jimlux at earthlink.net
Mon Aug 27 22:33:48 EDT 2012
On 8/27/12 4:15 PM, Rick Karlquist wrote:
> Several decades ago, the concept of the "smart clock" arose
> at what was then HP. The idea was as discussed here to
> "characterize" past aging, "predict" future aging, and
> then "correct" the aging. The goal wasn't to turn a quartz
> oscillator into an atomic clock replacement, but simply
> to get the oscillator through a 1 hour or so "holdover" time
> during GPS outages. It sort of worked for that very limited
> purpose, but in general, past performance of HP crystals wasn't
> a very good predictor of future results. Crystals would age
> in one direction for a while and possibly slow down as time
> when on, but then then might start aging in the other direction.
> There were also frequency jumps that were substantial and totally
> random. The reason why the HP crystals were unpredictable was
> that all the deterministic processes such as mass preferentially
> depositing on the crystal, so as to make the frequency age
> lower, had been eliminated by years of manufacturing improvements.
> The remaining processes were of the nature of quartz stress
> relaxation that were very random.
> Rick Karlquist
We see similar things in USOs (and other components as well) for
Things that people worried about 60 or more years ago just don't occur
anymore. People used to obsessively try to match diodes in HV strings,
for instance, because the process variability was high enough to make a
difference. These days, you get a reel of diodes and they're all pretty
much the same, and even reel to reel from month to month they don't
change much. That's what all that 6-sigma stuff is all about, after all.
All the low hanging, and even middle hanging, fruit has been picked..
(one big exception.. ICs which are not designed for radiation tolerance
seem to have large variability in radiation tolerance..it's just not
something that's controlled for in the process)
More information about the time-nuts