[time-nuts] Rb vs.Crystal OCXO

[Charles Steinmetz]

Magnus wrote:

>The PRS-10 have a nice little trick in it, it stores the previous 
>OCXO steering value, so on power-up it sets the OCXO to this

The PRS-10 has quite a number of nice tricks, in addition to 
particularly good engineering and high-quality construction of the 
basic physics package and support circuitry.  The OP (and others) 
should not expect the same level of performance from $30-$100 ebay 
Rubidiums (LPRO, FRS, FE-56xx, etc., etc.).

Very good to excellent OCXOs are available readily for $5 to 
$50.  IMO, those should be the standard of comparison for any 
aspiring time nut.  I'm not aware of any economy Rubidium that has 
close-in phase noise or low-to-medium-tau AVAR nearly as good as one 
of these very good OCXOs.  As mentioned by others, some Ru may do 
better than a TCXO close in and at low tau.  But so what?  The TCXO 
should not be a time nut's standard of comparison as far as a lab 
standard is concerned.

One quickly concludes that a good GPSDO, which includes a good OCXO, 
is the optimal solution for most time nuts.  The OCXO has excellent 
stability with respect to close-in phase noise and low-to-medium-tau 
AVAR, and is disciplined by the GPS for excellent stability at longer 
tau.  Probably the best turn-key solution is a Trimble Thunderbolt 
(although prices have risen in the last few years, so they are not 
the bargain they once were).  Other, less expensive Trimble units 
that are also supported by the Lady Heather monitoring program are 
available on ebay, and are probably the best bet today for 
bargain-hunters.  While I applaud the recent efforts to build simple 
DIY GPSDOs using inexpensive microcontrollers, from what I have seen 
so far most of them do not yet have the programming sophistication, 
particularly in the PLL loop filter and the houskeeping functions, to 
rival a good off-the-shelf GPSDO from a quality manufacturer.

Final thought for specifying/designing/buying a GPSDO for time nuts 
purposes:  Do not settle for a low-quality crystal oscillator (and 
especially not a TCXO).  You will never achieve best performance at 
tau < about 100 seconds that way.  Insist on a "10811-quality" OCXO 
(one of the many nice things about the Thunderbolt is that it has a 
very good OCXO).

Best regards,

Charles