kb8tq at n1k.org
Sat Sep 1 08:21:46 EDT 2018
> On Sep 1, 2018, at 3:06 AM, Magnus Danielson <magnus at rubidium.dyndns.org> wrote:
> On 08/31/2018 03:36 PM, Bob kb8tq wrote:
>> “Backbone timing” gets done by boxes buried deep in systems. Those systems take years
>> to design. The boxes that go in them similarly take years to get onto the market. Once designed
>> deployment is far from instantaneous. Operators are always pressed by cost constraints. Adding
>> anything beyond the minimums … not going to happen.
>> The result is that there are no systems out there that use WWVB or WWV other than wrist watches
>> and wall clock like devices. Utilities (cell phones, internet, finance ) run with something else. Converting
>> them to a secondary “something” is a many years sort of thing, even if it is technically feasible.
>> You can pull a bunch of spare GPS sat’s out of storage and get them in orbit *way* quicker than you can
>> rebuild every cell tower in the country. In fact, newer designs run their timing in a way that a GPS failure
>> is not that big a deal. How long it’ll take before that sort of design is common in the US…. years and years …
>> If you are going to come up with a time source at the ~ 10 ns level, that’s not going to happen from WWVB
>> or WWV. They never were good enough to get to that level and it’s not on the transmit end. You would need
>> a very different system. It’s been a long time since any of these services (internet, finance, cell ) were in the
>> millisecond or even the microsecond range. The modern stuff in all theses areas is < 100 ns.
> The actual requirements is usually on the 1-10 us level, but they are
> happy when they have the extra precision.
Well, they are and they aren’t. The newer systems (which rapidly become the only system) are at the one microsecond
level after being in holder for many hours ( days?). The assumption is that only one tower (or chunk) goes into holdover
at a time and the rest are still at least 10X better than that. Cut them all loose an the numbers would have to be tighter.
In order to get them all at the 100 ns level, you need a source that is around 5 to 10X better than that. The timing source
is not the only source of error and you need to “train” your holdover clock with something that is mighty good. The
holdover spec often applies after a very short period of training (a day to several days). Indeed, if you put a Cs standard
in every cell tower and every internet node you probably could back off a bit on the 10X. With quartz or Rb, you need
the accuracy in training.
>> How long would it take to change all this? Well first some random Senior Member of the IEEE would
>> have to start writing papers about the various issues. Various organizations in various countries would
>> need to hold meeting after meeting after meeting talking things over. Somebody eventually would have
>> to come up with funds to actually try a few things. Maybe they work in the real world / maybe they don’t
>> Once you prove you have a system that can do “good enough", you would need laws / regulations passed to
>> make the “new thing” part of the required designs. You also need funding bills to deploy the “source” end
>> of things and time to get that up and running. Once it’s running, you then give manufacturers some amount of time
>> to get it in the field ….. and extensions when that doesn’t happen. Twenty years? Thirty years? Maybe longer?
>> This stuff does not go very fast.
> It's been done for 10+ years now. Some 15+ countries have nation wide
> networks that makes them GPS independent for some applications.
> It has been a fun system to design and deploy.
> Getting more precision isn't all that hard, it just takes more effort in
> the details and hence money and time. If people need 100 ns or 10 ns
> system time, it can be done.
>> Best bet on what the “new thing” would be? Something like IEEE-1588 over fiber. It cuts out a bunch of this and
>> that in terms of experiments and testing the basic system. We know most of *how* to do it already. It’s just a matter
>> of a billions of dollars in tax money to get the gaps filled in and then a few tens of billions in tax money to get
>> the backbone gear in place. Once that’s done you ramp up to the really expensive part of the deal ….Is it paid
>> for by your tax return in April or by a higher price on every cell call / transaction you make? … who knows … it’s
>> a tax that you are paying either way.
> IEEE 1588 isn't working out at all for WAN, it's stuck in LAN
> environment, which is the dirty little secret of the industry. There is
> trials for dedicated wavelengths with using 1588 or it's extension White
> Rabbit that works great, but in any form of production environment it's
> a mess.
> Network based precision timing takes a number of hurdles to handle. I've
> done my fair bit of them.
> time-nuts mailing list -- time-nuts at lists.febo.com
> To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
> and follow the instructions there.
More information about the time-nuts