[alanning]

"Since ASKAP is an array of 36 dish antennas and the burst had to travel a slightly different distance to each dish, it reached each one at a slightly different time."

“From these tiny time differences—just a fraction of a billionth of a second—we identified the burst’s home galaxy and even its exact starting point..."

Simply amazing that they can get such precise measurements from so small a differential in distance.

[umvi]

> Simply amazing that they can get such precise measurements from so small a differential in distance.

How do they know the billionths of a second differences aren't due to the radio waves travelling through pockets of air that have slightly different densities/temperatures/moisture/refraction properties?

That just seems like incredible error tolerance

[ajford]

A lot of the atmospheric affects are on longer timescales than the sampling times, which means they can be calibrated out.

Most observations of this nature (interferometric observations) will point at a nearby calibrator source (usually a well characterized quasar or other "point source" star/object), and an atmospheric noise model will then be constructed based on this calibrator to apply to each dish to minimize/account for the atmospheric influence.

This is often repeated periodically to account for changing weather over time. And on top of that, a benefit of the location chosen for ASKAP was relatively dry air (being in the western desert climate of Australia).

[cmendel]

That's really interesting! Would you be able to point me towards some resources to learn more about this?

[ajford]

A lot of it I learned when I studied Radio Astronomy in college (my major). I'm a little rusty now after about 7 years.

However, here's a few resources I remember and was able to dig up.

1) A presentation from Max Plank University, which has one of the strongest radio astro programs around: https://www.mpifr-bonn.mpg.de/3244052/IMPRS_BB_HRK5.pdf

2) Presentation from ASTRON, an interferometric array in the Netherlands, part of the Netherlands Institute for Radio Astronomy: https://www.eso.org/sci/meetings/2015/eris2015/L6_Heald_cali...

3) George Moellenbrock's slides from the 14th Synthesis Imaging Workshop (hosted by the National Radio Astronomy Observatory) is pretty detailed on calibration specifically: https://science.nrao.edu/science/meetings/2014/14th-synthesi...

There was at one time a video of the last link, but it appears the hosting site is no longer live. There are other lecture slides on radio interferometry available from the Synthesis Workshop site: https://science.nrao.edu/science/meetings/2014/14th-synthesi...

[kawfey]

There is a system in place to account for atmospheric effects on radio signals. Here's one example at the Very Large Array, which uses the GPS to measure (via interferometry) and correct for disturbances in the ionosphere: https://www.aanda.org/articles/aa/pdf/2001/06/aa10277.pdf

[jakeinspace]

I'd assume they used some pretty solid statistical methods for reducing the atmospheric noise to a relatively negligible amount, based on covariance of their dish readings.

[sixplusone]

With 36 antennas it would average out to some extent, but since propagation in air is very near that of vacuum already, I don't think a °C or two makes much of a difference.

[mikeash]

I don’t think that’s quite right. Light travels about one foot in a nanosecond. Unless the signal was directly overhead, the distance differential would surely be at least several feet, and thus the time differential comfortably more than a billionth of a second. Maybe they meant a millionth.

[mturmon]

The wording is misleading. You need to compute the angle of incidence of the incoming wavefront to a high accuracy in order to localize the source.

The key quantity is not the absolute delay introduced by the absolute distance from antenna to antenna (which is tens of meters, http://www.atnf.csiro.au/projects/askap/ska.html). It's the differential delay as a function of incidence angle.

[mikeash]

Ah yes. The difference in arrival time was many nanoseconds but it needed to be measured to within a fraction of a nanosecond to accurately localize the source.

[BurningFrog]

I guess it's like any old heist movie:

The key is to synchronize your clocks beforehand!

[deepsun]

Poincare has proven that it's impossible to synchronize clocks exactly

[fit2rule]

By the time you count your clocks, they've gone askew.