[time-nuts] distirbuted sync

Magnus Danielson magnus at rubidium.dyndns.org
Tue Jul 23 12:07:28 EDT 2013


Hi Jim,

On 23/07/13 16:55, Jim Lux wrote:
> Starting a new thread...

Good idea!

> Brian wrote:
> Have you considered WWVB? Works fine within structures.
> Even though the carrier today is phase modulated one can probably glean
> 1 ms accuracy from it or the data transmitted.
>
> -- disasters occur world wide, any time day or night, so depending on
> WWVB won't work.
>
> Magnus wrote:
>
> The movie-business have similar problems, so a sync-ones and keep drift
> low system emerged to make field recordings easier.
>
> If it would be tolerable to have a "central" transmitter, putting a PN
> code over a voice radio system would suffice to keep the drift fairly
> well kept together for this form of system. If you choose to do it on
> the audio channel, then you can use of the shelf radios, and replace
> those or re-program those as needed. Also, they are dirt cheap nowdays.
>
> -- yes, I'd thought about that, but that's another piece of gear (the
> centralized transmitter). And if we added another radio into the system,
> you get into the whole size, weight, power aspect. We already have to
> have the GPS (for those places where GPS is available). We already have
> a comm link of some sort (TBD.. long range Bluetooth possibly) between
> the modules, so we could transmit a sync signal on that. The question
> would then be whether *that signal* has modulation and propagation
> characteristics that allow the frequency disciplining. BT is 2.45 GHz,
> and subject to all the multipath and other ills we encounter with the
> radar at 3 GHz.
>
> This is a bit down the road a bit, so what might wind up being the
> ticket for GPS denied is putting up GPS pseudolites at the site. that's
> back to the "extra piece of gear" problem, but maybe we could make a
> case that it isn't *our* piece of gear<grin>.
>
> Or have our modules have the ability to transmit a GPS-like signal that
> the GPS-18 would appropriately handle (oh yeah, I can see the regulatory
> issues looming for that one!)
>
> It *is* an interesting problem.. It's sort of weird, though, as I write
> the requirements..
>
> High frequency accuracy (1E-10, 1E-11) ideally.. or high stabiity over
> 1-100 seconds, with a way to get "knowledge".
>
> But relatively low timing accuracy: 1-3 milliseconds over the same 100
> second interval (1E-5)
>
> Often you have a time requirement that is commensurate with the
> frequency requirement.

Consider my audio channel over radio-sets.
Consider a 1 kHz sine being modulated.
Consider that lock your 10 MHz to that 1 kHz.
Keeping within a few degrees on that 1 kHz should not be too hard, 
unless you bump/turn the crystal too much.
A little intelligence to bridge the gaps when radio fading occurs is 
naturally needed.

Let's assume you maintain within +/- 1.8 degrees, that's 1/100th of the 
1 ms period. You have a maximum of 10 us drift in 1 s, which really is 
measly 1E-5, but as it is locked, it moves *tau which means 100 us at 10 
s and 1 ms at 100 s... for the peak error as you go into hold-over. It 
will naturally be better as it maintains track.

Seems feasible to maintain the needed tracking requirement on the 
back-of-envelope level of analysis.

Oh, if you have a number of these devices spread out, and coordinated 
time between them is important, just assign one of the devices as 
master. That way it disappears as "extra device".

Cheers,
Magnus


More information about the time-nuts mailing list