Poul-Henning,

I mentioned yesterday about integrator windup, this problem is similar but 
happens even without any I term present:

The problem is that the ocxo maintains its frequency even though the EFC 
control voltage is changing. Thus phase error is accruing making the efc larger 
and larger due to the P term.

Then at some point the crystal 'snaps'  and jumps in frequency, overshooting 
the desired frequency and causing the P term to start pushing in the opposite 
direction repeating the cycle.

Very similar to integrator windup, but not quite the same.

Main problem is the crystal is not following the steering input.

Most TCXOs and cheap ocxos have this problem, and there is no way to do 
anything about it since in the worst case the crystal simply refuses to run at 
proper frequency and thus the frequency will be approximated by cycling below 
and above the target frequency. Mind you we are talking about effects on the 
tens of parts per trillion levels. Enough to jump 10s' of ns back and forth 
over many minutes.

Bye,
Said

Sent From iPhone

> On Oct 20, 2014, at 23:41, "Poul-Henning Kamp" <p...@phk.freebsd.dk> wrote:
> 
> --------
> In message <766d6811-f733-4ab2-8574-24e4606e4...@aol.com>, Said Jackson via 
> tim
> e-nuts writes:
>> Thats exactly right Bob.
>> 
>> By the time your ocxo jumps to catch up to the efc voltage, you
>> have oversteered,  then the process starts in reverse and the ocxo
>> jumps in the opposite direction.
> 
> This is a well known PI effect called "windup".
> 
> The cause is a phase offset of opposite sign of the frequency offset.
> 
> The fix is simple:
> 
> Start running with only the P term, and engage the I term only after
> 
> 1. The input phase offset changes sign
> 
> or
> 
> 2a. The input phase offset levels off
> 
> or
> 
> 2b. Some calibrated amount of time has passed.
> 
> 
> -- 
> Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
> p...@freebsd.org         | TCP/IP since RFC 956
> FreeBSD committer       | BSD since 4.3-tahoe    
> Never attribute to malice what can adequately be explained by incompetence.
_______________________________________________
time-nuts mailing list -- time-nuts@febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.

Reply via email to