[time-nuts] Measuring ADEV using TBolt-Tic tester

WarrenS warrensjmail-one at yahoo.com
Sat Oct 8 01:05:12 UTC 2011

ws) responses below

From: "Tom Van Baak" tvb at LeapSecond.com

>> Long term, really low noise is all about the quality of the GPS signal 
>> and engine.
> And the antenna, and the multi-path, and ionosphere, etc. ...
ws) And all the other things that go into making a "Good GPS signal setup"

> With a Z3801A or TBolt (or any cheap single channel GPS) receiver
> you should expect maybe a 5 to 10 to 15 ns variation over a 12 or 24
> hour period. You should be able to see this with a 5065A or a good
> Cs reference.
ws) Are your Phase numbers RMS, Peak, or Peak to peak?
ws) On a good day 1/2 day my Tbolt's phase can be below 10 ns PP, on a not 
so good couple of days it can be 5 to 10 ns RMS

>> The Z3801's GPS engine is far inferior and not even close to the 10 to 
>> 100 ps, 1 sec noise that a Tbolt engine has.
> Careful, each does averaging. So to compare you have to include
> the time constant into the equation. The RS232 numbers the TBolt
> reports are heavily averaged so at 1 second they always look a lot
> better than they actually are. If you have hard data for your claim,
> I'd be interested in seeing it.

ws) My statements are when using raw 1 sec data without any averaging.
For the Tbolt, a test anyone can do is turn off LH filter and measure the 
"ppt" RMS freq offset value over a short time period.
For the Z3801, I think you have to take the difference between each two 
phase output values to be able to see freq changes in 1 sec.

ws) To validate the data, just need to cause the Osc to make a small known 
freq step in under 1 second and see what happens.
See how big of freq step is needed to be able to see a freq change above the 
noise and measure how long it takes for the new average freq value to 
Do the above with several different filter settings such as 0, 10, and 100 
The freq step can be done with a voltage step on the osc's EFC (using the 
manual DAC setting) or by tilting the osc and letting  "G" do the freq 

ws) Tbolt answers I got are:
5e-11 Peak freq change is visible and valid in the next 1 sec output with 
LH's filter set to zero
2 e-11 peak freq change setting is visible and settled within 10 seconds 
with LH filter set to 10 seconds
5e-12 freq change is measurable within 100 seconds with LH Filter set to 100 

ws) I've never measured a Z3801 but I have measure an oncore GPS engine 
which I think is what is used in them.
It had a couple decades worse resolution for the one second test

ws) Before picking at the fine details and getting totally off the subject 
which is how to use the Tbolt to make valid ADEV measurements.
How about someone verifying the method by comparing the ADEV noise readings 
using an external Osc that is known quieter than the Tbolt's Nose floor 
(such as a low noise 10811)
and one that is nosier  than the Tbolt's freq noise floor (like a LPRO).
For a plot of the Tbolt's noise floor I got see:

An important first step, and one simple way to verify that the Tbolt's total 
GPS antenna "system" is ALL setup and working well,
is to check that the Tbolt's phase noise is well under 3 ns RMS as reported 
by LH, over a one or two day time span with the TC set to 1000 sec and 
Damping set to 0.7 and with the display Filter off.
Also a good idea to check that the ppt RMS freq noise is around 3e-12 with a 
filter setting of 100 sec and 1.5e-11 with filter off with the above 


> /tvb

More information about the time-nuts mailing list