The Repository Absolute Magnitude Space

The Event Horizon Telescope Never Took a Photograph

September 8, 2026 · Runtime 11:33 · Watch on YouTube

The famous orange ring is not a photograph, and no instrument in the array ever recorded an image. Eight telescopes recorded arrival times; two conversions did the rest.

Chapters

Transcript

The picture that is not a picture 0:00

You have seen it. An orange ring, a dark centre, published in April two thousand nineteen as the first image of a black hole. It is not a photograph. No instrument in the Event Horizon Telescope ever recorded a picture of anything.

What eight telescopes actually recorded 0:17

What eight telescopes recorded was the arrival time of a wave. Nothing else. This episode ends on the one number in that famous image that a machine actually returned, and on the two conversions that stand between it and the headline.

The angle this episode ends on 0:33

Last episode ended by naming that number: forty two millionths of a second of arc. The width of a shadow, fifty five million light years away. Tonight, how you build a telescope big enough to see it, without building a telescope that big.

How small 42 microarcseconds is 0:50

Start with the angle, because everything here is downstream of how small it is. A second of arc is one thirty six hundredth of a degree. We are talking about forty two MILLIONTHS of one of those.

A hair at the height of the space station 1:04

Here is that angle with two objects you already size by eye. Take a single human hair, about eighty microns across. Put it at the altitude of the space station, four hundred kilometres up. That hair covers forty one millionths of a second of arc. The ring covers forty two.

Diffraction is not an engineering limit 1:25

Now the problem. A telescope cannot resolve detail finer than roughly its wavelength divided by its width. That is diffraction, and it is not an engineering limit you can spend your way out of. It is the shape of the wave.

The dish would be 6,400 km across 1:42

The Event Horizon Telescope observes at one point three millimetres. Divide that wavelength by the angle it needs to resolve, and you get the width of the dish that could do it. The answer is about six thousand four hundred kilometres. A single dish, the radius of the Earth.

You do not need the mirror, only its edges 2:02

Nobody built that dish. The trick is that you do not need the whole mirror. You need its edges. A dish works by collecting a wavefront across its width and adding it back together in phase.

Resolution comes from separation 2:16

So put a receiver where one edge of the mirror would be, and another where the opposite edge would be. Record what each one sees, and add them together afterwards. The resolution you get is set by how far apart they stand, not by how much sky they collect.

Very long baseline interferometry 2:34

That is very long baseline interferometry, and radio astronomers have done it since the nineteen sixties. What is new here is the frequency, and what the frequency costs.

A wave cycle lasts 4.3 picoseconds 2:47

To add two wavefronts together you have to know, for each one, exactly when it arrived. Not approximately. At one point three millimetres, a full cycle of the wave lasts four point three picoseconds.

The hydrogen maser makes it one instrument 3:02

So every station carries a hydrogen maser: an atomic frequency standard stable enough that two clocks on opposite sides of the planet stay in step for the length of an observation. The maser is not a detail of the array. It is what makes the array one instrument.

64 gigabits per second, per telescope 3:20

Then there is the data. Each station records at sixty four gigabits per second. Over a night that is roughly three hundred and fifty terabytes, from one telescope.

No network was going to move it 3:33

Eight telescopes, four nights, and the total runs to several petabytes. No network on Earth was going to move that from the South Pole. So it did not move on a network. It moved on hard drives, in crates, on aircraft.

Seven sites shipped. The eighth could not 3:51

The observations ran in April two thousand seventeen. Seven of the eight sites shipped their disks within weeks. The eighth is at the South Pole, and the South Pole had just entered its winter.

Eight months of Antarctic winter 4:05

There are no flights out of Amundsen Scott station during the polar night. The disks sat there for eight months. They reached the correlator in Massachusetts on the thirteenth of December, and only then could the array be assembled at all.

A data rate set by cargo flights 4:20

It is worth saying plainly what that means. The effective data rate of the largest telescope ever operated was set, for one of its eight eyes, by the schedule of cargo flights to Antarctica.

Eight telescopes, six places 4:34

Here is the array itself. Chile, twice. Hawaii, twice. Arizona. Mexico. Spain. Antarctica. Eight telescopes at six places, and the longest gap between any two of them is ten thousand seven hundred kilometres.

