Presets
Playback
Instrument
Detector and exposure
Sky
Layers
Model check
The science
Diffraction: the Airy pattern and the Rayleigh limit
A circular aperture of diameter D turns a point source into an Airy pattern, I(θ) = I0 [2 J1(x)/x]² with x = π D θ / λ. The first dark ring is at θ = 1.22 λ/D (the first zero of J1, x = 3.8317), and 83.8% of the light falls inside it. The Rayleigh criterion says two equal stars are just resolved when the peak of one sits on the first dark ring of the other, i.e. at θR = 1.22 λ/D; the sum of the two patterns then dips to 73.5% of the peak between them. Below about 0.78 θR (the Sparrow limit) the dip vanishes and the pair is one blob. Hubble (2.4 m, 550 nm) gives 0.058 arcsec; JWST (6.5 m, 2 µm) 0.077 arcsec; the Event Horizon Telescope uses an Earth-sized baseline at 1.3 mm for about 26 micro-arcsec.
The CCD equation and stacking
For a star whose light on the detector is S electrons in a measuring aperture of npix pixels, with sky B, dark current D (both in electrons per pixel per frame) and read noise R electrons rms per pixel per frame, the signal-to-noise ratio is SNR = S / √(S + npix(B + D + R²)). Photon arrival is Poisson, so the variance of S is S; the other terms add in quadrature. Averaging N independent frames multiplies the signal by N and the noise variance by N, so SNRN = √N · SNR1. Short frames pay the read-noise term npix R² once per frame, which is why many tiny exposures are worse than one long one when read noise dominates (graph 1), while for a source-limited or sky-limited image the two are equal.
What the panels show
Top left is the true sky (stars, a small galaxy, a close binary and dust). Top right is that sky after the telescope: each source is replaced by its pixel-integrated Airy pattern (3×3 sub-samples per pixel). Bottom left is one exposure with Poisson photon noise and Gaussian read noise drawn for every pixel; bottom right is the running mean of the frames so far. An arcsinh stretch lets faint stars and bright rings be seen together. Dust dims each star by AV (0.55 µm/λ)1.7 magnitudes, so near-infrared light pierces clouds that hide the same stars in the visible.
What is simplified
- Source rates, sky level, dark current and read noise are illustrative per instrument, not mission data; a larger or smaller aperture rescales the rates by its area (capped at 10 m for the EHT, which is an interferometer, not a filled dish).
- The PSF is a perfect Airy pattern. Real telescopes add obscuration, spiders, aberrations, jitter and (on the ground) atmospheric seeing, none of which are modelled. The Chandra mirror-figure floor of about 0.5 arcsec is added in quadrature to the diffraction limit and the pattern is scaled to that width, which keeps the Airy shape that a real X-ray mirror does not have.
- The EHT is an interferometer whose point-spread function is a "dirty beam" with many side lobes; treating it as one aperture of baseline length only reproduces the angular scale, not its imaging. X-ray and radio detectors count photons or measure voltages, so the electron picture is an analogy there.
- Noise is Poisson (Gaussian approximation above 18 e-) plus Gaussian read noise; no hot pixels, cosmic rays, flat-field or bias structure, and the stack is a plain mean with no registration or sigma clipping.
In-page checks
(1) A Monte Carlo of real Poisson and read-noise frames at the target pixel reproduces the CCD equation and the √N law for N = 1 to 64. (2) The Airy table is compared with an independent Bessel-integral J1: first dark ring at 1.21967 λ/D, 83.78% encircled energy against 1 − J0² − J1², and the 0.735 Rayleigh dip. (3) The pixel-integrated rendering conserves flux and the pixel scale equals the resolution limit divided by the sampling.


