Depth of Field: What Actually Controls It

Four variables, and one of them is a decision you make

Depth of field is the range of subject distances that render acceptably sharp in the final image. Exactly four quantities determine it: the f-number N, the focal length f, the distance s from the lens to the plane you focused on, and the circle of confusion c. The first three are settings and geometry. The fourth is not on the camera at all — it is a threshold you pick, and picking it is the step almost every explanation skips.

A point in the scene that is not exactly on the focused plane does not land as a point on the sensor. It lands as a small disc. The further that point sits from the focused plane, the wider the disc. Depth of field is simply the band of distances over which that disc stays under some diameter — and that diameter is c. Set c generously and the band is wide. Set it tightly and the band shrinks. Nothing about the lens changed.

The formulas, and the check that proves they are right

Everything runs through the hyperfocal distance, H = f²/(N·c) + f. From it the two limits follow: near = s(H − f)/(H + s − 2f) and far = s(H − f)/(H − s). When s reaches H the denominator of the far limit hits zero and the far limit becomes infinite — the classic definition of hyperfocal. Note that s is measured from the front principal plane, not from the sensor; for anything beyond a few feet the difference is of order f and vanishes into the rounding.

There is a one-line test that any correct implementation of these formulas passes. Focus exactly at the hyperfocal distance and the near limit must come out at exactly H/2. Substitute s = H into the near-limit expression: H(H − f)/(H + H − 2f) = H(H − f)/(2(H − f)) = H/2, with the (H − f) cancelling cleanly. Numerically, a 50 mm lens at f/4 on full frame gives H = 21.70 m, and feeding s = 21.70 m back through the near-limit formula returns 10.850 m. If a calculator does not do this, it is using a different definition of hyperfocal or has an error.

The table above is that machinery run eight times, on the same 50 mm lens, same full-frame sensor, same subject distance of 10 ft, with only the aperture moving. Read the last column: 11.6 in at f/1.4, 191.5 in at f/16. That is the entire aperture range of a fast prime, seven full stops, and it multiplies the sharp zone by 16.4.

Subject distance dominates, and it does so quadratically

Hold the 50 mm at f/4 and move instead. At 5 ft, depth of field is 8.2 in. At 10 ft, 33.9 in — 4.13 times more for a doubling. At 20 ft, 145.2 in — another 4.29 times. At 40 ft, 784.2 in, a further 5.40 times. Two doublings of distance did more than the entire aperture ring, and they did it without touching exposure, without touching motion blur, without touching anything else in the picture.

The reason is visible in the widely used approximation DoF ≈ 2s²Nc/f². Everything on the right is linear except s, which is squared — so doubling s should roughly quadruple the zone, and the measured 4.13 and 4.29 are exactly that. The approximation is excellent while s stays well below H: at 3 m it is 0.2% low against the exact formulas, at 6 m 6.8% low, at 12 m 30.1% low. The reason it degrades is that the far limit runs away toward infinity as s approaches H, which is why the third ratio came out at 5.40 instead of about 4.

The circle of confusion is a decision about viewing size

The value of c has to come from somewhere, and where it comes from is the human eye at a stated viewing distance. The classic basis is that a person with good vision resolves about 5 line pairs per millimetre at 25 cm, which corresponds to a blur disc of roughly 0.2 mm in the print. Divide that by the enlargement factor from sensor to print and you get c on the sensor. Rather than redo that arithmetic per print, the industry settled on a shortcut: c = d/1500, where d is the sensor diagonal.

For full frame, d = √(36² + 24²) = 43.267 mm, so c = 0.028844 mm — the value used throughout this article and usually quoted rounded to 0.029 mm. But 0.030 mm and 0.033 mm are also common, and the tighter d/1730 criterion gives 0.025 mm. Those are not errors; they are different answers to the question of how big the picture will be viewed. Swapping 0.029 mm for 0.025 mm shrinks every depth-of-field figure in the table by about 14%, because c sits in the denominator of H. That is the whole explanation for two calculators disagreeing on identical inputs — check what circle of confusion each one assumed before assuming one of them is broken.

This also means depth of field is not a fixed number for a photograph. The same file has less depth of field on a wall print than on a phone, because the enlargement is bigger and the same sensor-level blur disc grows past the eye's threshold. Nothing about the exposure changed; the viewing condition did. Any tool that reports depth of field without stating its circle of confusion is reporting half an answer.

Full frame does not have less depth of field — bigger apertures do

Here is the sentence people actually mean when they say a bigger sensor gives shallower depth of field, stated precisely: at the same framing and the same physical aperture diameter, depth of field is nearly identical on every format. The physical diameter is D = f/N, in millimetres, not the f-number. A 50 mm at f/4 has D = 12.5 mm. A 25 mm at f/2 also has D = 12.5 mm, and on Micro Four Thirds a 25 mm frames the same scene a 50 mm does on full frame.

