For deep-sky imaging, aim for roughly 1–2 arcseconds per pixel under typical seeing. Compute pixel scale as 206.265 × pixel size (μm) ÷ focal length (mm). Because atmospheric seeing (usually 2–4 arcseconds) — not your sensor — sets real resolution, matching your sampling to seeing matters more than chasing the finest possible scale.
What pixel scale is (and what “sampling” means)
Pixel scale is how much sky lands on a single pixel, expressed in arcseconds per pixel (arcsec/px, or ″/px). It’s set entirely by two things you already own once you’ve picked a camera and a telescope: the physical size of one pixel, and the focal length the light travels through before it reaches that pixel. Change either one and the pixel scale changes with it — nothing about the sky itself moves.
“Sampling” is what pixel scale determines in practice: how finely your camera divides up a star’s point-spread function (PSF) — the small blurred disc a point of light actually forms on your sensor, mostly from atmospheric seeing and optical imperfections rather than true point-source geometry. Too few pixels across that disc and you’re undersampling — stars come out blocky and detail gets lost between pixels. Spread the same disc across far more pixels than it needs and you’re oversampling — no extra real detail, just the same light divided into smaller, noisier pieces.
Pixel size itself is just one of six specs worth weighing before you ever get to this formula — see our guide to choosing a dedicated astrophotography camera for the other five.
The formula: 206.265 × pixel size (μm) ÷ focal length (mm)
Plug in your camera’s pixel size in microns and your telescope’s focal length in millimeters and the result is your pixel scale in arcsec/px. As AstroBackyard puts it, working the formula through a real example: “a camera with 3.76-micron pixels on an 800mm telescope produces an image scale of about 0.97 arcseconds per pixel.”
Why 206.265 (radians to arcseconds)
The constant comes straight from trigonometry, not astronomy. There are 206,265 arcseconds in one radian. The angle a single pixel subtends at the telescope’s focal plane is, for angles this small, just the pixel’s physical width divided by the focal length — a result expressed in radians. Multiplying that fraction by 206,265 converts it into arcseconds. Because pixel size is normally quoted in microns and focal length in millimeters, the constant is written as 206.265 (206,265 ÷ 1,000) so the two units cancel without a separate conversion step.
Sky & Telescope, Diffraction Limited and AstroBackyard all state the same underlying formula, though not always to the same precision — AstroBackyard’s worked example above rounds the constant to 206 rather than carrying the full 206.265. That’s exactly what you’d expect from universal geometry rather than a manufacturer-specific figure: the formula itself has nothing to reconcile between sources the way a sensor spec does, only the rounding varies.
Embedded Pixel Scale Matcher
Every worked example above uses one camera and one focal length at a time. If you’re checking your own gear — or shopping for a scope to pair with a camera you already own — running 206.265 × μm ÷ mm by hand for every combination gets old fast.
Open the Pixel Scale Matcher — free, every figure sourced, both positions shownEnter your sensor’s pixel size and your telescope’s focal length and the matcher returns your arcsec/px figure instantly, plus where it falls against both the Nyquist minimum and the wider practical range covered below — so you see the over/undersampling call laid out the same honest way this article makes it, not a single unexplained verdict.
Seeing is the real limit
A finer pixel scale doesn’t buy you finer real-world resolution once you run past what the atmosphere itself will resolve. Suburban seeing typically runs 2–4 arcseconds Stellarnomads AstroBackyard — meaning a point of light is already blurred to roughly that size before it ever reaches your pixels, regardless of how small those pixels are.
It’s tempting to read “finer pixel scale” as “sharper image,” but pixel scale only controls how finely you sample the light that reaches the sensor — it can’t undo blur the atmosphere already added. A rig sampling down at 0.3″/px does not resolve detail that 2–4″ of seeing has already smeared out; it just spreads that same blur across more pixels.
Nyquist’s 2× minimum vs. the practical 1–2″/px range
These are two related but different claims, and conflating them is a common mistake. Nyquist–Shannon sampling theory sets a hard floor: to fully capture a signal you need to sample at least twice per resolution element. Applied to a star’s PSF, that means at least 2 pixels across its full width at half maximum (FWHM) as an absolute minimum. In practice, Diffraction Limited puts the real optimum higher than that bare floor: “2.5 to 2.8 pixels is generally considered optimal.”
Separately, several sources translate typical seeing conditions into a practical arcsec/px target — and here the numbers genuinely differ by source, so hold them apart rather than blending them. Stellarnomads frames it as: “for typical seeing of 2 to 4 arcseconds, aim for roughly 1 to 2 arcseconds per pixel.” High Point gives a wider working range for average seeing, 1.5–3.0″/px. Both are legitimate practical guidance; they are not the same number, and this article doesn’t collapse them into one.
