For Interferometry to work the data from different baselines has to be collected at the same time so waiting 6 months does not help. Spacecraft could theoretically be used to extend the baselines but the volume of data to be transferred is prohibitive with current spacecraft comm tech.

Even on Earth they resort to shipping cases of hard drives instead of transferring over the Internet.

One technique where simply waiting 6 months works very well is in measuring parallax. The Gaia spacecraft takes advantage of this.

So I work at one of the participating institutes, and you're very much dead-on about transfer speeds/logistics limiting options here. I'm not an astronomer, so grain of salt, but the data is generated at a bit of an unwieldy pace: there are four collector nodes at each site, each of which generates ~16Gbps of raw data (though read speeds from disk after observation is more like 8Gbps). This, as you say, forces most locations to ship physical drives, and also makes centrally planning these observations rather tricky, as there's very little visibility into each station's observations until some time afterwards.

But I'm optimistic that some institutions (hopefully including mine!) will be able to transfer these data over the network after the next run of EHT observations. Exciting stuff for sure.

Where can I read more about how that volume of data is generated? I'm looking at the description of ALMA right now. So is each "node" a cluster of antennas? And each antenna is collecting a high-resolution snapshot across a wide frequency band. And each snapshot has to be timestamped. What's the sampling rate? It says the frequency range is 31GHz to 950GHz but how wide can a single snapshot be? Then to move it around are you using InfiniBand or something even faster?

So ALMA, being an array, is a bit of a different beast than single-dish telescopes like the one where I'm employed; they do indeed have an array of antennae and correlate all the collected data at a central correlator (at least, for normal observations). My institute has a single, much larger primary reflector (30 meter diameter) and does not require such a process during normal observation. However, during VLBI observations, which EHT is, the receiver is dumping data to four collector computers, which is what I was referring to as "nodes" generating ~16Gbps of data apiece.

I wish I could shed some more light on the ins-and-outs of exactly how these observations work, but I just run the computers, man. :) What I can tell you is that in order to move data off of the collector machines, they typically use m5copy, which is a part of the JIVE project (they have a Github repo: https://github.com/jive-vlbi/jive5ab). All communication between the control computer and the collectors happens within a private, physically-distinct network, but it's just standard commodity networking equipment between the control computer and the collectors. The folks in charge of the node's design are in the process of removing some of the bottlenecks to make electronic transfers more viable (the current spec doesn't even include a 10Gbps uplink!).

much faster than infiniband. A suitcase full of hard drives.

That's for moving between sites. I'm talking about the link between the antennas to each observatory's main computer.

How many seconds of observation data are required for one of these images?

For the entire image? I'm not sure, but a lot. Usually an EHT observing session is days in length, and the responsible astronomer(s) will reside at the telescope for a week or so...but I don't believe that 100% of that time is spent feeding data into the collectors. I can ask my coworker tomorrow for clarification on the actual observing time and post an update.

> Spacecraft could theoretically be used to extend the baselines but the volume of data to be transferred is prohibitive with current spacecraft comm tech.

Space-based very long baseline interferometry (VLBI) has been done at lower frequencies, most recently with the RadioAstron satellite[0]. There hasn't yet been a VLBI satellite observing at the same frequencies that the EHT uses, but there are mission concepts being discussed. [1] discusses some of the technical challenges.

[0] http://www.asc.rssi.ru/radioastron/

[1] https://arxiv.org/abs/2204.09144

I wonder what sort of range/throughput you could get with "Heavy" versions of the inter-satellite communications lasers which SpaceX is putting on their Starlink satellites...

Why not ship cases of hard drives then? With the plummeting cost of space launches, it seems reasonable to include enough ΔV(1) for a return mission.

(1) https://www.smbc-comics.com/comic/delta-v

edit: We could also send the (super)computer that will process these images in LEO, so we don't even have to worry about atmospheric re-entry.