen Chen, Marc Levoy · ACM SIGGRAPH Asia 2016 (Google Research)

This is the HDR+ pipeline behind Google's phone cameras. Instead of bracketing, it captures 2 to 8 raw frames at one constant exposure, deliberately underexposed so highlights do not clip, then aligns and merges them.

Its exposure schedule is a written-out exposure triangle: hold gain at minimum and let exposure time rise to 8 ms; then hold 8 ms and raise gain up to 4×; then raise both together to a maximum of 100 ms and 96× gain. Underexposure (dynamic range compression) is capped at 8×, three stops, because in their database of 26,071 scenes only 4% needed more. A 12-megapixel result took about 4 seconds on the phone. The limit: merging assumes frames can be aligned, so fast motion still blurs or falls back to a single frame.

Read the paper (free)

That 8 ms ceiling before gain rises is a design choice to fight hand shake, and it tells you the phone values sharpness over low ISO. If you want the longer story of how computation replaced glass, read how phone cameras use computational photography.

An early Weston Master handheld light meter with its dial calculator
An early Weston Master meter. Its dial is a mechanical EV calculator: set the film speed, read the light, and every aperture and shutter pair on the ring gives the same exposure.Photo: Colgill, CC BY-SA 4.0, via Wikimedia Commons

Calculate an equivalent exposure

Enter the settings that gave you a good picture. The calculator gives the exposure value they represent, the scene brightness at ISO 100 they correspond to, and the shutter speed you would need for the same result at f/2.8 and the same ISO.

Exposure value and equivalent exposure

Is the exposure triangle still useful?

Yes, as a ledger rather than a picture. Write every change in stops and the bookkeeping is exact: shutter from 1/250 to 1/1000 s is −2 stops, opening f/8 to f/4 is +2 stops, net zero.

What the triangle drawing gets wrong is treating ISO as a third source of light. It is a gain and metering setting, so the honest order of priority is:

  1. Get light. Move, add a lamp, or wait for it.
  2. Choose aperture for depth of field.
  3. Choose shutter for motion.
  4. Use ISO last to make the file bright enough.

Aperture also controls how much of the scene is sharp, which is where the next post picks up: what really controls depth of field. Colour is the other half of a correct exposure, covered in white balance and colour temperature.

Questions people ask

What is the Sunny 16 rule?

In bright sun, set the aperture to f/16 and the shutter time to 1 divided by the ISO, so 1/100 s (or the nearest 1/125 s) at ISO 100. It works because a front-lit sunny scene is about EV 15 at ISO 100. Open one stop for hazy sun, two for cloudy bright with no shadows, and three for heavy overcast or open shade.

Does doubling the ISO double the noise?

Not directly. Noise in the final image depends mainly on how much light the sensor collected. Doubling ISO while keeping the same brightness halves the light collected, which cuts SNR by about √2 (shot noise), so noise looks roughly 1.4 times worse, not twice as bad. At a fixed shutter and aperture, raising ISO can even make shadows slightly cleaner on many sensors.

Is one stop the same on every camera?

Yes, a stop is always a factor of 2 in light. What varies is calibration: two cameras may assign slightly different ISO numbers to the same gain, because ISO 12232 lets makers use the SOS or REI methods. Expect differences of a third of a stop or so between brands.

Why do my photos come out the same brightness at f/2 and f/8 in auto mode?

Because the camera compensates. In aperture priority it lengthens the shutter by four stops (16 times) when you go from f/2 to f/8, or raises ISO if it hits its shutter limit. The brightness stays put; the depth of field, motion blur and noise change instead.

Sources

  1. Wikipedia: Moonrise, Hernandez, New Mexico (Adams's 1943 and 1983 accounts of the exposure)
  2. Wikipedia: Exposure value (EV formula and ISO 100 scene table from ANSI PH2.7-1973 and PH2.7-1986)
  3. Wikipedia: Film speed (ISO 12232 methods, the 78/Hsat and 10/Hsos formulas, 1998, 2006 and 2019 revisions)
  4. ISO 12232:2019 Photography: Digital still cameras: Determination of exposure index, ISO speed ratings, standard output sensitivity, and recommended exposure index (2019)
  5. Hasinoff, Durand and Freeman: Noise-Optimal Capture for High Dynamic Range Photography, CVPR (2010)
  6. Hasinoff et al.: Burst photography for high dynamic range and low-light imaging on mobile cameras, SIGGRAPH Asia (2016)
  7. Photography Stack Exchange: What is the "exposure triangle"?
  8. Photography Stack Exchange: Why is distance absent from the exposure triangle?
  9. Photography Stack Exchange: What is the "Sunny 16" rule of thumb?
  10. Cambridge in Colour: Camera exposure
  11. YMCinema: Sony IMX925 global shutter sensor enters production (2026)
  12. YMCinema: Sony IMX947 global shutter sensor (2026)
exposure triangleexposure valueapertureshutter speedISO 12232Sunny 16noise
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