Does the GPS use PPS discipline?

I have an ntpd running on an RPi here with GPS and use its NMEA string for course clock and its PPS to get within uSec.  I note a delta between the Kiwi clock and that.

Comments

  • course and coarse!  Not a good thing to mix up in the GPS world...  I meant coarse clock.
  • There is some strange problem with ntpd running on the Beagle. It usually starts up okay and synchronizes the time properly. But on some Kiwis they start drifting anyway. There are no errors in the ntpd log indicating it isn't able to communicate with the external servers. Reboot the Kiwi and all is fine again. I'm not sure what to do about this.

    The Kiwi server does not currently try and set the time from the calculated GPS time.

  • I'll  point the ntpd in the Kiwi BB to my ntpd with PPS and see what happens

  • The GPS time displayed on the GPS tab isn't used for anything then?
  • Well, the GPS time difference is used to "calibrate" the ADC clock oscillator. You take a couple of consecutive GPS time solutions which are about 4-5 seconds apart and compute the different. Then compare that against how many ticks of the 66.7 MHz ADC clock there have been during that time. The resulting correction is the actual clock the rest of the code should use (e.g. in the RF mixer NCO and audio sample rate calculations).

  • Too bad you can't get PPS from the GPS chip. That would give you uSec accuracy. Maybe your technqiue yields the same though.

    -Jim
  • jksjks
    edited January 2017
    Yes, it's the same. Each GPS time solution is as accurate in time as the position solution is accurate in space since they are related (one implies the other). So even though the GPS code isn't producing a PPS, each time solution it produces every 4 or 5 seconds has the same accuracy as an equivalent PPS. It just isn't exactly on a one-second mark. But it has 1 usec or 100 nsec, or whatever, accuracy -- same as a PPS.

    You could generate a PPS since the ADC clock is now synched, more or less, to the GPS. You can just count however many ADC clock ticks are necessary to get to a second boundary because you know exactly what the GPS time was when the time and tick counter were captured. I'm now experimenting with something similar for time stamping the audio IQ samples.

Sign In or Register to comment.