Hi

They very much did not “just go away”. About the only way to get data is to measure a bunch of parts.
Since they typically increase in “time between events”, testing brand new parts likely would give you
the best data.

The main “issue” in terms of a lack of papers seems to be that the tie between cause and effect still
is in the “maybe” category. Over the years, the magnitude of the jumps has dropped. Folks are fixing
something. Most likely they are fixing a number of things that each can create a jump. Tying each change
to this or that maybe 5 or 10%  improvement …. good luck.

Bob

Some time-nuts readers may remember me from 25+ years ago circa 1997
with respect to the E1938A OCXO which used crystals that were basically
the same as the 10811 crystals except for (1) a low profile package and
(2) 10 MHz occurred at series resonance vs parallel resonance.  A small
group of technologists led by Jack Kusters had gradually improved HP
crystal processes over decades of time; by then the 10811 was 20 years
old.  Most of the work focused on aging, rather than jumps.  The method
used to "improve" aging consisted of burning in crystals on an "ager"
system and shipping oscillators (containing crystals) when they achieved
the aging spec they needed.  Certain HP divisions needed tighter specs.
(There were also phase noise specs to worry about).  They didn't
actually materially improve the average aging, but rather the same
division sold frequency counters that could be used to get rid of the
"junk" from the production line.  Jack told me that any sort of process
correlated aging that was discovered would be studied to find the root
cause and then eliminated.  I remember when I first got involved in
crystal oscillators around 1973, before I worked for HP, I was
introduced to crystals from "Colorado Crystal" that were packaged in
glass packages.  Their frequency aging was mostly or completely in the
positive direction (higher frequency as time passed).  Conventional
metal packages mostly were negative aging.  This was explained to be me
by the fact that glass didn't outgass anything that would land on the
resonator and lower its frequency.  Jack's philosophy was to figure out
how to prevent copper packages from outgassing in the first place, and
he was largely successful.  By the late 1990's, the low hanging fruit
had largely been picked, and there were no more papers to be written.
Also, Jack was busy giving private training classes to numerous crystal
fabricators so that we could establish alternative crystal oscillator
sources.  Once this was accomplished, the captive crystal fab at HP was
closed, more or less coinciding with the HP to Agilent spin off and the
retirement of Jack.  They later spun off the legendary 5071A, which I am
proud to say lives on 30+ years later as the 5071B.  Len Cutler would be
proud too.

I built up a lot of E1938A prototypes, and recorded aging data of many
of them.  While it was true that brand new crystals might age much
faster than 5 parts in 10^10 per day (the spec), they mostly improved
over time so that most got down to the spec.  However, they rarely
continued to get better than that with any consistency.  Crystals might
look promising for while, and then for no reason would change and often
go in the opposite direction for a while.  I also looked for jumps,
hoping to find crystals with less than normal jumps, either in size or
how often they occurred.  However, no "jump free" crystals were ever
observed.  The only advice Jack would offer was to say that both effects
were due to slow (or instantaneous) relaxation of stress in the crystal.
Of course, he had already applied any known remedy to this effect.  The
BVA crystal might have been a remedy, but we never attempted that at HP.
You can now buy an atomic standard for the price of a BVA oscillatorl

Meanwhile, mini Rb and mini Cs gas cell standards were getting better
and cheaper and smaller, so they became the defacto replacement for
crystals where jumps mattered (such as when locked to GPS).  Aging
didn't matter as much when GPS locking was in place due to the "smart
clock" concept that was at least named at HP AFAIK.  The smart clock
didn't really work that well in practice.  Len Cutler tried to construct
an ensemble of ten 10811's and operate them as a smart clock.  That
would have made a great paper at FCS, if only it could be made to work.
That is another discussion...

Hope that at least partially answered your question.  I think I can rule
out your theory that "they" figured out how to eliminate jumps and
simply kept it to themselves.  It's more likely they stopped trying due
to lack of a viable value proposition.

Rick Karlquist
N6RK

On 2025-03-12 09:34, Attila Kinali via time-nuts wrote:

Attila Kinali via time-nuts writes:

GPS disciplining ?

Almost nothing is free-wheeling any more.

--
Poul-Henning Kamp      | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG        | TCP/IP since RFC 956
FreeBSD committer      | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.

Out of curiosity, how much do swept quartz blanks help?

How about ion beam milling or other processes to remove defects from surfaces?

Plasma cleaning surfaces before sealing packages?