Run the numbers with each format's own circle of confusion — c = d/1500 on its own diagonal — and a subject at 10 ft. Full frame, 50 mm at f/4: 33.9 in. Nikon DX, 32.8 mm at f/2.62: 34.1 in. Micro Four Thirds, 25 mm at f/2: 34.2 in. A 1-inch sensor, 18.3 mm at f/1.47: 34.2 in. The spread across a 2.7× range of sensor sizes is 1.2%, which is smaller than the error in reading a focus scale. Depth of field does not know what sensor is behind the lens.

The real asymmetry is what each format lets you buy. A 25 mm f/1.0 for Micro Four Thirds would be needed to reach D = 25 mm, matching a 50 mm f/2 on full frame; a 25 mm f/0.5 would be needed to match a 50 mm f/1.0. Those lenses barely exist. So full frame does not have a different physics of blur; it has an easier route to large physical apertures at normal focal lengths. That is the whole of it, and it is why the crop-factor conversion has to be applied to the f-number as well as the focal length.

Using this on an actual shoot

Three practical consequences follow from the numbers above. First, if you need more depth of field and you are already at a working aperture, step back and crop later rather than stopping down further — one step back from 10 ft to 20 ft bought 4.29 times the zone, while a stop of aperture buys about 1.4 times. Second, if you need less depth of field, get closer before you open up, because the same relationship works in reverse and it works faster than the aperture ring.

Third, decide your circle of confusion before you trust any number. If the picture will be inspected at 100% on a monitor, the honest c is the pixel pitch, not d/1500 — on a 24-megapixel full-frame sensor the pitch is 6.0 µm, roughly a fifth of 0.0288 mm, and every depth-of-field figure collapses accordingly. If it will be a print seen at arm's length, d/1500 is the right convention and the table above applies as computed. Feed both into the depth of field calculator and read off the two answers; the gap between them is the honest uncertainty in the question.

Frequently asked questions

Does a longer lens really give shallower depth of field?

Only if you stay put. Keep the subject the same size in the frame and focal length almost drops out. At f/4 on full frame, framing a subject at a magnification of 1:60: 25 mm at 5 ft gives 36 in of depth of field, 50 mm at 10 ft gives 34 in, 100 mm at 20 ft gives 33 in, and 200 mm at 40 ft gives 33 in. An eightfold change of focal length moved the answer by 8%. What the long lens actually changes is the background: it is magnified far more, so out-of-focus areas look bigger and smoother even though the sharp zone is the same depth.

Is depth of field really one third in front of the subject and two thirds behind?

It is exactly one third in front only at one distance: s = H/3. Everywhere else the split is different, and the difference is large. With a 50 mm at f/4 on full frame, at 5 ft the split is 47% in front and 53% behind — essentially symmetric. At 10 ft it is 43/57. At 20 ft it is 36/64. The one-third rule is the value the ratio passes through on its way from symmetric at close range to infinite at the hyperfocal, and quoting it as universal is what makes close-up focusing go wrong.

Does stopping down always make more of the picture sharp?

No — diffraction eventually takes back what depth of field gives. A perfect lens spreads a point into an Airy disc whose diameter is about 2.44·λ·N; at a mid-spectrum wavelength of 550 nm that is 0.0215 mm at f/16 and 0.0295 mm at f/22. Compare those with the full-frame circle of confusion of 0.0288 mm: at f/22 the diffraction blur alone already exceeds your sharpness criterion, so the whole frame is soft even where it is nominally in focus. The crossover on full frame lands at about f/21.5, and it scales with the format — a smaller sensor has a smaller c and hits its diffraction wall at a wider aperture.

Why do two depth-of-field calculators disagree on the same inputs?

Almost always because they assumed different circles of confusion. Full frame is variously given 0.025, 0.029, 0.030 and 0.033 mm depending on which convention the author picked, and c sits in the denominator of the hyperfocal distance, so a 14% change in c moves every reported limit by roughly 14%. A second, smaller source is whether the tool uses the exact near/far pair or the 2s²Nc/f² approximation, which agrees to 0.2% at 3 m but runs 30% low at 12 m. Neither is a bug. Ask what c and which formula, and the two answers usually reconcile.

If I focus at the hyperfocal distance, is everything from half of it to infinity genuinely sharp?

Sharp by your chosen criterion, which is not the same as sharp at 100% magnification. With a 50 mm at f/8 on full frame, H = 35.7 ft, so focusing there puts the near limit at 17.9 ft and the far limit at infinity — but everything at those two extremes is blurred to exactly the circle of confusion you specified, 0.0288 mm. At d/1500 that is invisible in a normal print and clearly visible on a 24-megapixel file at 100%, where the honest criterion is closer to the 6.0 µm pixel pitch. If the frame will be scrutinised at pixel level, focus a little beyond the hyperfocal and accept a nearer limit further out.