Hello Stefan,
857 ns standard deviation in the 1-second pulse timing at 48 kHz sample
rate isn't spectacularly good, but considering the fact that the pulses
are sent over a 'lossy' Vorbis-compressed stream, it's not so bad.
You could possibly do better by running Paul's vlfrx tools on the RasPi,
and resample the data *there* (on the remote site) to exactly 48 kHz
before sending them over the line. Possibly easier said than done (since
vttime.c needs a differently shaped sync pulse), so before going that
way you could check the GPS module by connecting if directly to the PC
with a high-grade soundcard
(like the E-MU 0202 or similar 'large boxes' where the oscillators are
less affected by thermal drift than the miniature USB sticks).
Oh well, time to get back to the projected PIC-based, GPS-synced OXCO
with fast internal ADC and DSP. But the thing has neither WLAN nor
Ethernet :o( but the hardware is dead cheap and consumes far less power
than a Raspberry :o)
73,
Wolf .
Am 30.10.2016 um 21:37 schrieb DK7FC:
Hi Wolf,
Am 30.10.2016 21:15, schrieb Wolfgang Büscher:
Hi Stefan,
No significant improvement if the soundcard really supports 96 kHz
sampling or more,
and if the standard deviation in the GPS-based pulses is already
below 100 ns.
OK.
That is displayed on the SR panel as 'StdDev60' (standard deviation
in the recent 60 1-second pulses).
I know :-)
See the capture attached. This is the VLF stream from the tree. Could
it have to do with the vorbis stream? Or is it due to the module? The
module receives 7 satellites at the moment?
73, Stefan
|
