[time-nuts] Rubidium (Rb) or Caesium (Cs)
Tom Van Baak
tvb at LeapSecond.com
Tue Oct 18 22:32:30 UTC 2011
> I have a couple of suggestions for you; in another posting.
OK, now on the subject of how to measure the long-term
performance of a GPSDO, even a Rb-based GPSDO.
As you know one of the symptoms of this hobby is that you
search for ever greater stability. That's because to know for
sure how good your clocks are, you either to have build two
(or more) of them and compare amongst themselves, or you
have to compare your clock with someone who's got a better
The latter can be done a number of ways. One method is to
send your clock to a lab with a much better clock. You get
a report back and you're happy. Most national labs do this
as a (paid) service. I do it (for fun) for fellow amateurs.
I should add that clocks are sometimes "better" because of
phase noise, or because of frequency stability, or because
of time accuracy. Note that physical attributes like power,
weight, reliability, and cost are also factors. So really there
is no one "best" clock. Still, for the likes of us the usual
criteria is simply frequency stability.
The other method is with radio time transfer. This includes
WWVB, NTP, and GNSS. With the amazing performance
of modern GPS (and assuming the FCC doesn't destroy it)
this is what most people do these days for precision time.
GPS-based time transfer performance varies enormously.
Most hobbyist know about inexpensive GPS receivers that
have a 1pps output. You can now find these on eBay for
$10 to $50. It's just incredible; and one reason so many
people can play with precise time like never before.
Then there are all the commercial or homebrew GPSDO,
often built using these timing receivers at their core. They
vary quite a bit depending on the quality of their oscillator
and the quality of their implementation or firmware. The
famous hp Smartclocks were the darling of the amateur
community until recently when their supply dried up.
Then there are Trimble Thunderbolts which combine both
a high quality oscillator with a design that uses the LO as
the internal timing source. These are the new darling and
are still available on eBay.
Anyway, it appears the limit of any of these inexpensive,
surplus, single-frequency, C/A code-only devices is on
the order of a couple of ns per day. Although more than
for most people, it's never enough for some. I understand.
So off the top of my head here is a list of tricks to get to
the next level. They are not quantified or necessarily in
a specific order.
** Get a better local clock -- a cesium or couple of cesium's
will work quite well. An H-maser even better; but only a few
have shown up on eBay in the past decade. At first you
measure all the errors in GPS; but eventually, long-term, you
end up using GPS to measure the errors in the clocks. It's
all fun and highly educational, but be mindful of your goals.
** Clock ensemble -- Now that you can buy nice rubidium
oscillators for under $100 it would make a very interesting
project to combine 2 or 5 or 12 of them into one ensemble.
This would give you a local reference that is many times
better than any single OCXO or Rb. You can either steer
it with GPS or use it to better measure a GPSDO. A fun
project that no one has jumped on yet.
** Common View GPS -- developed in the late 90's this is
still a very useful way to go. Over than range of hundreds
to thousands of miles you can eliminate some of the errors
inherent in GPS.
** All-in-View Common View -- it was thought that using
one SV would be a good solution but there are so many
SV out there now that simply using the default all-in-view
algorithm in a common view setup gives good performance
with a lot less hassle.
** WAAS -- I don't know where this fits in; some modern
receivers get corrections from WAAS channels. Someone
can tell me what the net timing performance gain is, or why
I hardly hear about it except for landing airplanes.
** Dual-frequency receivers -- a large gain in performance
since it allows the receiver, in real-time, to estimate and
remove propagation errors. Requires L1/L2 GPS antennas.
** Carrier Phase -- also a large gain in performance. There
are dozens of papers in the last decade that show graphs
of code vs. carrier performance. Note all the cheap GPS
receivers we use, like the Z3801 and TBolt are code only.
Carrier phase is usually found on expensive receivers.
** Choke ring, ground plane -- this has been covered before.
Along with temperature stabilized antenna, cables, and/or
receiver, it's an inexpensive way to gain some performance.
I don't have numbers but it seems at the point people use
dual-frequency and carrier-phase receivers they start to
pay this level of attention to the antenna.
** Post-processing -- there is a penalty for knowing what
time you think it is right now vs. knowing what time it really
was an hour or day ago. Most high-end navigation/timing
systems separate data collection from precise results. If
over time you can combine the results of tens of hundreds
of other labs everyone can benefit from greater accuracy.
When you get below a nanosecond, precise orbits and
signal propagation are very complicated and variable.
** Multiple GNSS solutions -- many high-end receivers
combine GPS, GLONASS, and Galileo: the more sats
** TWSTTS -- two-way satellite time transfer systems. I
mentioned this in the earlier email.
More information about the time-nuts