[time-nuts] Thunderbolt? (re simple gpsdo.)

David davidwhess at gmail.com
Fri Dec 30 18:07:02 UTC 2011


There is no reason you can not do that.

It is tricky because the low comparison frequency limits the loop
bandwidth like any sampled data system and the analog requirements for
the low frequency design become an issue do to leakage and the
impedance levels needed.  The long time constants involved could make
tuning the loop response tricky.

I wonder though how much of a problem jitter in the GPS 1pps signal
would be.  I understand that some receivers are useless for this type
of application because of excessive jitter.  Would something like a
Garmin 16x or 18x work?

On Fri, 30 Dec 2011 09:26:52 -0800, Chris Albertson
<albertson.chris at gmail.com> wrote:

>On Thu, Dec 29, 2011 at 8:24 PM, Mark Spencer <mspencer12345 at yahoo.ca> wrote:
>> One of the designs using the 10 kHz output from a Jupiter gps engine and a simple PLL to discipline an ocxo might be good starting point if suitable gps engines are still available.  There won't be much to tweak but the performance could be quite good.
>
>I just finished reading about that one.  It requires no longer
>available GPS reciever.   Maybe I should r-phrase the question:
>
>"What is the simplest design for a GPSDO that uses only the PPS signal
>from a modern GPS?"
>
>Half of the reason for the question is academic.  Then if a simple
>enough design presented itself it would be fun to try it.
>
>The simplest design I can think of now is based on a flip-flop.  The
>PPS "sets' the FF and the next raising edge of the local oscillator
>"resets" it.   (The local oscillator might need to be divided down or
>a slower 1Mhz oscillator used so the FF remains "on" for a reasonable
>time.)
>
>Next the FF gates a current source to a capacitor.   The voltage in
>the cap is amplified and controls the local oscillator frequency via a
>low pas filter.  Likely the "low pass filter" would be an active
>device that we call an "integrator"
>
>You need to discharge the cap for the next cycle.  One could rig a
>one-shoot timer to discharge the cap.  Actually there are about four
>states that need to be cycled every second.



More information about the time-nuts mailing list