[time-nuts] Two types of GPDSO / Rubidium

[Chris Caudle]

On Wed, September 25, 2019 5:45 pm, Taka Kamiya via time-nuts wrote:
> discussions take place both for GPSDO with Crystal Oscillators and
> Rubidium modules.  It appears there are two types of each.
> 1)  fixed frequency type (less jitter)
> 2)  frequency agile type (more jitter)

That is a bit oversimplified.  GPSDO is "GPS disciplined oscillator,"
meaning that the output frequency of the oscillator is forced to follow
the inverse of the time measurements derived from the Global Positioning
System.
Obviously if the oscillator has to follow another reference it cannot
truly be fixed frequency on the output, but there are various different
ways that the frequency can be varied.  Depending on the important
parameters for your needs the differences in mechanism may or may not be
important.

> I've read frequency agile Rb modules (ones you can change output
> frequency) is one kind of Rb (sa.22c and fe5650, etc), and there is
> another one that you cannot change frequency. (ie. T-bolt, PRS10, etc).

You are mixing different types of equipment together, so it is a little
difficult to know what you are asking.
FE5650, PRS10 and SA.22C are rubidium frequency standards.  A rubidium
frequency standard is fundamentally an OCXO which is slaved in a frequency
lock loop to a way to measure atomic transitions in rubidium vapor.  There
are various systematic errors which pull the frequency of that atomic
transition measurement, so a rubidium standard will commonly have an
adjustment mechanism that can adjust the output frequency far enough to
compensate for those systematic errors, but you need a more accurate
device for comparison for that adjustment to be worthwhile.

A Thunderbolt is a complete GPS disciplined oscillator system which
accepts GPS signals from an antenna, and outputs 10MHz which is slaved so
that on average there are always 10 million output transitions per
GPS-derived second estimation, and outputs a pulse once per second which
occurs on  the transition of the GPS-derived second estimation.

When the GPS signal is not available for some reason (antenna failure, GPS
system problems, etc.) the GPS disciplined oscillator has to run in
undisciplined mode.  Over that time the corrections to the oscillator
frequency are not available, so only the intrinsic stability of the
oscillator used sets the long term stability of the output frequency.  In
that particular case (no GPS available) a GPSDO using a rubidium
controlled oscillator as part of the design will have more  stable long
term frequency than an oven quartz oscillator.
While GPS signal is available to measure and control the output frequency
there is little difference between an ovenized quartz and a rubidium
controlled (which recall also has an ovenized quartz oscillator driving
the output, but has a control loop using rubidium vapor transitions to
correct any frequency drift of the quartz oscillator).

> Words like phase noise and PLL are thrown out often in discussions.

A very good place to start is searching for John Vig oscillator tutorial.
John Vig has a lengthy presentation about all things relating to
oscillators, if you can find the version which includes the notes along
with the presentation slides you can spend a couple of days studying that

> Is this because frequency agile type has the ultimate output from PLL
> (subject to jitter) and fixed frequency type is from OCXO?

Mixing different types of things into a single question again.
The term PLL is phase locked loop, it is a particular type of control loop
design which measures the edge transitions of a reference frequency input,
the edge transitions of an output signal, and controls the output so that
the edges match the reference.
You can phase lock quartz, rubidium, pendulum.   Maybe an hour glass, but
the only transition I can think of is when you turn the hour glass over,
so that  would be an impractical PLL.

> Even in main well known brands, I understand PRS10
> and sa.22C and fe5650 are fundamentally different.
> I guess they are all "GPS disciplined" in some way

No, SA.22C and FE5650 have no connection to GPS,  they are not GPS
disciplined in any way as stand alone devices, but can be used as part of
a full GPS disciplined oscillator design. The SA.22C has a
pulse-per-second (PPS) input, so it can be connected to GPS, but that gets
into a lot of secondary questions relating to how that PPS is derived and
whether it helps or hurts overall stability.

-- 
Chris Caudle