[time-nuts] Tbolt time constant

WarrenS warrensjmail-one at yahoo.com
Tue Oct 13 16:17:45 UTC 2009


> I gather from your referenced web page that it seems that a 
> higher damping number was better. Is this a correct assumption ?
Well Not exactly, It depends on the type and source of the noise error.
The simple answer is that the default damping setting of 1.2 on the Tbolt  
is a good compromise value (when the Dac gain is set correct)

For the Tbolt, all else being equal; the effect of damping is:  
As the damping setting goes down toward about 0.25, 
the Freq noise due to it ringing near the TC freq will go up, 
and the DC type Phase error and its correction time will go down.
Depending on the damping accuracy and the TC setting, 
somewhere around a damping setting of 0.25, the Tbolt loop will become unstable.

As the Damping goes up >1, the AC  phase noise caused by the GPS will go down, 
and the 'DC' Phase offset error caused by the Osc will go up. 
and the time to correct for slow type of Phase errors goes up.

The Tbold has some non typical responses to loop errors.
It depends if the error is Phase or Freq errors and
if the Error is due to the GPS signal it's self or the OSC.

The effective TC to correct for any differences between the OSC and the GPS
depends if you're considering Freq errors OR phase errors.
While Freq errors are pretty much corrected with the time constant set by the TC value (when Dac gain is correct)
A phase error takes the Tbolt more like 4 or 5 TCs to correct.

When I set the damping to 0.5 to 0.7, the Phase error correction is faster,
but the trade off is the Freq noise will increase because of the Freq correction overshoot.

In the case of a Tbolt in a good environment, the biggest source of error I've seen is 
Freq error cause by the loop responding too quickly and non linearly  to GPS noise. 
I find that an over damped setting of 1.5 to 2.5 helps the noise at TCs around 400 when the Osc is stable.

At a 400 TC setting, when I care more about Phase error than Freq noise, 
I use an under damped setting of 0.5 to 0.7. 
When I care more about Freq noise than phase error 
I use an over damped setting of 1.5 to 2.5, 
and for a compromise, the default damping  of 1.2 works good.

Have fun,
ws

***********************
From: <bg at lysator.liu.se>
>Hi Bill,
>
>While certainly not beeing Tom...
>Damping is used in control theory to describe how the time response of a
>second order system behaves. Depending on your feedback regulator you
>might create a second order system.
>
>   http://en.wikipedia.org/wiki/Damping_ratio
>   http://en.wikipedia.org/wiki/File:2nd_Order_Damping_Ratios.svg
>
>Higher number means slower reaction to a change in the system. Lower
>number equals a quicker response but with "ringing" overshoot.
>
>The time constant is related to how quick changes in the error signal the
>regulator is capable of following. Long time constant means the regulator
>will attenuate high frequency content in the error signal. We can have
>longer time constants when we trust the OCXO is very stable at these time
>scales.
>--
>   Björn
*****************
>Hi Tom,
>
>Can you explain what is meant by "damping ?"
>How is the "damping" related to the time constant ?
>I gather from your referenced web page that it seems that a higher number was
>better. Is this a correct assumption ?
>thanks,
>
>Bill.....WB6BNQ
>
***********************
>> Tom Van Baak wrote:
>>
> Does the longer time constant actually make a "better" frequency standard, or 
> does it just push the noise from one frequency region to another?  Has 
> anybody with gear good enough to measure close in noise looked at the 
> spectrum for different time constants?
>
>Hal,
***
>>
>> Hal,
>>
>> Last I checked - phase noise, as in L(f), is the same for all
>> time constants.
>>
>> Frequency stability, as in sigma(tau), ADEV, etc. typically
>> improves as the TC increases, out to 1000 s. The default of
>> 100 s works pretty well for normal environments. But if your
>> sat visibility and voltages and temperature and air currents and
>> local seismic (vibration, shock, kids playing, doors closing, etc.)
>> environment is unusually stable then you could bump the TC
>> beyond 100, maybe even near 1000. I run my best TBolts
>> closer to 300.
>>
>> For some time constant plots, see:
>>
>> http://www.leapsecond.com/pages/tbolt-tc/
>>
>> /tvb
***************





More information about the time-nuts mailing list