The gap between the dishes matters
Imagine two stars so close together that a radio telescope sees one blurred spot. Making the dish wider would help: a broad opening can compare the arriving wave across a larger distance. Now put a second dish far away and compare their recordings. The separation between them supplies another long distance over which the wave can be compared. This separation is called a baseline.
The two dishes need not send their waves into the same physical mirror. Each records a rapidly changing electrical signal. A correlator multiplies appropriately delayed recordings and averages them. Waves from a chosen direction line up; waves arriving at a different angle slip in phase, the position within their repeating cycle. Changing the direction changes that slip.
A longer baseline makes the slip change faster across the sky. It is therefore sensitive to finer angular detail. The dishes still collect light only through their own apertures. A continent-wide separation supplies angular discrimination, while the collecting area and receiver noise determine how faint a signal can be measured.
Reading: NRAO Essential Radio Astronomy, Chapter 3, aperture synthesis
One pair sees a stripe pattern
A useful picture is a transparent sheet of alternating bright and dark stripes laid over the sky. Add the brightness under the bright stripes and subtract the brightness under the dark ones. Then shift the stripes by a quarter cycle and repeat. These two answers describe one complex visibility: a measured amplitude and phase. “Complex” means that both parts are retained, so the position of the stripes matters as well as their spacing.
A single bright point gives a strong signal on every baseline. Two points can reinforce or cancel one another, depending on where they fall within the stripes. An extended disk averages over many bright and dark stripes; on sufficiently long baselines its visibility becomes small. A small signal can therefore reveal resolved structure even when neither dish has made a sharp picture by itself.
Rotating the baseline rotates the stripes. Lengthening it narrows their spacing. Measuring only east–west pairs leaves north–south structure poorly determined, much as a shadow seen from one side hides depth. An array needs several lengths and directions to separate shapes that otherwise give the same answers.
Reading: NRAO Essential Radio Astronomy, Chapter 3, aperture synthesis · EHT Collaboration, First M87 Results IV, arXiv:1906.11241v1
Earth supplies more viewing directions
The ground carries every telescope around Earth’s axis. During a night, the same physical pair presents a changing separation when projected perpendicular to the source direction. Its measurements trace an arc in a map of baseline lengths and orientations. Astronomers call this the spatial-frequency plane: it records which stripe patterns the array has tested.
This is why a long observation can improve an image even after the random noise has become small. The array has asked the sky new geometric questions. Adding a station is especially powerful because it creates a new pair with every station already operating. Four telescopes form six distinct pairs; a fifth creates four more.
The sky must remain sufficiently steady during the combined observation. If a source changes before the array has finished gathering its directions, an apparent asymmetry can mix time variation with spatial structure. Dynamic imaging tackles that problem by fitting a sequence of images and making its assumptions about changes between frames explicit.
Reading: NRAO Essential Radio Astronomy, Chapter 3, aperture synthesis · Johnson et al., Dynamical Imaging with Interferometry, arXiv:1711.01286v1
The blank spaces still influence the picture
Real arrays leave most stripe patterns unmeasured. Simply placing the known measurements into a Fourier sum produces a dirty image. Each point of the sky comes with a structured blur, the dirty beam, whose sidelobes can create bright companions and dark patches. Those patches belong to the incomplete sampling, so a negative patch in this first map need not represent a physically negative source.
Image reconstruction asks which positive brightness distribution could have produced the measured signals. Many candidates may fit. A smoothness preference favors gradual changes; a compactness preference favors emission near the center. With sparse data, those choices can change visible structure. Inspecting more than one reconstruction helps reveal where the observations are persuasive and where assumptions do more of the work.
The Event Horizon Telescope’s 2019 M87 imaging paper made this comparison on a much larger scale. Teams used independent methods, tested them on synthetic sources and surveyed reconstruction settings. Their agreement supported a persistent ring-scale structure. A published picture carries the result of those comparisons, calibration tests and source modeling alongside the measurements.
There is a way to recover some shape information even when the atmosphere scrambles the phase at each station. Multiply the signals around a triangle in the same direction: each station’s phase error appears once positively and once negatively. The remaining angle is the closure phase. It is one of the reasons a widely separated array can learn about an asymmetric source despite imperfect atmospheric calibration.
Reading: EHT Collaboration, First M87 Results IV, arXiv:1906.11241v1 · EHT 2019-D01-01 M87 data release, pinned commit 27ae76a