From: Bob Camp via time-nuts time-nuts@lists.febo.com
Sent: Wednesday, March 12, 2025 3:21:33 PM
To: Discussion of precise time and frequency measurement time-nuts@lists.febo.com
Cc: Attila Kinali attila@kinali.ch; Bob Camp kb8tq@n1k.org
Subject: [time-nuts] Re: Quartz crystal frequency jumps

Hi

They very much did not “just go away”. About the only way to get data is to measure a bunch of parts.
Since they typically increase in “time between events”, testing brand new parts likely would give you
the best data.

The main “issue” in terms of a lack of papers seems to be that the tie between cause and effect still
is in the “maybe” category. Over the years, the magnitude of the jumps has dropped. Folks are fixing
something. Most likely they are fixing a number of things that each can create a jump. Tying each change
to this or that maybe 5 or 10%  improvement …. good luck.

Bob

time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe send an email to time-nuts-leave@lists.febo.com

Hello,

We will probably regret that one day.

That's why I'm interested (as a mere amateur) by self contained
oscillators (with manual periodical recalibration while the possibility
lasts, and not just via gnss).

Sebastien

On 13/03/2025 08:07, Poul-Henning Kamp via time-nuts wrote:

Ultimately, it's all driven by economics.  GNSS disciplining got so good for most applications, and inexpensive atomic references also fill in the gap. 
A Rb in a box is a few thousand $. The PRS10 is $2k. A CSAC is about the same from digikey. Syrlinks has the MMAC (I don't know what they cost) which appears to be CSAC like

It's hard to justify a big physics research project to dive deeper, unless you've got a customer willing to fund it.  After all, one of the reasons DARPA funded the CSAC development was to avoid having to invest in that physics research project on crystals.  (FWIW, DARPA is quite proud of the CSAC, they've got pictures of it in their lobby wall displays).

I suspect that most development dollars are going towards atomic sources or non-quartz-crystal resonators. 

There is a need for good clocks without GNSS - Timekeeping and time distribution at the Moon, for instance. If you're in low orbit, you're going to lose visibility back to Earth, so you need to have good holdover. Or, at Mars.  

If you want 1 ns timing (to pick a round number), over a 1,000 second gap (~15-20 minutes), you need 1E-12 kinds of adev.  That's a few orders of magnitude better than a CSAC (at least from the datasheet numbers, in reality, they do somewhat better, but not that much better).

CAPSTONE is flying a CSAC connected to a Iris X-band transponder to experiment with this in a cis-lunar environment and it's been flying for about 2 years. One of the claimed benefits of the NRHO orbit is that maintains Earth visibility most of the time (at the cost of having enormous ranges (>100k km) at apolune, which makes it less suitable for time/nav)

On Thu, 13 Mar 2025 10:24:18 +0100, Sebastien F4GRX via time-nuts time-nuts@lists.febo.com wrote:

Hello,

We will probably regret that one day.

That's why I'm interested (as a mere amateur) by self contained
oscillators (with manual periodical recalibration while the possibility
lasts, and not just via gnss).

Sebastien

On 13/03/2025 08:07, Poul-Henning Kamp via time-nuts wrote:

time-nuts mailing list -- time-nuts@lists.febo.com
To unsubscribe send an email to time-nuts-leave@lists.febo.com
 

Strange frequency jumps, I had much fun with it  ......

We manufactured DSL modems which were kicked out of sync with a
temperature change, namely in the exhaust of vacuum cleaners :) . The
test in the climate chamber showed a frequency jump  at a specific
temperature for each crystal.  It was a jump, way beyond the the
specified deltaf/deltaT . We solved it with a smarter PLL and we
substituted the crystal supplier. The root cause was of no further interest

Cheers Detlef

Am 13.03.2025 um 18:05 schrieb Jim Lux via time-nuts:

--
Diese E-Mail wurde von Avast-Antivirussoftware auf Viren geprüft.
www.avast.com

On 3/13/25 08:07, Poul-Henning Kamp via time-nuts wrote:

There is just immense amount of oscillators which will never be steered
or resynthesized (just different frequency adjustment method).

A frequency jump will upset any PLL and depending on the order and
bandwidth, the jump may be within tolerance or not. If not within
tolerance, the lock will be lost and you need to recapture it, or that
happens automatically but you may choose potentially better ways.

Regardless, frequency jumps remains an issue that needs to be handled,
even if this is to ensure that most likely jumps is within PLL tolerance.

Cheers,
Magnus

Hi

Typically that sort of thing would be called a perturbation. They are very common and the reasons behind most of them are pretty well understood.

Bob