10,700 km, from both sides 4:56

Now put that number next to the one from a minute ago. The dish you would need for the nominal resolution of this array works out to ten thousand seven hundred kilometres across. Those are the same number, seen from opposite sides. That is what Earth sized means here.

Eight dishes are not a mirror 5:15

But eight dishes are not a mirror, and this is where the honest version of the story starts. A filled mirror samples the wavefront everywhere across its surface. Eight telescopes sample it in eight places.

What correlation actually gives you 5:29

What correlation gives you is not a picture. For each pair of telescopes it gives one complex number: how strongly the two signals agree, and by how much they are shifted. Astronomers call it a visibility. Eight stations give you at most twenty eight pairs.

More possible images than data 5:48

Earth's rotation drags those pairs across the sky through the night, so each one traces an arc rather than a dot. It still leaves most of the measurement empty. There are far more possible images that fit the data than there is data.

The gap is filled by an assumption 6:04

That gap is not filled by the telescope. It is filled by an assumption about what images are plausible. Which is exactly the place where a team can talk itself into seeing what it expected.

Four teams, blind to each other 6:17

So the collaboration split into four teams, each one blind to the other three, using two different families of algorithm. Nobody compared results until every team had finished.

The agreement is the evidence 6:30

All four came back with a ring, of about the same size, brighter on the same side. That agreement is the evidence. Not the picture being beautiful. Four independent reconstructions failing to disagree.

42 plus or minus 3 microarcseconds 6:46

And here is the measurement. The emission ring has a diameter of forty two, plus or minus three, millionths of a second of arc. Stable across four different nights, recovered by every calibration scheme they tried.

That is all the instrument returned 7:01

That is the whole of what the instrument returned. An angle on the sky, and a brightness that is more than ten times higher on the ring than in the middle. Everything else you have heard about this object is conversion.

The dark centre is not the horizon 7:15

Take the dark centre first, because almost every retelling gets it wrong. That dark circle is not the event horizon. It is the shadow, and the shadow is substantially larger than the horizon that casts it.

The shadow is larger than what casts it 7:30

Light passing close to the hole is bent around it, so the black patch you see is the region where every ray that reaches you has already been captured. Depending on how fast the hole spins, that patch is between roughly two and a half and five times wider than the horizon itself.

The gravitational scale: 3.8 microarcseconds 7:48

Now the first conversion. There is one natural angular scale for a black hole: its gravitational radius, divided by its distance. For this object the collaboration puts that at three point eight, plus or minus zero point four, millionths of a second of arc.

The factor of eleven comes from simulations 8:07

Which means the ring is about eleven times that scale. Forty two divided by three point eight. But that factor of eleven does not come from the telescope. It comes from simulations of how hot gas glows around a spinning hole.

7 per cent in, 10.5 out 8:24

And you can see the price of it in the numbers. The ring itself is measured to seven per cent. The gravitational scale derived from it is known to ten and a half, and the extra came from the modelling, not from the array.

The second conversion is distance 8:39

The second conversion is distance, and this is where two earlier episodes come back. That angular scale is a radius divided by a distance. To turn it into a mass, you need to know how far away M eighty seven is.

16.8 megaparsecs, from the ladder 8:55

The Event Horizon Telescope does not measure that, and cannot. The figure used is sixteen point eight megaparsecs, and it comes from the distance ladder: stellar populations and surface brightness, calibrated on rungs below them.

A borrowed distance 9:11

So the most precise angular measurement ever made of anything is multiplied by a distance borrowed from an entirely different chain of calibrations. That is not a flaw. It is just true, and it is rarely said out loud.

Doing the arithmetic 9:26

Put the two together and the mass follows. The angular scale, times the distance, times the speed of light squared, divided by the gravitational constant.

6.47 against a published 6.5 billion 9:38

Doing that arithmetic from the published inputs gives six point four seven billion solar masses. The collaboration publishes six point five, plus or minus zero point two statistical, plus or minus zero point seven systematic. The chain closes.

Where the precision went 9:57

Which brings us to the only comparison this episode really wanted to make. The angle is known to seven per cent. The mass is known to eleven.

Not from the telescope 10:08

And that lost precision did not come from the telescope. Three of those four extra points came from the simulations that set the factor of eleven, and most of the rest from a distance measured by somebody else, decades earlier.

