5061B ADEV curves and questions
The attached is a plot of some ensemble ADEV measurements of several 5061B cesium beam oscillators.
Two outlers are interesting to me, and I am wondering if anyone has ideas about them:
a) 5061B4U was the same unit as 5061B4, but it was obtained when I accidentally tested the unit with the cesium beam turned off. I expect that this curve (brown) is the behavior of the undisciplined OCXO. What I am wondering about is why the unit can't do better than it does up to 300 seconds with the cesium beam turned on. Actually, the same question applies to all of these units. Ideas?
b) 5061B6 is a unit that, according to all the various meter readings, is working fine. However, it is evident that it doesn't lock well. Any thoughts about this one?
Thanks-
Jim
It's possible that 5061B4 and 5061B6 have noisier tubes, but they are still
basically in spec if you are running in OPER mode. The PLL does not look
like a brick wall, either in terms of its loop response or the additional
noise contributed by the rest of the lock-in hardware, so the extra
short-term noise you are seeing with the beam turned on looks OK to me. You
will probably get better performance in LTC mode assuming the units remain
locked. If you are already running in LTC mode, then I'd say 5061B4 and
5061B6 definitely look noisier than expected between t=1s and t=100s.
ADEV traces should always be inspected in conjunction with the underlying
data record when you're not sure what's going on. So, to diagnose the
behavior of 5061B6 past t=1000s, you can (hopefully) look at the
frequency-difference trace to identify when the jumps occurred, how often
they occurred, how long they lasted, and (again, hopefully) what temporary
disturbances might have caused them.
They are all about 10x better than the HP specifications at t=0.01s. This
is par for the course, likely because HP didn't update the specs when they
started using 10811s instead of the older cylindrical OCXOs. You can tell
that 5061B4U has a really nice rock; its ADEV is typical of the best
10811-60109 examples, below 4E-13 @ t=1s and staying in the 13s out to
t=1000s or so. Around one in ten will do that.
-- john, KE5FX
-----Original Message-----
From: AC0XU (Jim) via time-nuts time-nuts@lists.febo.com
Sent: Monday, July 29, 2024 9:05 PM
To: time-nuts@lists.febo.com
Cc: AC0XU (Jim) James.Schatzman@ac0xu.com
Subject: [time-nuts] 5061B ADEV curves and questions
The attached is a plot of some ensemble ADEV measurements of several 5061B
cesium beam oscillators.
Two outlers are interesting to me, and I am wondering if anyone has ideas
about them:
a) 5061B4U was the same unit as 5061B4, but it was obtained when I
accidentally tested the unit with the cesium beam turned off. I expect that
this curve (brown) is the behavior of the undisciplined OCXO. What I am
wondering about is why the unit can't do better than it does up to 300
seconds with the cesium beam turned on. Actually, the same question applies
to all of these units. Ideas?
b) 5061B6 is a unit that, according to all the various meter readings, is
working fine. However, it is evident that it doesn't lock well. Any thoughts
about this one?
Thanks-
Jim
AC0XU (Jim) via time-nuts writes:
What I am wondering about is why the unit can't do better than it does up to 300 seconds with the cesium beam turned on. Actually, the same question applies to all of these units. Ideas?
The PLL/LF/Mod/Demod circuitry in the 5061 is designed to handle the full envelope of the specification, which includes horrors like "30G 11ms" (Table 1-1).
(That sounds like "submarine + near-miss depth-charge" to me ?)
To stay locked through that, the PLL needs to be pretty insistent.
Both of my 5061As had a "TIME CONSTANT" switch, where you could select either "1s" or "60s" "loop time", and even th 60s setting was clearly too trigger-happy in my lab. (I think this switch disappeared in later models ?)
--
Poul-Henning Kamp | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG | TCP/IP since RFC 956
FreeBSD committer | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.
Hi Jim,
Consider that ADEV is not a very good tool to measure systematics, as it
is designed to characterize noise in absence of systematics.
So, look at how the phase varies over time, use the linear regression to
remove phase and frequency offsets and look at the residue (that is push
r for the time-plot in TimeLab) and see what that gives you. For
non-locked cesiums you will naturally have a drift component creating a
parabola, but for OCXOs being heated for a long time, this should be not
so dominant that you can't see other things happening too.
Sometimes it can be fruitful to flip between phase and frequency plot.
Since some modulation is on frequency and not on phase, viewing that in
frequency may be easier to spot, but phase tends to help filter out
quite a bit of noice, but with a 90 degree phase-shift of phase compared
to the modulation waveform.
Considering the low ADEV for the unlocked OCXO, one can wonder if not a
longer time-constant would be beneficial for the lock, but that comes at
the cost of higher sensitivity to other perturbations, as already noted.
Cheers,
Magnus
On 2024-07-30 06:04, AC0XU (Jim) via time-nuts wrote:
The attached is a plot of some ensemble ADEV measurements of several 5061B cesium beam oscillators.
Two outlers are interesting to me, and I am wondering if anyone has ideas about them:
a) 5061B4U was the same unit as 5061B4, but it was obtained when I accidentally tested the unit with the cesium beam turned off. I expect that this curve (brown) is the behavior of the undisciplined OCXO. What I am wondering about is why the unit can't do better than it does up to 300 seconds with the cesium beam turned on. Actually, the same question applies to all of these units. Ideas?
b) 5061B6 is a unit that, according to all the various meter readings, is working fine. However, it is evident that it doesn't lock well. Any thoughts about this one?
Thanks-
Jim
time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe send an email to time-nuts-leave@lists.febo.com