[time-nuts] Advantages & Disadvantages of the TPLL Method

Charles P. Steinmetz charles_steinmetz at lavabit.com
Thu Jun 17 02:10:31 UTC 2010

Warren wrote:

>Charles Posted:
>>How much EFC is required depends, in part, on the strength of the pulling.
>>There are three varying inputs.
>NOT at ALL what my test have shown so I guess we do NOT agree on this.
>The point you missed, is only the EFC is changing significantly 
>because of the high gain and BW.
>It insures the two Oscillators are held to within  femtoseconds of 
>each other, to at least out to the e-16 at large taus.
>So other things are held constant enough that their effects are kept 
>below any ref Osc effects.

Why must everything be a matter of other people missing something?  I 
understand how oscillators behave with respect to injection 
pulling/locking, and how that might affect the operation and accuracy 
of a system such as you are using.  I myself noted that Magnus had 
suggested the effect may not be significant in such a system, but 
that drawing that conclusion for any particular design would require 
careful experiments and, hopefully, backup by mathematical 
analysis.  How is that missing anything?

>The "carefully constructed experiments", that show it works as 
>advertised have been done, and the most important ones have been posted.

Forgive me if I missed something, but all I saw regarding the 
relative gains of the error loop and the injection loop were (i) that 
you "increased the coupling by a factor of 1000" and (ii) that you 
used a variable attenuator.  If you did carefully designed 
experiments, nothing I saw posted suggested it.

This is a potentially important point because some oscillators one 
might want to test (or use as a reference) may be very much more 
sensitive to injection locking (pulling, actually) than the ones you 
are using.  Therefore, the behavior should be characterized so users 
can determine whether it might affect their results.  "It didn't seem 
to have any effect using the oscillators I had" is not really a very 
useful characterization of the behavior.  [I do see that in a 
subsequent message you asked for suggestions for further tests.]

I'm curious how you determined that the oscillators are being held to 
within femtoseconds of each other.  And, how many femtoseconds?

Best regards,


More information about the time-nuts mailing list