[time-nuts] Frequency reference

Jim Lux james.p.lux at jpl.nasa.gov
Mon Apr 21 14:22:31 UTC 2008


Quoting Attila Kinali <attila at kinali.ch>, on Sun 20 Apr 2008 04:23:38 PM PDT:

> On Sun, 20 Apr 2008 09:25:37 -0700
> Jim Lux <james.p.lux at jpl.nasa.gov> wrote:
>

> Well.. then you have to build multiple frequency sources
> that exhibit different physical behaviour, otherwise
> slight changes in the enviroment that degrate your
> precision will go unnoticed (ie, if all sources have
> the same temperature coefficient then temperature
> changes will affect all of them the same way making
> you unable to measure this effect)

A good point..  One might, for instance compare your OXCO source  
against GPS 1pps ticks.


>
>> or, take it to somewhere that has a higher quality standard and compare
>
> Which is quite difficult if you don't have access to a physics
> lab which you can use for a few days to weeks.

Indeed..

>
>> or, just trust that the performance is inherent in the design, and if
>> it works at all, it's good enough. Typically, if you are building a
>> copy of a known good design, this is a good start.
>
> I'm an engineer, i don't trust anything i cannot measure,
> because i know that errors and mistakes are inherent in any design :-)


True enough, but I suspect you are willing to trust a design to some  
point (perhaps not to it's design precision, but some lesser level),  
as long as the design is such that errors in function are knowable(as  
opposed to just larger measurement uncertainty).

And, you probably trust others to make some of the measurements. For  
instance, you probably trust the mfr of your inch/cm scale to some  
precision (i.e. you don't have a primary length standard sitting in  
your garage.. although on this list, one can't be too sure..).

>
> 				Attila Kinali
>
> --
> The true CS students do not need to know how to program.
> They learn how to abstract the process of programming to
> the point of making programmers obsolete.
> 		-- Jabber in #holo
>






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