Author Topic: GPSDO frequency error and Allan Deviation  (Read 1974 times)

0 Members and 1 Guest are viewing this topic.

Offline kg4arnTopic starter

  • Supporter
  • ****
  • Posts: 271
  • Country: us
GPSDO frequency error and Allan Deviation
« on: July 17, 2019, 02:25:25 pm »
 Forgive me if this is common knowledge or discussed elsewhere.  I can’t seem to find the answer to this question and everything I have web searched seems to stop just short of explicitly discussing it.
I have a surplus communications GPSDO :  LUCENT RFTG-u.  I am using it as a frequency standard for my home lab.  The OCXO inside is an MTI 260 series 5MHz.

The short term stability of the OCXO is specified 1E-12 for 1 second and 3.5E-12 for 10 seconds.

The software polls the GPSDO every 10 seconds.

My question is:  what is the Allan Deviation graph telling me? 

 1. Is the reference frequency output error for 2 measurements taken 10seconds apart is 3.09E-9 (the tau=10 second Allan deviation value)? And that I have to average 1024 measurements taken 10 seconds apart to achieve an error of 2.38E-12?
2. Or I am misunderstanding completely and if the GPSO is allowed to run for 40960 seconds the noise will have averaged out and the error is now limited by the OCXO short term error of 3.5E-12?
 
 
 

Offline awallin

  • Frequent Contributor
  • **
  • Posts: 694
Re: GPSDO frequency error and Allan Deviation
« Reply #1 on: July 17, 2019, 02:51:46 pm »
from your post I did not understand what your software is measuring. clock measurements are always comparisons - do you have two identical GPSDOs?

My question is:  what is the Allan Deviation graph telling me? 

roughly: average the frequency data into 'bins'/windows of length tau, and the ADEV is the 'rms' noise of the resulting time-series.
 

Offline kg4arnTopic starter

  • Supporter
  • ****
  • Posts: 271
  • Country: us
Re: GPSDO frequency error and Allan Deviation
« Reply #2 on: July 17, 2019, 10:47:31 pm »
Thanks awallin
 The software documentation states that the Allan Deviation is calculated by comparing:
The software documentation states that the Allan Deviation is calculated by comparing the:
 1pps signal locally generated from the OCXO by means of a frequency divider
  With
 The receiver’s 1pps (derived from the satellite constellation).

However, when I read (again) further it appears to answer my question. :palm:

“The comparison of these two PPS signals is the input information to the unit’s frequency standard
control loop. Because this input information is dominated by the GPS 1pps jitter (resulting largely
from GPS receiver and propagation limitations), do not be fooled into thinking that the Time Stability
Measure presented here is that of your frequency standard. It is only an internal value which
indicates overall system performance (including the GPS reception)
. For reliable stability
measurements, you should make independent measurements against a local high stability reference.
The task of the frequency standard’s control loop is to ‘marry’ the high short-term stability of the
internal OCXO with the high long-term time stability of the GPS system. For this reason the 10MHz
output of your unit will have significantly better stability (for short observation times) than the
values indicated in the Time Stability Measures chart.”
[size=0pt][/size]
 

Offline awallin

  • Frequent Contributor
  • **
  • Posts: 694
Re: GPSDO frequency error and Allan Deviation
« Reply #3 on: July 18, 2019, 11:15:54 am »
right. you really want two or three clocks to do comparisons..

SRS has a graph of how they think the ADEV for GPSDOs with different LOs should behave:
https://www.thinksrs.com/images/instr/fs740/FS740_lockedLG.gif

so with an OCXO or Rb LO the GPS-steering should take over at time-constants of 1000s (16 minutes) or more.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf