|
Dear LF group,
since DCF39 got a new transmitter in
April 2007, the 137 kHz band has been plagued by increased
interference from it's excessively wide FSK switching
sidebands. Viewed from here, they extend down to below 136.9 kHz on a
quiet day. There are a few narrow gaps around baudrate multiples from the
carrier, which can provide clean lanes for narrowband modes.
Recently I noticed a subtile change in the shape of
the spectrum (see screenshot). In addition to the 10 Hz wide gaps at n x
200 Hz from the active (space) frequency 139.17 kHz (ie. at 137.77, 137,57,
137.37 ... kHz), there are now additional, slightly wider gaps grouped
around the idle (mark) frequency 138.83 kHz (ie. at 137.83, 137.63, 137.43
...). This is apparently related to a small modification in the timing of the
transmitted serial data bytes (one start bit, nine data bits, one stop
bit). The data bits have always been exactly 5.0 ms long, but the duration of
the stop interval used to be slightly longer (5.31 ms). This has now been
changed to 5.0 ms as well.
Most of the transmitted data actually are 16 byte
timestamps, which are repeated every 11 seconds and used to synchronize the
clocks in the remote receivers. They now seem to be aligned to UTC
time, with the start bit commencing exactly 875 ms before the
second, and the last data bit ending on the GPS second pulse. Klaus Betke (DL4BBL) has derived the binary telegram format
and published it on his former website www.longwave.de. With this information, it
should be possible to build a DCF39 (as opposed to DCF77 or MSF) clock.
There may also be new opportunities for DCF39 phase extraction and sideband
cancellation.
The software noise blanker in Spectrum Lab can be
configured to reduce the impact of the DCF sidebands on QRSS reception. I
found it advantageous to not suppress the complete telegrams, but
only a short time around each of the individual FSK transition clicks.
To do this, you will have to feed the soundcard without enough bandwidth (at
least 1 kHz) to contain the clicks, but not the carrier frequency
(138.83 kHz). Then you can optimize the noise blanker threshold (9 dB) and
ramptime settings (2.5 ms, short enough to retain most of the signal,
but long enough to restrict the generation of new sidebands).
Kind regards,
Markus (DF6NM)
|
