[time-nuts] PLL Tester performance

[Steve Rooke]

On 28 May 2010 08:54, WarrenS <warrensjmail-one at yahoo.com> wrote:
> Nuts,
> I have been unable to explain to some of the upside down experts out there, in a way that they can understand, why the PLL Tester method works good enough for many.
> Maybe now that they can see that the simple PLL tester gives nearly the exact same answers as the TSC 5120A over the whole range of taus,

Warren, how can you expect a $10 tester to perform as well as a
hi-falutin $$$$ like a TSC 5120A. I mean it stands to reason that this
can't be the case so I can see why the experts on this list have
poo-pooed it! I mean, are you really serious!!!

> They will be able to set aside their prejudices long enough to see why it works as good as it does when using over sampling,
> and why I say "It can work as good or better than Anything else out there".  (with the performance limited by it's reference oscillator)

But surely the performance of the reference oscillator doesn't come
into it when you have such quality methods like DMTD out there, or
does it...

"One must also realize that any frequency measurement involves two
oscillators. In some instances, one oscillator is in the counter. It
is impossible to purely measure only one oscillator. In some instances
one oscillator may be enough better than the other that the
fluctuations measured may be considered essentially those of the
latter." - NIST

Agreed the instabilities in the transfer-oscillator cancel out but
what about the actual ref osc.

> When everything is set up correctly,
> IF you feed an ADEV program like PLOTTER an over sampled raw ADC frequency data set,
> The program will calculate the correct ADEV values from about 1/10 the raw data rate out to about 1/3 of the length of the data run.

So your saying that this method allows you to make, say, 10 samples
over, say, a 1Hz period to better average the result. Well, that
sounds a whole lot better than just getting a single sample using
time-stamping of the measured wave period.

> setup correctly means:
> PLL_Bandwidth > ADC_sample_rate > ADC_LP_Filter > Tau0.
> Using a factor of 3 to 5 for each ">" works OK, (more is generally better).
>
> an extreme example: using a factor of 10 for every  ">"
> PLL_Bandwidth = 10 KHz
> ADC_sample_Rate = 1 KHz
> ADC_LPF = 100 Hz
> Valid Tau data = 10 Hz and below (with 100 Hz BW noise)
>
> If one wants to reduce the long 1KHz rate frequency data log that the tester produces in this example to say a more normal 0.1 sec tau0 data set, then down converting the longer freq data set using a simple averaging program that does (sum of n_samples / n)  does not change any of the remaining ADEV answers and reduces the length of the data set by n.

Agreed, it just looks like the final samples will be averaged and
better reflect the true state of things instead of suffering errors
caused by instability in the sampling system as this averaging should
improve certain types of stability in the measurement system by a
factor of n.

> PS) If one wants to reference some obscure write up that they think says why this should not work, if they will also include an example of a data set that it does not work on, I'll try to show what they did wrong.

Well, I for one would like to see any papers indicating why this would
not work as it seems like a system that is simple enough to be
implemented by a joe-average time-nut, even if it is not up to the
high standards of the experts on this list. I'd also like to see the
same DUT tested using this tight-pll and acceptable-experts method,
given the same reference oscillator. Of course, one test will not
completely prove this so it should be done with a number of sources. I
think that what has to be understood is what exactly defines a DUT as
there has been some hypothetical suggestions of scenarios that it
would supposedly not work but are they at all likely to be something a
time-nut would use. This is really an open-ended question as we look
at xtals, whether they are disciplined by GPS, Rb, Cs...

Steve

> ws
>
> _______________________________________________
> time-nuts mailing list -- time-nuts at febo.com
> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> and follow the instructions there.
>



-- 
Steve Rooke - ZL3TUV & G8KVD
A man with one clock knows what time it is;
A man with two clocks is never quite sure.