[time-nuts] Re: Low Phase Noise70 10 MHz bench signal source sought

[Richard (Rick) Karlquist]

Yes, Jim Johnson (who worked for Len Cutler) worked on this hero
experiment.  IMHO, it was a boondoggle from the get go.  BTW,
this wasn't Jim's fault.  I guess you could blame Len if you
needed to.

1.  You might think that we had the capability to select 10811's
as they came off the line.  Number of problems with this.  I don't
believe they 100% tested for ADEV.

2.  The tester they used for ADEV consisted of a special 10811
that was 500 Hz off frequency driving a dual mixer box (Model 10514?)
which produced a 500 Hz beat note, which then drove an ADEV system 
(model 5490?).  Only a few of the offset
10811's were produced due to the obvious disruption to the production
line because of the offset frequency.  This small group of offset 
10811's were measured in a round robin set of measurements to select the 
best one of that small group.  Note that is was definitely NOT in the 
top 1% (or whatever) of the total production of 10811's.  Anyway, the
winner of the round robin became the "golden" 10811 and the runner up 
became the backup for the golden one.  So the best you would be able to 
do is come up with n 10811's that were somewhere between almost as good 
as the golden one, and much better than the golden one, but they were 
obviously not measurably better.

4.  I tried to propose a measuring scheme that would use the offset 
10811 as the common mode LO for two DUT's, with the offset osc noise 
cancelling out.  Or something like that.  But I was not on the project 
team and so I never got any buy in.

5.  They should have made a proof of concept system with only 2 or 3 
oscillators, but instead they started with 9 (or was it 10?).

6.  Time-nuts reading this must be wondering by now how they isolated 
the individual oscillators to prevent injection locking.  Well, that's
"nontrivial :-)  The HP distribution amplifier products were a joke, so 
they didn't provide any starting point.  I happened to be the most 
knowledgeable person at HP when came to this field, because of my work 
on the 5071A output amplifier patent.  However, that design used a trick 
that couldn't be applied to the oscillator ensemble.  The trick involved 
starting with 80 MHz and dividing down, etc.  For more details, read my 
patent or FCS paper.  The whole reason for the trick was that I 
convinced myself of the extreme difficulty of doing it the conventional way.

5.  They never come up with an algorithm that made sense.  You would 
somehow have to combine a group of 8 oscillators and then compare it to 
the 9th.  They complete a round robin going through the other 8.  Then
you could rank them.  Then you would have to run them through a weighted 
combiner or something.

Anyway, the project was never completed.  I am thinking it would have 
made a great PhD dissertation for someone who could find a thesis 
advisor and who could waste several years of their life on it.
Maybe at Univ. of Colo.

In conclusion, in case you are thinking of doing anything like this ... 
don't :-).

Rick N6RK

On 4/2/2022 9:42 AM, Tom Van Baak wrote:
> Yes, I believe hp did what you propose in their Santa Clara cesium test 
> lab. It was an ensemble of selected 10811 oscillators tightly phase 
> locked to improve short-term stability and phase noise. Rick might know 
> the details.
> 
> During a tour I saw the instrument rack but was unable to peek inside. 
> Performance details were not given and I didn't get to count the 
> oscillators but I think it was on the order of half a dozen. I figured 
> that using 2 or 3 probably isn't worth all the effort. And using ten or 
> more is diminishing returns or maybe beyond requirements.
> 
> You can imagine if you were allowed to select the best of the best 
> 10811's as they came off the manufacturing test line and then combined a 
> bunch of them as they did, the results would be quite impressive.
> 
> 
> /tvb
>