[time-nuts] Brand new OCXO/TCXO reccomenations

[Forrest Christian (List Account)]

I accidentally only replied to an off-list reply with the following
information.   It was also meant to go to the entire list:

  Let me clarify here a bit:

The application is:

1) In an unconditioned/semi-conditioned space.   The ambient
temperature can slew 20*C easily over the holdover period.   (Think sun
hitting an outdoor enclosure first thing in the morning - the ambient
temperature tends to rise quickly).   The overall temperature range is -40
to +60C, although those edges are rarely hit, and usually only when
fans/heaters fail (more typical is 0-50C).

2) I need to stay within a 'few' microseconds of UTC alignment.   No it
isn't specced better than that.  4-5 uS is probably a good target.
However, 10uS is probably ok and 30uS has been shown to not cause a lot of
problems.  Beyond that things get bad fairly quickly.   Nominal 1PPS UTC is
coming from a GPS receiver.   The application is ok with a uS or so of
cycle-to-cycle jitter.

3) 10 minutes would be a good target.   Longer is better, like in any
holdover.

My basic math has been:   5 uS over 10 Minutes is ~8ppb.    (5uS/600).
If I can tolerate worst-case of 30uS, then I can get away with 50ppb.

The implementation idea today is to clock a FPGA with an appropriately
stable clock, then do either a 1Hz PLL or some other similar arrangement in
the FPGA designed to track the GPS 1PPS when it's operational, or switch to
holdover when it's not.   I don't particularly see a good reason to
discipline the oscillator when I can effectively measure it's frequency in
the FPGA - If I know the measured frequency of the oscillator is
10.0000033425 Mhz, I can output the 1PPS every 10000003.3425 cycles, with
each 1PPS aligned to the next clock edge.   The jitter introduced by this
method isn't going to be a problem in this application.   Or put
differently:  I'm not steering the oscillator, I'm steering the divider in
the FPGA.

So my spec when I get right down to it is that I could tolerate a clock
source which is fine with a worst-case drift of 50ppb over 20*C, but I
would be more comfortable with something close to 10ppb.

I note that Abracom and probably others make TCXO's which are rated at
50ppb over their entire temperature range, and the charts show that  I
probably can expect better than that over almost any given fairly narrow
temperature range - and the lower power and ease of use are very much
plusses.   But, they cost more than an low-end OCXO.

The main disadvantage of the OCXO's is the higher power needs.   And my
tendency to say 'well, I've already got a OCXO in the circuit, maybe we
should go up to the expensive parts...

I apologize for the softness of the specs here and (even worse) in my
original message.  Right now I'm sort of just playing with the 'is this
doable for a reasonable cost, and what are the cost/benefit tradeoffs'
game.   And trying to get something I can play with and throw into the
"environmental chamber" (aka converted old freezer) for testing.


On Wed, Sep 23, 2020 at 6:20 PM Forrest Christian (List Account) <
lists at packetflux.com> wrote:

> I'm hoping some of the list are familiar with the newer
> currently-available oscillators available from the distributors.
>
> Some background:
>
> I'm starting to play with a short-term 1PPS holdover application, mainly
> to cover up brief GPS signal loss events (a few minutes at most).
>
> Since I may at some point want to make numerous copies of this, I'd rather
> start with a new oscillator as opposed to digging through the junkbox or
> buying surplus.   Lower cost (<$100 ideal, lower is even better) and power
> consumption is good.   My short-term stability requirements indicate to me
> that I'm probably looking for an OCXO instead of a TCXO, although it looks
> like some of the best TCXO's out there are in the range which will work in
> most cases (50ppb over temperature (or better)).
>
> I suspect I'm going to handle the holdover in a FPGA.  As such, I don't
> think I'm going to bother trying to voltage control the oscillator.  Seems
> like it will make the overall circuit simpler and I don't have to worry
> about a whole bunch of control circuit temperature compensation.
>
> I've dug through enough datasheets at this point, my eyes are glazing
> over.   I also know that the spec sheets are only part of the story when it
> comes to oscillators.  I also have found over the years that often the part
> I'm looking for (that 20ppb Temperature-stable TCXO for a reasonable price)
> exists if you know the vendor to look at...
>
> So any specific recommendations or pointers toward a brand/type would be
> appreciated.
>
> --
> - Forrest
>


-- 
- Forrest