[time-nuts] leap second too early on TDF timesignal
Pieter-Tjerk de Boer
ptdeboer at cs.utwente.nl
Tue Jan 3 15:42:58 EST 2017
It seems France's TDF timesignal (phase modulation of a 162 kHz carrier,
which until the end of 2016 also carried France Inter radio sound as
amplitude modulation) inserted their leapsecond a minute early.
Like MSF and DCF77, TDF transmits one bit per second, together composing
the exact time at the beginning of the next minute, with a special marker
in the last second of each minute.
This is the data received around the turn of the UTC year, with M denoting
010000100000001000101 1110101 1 000000 010000011110000111010001M -> 17-01-01 00:57
010111000000001000101 0001101 1 000000 010000011110000111010001M -> 17-01-01 00:58
0000111000000001000101 1001101 0 000000 010000011110000111010001M -> 17-01-01 00:59
000011000000001000101 0000000 0 100000 110000011110000111010001M -> 17-01-01 01:00
000111000000001000101 1000000 1 100000 110000011110000111010001M -> 17-01-01 01:01
000111000000001000101 0100000 1 100000 110000011110000111010001M -> 17-01-01 01:02
| ------- ------
| minute hour
`--- leap second announcement bit
Clearly, there's an anomaly in the third of these lines: it has one bit
(and thus one second) more. I've assumed the extra bit is the first one,
since that gives sensible decoding results for the time and date code.
Thus, apparently the leap second was inserted in the 00:58:xx minute (during
which the time code for 00:59 is broadcast); that's in France's time zone,
so 23:58:xx UTC, rather than in the 23:59:xx UTC minute as it should.
Also, the leap second announcement bit was never set.
I then decoded TDF from my recordings of the 2005 and 2008 leap seconds, and
found that also in those, there was no leap second at the appropriate moment
in TDF's timesignal.
Some more details, as well as a view of the entire 0 - 29 MHz radio spectrum
around the leap second, are on my website: http://pa3fwm.nl/signals/leapsecond-2016/
More information about the time-nuts