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

[Magnus Danielson]

On 06/15/2010 03:28 PM, WarrenS wrote:
> Bruce posted
>
>> But Adler's equation indicates that an oscillator is much more
>> susceptible to injection effects when the injected signal frequency is
>> very close to the oscillator frequency.
> No argument,
> BUT
> The thing that you (and maybe Adler?) are missing is that effect goes
> away when the two frequencies ARE exactly the same.
> I'm not talking close, I'm talking the exact same freq with phase held
> in quadrature within single digit femtoseconds.
> BIG difference, Once that is understood, then that sort of answers your
> other comments.

Actually, as we have this in a PLL, it doesn't work out quite like that. 
Wolaver shows how the PLL is modified with an additional proportional 
path (his analysis is on active PI loop, but the additional proportional 
path also applies to active lag loops).

For the active PI loop, the injection locking will modulate the PLL 
bandwidth. I would need to analyse the details for an active lag filter, 
but since the gain increases in the loop, so will the loop frequency. 
The implication for the EFC is that not all the gain goes through the 
amplifier, so the EFC deviations will become less. However, a tight PLL 
has high gain for starters, so unless there is very high injection gain 
the effect will be very modest. If the gain is calibrated during closed 
loop conditions and the injection is relatively stable, then the effect 
is essentially cancelled anyway. I would assume that the gain-change of 
typical injection is a small fraction of loop gain, so then it is a 
small effect that should not affect the results to much. It may be good 
to measure the injection gain and compare it to the loop gain.

For a loose PLL, as being used for phase noise measurements, injection 
locking is a much bigger concern.

But, the effect is there even when the frequency is the same. It is just 
that the application in tight PLL makes it very small.

>> For each and every oscillator pair someone may try?????
> Can't say for sure, I've only tried the ones I've tried, but even the
> ones that are highly susceptible were OK.
>
>
>> At best you've only shown this to be true for the particular
>> oscillator pair being compared.
> Yep, maybe I'm just real lucking again and it only applies to all the
> ones I've tested.
>
>
>> That's descending into the murky realms of pseudoscience.
> OR as I see it, it is using just a little common sense.
> When is the last time you heard of a problem with an oscillator
> injection locking to it's self?

I think these can be answered by much better arguments than 
pseudoscience or pure luck. The loop gain dominates over the injection 
gain and thus makes it a minor effect. and that by applying analysis.

>> Not only must the effect of injection locking be insignificant for the
>> reference, it has to be insignificant for the test oscillator as well.
> NO argument, If you are testing a DUT what this does effect, then one
> should buffer it or take other precaution.
> If you want to make sure the tester is not effecting the osc there is
> another choice besides the Buffer.
> DON'T connected it. The tester will work at better than -60 dB signal
> levels and if one just gets a couple of small wires close that is enough
> signal coupling that one can made 1 sec and slower tau readings.
> OR you do both the buffer and antennas you can test the OSC from across
> the room.
> Which side of the list do you think that should go on, advantages or
> disadvantages?

Neither. It's a characteristic, it needs to be analyzed. If the DUT is 
very sensitive, then additional care may be taken or maybe it just isn't 
a very good solution.

>> If injection locking is an issue the efc gain with the loop
>> open will differ from the efc gain with the loop closed.
>> It will change the loop parameters in particular the efc gain.
>> Its just a matter of how much it affects the efc gain.
> NO argument, The PLL loop is never opened. THAT will screw up everything
> and cause the injection, delta gain, etc.
>
> Sounds like it is time for someone to find or write another one of the
> fancy math papers that covers this case.

Page 102-104 in Dan Wolavers book covers this with sufficiently clarity 
that I could answer this. I have been trying to hint to this before. 
Anyway, page 103 is a very instructive conversion from different 
representations of a PI-loop with injection into a simpler form. The 
math follows up and remaining analysis becomes very easy to do.

Cheers,
Magnus