Hi Magnus –
To resolve the cliff hangers: 😊
There really wasn’t a way to “fix” the top cover effect. Looking back
now as a career engineer, rather than from my inexperienced view from
1986, what I realize is that in any precise application, there will
always be something holding you a step away from perfection. You
can’t tune this out in every individual product. One challenge with
the 5061B is that it was a **primary** standard, meaning that it didn’t
need calibration, by definition. If you spent a bunch of effort
calibrating a unit, is it still a primary standard? With careful
enough measurements, you could detect that all were not identical.
(We would do this by comparing the phase drift of the 10MHz signal
against our house standard. Over a few hours, an instrument would drift
a few nano-seconds. That is, the unit under test was different than
the house standard.
But it then begs the question, which one is right? Time keeping labs,
like the one with NBS in Colorado, or the west coast standard in our
building in Santa Clara CA, or the one in Geneva, would determine the
“best” time by averaging a bunch of primary standards. The unit that
was closest to the average would be declared the house standard. (At
least that is how I remember it worked.) Essentially establishing
truth by taking a vote. (As many on this list know so well.)
For a specific instrument shipping to an isolated customer, where they
didn’t have another standard to compare against, you had to take your
cesium standard as perfect. It was the primary standard after all. As
the instrument would get handled or power cycled, it could shift a touch
each time. And God forbid if you took the top cover off for some
reason! All of Felix’s fine tuning, screw by screw, would be lost.
(I don’t think he actually hand tuned each screw on every product. More
likely, he pointed out that if the screws were removed or tightened
differently, the frequency offset would change. Not an ideal behavior. )
From a practical “factory” approach to the 5061 products, we guaranteed
that every cesium standard was within specs. We couldn’t guarantee
that they would be exactly centered (if we could, we would tighten up
the specs), or would always remain centered (or even the same) if you
messed around it with it. Just comfortably within specs.
So, with no disrespect to Felix and his efforts to make things better,
at some point, work to perfect each unit becomes silly. You could fool
yourself into thinking it was better, but then only if the instrument
was never touched, moved, power cycled, etc. And even then, would it
still be exactly the same months later?
Fortunately, the managers and senior engineers had a realistic
perspective on how perfect was achievable. Felix was good in that he
kept us from being sloppy, and sometimes would find real things that we
screwed up. But torqueing screws carefully (and uniquely for each
product), or twisting cables left vs. right was a step too far. To
really make the cesium standards better, they needed a better design.
And that was what the 5071A was all about. But that is a story for
another day.
Hugh Rice
Date: Fri, 21 Dec 2018 08:22:05 +0100
From: Magnus Danielson <[email protected]
<mailto:[email protected]>>
To: [email protected] <mailto:[email protected]>
Cc: [email protected] <mailto:[email protected]>
Subject: Re: [time-nuts] HP Cesium Standards in the International
Atomic Time Scale, the legend of Felix Lazarus, and the "top cover
effect".
Message-ID: <[email protected]
<mailto:[email protected]>>
Content-Type: text/plain; charset=windows-1252
Dear Hugh,
I really enjoyed reading this! You have several cliff-hangers in there:
Did you (HP) fix/reduce the top cover issue? Did you alter the setup to
meet tighter specs? Did you fix the oven controller cable offset?
What else war-stories do you got?
It is by war-stories one shares knowledge, lessons learned is not
without its background and at least you have a great story.
Cheers,
Magnus
On 12/20/18 12:36 AM, Rice, Hugh (IPH Writing Systems) wrote:
Hello Time Nuts,
I found this HP Application note in my archives, and attached a
scanned copy:
Application Note 52-4. Contribution of HP clocks to the BIH's
International Atomic Time Scale (IATS).
I also found a couple of archives for HP application notes for anyone
who may be interested:
http://hparchive.com/appnotes
https://www.keysight.com/main/editorial.jspx?cc=US&lc=eng&ckey=1127547&id=1127547&cmpid=zzfindclassic-app-notes
It is an interesting snapshot at the method of keeping the official
IATS time, and how HP Cesium standards are a major part of it, published
in 1986.
The author, Felix Lazarus, was a legendary Field Application Engineer
(or something like that) for HP in Europe, based in Geneva Switzerland.
He was obsessively fussy, and insisted that any Cesium Standard shipped
to key customers in Europe were first shipped to him, so he could verify
acceptable performance before the customer received the instrument.
He would fire up the product, re-tune and re-align all the settings,
and then compare it to his house standard. If it wasn't up to his
exacting standards, he would keep tuning and testing until it was
acceptable - to him. He was looking for performance several times better
than our published specifications, which were 5 x 10e-12. He wasn't
satisfied until is was less than 2, or something like that. It drove us
factory guys crazy. He was a well-respected figure in the time keeping
world, and would bash us for shipping product that were not beating the
specs by enough margin.
I think he is the one that discovered the "top cover effect". If you
removed the top sheet metal cover from the instrument, the offset would
shift by a part in 10 to the 12th or so. If you put the cover on, and
changed how tight the screws were tightened, it would shift differently.
I recall he wanted us to fix this.
I was the "Production Engineer" on the Cesium standards, a young BSEE
college graduate. I barely knew how a basic op-amp amplifier worked, and
was completely overwhelmed by the complexity of the Cesium Standards.
"Go fix the problem on the most accurate commercial atomic standard for
sale in the world, where if you change how tight a screw is, the
performance shifts a touch." It is safe to say that I didn't make this
my highest priority. There were theories that the root cause was subtle
changes to the ground loops with a change like this. The whole product
used all the sheet metal as a common ground, meaning that the ground
return paths were not designed at all, just left to chance.
A related issue that I didn't work on was the "oven controller cable
offset." There was a big multi wire cable o the cesium oven heater
controller, and if you twisted it left vs. right before plugging it in,
the offset of the standard would change.
Working on the 5061B destroyed my confidence in my engineering
abilities. I didn't think I could solve "real" engineering problems,
because of issues like this. After working on the 5061B product for
several years, I applied for a job as an engineering manager over the
frequency counter production product line. During the interviews, my low
technical self-confidence came through, and the R&D management partners
to this position were worried I couldn't provide technical leadership to
the other engineers. So, in true HP fashion, my they sent me through the
full scale HP R&D engineering interview -about a half dozen deep 1:1
technical interviews with EE experts in the lab. Turns out that I wasn't
a dunce after all, just scarred from my experience working on the cesium
standards. I got the job.
I have a handful of other stories like this from my days inside HP
frequency and time division. Let me know if you want to hear more. Maybe
Rick Karlquist will tell some stories of developing the 5071!
Hugh Rice
