[time-nuts] FMT on October 13
magnus at rubidium.dyndns.org
Wed Sep 26 07:15:15 EDT 2007
From: "Didier Juges" <didier at cox.net>
Subject: Re: [time-nuts] FMT on October 13
Date: Wed, 26 Sep 2007 05:57:14 -0500
Message-ID: <006201c8002c$00044520$6501a8c0 at didierhp>
> I just wanted to point out that with reciprocal counters, you can get
> resolution much better than the 1Hz/s you would get with conventional
> frequency counters, even though the actual accuracy of the measurement may
> be way off.
These days conventional counteras are reciprocal counters. It is only the
old-school counters which is not reciprocal. Nothing wrong with old-school,
but a conventional counter of the shelf today is probably a reciprocal jobbie.
> The original question seemed to imply that with a short transmission time,
> you could not guarantee a frequency accuracy of 1e-6 Hz, which you probably
> can't anyhow, but the limit is not the resolution of the instrument or the
> measurement method.
> I do not know how far off calibration my HP 5370s are, but the 20pS
> resolution is at best only usable under some circumstances that I have not
> isolated yet, due to jitter.
> When measuring a 3.5 MHz signal (@1dBm) from my HP 8657B through 1 meter of
> good coax cable (with counter and generator phase locked to the Thunderbolt
> GPSDO) in Frequency mode with a 1s gate time, the resolution is 1e-5Hz, with
> about 1e-3Hz p-p variation. When measuring over 1 period with 10,000 periods
> sample size, the resolution is only 1e-1Hz with a standard deviation of ~400
> Hz (or about 0.1%). Of course, over the air, it will be much worse due to
> noise, let alone propagation, fading and multipath.
When measuring over a longer period you see a different spot on the ADEV/MDEV
curve. Chances are that you are more unstable there for an OCXO. Both linear
and noise products will make things harder. It can be a challenge to separate
the drift rate due to signal path shifts and that of the OCXO.
More information about the time-nuts