[time-nuts] A man with one used rubidium standard knows what 10MHz is, more than one, not so much...

Chris Albertson albertson.chris at gmail.com
Wed Feb 22 20:31:42 UTC 2012


On Wed, Feb 22, 2012 at 12:15 PM, Don Barr <don.barr at gmail.com> wrote:
> Hi,
>
> (sorry to Tvb for stealing a good subject line)
>
> I'm just getting into clocks and oscillators, so please excuse any
> ignorance.  I like probably many new list watchers saw Dave Jones of
> EEVBlog fame with his used FE-5680A and thought - geez I didn't realize
> those were so cheap, I'd like an "atomic clock" too, and promptly ordered
> three from three different vendors off of ebay.   I figured I'd get one out
> of three that showed up and worked, but I actually got all three, and all
> three appear to operate.  I also saw on leapsecond that many time-nuts use
> the Trimble Thunderbolt as a GPSDO, and as they were cheap, I ordered one
> of those from ebay as well.
>
> Now that I have all of these, and I've cobbled together a power supply to
> power everything I hooked the Thunderbolt and the 5680's to the one piece
> of test equipment I currently have - an oscilloscope.   Maybe out of
> ignorance I decided that the GPSDO would be the "most accurate", so I set
> the trigger of the scope to the channel that the thunderbolt was connected
> to.   I figured the changing relative waveform phase of each unit would
> give me idea of how in line each unit was at a gross level.
>
> To my delight the GPSDO and one 5680 is very close, and the relative phase
> of the GPSDO and that unit remain relatively constant.   Two of the other
> units are "off", and I'm curious as to why.  Each of the "off" units goes
> in and out of phase with the GPSDO (i.e., 360 degrees goes by) in about 20
> seconds, which I think is +/- 0.05Hz,  which I think, is 5x10e-9 at 10MHz
> (correct me if I'm wrong).    (All units left to "warm up" over 12 hours,
> the GPSDO has clear view of the sky and self survey was completed, etc.)
>
> I've tried to do a bit of research on Rubidium aging, which seems to cause
> the oscillator to slow over time (
> http://www.dtic.mil/cgi-bin/GetTRDoc?AD=ADA485423) when exposed to air
> (helium), so I expect the 10 years or so of service would make sense
> assuming about 5*10e-11 a month, over that period of time.  I presume
> that's what the DDS in the unit is there for to allow the units to be
> corrected for aging (
> http://www.ko4bb.com/Manuals/05%29_GPS_Timing/FEI/FE-5680A/FE-5680A-tuning.pdf)
>
>
> So I guess my questions are:
> Does my line of reasoning make sense?
> Should the GPSDO, for lack of a better standard, over the "eyeball averaged
> over 20 seconds" be the best frequency source?

The FE5680 units you bought have a way to make a fine adjustment to
the frequency and then the save the adjustment in eprom memory.   So
if the units you bought track the GPS simply depends on whatever the
previous own stored in the eprom.  You can change it via RS-232
commands.

Why would someone set the eprom to anything other than exactly 10MHz?
I don't know.  Perhaps the unit was last used inside a cell tower in
the desert when it was way-hot outside?  Maybe you have a better
calibration signal than the tech who last set it?  No way to know.

What you need is to build a phase locked loop controller to lock your
FE5680 to your Thunderbolt.   You can use a very long time constant.
A controller like that is on my do-to list.



Chris Albertson
Redondo Beach, California



More information about the time-nuts mailing list