Why there is a shutter limit at all
The Earth turns once on its axis every sidereal day, 86,164 seconds. Divide 360 degrees by that and the sky drifts past at 15.041 arcseconds per second — a little over four thousandths of a degree. On a camera on a fixed tripod, every star draws a short arc across the frame for as long as the shutter is open, and the only question is whether that arc is short enough to look like a point.
Two things decide the answer. The first is the image scale: how much sky one pixel covers. A pixel of pitch p micrometres behind a lens of focal length f millimetres subtends 206,265 × p/1000 ÷ f arcseconds. On a 24 MP full-frame body — 5.98 µm pixels — behind a 24 mm lens, that is 51.4 arcseconds per pixel, so 15 arcseconds of motion is less than a third of a pixel per second.
The second is declination. A star on the celestial equator travels the full 15.041 arcseconds a second; one at declination δ travels 15.041 × cos δ. Polaris, at nearly 90°, barely moves at all. The Milky Way core sits near declination −29°, so it moves at about 87% of the equatorial rate — close enough to the worst case that both rules simply assume it.
Both rules exist because doing the calculation properly at 2 a.m. in the dark is not practical. They are compression: a whole optics and perception problem squeezed into a number you can do in your head.
The 500 rule and the NPF rule compared
The 500 rule says the longest usable shutter in seconds is 500 divided by the effective focal length, which means the real focal length multiplied by the crop factor. A 24 mm lens on full frame gives 20.8 s; the same lens on a 1.5× APS-C body gives 13.9 s, because the smaller sensor magnifies the same angular motion across a smaller frame.
Where the 500 comes from is the interesting part. Its ancestor is the 600 rule from film days, calibrated against what looked sharp in an 8×10 print viewed at arm's length. It contains an implicit assumption about how big a smear can be before you notice, and that assumption was made with film grain roughly 10 to 20 µm across. Modern pixels are two to four times smaller, which is why the rule looks generous today and why photographers have progressively tightened it to 500, then 400, then 300.
The NPF rule, from Frédéric Michaud of the Société Astronomique du Havre, rebuilds the estimate from what actually blurs a star: t = (35N + 30p)/f, with N the f-number, p the pixel pitch in micrometres and f the true focal length. The two terms are the two competing sources of a star's finite size. The 35N term grows with f-number because diffraction spreads a point source in proportion to the aperture ratio; the 30p term grows with pixel size because a larger pixel hides more motion. Note that crop factor does not appear — it does not need to, because the pixel pitch already encodes the sensor's real sampling.
Which one is stricter is not fixed. Setting them equal, NPF is the tighter constraint exactly when k(35N + 30p) < 500. For a 24 MP full-frame body with 5.98 µm pixels, that means 35N < 500 − 179.5, so NPF is stricter at any aperture below f/9.2 and looser beyond it. At the wide apertures night photography actually uses, NPF is always the tighter of the two — at f/2.8 it gives 11.6 s against the 500 rule's 20.8 s, a factor of 1.8.
Worked example: 24 mm f/2.8 on a 24 MP full-frame body
You are shooting the Milky Way with a 24 mm lens at f/2.8 on a 24 MP full-frame camera: 35.9 mm sensor width, 6,000 pixels across.
- 500 rule. 500 ÷ (24 × 1.0) = 20.83 s.
- Pixel pitch. 35.9 mm ÷ 6,000 px = 0.0059833 mm = 5.983 µm.
- NPF numerator. 35 × 2.8 + 30 × 5.983 = 98 + 179.5 = 277.5.
- NPF rule. 277.5 ÷ 24 = 11.56 s.
- Image scale. 206,265 × 0.0059833 mm ÷ 24 mm = 51.42 arcseconds per pixel.
- Trail at the 500-rule shutter. 15.041 × 20.83 s = 313.4 arcseconds of sky motion; divide by 51.42 arcsec/px = 6.09 pixels.
- Trail at the NPF shutter. 15.041 × 11.56 = 173.9 arcseconds, or 3.38 pixels.
- Frames for 30 minutes. ⌈1,800 s ÷ 11.56 s⌉ = 156 frames.
That trail figure is what makes the disagreement concrete. A six-pixel streak is obvious when you inspect the file at 100%, and invisible in a 1,000-pixel-wide web image. A three-pixel streak looks like a slightly soft star even at full magnification. Which rule is right depends entirely on how the image will be viewed — which is exactly the judgement the original 600 rule made for you, using 1970s viewing conditions.
Maximum shutter by focal length and format
| Focal length (mm) | 500 rule, full frame | 500 rule, APS-C 1.5× | 500 rule, MFT 2.0× | NPF, 24 MP FF at f/2.8 |
|---|---|---|---|---|
| 14 | 35.71 | 23.81 | 17.86 | 19.82 |
| 16 | 31.25 | 20.83 | 15.63 | 17.34 |
| 20 | 25.00 | 16.67 | 12.50 | 13.88 |
| 24 | 20.83 | 13.89 | 10.42 | 11.56 |
| 35 | 14.29 | 9.52 | 7.14 | 7.93 |
| 50 | 10.00 | 6.67 | 5.00 | 5.55 |
| 85 | 5.88 | 3.92 | 2.94 | 3.26 |
| 135 | 3.70 | 2.47 | 1.85 | 2.06 |
| 200 | 2.50 | 1.67 | 1.25 | 1.39 |
Every NPF figure in the last column is 277.5/500 = 0.555 times the full-frame 500-rule figure in the same row, because both are constants divided by the same focal length. That fixed ratio is why NPF and the 500 rule never cross for a given camera and aperture — only changing N, p or the crop factor moves the crossover.
