[time-nuts] GPS lock of the FE5680. Current experiment and question
davidwhess at gmail.com
Sat Feb 11 01:54:11 UTC 2012
Flip-flops hardened against metastability are available.
I would try a track and hold before a sample and hold but I wonder how
accurate it would be having to rely on the oscillator waveform. Fast
sample gates are non-trivial. I believe better than either would be a
time to voltage converter which just needs a set/reset flip-flop and
switched current integrator. Dig up the schematic for a Tektronix
2232 oscilloscope for a "simple" example that yields 50pS resolution
using late 1980s technology.
I just recently ventured into the sampling oscilloscope world with a
7S11/S-4/7T11 setup. The S-4 uses a traveling wave gate which avoids
the fast sample gate problem but I would consider that the far side of
esoteric since I have not studied it in detail yet.
On Fri, 10 Feb 2012 17:23:46 -0800 (PST), Robert LaJeunesse
<rlajeunesse at sbcglobal.net> wrote:
>Sampling unsychronized signals with a DFF is problematic, since if setup and
>hold times are not met the output could oscillate and maybe settle to some noise
>driven value. I can't help thinking that if you are sampling the 10MHz signal at
>1Hz the only way to get reasonable resolution is to sample the 10MHz sinewave
>signal's fastest part with tight analog sample & hold. Looking at the result wih
>a slow, low-cost 24-bit A to D chip would give tremendous resolution - if the
>drift was low enough. Use the ADC over a sub-range and add a small micro for
>noise filtering and averaging and one can achieve measurements result in the 10s
>From: David <davidwhess at gmail.com>
>To: Discussion of precise time and frequency measurement <time-nuts at febo.com>
>Sent: Fri, February 10, 2012 8:10:56 PM
>Subject: Re: [time-nuts] GPS lock of the FE5680. Current experiment and question
>All you need for this is the flip-flop. Clock the flip-flop with the
>1 PPS signal and capture whether the oscillator is leading or lagging.
>This requires the 10 MHz oscillator to be within 1 Hz but if you
>divide it down before the comparison, you can extend this range as
>needed to handle wider initial oscillator frequencies and larger
>amounts of PPS jitter.
>The simple GPSDO design in QST a couple years ago did something like
More information about the time-nuts