wwvbpy generates WWVB timecodes for any desired time. These timecodes may be useful in testing WWVB decoder software.
Where possible, wwvbpy uses existing facilities for calendar and time manipulation (datetime and dateutil).
It uses DUT1/leap second data derived from IERS Bulletin "A" and from NIST's "Leap second and UT1-UTC information" page. With regular updates to the "iersdata", wwvbpy should be able to correctly encode the time anywhere within the 100-year WWVB epoch. (yes, WWVB uses a 2-digit year! In order to work with historical data, the epoch is arbitrarily assumed to run from 1970 to 2069.)
Programs include:
wwvbgen
, the main commandline generator programwwvbdecode
, the main commandline decoder programwwvbtk
, visualize the simulated WWVB signal in real-time using Tkinterdut1table
, print the full history of dut1 values, including estimated future valuesupdateiers
, download the latest dut1 data including prospective data from IERS and NIST
The package includes:
wwvb
, for generating WWVB timecodeswwvb.decode
, a generator-based state machine for decoding WWVB timecodes (amplitude modulation only)uwwvb
, a version of the decoder intended for use on constrained environments such as CircuitPython.
The author (@jepler) occasionally develops and maintains this project, but issues are not likely to be acted on. They would be interested in adding co-maintainer(s).
The National Institute of Standards and Technology operates the WWVB time signal service near Fort Collins, Colorado. The signal can be received in most of the continental US. Each minute, the signal transmits the current time, including information about leap years, daylight saving time, and leap seconds. The signal is composed of an amplitude channel and a phase modulation channel.
The amplitude channel can be visualized as a sequence of (usually) 60 symbols, which by default wwvbgen displays as 0, 1, or 2. The 0s and 1s encode information like the current day of the year, while the 2s appear in fixed locations to allow a receiver to determine the start of a minute.
The phase channel (which is displayed with --channel=phase
or
--channel=both
) consists of the same number of symbols per minute. This
channel is substantially more complicated than the phase channel. It encodes
the current time as minute-of-the-century, provides extended DST information,
and includes error-correction information not available in the amplitude
channel.
Usage: python -m wwvb.gen [OPTIONS] [TIMESPEC]...
Generate WWVB timecodes
TIMESPEC: one of "year yday hour minute" or "year month day hour minute", or
else the current minute
Options:
-i, --iers / -I, --no-iers Whether to use IESR data for DUT1 and LS.
(Default: --iers)
-s, --leap-second Force a positive leap second at the end of
the GMT month (Implies --no-iers)
-n, --negative-leap-second Force a negative leap second at the end of
the GMT month (Implies --no-iers)
-S, --no-leap-second Force no leap second at the end of the month
(Implies --no-iers)
-d, --dut1 INTEGER Force the DUT1 value (Implies --no-iers)
-m, --minutes INTEGER Number of minutes to show (default: 10)
--style [bar|cradek|default|duration|json|sextant]
Style of output
-t, --all-timecodes / -T, --no-all-timecodes
Show the 'WWVB timecode' line before each
minute
--channel [amplitude|phase|both]
Modulation to show (default: amplitude)
--help Show this message and exit.
For example, to display the leap second that occurred at the end of 1998,
$ python wwvbgen.py -m 7 1998 365 23 56
WWVB timecode: year=98 days=365 hour=23 min=56 dst=0 ut1=-300 ly=0 ls=1
'98+365 23:56 210100110200100001120011001102010100010200110100121000001002
'98+365 23:57 210100111200100001120011001102010100010200110100121000001002
'98+365 23:58 210101000200100001120011001102010100010200110100121000001002
'98+365 23:59 2101010012001000011200110011020101000102001101001210000010022
'99+001 00:00 200000000200000000020000000002000100101201110100121001000002
'99+001 00:01 200000001200000000020000000002000100101201110100121001000002
'99+001 00:02 200000010200000000020000000002000100101201110100121001000002
(the leap second is the extra digit at the end of the 23:59 line; that minute consists of 61 seconds, instead of the normal 60)
wwvbpy stores a compact representation of DUT1 values in wwvb/iersdata_dist.py
or wwvb_iersdata.py
.
In this representation, one value is used for one day (0000UTC through 2359UTC).
The letters a
through u
represent offsets of -1.0s through +1.0s
in 0.1s increments; k
represents 0s. (In practice, only a smaller range
of values, typically -0.7s to +0.8s, is seen)
For 2001 through 2019, NIST has published the actual DUT1 values broadcast, and the date of each change, though it in the format of an HTML table and not designed for machine readability:
NIST does not update the value daily and does not seem to follow any specific rounding rule. Rather, in WWVB "the resolution of the DUT1 correction is 0.1 s, and represents an average value for an extended range of dates. Therefore, it will not agree exactly with the weekly UT1-UTC(NIST) values shown in the earlier table, which have 1 ms resolution and are updated weekly." Like wwvbpy's compact representation of DUT1 values, the real WWVB does not appear to ever broadcast DUT1=-0.0.
For a larger range of dates spanning 1973 through approximately one year from now, IERS publishes historical and prospective UT1-UTC values to multiple decimal places, in a machine readable fixed length format.
wwvbpy merges the WWVB and IERS datasets, favoring the WWVB dataset for dates when it is available. There are some caveats to this, which are mostly commented in the wwvb/updateiers.py
script.
wwvb/iersdata_dist.py
is updated monthly from github actions or with iersdata --dist
from within the wwvbpy source tree. However, at this time, releases are not regularly made from the updated information.
A site or user version of the file, wwvb_iersdata.py
can be created or updated with iersdata --site
or iersdata --user
. If the distributed iersdata is out of date, the generator will prompt you to run the update command.
Leap seconds are inferred from the DUT1 data as follows: If X and Y are the
1-digit-rounded DUT1 values for consecutive dates, and X*Y<0
, then there is a
leap second at the end of day X. The direction of the leap second can be
inferred from the sign of X, a 59-second minute if X is positive and a
61-second minute if it is negative. As long
as DUT1 changes slowly enough during other times that there is at least one day
of DUT1=+0.0, no incorrect (negative) leapsecond will be inferred. (something
that should remain true for the next few centuries, until the length of the day
is 100ms different from 86400 seconds)
This should be considered more experimental than the AM channel, as the tests only cover a single reference minute. Further tests could be informed by the other implementation I know of, except that implementation appears incomplete.
Run the testsuite, check coverage & type annotations with gmake
.
There are several test suites:
testwwvb.py
: Check output against expected values. Uses hard coded leap seconds. Tests amplitude and phase data, though the phase testcases are dubious as they were also generated by wwvbpy.testuwwvb.py
: Test the reduced-functionality version against the main versiontestls.py
: Check the IERS data through 2020-1-1 for expected leap secondstestpm.py
: Check the phase modulation data against a test case from NIST documentationtestcli.py
: Check the commandline programs work as expected (limited tests to get 100% coverage)