[time-nuts] GPS and PLL's...

[Bruce Griffiths]

Bales, Tracy wrote:
> I just started receiving the Time-Nuts emails...I admit that I've got a
> lot of learning ahead of me.
>
>  
>
> I've got a Trimble GPS module that outputs a 10 uSec wide pulse every
> second.  Now this is where I need some education...
>
>  
>
> How would you use the 1 PPS output to steer a phase locked loop running
> at 10MHZ?  Are you dividing the 10MHZ clock by 10 million?
>
>  
>
> If you are dividing by 10 million, wouldn't it take forever for the PLL
> to lock?
>
>  
>
> See, I gave everyone warning about my learning curve!!!  Thanks...
>
>   
Tracy

If one only has the PPS output from the receiver, then indeed this is
all you can use to discipline the 10MHz oscillator.
A PLL is generally used for this, and yes the time to lock can be
relatively long.
The lock time is governed by the PLL dynamics (principally the loop time
constant).
Typical loop time constants can range from 100 sec to 1000 sec or more
for a good OCXO.

Whilst one can divide the 10MHz down to 1Hz, this isn't necessary.
All one really needs is a phase detector with a range somewhat greater
than the timing noise in the GPS receivers PPS output.
For example one can use the 10MHz signal directly and measure the time
delay from the leading edge of the PPS pulse to the next positive slope
zero crossing of the 10MHz signal.
This will suffice if the PPS timing noise is somewhat less than about
50ns or so.
However if the EFC tuning range of the OCXO is too great the OCXO could
lock to a frequency like 10MHz + N Hz where N is a small integer with a
range of say (-10, +10) - includes 0.
To avoid this its usually better to divide the OCXO frequency down to
1MHz or so and measure the time delay from the leading edge of the PPS
pulse to the next positive slope threshold crossing of the 1MHz signal.
The possible frequencies that the 10MHz oscillator can lock to are then
10MHz + 10*N Hz. The EFC range will likely only allow locking to 10MHz,
if not divide down to  a lower frequency such as 250kHz, 100kHz etc.

Various techniques (analog and/or digital) of varying complexity can be
used to measure this time delay.


Bruce