[ntp:questions] demonstrate traceability to UTC

jlevine jlevine at boulder.nist.gov
Mon Mar 16 14:58:49 UTC 2009


Hello,

> I'd be very happy to see this discussed here, and I'm happy to work on
> getting this information onto appropriate pages at support.ntp.org .

   As used by the standards community, "traceability" means that there
is an unbroken chain of measurements from the end-user application
back to a national measurement institute or timing laboratory. In the
US,
this means either NIST or the US Naval Observatory. Each link in the
chain must be evaluated and characterized with an uncertainty.
   In the case of timing, this evaluation must included the offset and
jitter in the application that receives a time stamp from a system
clock and applies it to some event. Any time jitter in the detection
of the
event must also be included. This evaluation must generally be
ongoing,
since the characteristics of the links may change with time.
   The overall uncertainty in the process is typically computed as the
square root of the sum of the squares of the individual uncertainties.
The consequence of this method is that the largest uncertainty tends
to dominate the result, and improving links whose uncertainties are
already small tends to have only a small effect on the result.
   "Technical" traceability is a weaker concept. It generally means
that a user has followed generally accepted practices and is using
a method that has been tested in other configurations. However, the
actual system that links the final application to the national
metrology
insittute or timing laboratory has not been evaluated. The implicit
assumption is that the user has implemented the method correctly,
that all of the links are functioning properly, and that the system is
typical of others using the same method.
   "Legal" traceability is a stronger concept. It means that the
user has implemented traceability and in addition has added
reporting and monitoring functions so that it is possible to
prove traceability in a legal adversary procedure. In general,
the user will have to maintain log files to prove traceability
at some point in the past.
    It is also possible to establish traceability using a
disinterested
third party to monitor the system. For example, NIST has a number
of agreements to monitor remote systems and provide periodic
reports. These agreements are based on the individual requirements
of the user, and the cost depends on the details of the procedure.
   The question of how much of this any user should implement must
be decided based on the user's requirements. In particular, a user
who needs "legal" traceability should get professional legal advice.
   I am happy to answer more specific questions either through this
forum
or by direct e-mail. However, it may be difficult to provde
quantitative answers
about a configuration without a more detailed analysis. If you need
this sort
of detailed study, you will probably have to pay for it.

Judah Levine
Time and Frequency Division
NIST Boulder





More information about the questions mailing list