[time-nuts] Leap Second results: cheap GPS/1PPS receivers

Thu Jul 2 11:37:11 UTC 2015

Hi

You don’t *have* to go to the “many days” approach if that’s beyond the accuracy you require. 
You *can* go there if you need to. If you don’t have a GPSDO (or something similar) it may not
make much sense to go quite that far. 

Even with less accurate sources, the same approach still works. You can compare a pair of 
XO’s with the same approach. You may need to check them every hour. You probably have 
all the data you need in < 10 seconds….

======

The next layer to this is the computing counter. These gizmos came out in the late 1960’s and have
been pretty common ever since. They look at phase (period) as well as frequency. That lets them 
have a constant number of digits regardless of input frequency. Before they came out, you needed
to do the period stuff manually (been there done that …). As long as you are running one of them, 
about any square wave frequency is as good as any other one. 

I’ve been out of school for “quite a while” (gulp). Despite that, I’ve never used anything other than
a computing counter in an industrial setting. Even here at home, I’d have to look pretty hard to
find a counter that does not run that way. The auction sites will sell you one for < $100 delivered.
You may not be able to find a 10811 OCXO for the price of one of these counters ...

Bob

> On Jul 1, 2015, at 11:59 PM, Bob Albert <bob91343 at yahoo.com> wrote:
> 
> Oh boy Chris, some good information there.  I guess my first step would be to get a GPS disciplined source.  My preference would be to use an external 10 MHz source.  And even if I did, I'd lose it some times of day depending on propagation.  Maybe the 1 pps to set my OCXO as close as I can get it, then recheck every few weeks or months to look for long term drift.  But the gear is already old so aging has likely stabilized.
> 
> I assume I can buy a 1 pps source that is practically dead-on accurate and use it when I want to verify my master oscillator.  Maybe you can guide me as to the way to do that with minimum fuss and maximum precision.
> 
> The bottom line here is that I have absolutely no use for such precision.  It's just for fun and education.  I am one of those nerds who likes to stare at a counter and see that each successive reading is within one count and even do a mental statistical analysis to decide which digit is closer and by how much.
> 
> It's safer and cheaper than hanging out in bars.
> 
> Bob
> 
> 
> 
> On Wednesday, July 1, 2015 8:11 PM, Bob Camp <kb8tq at n1k.org> wrote:
> 
> 
> Hi
> 
> Take your OCXO and divide it down to 1 pps.
> Take the PPS out of your GPSDO
> Feed both of them into your counter
> Time the “delta” between the two to the resolution of a 40 year old 5335 ( low ns). 
> Wait 10 seconds
> Repeat
> 
> You now have data to a resolution of ~2x10^-10
> 
> Shut down the counter, go to bed, come back 100,000 seconds later (a bit over a day)
> 
> Repeat the same measurement
> You now have data to ~2x10^-14.
> 
> Come back each day for week
> Each day you get another point that is in the low parts in 10^-14 range
> 
> There are not a lot of other cheap ways you can do that in your basement. 
> 
> Now if you fire up and try this, there are a few things to watch out for. If the two sources are not fairly close
> in frequency, you will have the PPS signals roll past each other. That’s one of the reasons you take that
> first 10 second reading. It lets you see if things are close enough for this to make sense without correction. 
> 
> Another subtle issue are sources that drift more than you might like. Unless you can count on the source to 
> be stable to << 10 ppm, stretching the points out to a full day apart probably is not a good idea. 
> 
> Lots of fun !!
> 
> Bob
> 
> 
> > On Jul 1, 2015, at 6:47 PM, Bob Albert via time-nuts <time-nuts at febo.com> wrote:
> > 
> > Okay that makes some sense.  I will have to ponder over this and see how it will help.  But in any case (which I sort of expected) it's a time consuming procedure, during which the counter isn't available for other use.  And when I go to bed I won't know the next day if I might have missed a count.
> > This business of precise time is amazingly interesting, especially since an amateur like me can make some pretty impressive calibrations.
> > The OCXO in the 5328A runs 24/7 but I don't know if it drifts more when the counter is on than it does with the counter off (due perhaps to internal temperature changes).  Of course, any drift while the counter is off is not important.
> > Luckily I can often receive 25 MHz WWV so that is handy for calibration.  I also have an 8657B with its OCXO although that one isn't running with the unit switched off.  (The oven is holding temp 24/7).
> > All advice here eagerly sought and appreciated.
> > On a side note, when I measure the 8657B with the 5328A at 700 MHz it's off by just a few Hz, and stabilizes after a half hour or so.  It's the generator that needs to stabilize, not the counter.  I am still unsure which unit is closer to being correct but I suspect the counter is closer.
> > 
> > Bob
> > 
> > 
> > 
> >    On Wednesday, July 1, 2015 3:36 PM, Bob Albert <bob91343 at yahoo.com> wrote:
> > 
> > 
> > So I can measure the time interval, which should be one second.  If I measure 10 intervals, my resolution would be 1 part in 10 million.  I think I can set the master oscillator closer than that with beating against WWV. So maybe buying a 1 pps source won't improve anything for me.
> > 
> > 
> > 
> >    On Wednesday, July 1, 2015 2:39 PM, Hal Murray <hmurray at megapathdsl.net> wrote:
> > 
> > 
> > 
> > subscriptions at burble.com said:
> >> My linux boxes also did a double 59:
> >> Wed Jul  1 00:59:58 BST 2015
> >> Wed Jul  1 00:59:59 BST 2015
> >> Wed Jul  1 00:59:59 BST 2015
> >> Wed Jul  1 01:00:00 BST 2015
> >> Wed Jul  1 01:00:01 BST 2015 
> > 
> > That's to be expected (unless you have fancy software).
> > 
> > Internally, Linux (and most other modern OSes) use UTC.  The API doesn't 
> > include the TAI offset.  There is no way to talk about the extra second.
> > 
> > The same problem happens with the switch to summer time.  You need the 
> > time-zone to tell if the time within the magic extra hour is summer time or 
> > winter time.  Without the zone, you have to pick one and then there is no way 
> > to talk about the other one.
> > 
> > To insert the extra second, Linux just sets the time back 1 second, the same 
> > way you set your clocks and watches back in the fall.
> > 
> > 
> > 
> > -- 
> > These are my opinions.  I hate spam.
> > 
> > 
> > 
> > _______________________________________________
> > time-nuts mailing list -- time-nuts at febo.com
> > To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> > and follow the instructions there.
> > 
> > 
> > 
> > 
> > 
> > _______________________________________________
> > time-nuts mailing list -- time-nuts at febo.com
> > To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> > and follow the instructions there.
> 
> 