[time-nuts] Did a member of time-nuts buy this?

Gregory Maxwell gmaxwell at gmail.com
Sat Dec 6 16:45:21 UTC 2014


Bob Camp wrote:
> Unless you are making a GPS receiver from scratch (which you might be), there is a certain “trust factor” that comes into using a GPS for timing. Since you can’t play with the firmware, you trust that the guy who wrote it did a good job.

As compared to internet facing software embedded systems seem to be
unusually fragile, consider this paper on GPS receivers with
adversarial signals:
http://www.andrew.cmu.edu/user/tnighswa/GPS_CCS.pdf

And the trust with using GPS goes beyond the quality of the
construction of the receiver:  You're trusting the the GPS
constellation is working and correct (see the recent GLONASS failure)
and you're trusting that there aren't random jammers going by, you're
trusting that there isn't someone in physical proximity manipulating
the signal intentionally (see the paper above), or even just random
truckers going by with jammers (there have been past threads on
time-nuts) about this.  IIRC the stated US policy with respect to GPS
signal integrity is that it may be intentionally degraded (and can be
degraded in a geographically targeted manner) for e.g.
political/military objectives, so you trust that you won't be the
target or collateral damage of any such degradation or that it won't
be severe enough to effect you.

GPS driven timing works amazing well under most conditions most of the
time and at a very low cost. The trade-off is that you're taking more
fringe risk and greater trust. I sometimes worry that we're building
too much public infrastructure which is depends on a single system (or
on space based timing in general, since Kessler syndrome, while
unlikely, is a risk that exists) now that loran is gone in the US. Of
course, the attractiveness of GPS makes this self-fulfilling: Solid,
long living, CS primary frequency sources would probably be much less
expensive of GPS didn't cover so much of the commercial demand for
them.  There are newer receivers (e.g. ublox m8) that are concurrent
mult-gnss which might help, or maybe not: who knows what the receiver
will do if one system starts emitting crap?  I am not especially
confident that the software in these systems is well baked under
exceptional conditions.

If you're working on things with no availability requirements, no
real-time requirements (e.g. able to go download after-the-fact GPS
reliability and precise ephemeris from NGS), and aren't doing anything
where your timing is likely to be intentionally attacked, say for
test-lab purposes... then these issues may be less of a consideration.

In the context of time-nuts though many people are interested in the
art and science of precise time/frequency for pretty much its own
sake... and the driving need for the lowest phase noise or best adev
at some window might just be because it's possible.  In that light,
the extremes of autonomy, reliability, avoidance of systemic risk, and
surviving attacks are also interesting parameters that I find to be
interesting to explore, and they're ones which perhaps have inadequate
commercial attention on them these days since it seems people are
often (a little too) willing to trust and then point fingers when
things fail.

[Or at least this is an area I personally find interesting ... I wrote
this back in 2011 not so long after I started reading time-nuts:
https://people.xiph.org/~greg/decentralized-time.txt  before I knew
common-view time-transfer was already a thing, and when I knew a
little less nothing than the nothing I know now about time/frequency
standards.]

In terms of the 5061A  at least some of the old surplus units floating
around out there are "non-working" for silly reasons,  e.g. left
sitting for a long time, and they'll actually lock up fine if left
with the ion pump running for a few days, or the OCXO put back on
frequency, or the gain adjusted.... though I wouldn't spend $1k just
to find out. I picked up a 5061B for basically shipping costs a while
back and it was up and running reliably after some minor repairs...
though the beam current is low and it likely doesn't have much life
left in the tube. It's hard to deny how interesting and finely built
these devices are, objects of techno-lust in their own right, even in
surplus-and-maybe-not-reliable and impossibly-expensive-to-refurbish
condition.

As an actual lab tool-- rather than a science project, sadly, I do
have to agree that you're better off with a GPSDO than a surplus CS
unless you happen to get really lucky in the surplus gear lottery. Of
course, none of this is mutually exclusive. It's possible and
reasonable to have both.



More information about the Time-nuts_lists.febo.com mailing list