[time-nuts] FMT on October 13

Tom Van Baak tvb at LeapSecond.com
Mon Sep 24 20:57:22 EDT 2007


>> Plans are to transmit two 10 minute test periods, and a third if the
>> transmitters aren't melting by that point. 
> 
>> Our goal is to transmit a signal known in frequency to parts in 10e-12
>> (i.e., less than 0.0001 Hz error at 10 MHz) and stable to a similar
>> level during the course of the transmission.  Frequencies will be
>> measured at the transmitter site with a system capable of microHertz
>> resolution referenced to a GPS disciplined oscillator, and will also
>> be monitored by another station in groundwave range that can measure
>> the frequencies with similar accuracy. 
> 
> Suppose I have a pile of good lab gear, and it gets N seconds of signal.
> 
> How accurately can it measure the frequency?

Hi Hal,

If you have a low noise CW signal, a cheap, legacy 1 ns
resolution counter will give you 9 digits of resolution per
second. So to measure to parts in 10^12 requires gate
times on the order of a thousand seconds.

A fancier, modern counter like a HP 53132A is almost ten
times better than that so 100 s gate times are all you need
for 12 digits. Further, if it's an oddball frequency (i.e., not
a nice multiple or fraction of 10 MHz) even 10 second gate
times are sufficient with this counter (it does clever CW
oversampling inside).

For extreme counters like HP 5370 or SR 620 with resolution
well under 50 ps you can measure any frequency to 12 digits
in a matter of tens of seconds.

The main problems at this level are often that neither your
frequency reference nor the frequency you are measuring
are stable to parts in 10^12th. So the measurements you
get will contain the sum of noise in both sources and the
counter itself. And this noise is often well above parts in
10^12th. It takes time, statistics, or other tests to determine
the noise contribution of each.

I would think this is especially true for non-local frequencies,
such as one received over the air. I'll leave it to you FMT
guys to comment on the magnitude of degradation due to
transmission and reception noise.

While were at it, in the case mentioned above I'm a curious
about their FMT frequency standard -- if it's really accurate
to parts in 10^12, as they imply, over 10 minutes. I could
believe this if it were an Rb or Cs-based GPSDO.

Usually the accuracy of GPS disciplined oscillators are spec'd
for averaging times over a day. And at one day, parts in 10^12
is very easy (many are down in the low 13's or 14's). But over
a short span like 10 minutes most quartz-based GPSDO wander
in frequency by many parts in 10^11. See, for example, these
two nice quartz GPSDO over 10 minutes and note the scale is
1e-11 per *division*; which is almost 1e-10 full-scale.

http://www.leapsecond.com/pages/fury/#6

/tvb




More information about the time-nuts mailing list