[time-nuts] GPS clock to RT-Linux for pulsar machine
Tom Van Baak
tvb at leapsecond.com
Fri Jul 1 20:48:08 EDT 2005
To avoid a large number of administration and
synchronization problems, how about not directly
linking the PC or OS clock to the ADC sample
clock at all. Instead, inject a full-scale <100 us
wide GPS 1 PPS into the ADC input itself. That
way your continuous serial ADC data stream also
contains embedded UTC "ticks", for free.
With 100 us sampling you lose at most 1 or 2
valid analog data points out of every 10,000. But
you gain perfect synchronization, including all
long-term information on phase and rate changes
between GPS and the maser.
This eliminates any dependency on the quality
of the PC or the OS timing jitter or latency from
the equation. Detecting the ticks for digital timing
or excluding the ticks for analog measurement in
software is not much more than one line of code.
> Hello all,
> As part of my radioastronomy job, I'm being asked to devise a low-cost
> pulsar detection machine for very high frequencies (~80 GHz), which
> requires many tens of hours of integration time staring at the pulsar
> while maintaining absolute timing accuracy of ~50 microseconds.
> I have GPS time and a maser available for absolute and relative timing
> respectively. I expect to be taking the data via a low-cost A/D card
> clocked by the maser at about 100 us per sample. I plan to use RTLinux
> unless I can be convinced there's a better way, as I already use
> RT-Linux successfully for our various spectrometers.
> My question is: What's a good way to synchronize the RT-Linux OS clock
> to the GPS clock? I am considering the use of the A/D card's interrupt
> with the 1PPS from GPS. I expect to use NTP to get the correct second.
> Does NTP provide a way to tie an interrupt from some user-provided
> hardware to sync up the fast clock to ~50 microsecond accuracy, or is
> there some other method?
> Thanks in advance for any help.
> time-nuts mailing list
> time-nuts at febo.com
More information about the time-nuts