Hi So, backing off a bit:
Assuming you are going to compare two data points a bit over 24 hours apart (100,000 seconds): Accuracy of reading pair Result pp Result sci 100 ms 1 ppm 1x10^-6 10 ms 100 ppb 1x10^-7 1 ms 10 ppb 1x10^-8 100 us 1 ppb 1x10^-9 10 us 100 ppt 1x10^-10 1 us 10 ppt 1x10^-11 At 100 ms, you need to go for a *lot* of days before you get out of the “that’s interesting ….. yawn …” category. Since it’s a pair you are concerned about, each might have a 50 ms error to get you to the 100 ms total. It’s still plenty good enough for a typical wall clock. At one day you don’t cross into “as good as a free running Rb” territory until you get below 100 us. That’s also the point you hit roughly the typical GPS module’s performance at 1 second. If you want to get a “Time Nuts” grade solution, microseconds are indeed the units to worry about. Assuming you hit can 100 us per pair, and want to get into the 1 to 10 ppt range, you are out around 100 days for each “run”. That seems like a long time to wait. Since one period is 16.667 us, hitting well below 100 us implies keeping track of which cycle is which. Getting to 10 us is just above a 180 degree phase error and 1 us is still above 10 degrees. I’d say that for a practical “Time Nut" device … you need the equivalent of cycle tracking at the very least. Indeed, as mentioned by … somebody …. this all can be done and … ummm …. has been done for similar signals by …errr …somebody :) Bob > On Dec 4, 2018, at 3:08 PM, David G. McGaw <david.g.mc...@dartmouth.edu> > wrote: > > That is the specified jitter. They have also said in communications that it > has about 50mS resolution. That is as close as they are willing to say a > system can be synchronized with it. Perhaps someone will discover a clever > way to enhance that. > > BTW, I have been told it has also been successfully tested for lock in > Brazil. Is there anyone in Australia want to give it a try? Perth is almost > directly opposite Fort Collins. :-) > > David N1HAC > > > On 12/4/18 2:00 PM, Hal Murray wrote: > > > > > As was said, IRQ delay is +/-100 mS from the second edge, hardly what a > Time-Nut is looking for. > > > > There is no problem with a delay as long as it is constant. If I know what it > is, then I can correct for it. > > The problem is the noise/jitter on the delay. I'm sure somebody will have > data soon. > > > > > _______________________________________________ > time-nuts mailing list -- time-nuts@lists.febo.com > To unsubscribe, go to > http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com > and follow the instructions there. _______________________________________________ time-nuts mailing list -- time-nuts@lists.febo.com To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com and follow the instructions there.