Choosing between the two, and what to do next
Start from how the image will be seen. For a web-sized nightscape where the sky is one element among foreground and composition, the 500 rule is genuinely adequate, and the extra 80% of light it buys is worth more than the trailing costs you. For a print, a large-screen display, or anything where a viewer will look closely at the stars, use NPF or a 300 rule.
The frame-count output points at the better answer for most people: stack. Twelve exposures at the NPF limit, aligned on the stars and averaged, give roughly the signal-to-noise of one exposure 12 times as long while keeping every star as tight as a single frame. Free software will do the alignment. It converts the shutter limit from a hard ceiling on image quality into a mild inconvenience.
Beyond about 200 mm of effective focal length the untracked exposures become so short that a star tracker changes the game entirely. A tracker that follows the sky removes the limit completely for the sky and introduces a new one for the landscape, which is why tracked nightscapes are usually shot as two blended exposures.
Three practical checks before you trust either number. Shoot one frame and inspect a star at 100% near the centre of the frame — not at a corner, where coma and astigmatism from a fast wide lens will smear stars far more than the sky's motion does. Confirm the tripod is genuinely still, because a light breeze on a large lens hood produces exactly the same signature as trailing. And be aware that stopping down to reduce lens aberrations lengthens the NPF limit, since the 35N term grows with f-number — going from f/2.8 to f/4 changes the numerator from 98 + 179.5 = 277.5 to 140 + 179.5 = 319.5, raising the NPF time on the worked example from 11.56 s to 13.31 s, though it costs a full stop of light.
The crop factor appears in one rule and not the other
The 500 rule needs the crop factor because it works from effective focal length, which is a proxy for how much the frame magnifies angular motion. The NPF rule does not, because it works from the pixel pitch, which already tells you how finely that motion is sampled. Feeding an equivalent focal length into the NPF rule double-counts the crop and gives an answer that is too short by exactly the crop factor. Enter the number printed on the lens barrel.
What neither rule accounts for
- Lens aberrations. Coma, astigmatism and lateral chromatic aberration in the corners of a fast wide-angle lens routinely smear stars further than the sky's motion at the calculated shutter.
- Atmospheric seeing. Turbulence blurs a point source by 1 to 4 arcseconds even in good conditions. At wide focal lengths that is far less than one pixel and irrelevant; past about 300 mm it starts to matter.
- Declination. Both rules assume the worst case at the celestial equator. Shooting toward the pole, you can exceed both by roughly 1/cos δ — which is why circumpolar star-trail shots are a different discipline entirely.
- Field rotation. Even the sky's motion is not a pure translation across a wide frame; stars at different declinations move at different rates, so a very wide field has a small amount of shear that no single shutter time removes.
- Vibration. Wind on a lens hood, a loose ballhead or shutter shock on a mirrored camera produce blur that looks identical to trailing but is not fixed by a shorter exposure alone.
- Whether the exposure is correct. These rules cap the shutter. Whether that shutter, your aperture and your ISO collect enough light is a separate question, and the usual answer at the shutter limit is to open the aperture as far as the lens allows and raise ISO.
Key terms
- Sidereal rate
- The apparent rotation rate of the sky: 360° per sidereal day of 86,164 s, which is 15.041 arcseconds per second at the celestial equator.
- Pixel pitch
- The physical centre-to-centre spacing of pixels on the sensor, in micrometres. Sensor width divided by pixel count.
- Image scale
- How much sky one pixel covers, in arcseconds per pixel: 206,265 × pitch (mm) ÷ focal length (mm).
- Declination
- Angular distance north or south of the celestial equator, the sky's equivalent of latitude. Apparent motion falls as its cosine.
- Integration time
- Total accumulated exposure across all stacked frames. Twenty 15-second frames give five minutes of integration.
Where this sits in a night-shooting workflow
The shutter limit is one of three constraints that fight each other on a nightscape. The second is depth of field: the foreground has to be acceptably sharp too, and at f/2.8 on a 24 mm lens the near limit of acceptable focus is closer than people expect. Work it out with the depth of field calculator and the hyperfocal distance calculator, remembering that focusing at the hyperfocal distance puts infinity — where the stars are — at the very edge of acceptable, which is why most astro shooters focus on a bright star instead and accept a softer foreground or blend two exposures.
The third is framing. How much sky a lens takes in determines whether the Milky Way core fits above your foreground, and the field of view calculator answers that directly. If you are working across formats, the crop factor calculator converts between them — and note that the same crop factor that shortens your 500-rule shutter also narrows your field, which is the same physical fact seen twice.
Star trails are the other half of this subject, and they invert every assumption here. A trail image wants motion, so it is built from hundreds of consecutive frames at whatever shutter the noise floor allows, stacked with a maximum blend rather than an average. The interval arithmetic for that sequence — frames, total duration, card space — is the same as for any interval sequence, which the timelapse interval calculator handles.