[time-nuts] 9390 GPS RX
ed_palmer at sasktel.net
Tue May 28 21:15:10 EDT 2013
I have some good news. I decided to copy your test by powering up an
FRK that's been sitting in a box for over a year. Once it locked, I
measured the frequency and was surprised to find that it was about .06
Hz (6e-9) low. I added about 890 ohms between the two external
frequency adjustment leads and got an error of about 7e-11. So your
unit might be in much better shape than I thought. Sorry if I misled you
I'm also looking at the Allan Deviation. It's only been running for a
few hours, but one thing that I noticed is that the frequency is
drifting as the unit warms up and works the kinks out. This drift
degrades the AlDev numbers by looking like aging or a high noise floor.
Try using Hadamard Deviation instead of Allan Deviation. Hadamard is
insensitive to aging. So far, my AlDev values aren't much different
than yours but as I run the graph out beyond 800 sec., the numbers do
get a little better. Also, remember that most Rb standards aren't
optimized for low AlDev values since most applications don't need that.
On 5/28/2013 4:10 PM, Ed Palmer wrote:
> The frequency offset of 3e-9 is still very high. Since the 9390 will
> discipline the FRK, the C-field adjustment of the FRK doesn't really
> matter. Before you open up the FRK again, see if the C-field can
> change the frequency. If it can, and if the range is similar to the
> specified range, then the C-field is working and there's no point in
> opening the FRK.
> As for your AlDev numbers, in no particular order.....
> Have you measured the noise floor of your measurement system?
> Check the power supply for hum and noise.
> Is the crystal control voltage stable or is it noisy?
> A weak Rb lamp might give you a noisy signal, even though it's
> locked. Is the lamp voltage noisy?
> Check for a bad cable or some other bad connection. If possible, use
> only double-shielded cables.
> Running down problems like this can drive you crazy. The levels are
> so low that it could be almost anything from the computer you're
> running Timelab on to your neighbor's wireless phone!
> A good OCXO often takes a month or more to stabilize after you pick it
> up in an auction. Rb standards aren't usually that leisurely about
> settling down but I'd just let it run and see what happens. This might
> be a good time to put it back in the 9390 and see what happens. Does
> the 9390 have commands that will show you how well the FRK is performing?
> You also might have a unit that's working properly but is a little
> noisy. The specs don't cover something like this. You're well
> outside the time for 'short term stability' but haven't yet got to the
> 'aging' spec. If you're using the 5370 by itself, you can't measure
> the short term stability at all.
> On 5/28/2013 11:37 AM, Mark C. Stephens wrote:
>> Ed and Bob,
>> The FRK has settled down to 0.03Hz out after 2 days undisturbed (back
>> in its black box).
>> Allan deviation @ 800 seconds is 4.88E-12 which I am not terribly
>> happy about.
More information about the time-nuts