[time-nuts] Tale of Two GPStars...

[Bob kb8tq]

Hi

Ok, let’s put some numbers on this.

What goes into your typical “Rb GPSDO” gizmo is a telecom Rb. They normally 
start off around 2x10^-11at 1 second. They drop by the square root of tau as you 
go on out. Indeed the PRS-10 has an odd bump in the 1 to 10 second range, there
is an exception to every rule :).

So:

At 100 sec the Rb is at 2x10^-12. A good OCXO can indeed beat this number.

At 1,000 sec the Rb is at 6x10^-13. It is better at that point than 99.99% of the OCXO’s out there.

At 10,000 sec the Rb is trying to hit 2x10^-13 ….

The first gotcha is that the loop adds noise to all these numbers. Your typical telecom GPSDO
box struggles to stay below 1x10^-11 at 10 to 100 seconds. 

The second issue is that these all are impacted by temperature. If your lab moves a couple degrees 
C with a 30 to 90 minute cycle, that’s pretty normal. Temperature sensitivity coefficients 
of 1x10^-11 to 1x10^-12 / C are not unheard of ( 1x10^-11 x 50 = 5x10^-10). That’s going to impact
both the OCXO and the Rb as you go from 1K to 10K seconds…..

The good news (I guess) is that most control loops turn things over to the GPS past a few hundred
seconds. How well the OCXO does past that is sort of irrelevant. There’s not enough data on the
Rb versions to really know what they do. 

Some numbers on Rb vs OCXO in a GPSDO can be found  in the SRS 740 spec sheets. If somebody 
wants to send me an Rb based one I’d be happy to run some numbers. I already have the data on the 
OCXO version. 

https://www.thinksrs.com/downloads/pdfs/catalog/FS740c.pdf <https://www.thinksrs.com/downloads/pdfs/catalog/FS740c.pdf>

The unit is aimed at a lab bench rather than telecom environment. It does indeed do a bit better
than the typical $70 telecom GPSDO. The data sheet shows an Rb to OCXO “crossover” in the
vicinity of 60 seconds.  At least on the OCXO version the data shown should be interpreted as 
“typical” rather than a spec limit …. That may impact the real world crossover point.

Bob

> On Jun 5, 2020, at 12:50 AM, Charles Steinmetz <csteinmetz at yandex.com> wrote:
> 
> Bruce wrote:
> 
>> And I'm still amazed I lucked out with the rubidium module.
> 
> Actually, the unit's PN and ADEV performance would almost certainly be considerably better with a good OCXO instead of the Rb.
> 
> At low tau, the PN and ADEV performance of a GPSDO is determined by the local oscillator (generally, an OCXO or Rb).  The OCXO is better with respect to PN and ADEV than a Rb out to tau >~1000 seconds, so in the very important area of tau <1000 seconds, the Rb unit loses to the OCXO unit.  Above tau >~1000, the averaged GPS signal governs PN and ADEV performance, so you never really benefit from the Rb oscillator.
> 
> Unless holdover performance is the most important criterion (which it almost certainly will *not* be for a time nut), a GPSDO with a good OCXO is superior at all tau to one with a Rb.
> 
> Best regards,
> 
> Charles
> 
> 
> 
> _______________________________________________
> time-nuts mailing list -- time-nuts at lists.febo.com
> To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
> and follow the instructions there.