[hpsdr] GPS 1PPS board

[Graham / KE9H]

Chris:

Yes that is the basic concept, and it will work, but then you need
to start thinking about what accuracy you want and the effect of
the jitter.

I have a homebrew GPSDO using a UT+ receiver, with about
+/- 80 ns jitter on the 1 PPS signal, gating a simple counter,
managing a 10 MHz HP10811 ovenized oscillator.

A cycle at 10 MHz is 100 ns, so 100 ns jitter is a variable error of
up to one cycle on top of any other implementation errors.  Your receiver
with 1 us jitter would be ten times as bad.

So with 100 ns jitter, you get a variable count error of +/- 1 cycles
on top of whatever the real error is.

If you count for one second then the system will hover at 10 MHz
plus minus 1 cycle, or about 0.1 ppm accuracy.

If you count for 10 seconds, then the system will hover at +/- 0.1 cycles
or 0.01 ppm accuracy.

I count for 2048 seconds and the system hovers at +/- 2 milliHz
or 0.2 ppB accuracy.  Plenty good enough for what I need.  The
HP oscillator ages at 5 milliHz per day, so plenty good enough to
actually watch the crystal age over the course of a single day,
and watch the system steer to correct it out.

But not good enough to win a frequency measuring contest, because
those guys measure sub 1 milliHz.

But, measuring phase instead of simple counting would improve
my accuracy by another order of magnitude.

Performing the granularity/sawtooth correction would add another
order of magnitude of accuracy.

And there are probably better ways to steer it, rather than a
simple proportional error correction in such a noisy system.
Perhaps adding some trend analysis for the aging and mathematical
averaging of the remaining noise in the correction signal.

The best GPSDOs are measured at several orders of magnitude
better than mine.   Hmmmmm. The slippery slope to becoming
a "Time Nut" is easy to slide down.

If you use your Rockwell receiver, depending on how you run it,
it will likely be on the order of one magnitude worse than mine, but your
system might meet your needs just fine.  And for $9, you will
have a lot of fun learning, and can upgrade anytime you want.

--- Graham / KE9H

==

Chris Albertson wrote:
> ***** High Performance Software Defined Radio Discussion List *****
>
> On Mon, Apr 6, 2009 at 9:11 AM, Graham / KE9H <KE9H at austin.rr.com> wrote:
>
>   
>> Although a cheap way to get started, this is a navigation receiver, not a
>> timing receiver,
>> and it will not be a good choice for frequency control.  This one's spec is
>> that the 1 pps
>> signal has a jitter spec of +/- 1 microsecond.  Timing receivers, such as
>> the Motorola
>> Oncore UT+ have a timing mode, that provides jitter around +/- 50
>> nanoseconds,
>> without sawtooth correction, and sub 10 nanoseconds with sawtooth
>> correction.
>>     
>
> It's not that bad, the purpose of the 1PPS signal is to correct for
> long term drift by providing a direct real time link to a national
> standard frequency reference.
> What the controller might do is count how many times the signal from
> your 10Mhz oven-ized
> XTAL oscillator crosses zero in a one second period.   Subtract this count from
> the expected number.  You will have a small positive or negitive
> number.  The idea is that you very slowly adjust the oscillator so
> that tens of thousands of these small numbers add up to exactly zero.
> This corrects for effects like the age of the crystal or the
> temperature sensor on the crystal oven slowly drifting over time.
> These effects are slow enough that you can average over many thousands
> seconds.
>
> Basically what you are doing is never "jerking" your 10Mhz OSC.  You
> are instead using the 1PPS signal to slowly adjust the "faster or
> slower" knob that we used to see on the back of clocks.
>
> Yes a better GPS unit will be better because your 10Mhz oscillator
> will converge faster because it needs to average few 1 second periods.
>   


 1239059128.0