Radio Interferometry Image Reconstruction Lab

Array Configuration

Antennas 6

Synthesis duration 8 h

Thermal noise 0.05

CLEAN loop gain 0.20

Live Readouts

uv samples collected0 / 0

CLEAN iterations0

Dynamic range—

Phase: Earth-rotation synthesis — accumulating baselines…

How it works

A radio interferometer never photographs the sky directly — each antenna pair (a baseline) measures one complex sample of the sky's spatial Fourier transform, called a visibility V(u,v). As Earth rotates, a baseline's projection onto the sky sweeps an elliptical arc through the Fourier (u,v) plane, gradually filling in coverage:

V(u,v) = ∫∫ I(x,y) e^(-i2π(ux+vy)) dx dy   (van Cittert–Zernike theorem)

Because only a sparse set of (u,v) points is ever sampled, the naive inverse transform gives a dirty image — the true sky convolved with a dirty beam B(x,y), the point-spread function of the sampling pattern itself:

I_dirty = I_true ⊛ B,   B(x,y) = FT⁻¹{ sampling function W(u,v) }

The Högbom CLEAN algorithm (1974) removes this sidelobe contamination iteratively: find the brightest pixel in the residual, subtract a small fraction (the loop gain) of the dirty beam re-centred on that pixel, record a "clean component" there, and repeat. Summing the clean components and adding back the leftover residual reconstructs a much sharper image than the raw dirty map ever shows.

Antennas — more baselines (pairs) fill the uv-plane faster and denser.
Synthesis duration — longer Earth-rotation tracking sweeps longer uv arcs, improving angular resolution.
Thermal noise — random receiver noise added to every visibility; it raises the CLEAN residual floor and lowers achievable dynamic range, exactly like real telescope sensitivity limits.
CLEAN loop gain — the fraction of each peak removed per iteration; low gain is slow but stable, high gain converges fast but can overshoot.

This is the real algorithm class (CLEAN and its descendants, plus regularised imaging like the EHT Collaboration's eht-imaging package) used to turn the Event Horizon Telescope's sparse global baselines into the 2019 image of the M87* black hole shadow — the same "revolutionary scientific image" this article discusses, reduced here to a toy but mathematically real 22×22 model you can watch converge.

uv-plane samples (Fourier coverage) True sky · Dirty image · CLEAN result

⚠ Couldn't load the 3D engineThree.js failed to load from the CDN. Check your connection and reload.