The table below applies the sourced formula to the 3.76μm pixel shared by the IMX571 and IMX533 sensors — see our IMX571 vs. IMX533 comparison for how the two otherwise differ — across a range of focal lengths, with each verdict judged against Stellarnomads’ roughly 1–2″/px target for 2–4″ seeing.
| Focal length | Typical setup at that length | Pixel scale at 3.76μm | Verdict (vs. 1–2″/px target) |
|---|---|---|---|
| ~250mm | Small travel refractor (e.g. a fast 50–60mm apo) | 3.10″/px | Undersampled |
| ~400mm | Small-to-mid apo refractor (roughly 70–80mm) | 1.94″/px | Well-matched |
| ~600mm | Mid apo refractor (roughly 100–130mm), unreduced | 1.29″/px | Well-matched |
| ~800mm | Reduced SCT/RC, or a longer apo refractor | 0.97″/px AstroBackyard | Oversampled |
| ~1200mm | 8″ SCT with a 0.63× focal reducer, or a large reduced RC | 0.65″/px | Oversampled |
| ~2000mm | Unreduced 8″ SCT or RC, or larger Cassegrain-class optics | 0.39″/px | Oversampled |
“Typical setup” describes the general class of telescope that commonly runs at each focal length, not a specific sourced product — use it to sanity-check which row matches your own gear, not as a buying recommendation.
The 800mm row matches AstroBackyard’s own worked example above almost exactly (0.97″/px), which is a useful sanity check on the arithmetic rather than a coincidence — it’s the same formula and the same pixel size. If you’re shopping for a camera to run these numbers against your own scope, the ZWO ASI2600MM Pro and ZWO ASI533MM Pro both use this same 3.76μm pixel — check current price on the ASI2600MM Pro at Agena.
Oversampling vs. undersampling: which is more forgiving?
This is the genuine, unresolved dispute at the center of this topic, and the forum threads that currently rank for it only ever show one side of it. It deserves both positions stated plainly.
Position A (Stellarnomads) says err toward undersampling: “a well-dithered, drizzled 2″/px dataset beats a noisy 0.7″/px one on almost every night you will actually get.” The reasoning is practical, not theoretical — seeing, tracking and guiding rarely cooperate enough on a given night to actually use a very fine scale, and a technique called drizzle can recover resolution from undersampled, dithered data after the fact Wikipedia Telescope Live. Position B holds that modern processing rewards oversampling instead. AstroWorldCreations found an oversampled Bin1 capture “clearly better” once run through modern deconvolution tools like BlurXTerminator, and Stan Moore argues that true critical sampling sits closer to ~3.33 pixels per FWHM — well above the bare Nyquist floor. What both sides agree on: beyond roughly 2.5–3 pixels per FWHM, you’re spreading the same light over more pixels and losing signal-to-noise for no added detail Diffraction Limited.
Neither position has settled this. If your nights are typically compromised by seeing, tracking or guiding — which is most nights, for most imagers — Position A’s case for erring toward undersampling and leaning on drizzle is worth taking seriously. If you consistently get steady seeing and good tracking, and you process with modern deconvolution tools, Position B’s case for deliberate mild oversampling has real practitioner support too. This article isn’t going to pick a winner where the evidence itself doesn’t.
Fixing a mismatch: binning and drizzle
If your setup lands well above roughly 2.5–3 pixels per FWHM — oversampled — binning is the standard fix. 2×2 binning combines four physical pixels into one effective pixel, doubling the effective pixel size and therefore doubling your arcsec/px figure. That deliberately trades away resolution you weren’t using in exchange for combining photon counts per effective pixel, which is precisely the SNR you were losing per the shared ground above.
If your setup lands below the practical target — undersampled — drizzle is the corresponding tool, though it works differently: rather than changing anything at capture time, it reconstructs a finer output grid from many dithered subexposures, using the small sub-pixel offsets between them to recover detail the native sampling alone would have missed Wikipedia Telescope Live. It’s the technical basis for Position A’s argument above — it doesn’t require dithering guesswork, just enough dithered subs to reconstruct from.
FAQ
What is a good pixel scale for deep-sky astrophotography?
For typical suburban seeing of 2–4″, Stellarnomads puts the target at roughly 1–2″/px; High Point gives a slightly wider working range of 1.5–3.0″/px for average seeing. Both agree the right number moves with your local seeing, not a fixed universal figure.
What is the pixel scale formula?
Pixel scale (arcsec/px) = 206.265 × pixel size (μm) ÷ focal length (mm) Sky & Telescope — see the formula section above, or skip the arithmetic with the embedded pixel-scale matcher.
Is a smaller pixel scale always better?
No. Past roughly 2.5–3 pixels per star FWHM, Diffraction Limited notes you’re spreading the same light over more pixels without resolving extra detail — you lose signal-to-noise for nothing in return. How much that actually costs you in a finished image is itself disputed; see the oversampling section above.
What happens if my pixel scale is too coarse (undersampled)?
Stars render as blocky, square-edged blobs instead of smooth points, and fine structure can get lost between pixels. Stellarnomads argues this matters less than it sounds, since a well-dithered, drizzled undersampled dataset can beat a noisy oversampled one — but that’s one side of a genuine dispute, not settled fact; see the both-sides breakdown above.
Does binning change my pixel scale?
Yes. 2×2 binning combines four physical pixels into one effective pixel, doubling the effective pixel size and therefore doubling your arcsec/px figure — the standard fix for an oversampled setup.
Is the Nyquist limit the same as the “1–2″/px” rule of thumb?
No, and conflating them is a common mistake. Nyquist theory sets a hard minimum of at least 2 pixels across a star’s FWHM. The 1–2″/px figures are a separate, seeing-based practical corridor, and Diffraction Limited puts the actual sweet spot higher still, at 2.5–2.8 pixels across FWHM.