[time-nuts] Aircraft ping timing
David J Taylor
david-taylor at blueyonder.co.uk
Sun Mar 23 06:11:49 EDT 2014
Note that the ADS-B mentioned is just a fancy version of the
transponder that was turned off.
I ran into your R Pi page about halfway in the process of doing my
Beaglebone Black RTLSDR page. I have RTLSDR and dump1090 running on
Angstrom Linux. I bought the GPS you suggested and will attempt to get
NTP going with it, so expect a question or two off line. ;-)
I wonder if your problem with the timing is related to the R Pi being
nearly maxed out. The Beaglebone Black is about 4x the processing power
of the R Pi, with dump1090 running at 30% CPU.
Seems to me that MLAT is a function of the distance between receivers,
At 0.5uS, the distance a radio wave travels is 150 meters. Throw in
some error and worst case it, you are still talking less than a km.
Good enough for MLAT considering the option is no location information.
The flightradar24 decoders are using a Beaglebone Black with the "radar
cape". That unit claims a 100MHz clock, so the timing resolution is
10ns. But my recollection is the GPSDO units only do around 100ns when
the dust settles. The radarcape uses an FPGA, so it doesn't have
the computer delay issue and is probably limited by the GPSDO accuracy.
Personally for the cost, I'd take the 1km error from the DVB-T stick
and call it a day.
No problem with off-line questions - providing it's not already covered on
the Web page!
The Raspberry Pi is running at 35% CPU for decoding mode-S data, and 45% if
you add mode-A/C data, so it doesn't appear to be maxed out - at least with
the load here. Interesting that even with the extra CPU power of the
BeagleBone Black you still see as high as 30% CPU. The author of the
software once said that all the memory movement was the main CPU load, to
which I asked whether using the DMA controllers might help, but apparently
that's not easy to do, and the DMA registers are shared with other
functions, IIRC. Malcolm Robb, the author, is very open to comments,
It seems that the present software (either timing the reading the DVB-T
stick or the Mlat processing) cannot produce "good enough" results from the
DVB-T stick, something I don't quite understand. The method used is to
compare the timings with aircraft of known position - as this may be
off-topic there's are references here:
PlanePlotter does /not/ rely on the PC being synced to within fractions of a
millisecond, as it uses relative timing with a quantisation interval
determined by the receiver (50 ns for the SBS receiver, 500 ns for the DVB-T
stick). The RadarCape with its GPS not only provides accurate absolute time
(within 100 ns), but absolute position as well. Of course, it costs about a
hundred times more than a DVB-T stick!
SatSignal Software - Quality software written to your requirements
Email: david-taylor at blueyonder.co.uk
More information about the time-nuts