3 ms·

There was an interesting event a few y̶e̶a̶r̶s̶ months ago around Long Island, where it appeared that the Air Traffic Control software got confused by the magne

There was an interesting event a few y̶e̶a̶r̶s̶ months ago around Long Island, where it appeared that the Air Traffic Control software got confused by the magnetic declination. Basically it was generating false radar targets in the wrong place: They were rotated 13 degrees, the magnetic declination in the area, from the actual aircraft. Thread with details here: https://twitter.com/lemonodor/status/1508505542423064578 https://twitter.com/lemonodor/status/1508505542423064578 A short summary is that sometimes when ATC radars see an aircraft, if it doesn't seem to have ADS-B (is in a position where no known ADS-B-equipped aircraft is), it will create a synthetic target, broadcasting it out so that aircraft with ADS-B In can see it on their traffic displays. That system is called TIS-B. Somehow the radar in Islip got confused by the magnetic declination: it was generating TIS-B targets that were rotated 13 degrees from where the aircraft really were. And it was doing it for pretty much every aircraft, whether or not it had ADS-B. I assume it wasn't able to correlate the known ADS-B position with the secondary radar returns because they were offset by 13 degrees. Very weird way for a pretty important piece of ATC infrastructure to fail.

defrost 4y ago
"A few years ago" would have made perfect sense .. the tweet, however, is dated Mar 29, 2022 and states "Seems like something weird was going on with ATC systems around New York a few days ago." None the less I'd (semi educated) guess that someone at the rader site in question "updated" | "improved" the implementation of The World Magnetic Model [1] in use there and effectively dropped the correct corrections from the end result (for a few days until it was noticed and fixed). Normally coefficient sets are updated every five years .. and frequently sites and software that use the WMM find themselves with a current IT support person that has not previously encountered a WMM update and then backtrace their way into an "understanding". This can cause errors until noticed. Critical infrastructure places (Military, Air Traffic Control, etc) tend to fare better and have longer periods of staff retention, etc. but it can still happen that "we no longer have anyone that was here the last time (five years ago) this happened" However; given the last epoch update (WMM2020) was released 12/10/2019 (in 'merc date speak MM/DD/YYY (barbarians)) and remains current until expiring on December 31, 2024 we can rule out a simple epoch update FUBAR. This leaves the other semi regular source of WMM errors - programmers that don't like FORTRAN (esp. super opaque single letter variable 25th degree spherical harmonic functions that date from the 1960s) and feel the need to "do it better". Having barely survived the prior epoch update and knowing that another will happen again, they carve out some space to build a better mousetrap so that next time will be easier. This is often a, uhh, interesting intersection of ego and hubris. [1] https://www.ngdc.noaa.gov/geomag/WMM/ https://www.ngdc.noaa.gov/geomag/WMM/

defrost 4y ago

"A few years ago" would have made perfect sense .. the tweet, however, is dated Mar 29, 2022 and states "Seems like something weird was going on with ATC systems around New York a few days ago." None the less I'd (semi educated) guess that someone at the rader site in question "updated" | "improved" the implementation of The World Magnetic Model [1] in use there and effectively dropped the correct corrections from the end result (for a few days until it was noticed and fixed). Normally coefficient sets are updated every five years .. and frequently sites and software that use the WMM find themselves with a current IT support person that has not previously encountered a WMM update and then backtrace their way into an "understanding". This can cause errors until noticed. Critical infrastructure places (Military, Air Traffic Control, etc) tend to fare better and have longer periods of staff retention, etc. but it can still happen that "we no longer have anyone that was here the last time (five years ago) this happened" However; given the last epoch update (WMM2020) was released 12/10/2019 (in 'merc date speak MM/DD/YYY (barbarians)) and remains current until expiring on December 31, 2024 we can rule out a simple epoch update FUBAR. This leaves the other semi regular source of WMM errors - programmers that don't like FORTRAN (esp. super opaque single letter variable 25th degree spherical harmonic functions that date from the 1960s) and feel the need to "do it better". Having barely survived the prior epoch update and knowing that another will happen again, they carve out some space to build a better mousetrap so that next time will be easier. This is often a, uhh, interesting intersection of ego and hubris. [1] https://www.ngdc.noaa.gov/geomag/WMM/ https://www.ngdc.noaa.gov/geomag/WMM/