This does not weaken the result 10:22

None of this makes the result weaker. It was checked four ways, blind, and every one of those conversions is published with its own error bar, in the collaboration's own papers. The strength is that you can take the chain apart at all.

Ask which link the instrument touched 10:39

So the practical version. When a headline gives you a mass, a size or an age, ask which link in the chain the instrument actually touched. Usually it touched an angle, a brightness or a delay, and everything else was multiplied in afterwards.

Next: the image that moved 10:57

Forty two millionths of a second of arc. A hair at the height of the space station, measured by eight telescopes that had to agree with each other to four picoseconds, and then wait out an Antarctic winter. The same array pointed at our own galaxy's black hole on those same nights. That picture took five more years, because Sagittarius A star changes while you are looking at it. Next time: the image that moved. Absolute Magnitude Space. Every episode ends in a measured number.

Description and sources

In April 2017 eight radio telescopes on four continents observed M87 at 1.3 mm, each one time-stamped by a hydrogen maser and recording at 64 gigabits per second. None of them produced a picture. Correlation returns visibilities -- amplitude and phase for each pair of telescopes -- and eight stations give at most twenty-eight pairs, so the image is a reconstruction from a sparsely sampled measurement. That is why four teams built it independently, each blind to the others, across two families of algorithm. All four returned the same ring.

What was measured is an angle: 42 plus or minus 3 microarcseconds, a human hair seen from the altitude of the space station. Everything after that is conversion. The factor of eleven relating the ring to the black hole's gravitational radius comes from simulations of glowing plasma, not from the array; the distance of 16.8 megaparsecs comes from the ladder of episodes two and three. Together they give 6.5 billion solar masses -- known to 11 per cent, against 7 for the angle. The precision lost in between did not come from the telescope.

PRINT-READY, FROM THIS CHANNEL

The Conversion Chain, Vol. 01 — the method, and four numbers taken apart with it

https://therepository.gumroad.com/l/the-conversion-chain

PRIMARY SOURCES

EHT Collaboration (2019). First M87 EHT Results. I. The Shadow of the Supermassive Black Hole. ApJL 875, L1. arXiv:1906.11238 -- the 42 +/- 3 microarcsecond ring.
EHT Collaboration (2019). ... II. Array and Instrumentation. ApJL 875, L2. arXiv:1906.11239 -- 1.3 mm, hydrogen masers, 64 Gbit/s, baselines 160 m to 10,700 km, ~25 µas nominal resolution.
EHT Collaboration (2019). ... IV. Imaging the Central Supermassive Black Hole. ApJL 875, L4. arXiv:1906.11241 -- four teams, each blind to the others; CLEAN and RML.
EHT Collaboration (2019). ... VI. The Shadow and Mass. ApJL 875, L6. arXiv:1906.11243 -- GM/Dc2 = 3.8 +/- 0.4 µas; M = 6.5 +/- 0.2(stat) +/- 0.7(sys) x 10^9 Msun; D = 16.8 Mpc.
Matthews, L. D., et al. (2018). The ALMA Phasing System. PASP 130, 015002. arXiv:1711.06770 -- ALMA as a phased array of ~37 antennas.
EHT Collaboration (2022). First Sgr A* EHT Results. I. ApJL 930, L12. arXiv:2311.08680 -- the 51.8 +/- 2.3 µas ring named in the closing.
ESO science release eso1907 (10 April 2019) -- the eight telescopes and their sites; ~350 TB per telescope per night, flown to the correlators on disk.
EHT status update, 15 December 2017 -- the South Pole disks reached MIT Haystack on 13 December.

ARCHIVE PLATES: all 17 are credited on screen in the video and listed in this project's manifest.

Absolute Magnitude takes one number the internet repeats and asks what an instrument actually returned. Every episode ends on a measured figure with its source on screen, and says plainly which part of it was converted afterwards.

No AI-generated imagery is used on this channel. Every frame of archive footage is real, publicly released material, credited on screen with its licence; every diagram is drawn in code from the numbers in the cited papers.

Corrections are pinned, not quietly edited.

#EventHorizonTelescope #BlackHole #M87 #VLBI #RadioAstronomy