[time-nuts] Behavior of a disciplined rubidium oscillator?

[John Miles]

Said at JLT isn't currently on the list so I'll forward his input:

----------------------
.	It looks like you are measuring the output of the OCXO or TCXO
inside the LN Rb. That unit has a tempco spec of 20ppb over its operating
temperature, the Rubidium is much much better of course (up to 0.07ppb on
the Ultimate unit!). With a spec of 20ppb, the unit will react to sudden and
large temp changes.

.	The OCXO does get disciplined by the Rubidium, however the time
constant is user-selectable, and thus the speed it takes to recover
frequency/phase errors after a temperature change depends on the user
settings inside the unit. Time constants can vary up to many 100's of
seconds, and this may be the reason why your unit reacts the way it does

.	These units are designed to provide down to 4E-013 performance from
0.1s to 10+ seconds just sitting on a bench typically. You cannot achieve
that if you thermal-shock this unit due to the single oven OCXO inside.
However we do achieve that performance just sitting on a bench here without
any special shielding, with normal AC cycling. So something is potentially
not working properly or not set up correctly with your unit. To debug, we
would need to get a GPSCon log file, see below.

.	If you require very stable frequency/phase over temperature, then
you should use the direct ("unfiltered") 10MHz and 1PPS outputs from the
Rubidium oscillator, and not their filtered versions from the SOCXO. The
manual describes how to switch to the Rb direct 1PPS and 10MHz outputs via
the command line. You should get up to 0.07ppb over temperature on those
outputs if you have the Ultimate version and the unit is properly
configured.

.	If you ordered the lowest cost version, then you only have a good
TCXO as the noise filter. That has a tempco of 75ppb(!) but the time
constants are only 10's of seconds for that unit. It is only used to improve
the phase-noise of the Rubidium. Please put this into perspective: the TCXO
has > 1070x the thermal sensitivity of the Ultimate Rubidium oscillator
itself

.	Please provide log files using the free JLT GPSCon application that
can be downloaded from the JLT website, otherwise it is very hard to figure
out what is happening

Unfortunately there are no DOCXO's on the market that would a) fit into the
1.00 inch high enclosure, and b) have even anywhere near as good an ADEV or
Phase Noise performance of the SOCXO we use in the LN Rb module. It's simply
physics.

We opted for the incredible stability and phase noise performance this
oscillator provides, but that does have drawbacks. If we had used a DOCXO,
the thermal stability would have been 100x better than the 20ppb we have
now, however the ADEV and Phase noise would have been an order of magnitude
and more than 10dB worse. 

So in summary, if you need great phase and frequency performance over
thermal shocks, then please use the unfiltered, Rb-direct output option,
that is why they are available via software command.

Hope that makes sense,
Said

----------------------

-- john, KE5FX
Miles Design LLC


> -----Original Message-----
> From: time-nuts [mailto:time-nuts-bounces at lists.febo.com] On Behalf Of
> AC0XU (Jim)
> Sent: Sunday, February 03, 2019 6:52 PM
> To: time-nuts at lists.febo.com
> Subject: [time-nuts] Behavior of a disciplined rubidium oscillator?
> 
> Time Nuts-
> 
> Here is a reference to the LN RB spec sheet:
> 
> http://jackson-labs.com/index.php/products/ln_rubidium
> 
> It doesn't seem to give a temperature coefficient. It does say the
baseplate
> temp range is -20 to +70C.
> 
> Again, my issue is that a slight change in temperature leads to a huge, if
> short term, time shift. That seems to me like very strange behavior. I
will
> contact JL to see if they have an explanation. It does suggest that the
unit
> should be installed in a temperature-controlled enclosure.
> 
> Jim