I have tweaked mine to my satisfaction now. An offset of -768 gets it as close as I can measure under sub-optimal conditions. See results at bottom.
I am still curious as to whether Rubidium exhibits the same resonance peaks that Cesium does, per Tom Van Baak's graph at http://leapsecond.com/images/cfield.gif . > Hi > > A few generic Rb information points: > > 1) All the telecom Rb’s need an external heatsink. To just mount them on > a PC board, you need a lot of airflow (like fans …). > > 2) Heat sinking improves the lifespan of these devices. Without a heatsink, > a year or two is doing pretty well. With a heatsink that gets the base to 40C > (possibly heatsink + small fan) a decade or more is possible. It sits on a square-foot slab of steel to the corner of which is mounted a Panavise base. > > 3) If your Rb has a PPS output, cal is pretty easy. Just compare the pps > to a < $20 GPS module with a scope. A day or three of tweaking should > get you pretty close. It does, but the generic problem with this technique is that there is an arbitrary constant offset between the two pulses. The first time I did this some years ago, I set an 8-digit counter to show the time difference between the two pulses (~735ms that time), then took a timestamped photo every ten minutes or so, did some math, and found the Rb and the GPSDO were within 2e-10 of each other. You want the delta as stable as possible. There's also the Lissajous method. I never run mine for more than a week, and often less than a day. It's more of a Because I Can sort of thing... > > 4) With very few exceptions telecom Rb’s either have rotten phase noise / > spurs > or they have horrible phase noise / spurs. If you are going to do anything > “fancy” (like microwaves) with one, you will need a cleanup loop. Fortunately for me, I am not.... > > 5) Many outfits produced a wide range of parts, all with the same part number. > What you have with a XXXX model number on it may be *very* different than > what I have with the same … errr ... FE-5680 model number on it. Basically, the two [non-]programmable units mentioned earlier. I have the non- version. It has only the DE9 connector. > > 6) Most Rb’s have some sort of crystal oscillator in them. There often is an > adjustment (re-centering) needed on surplus parts. On tindie, there's also Nick Sayer's GPS disciplining board for the FE-5680A. > > 7) Like OCXO’s Rb’s do have a warmup / retrace process that runs into > days. They also are sensitive to temperature, voltage, pressure, and > humidity. Compared to an OCXO, most of these sensitivities are pretty > small. > > As mentioned in a number of posts, in a lot of ways, an Rb makes a pretty > good basement lab standard. Compared to a GPSDO, there are a lot > fewer things to go wrong. > Other than the lamp life... I've heard of ways to rejuvenate it. Mine spends 99.44% of its time powered off, though. == Here's the output from a Python script I wrote to evaulate oscillator performance (I'm too impatient to wait for ADEVs) using vxi11 to talk to a LAN-connected GPIB controller. And while this counter does do statistics, it doesn't do linear regression. This is from this morning; it was behaving almost as well last night. bash-4.3$ python count-one-adhoc.py 1e7 Given ideal 1e7 HEWLETT-PACKARD,53131A,0,3944 Reference is EXTERNAL at +1.00000E+007 Hz 3.452 Previous gate time Suggested gate time 1.79330192372 1.793 Gate time 1589572346.83 first time 0 bias 10000000.0 expected value 12 readings to take 1 9999999.9998 Hz 1.93545103073 s 1.93545103073 delta s 2 10000000.0 Hz 3.87260103226 s 1.93715000153 delta s 3 9999999.9998 Hz 5.80996990204 s 1.93736886978 delta s 4 10000000.0002 Hz 7.74672698975 s 1.93675708771 delta s 5 10000000.0002 Hz 9.68287706375 s 1.93615007401 delta s 6 9999999.9999 Hz 11.6207020283 s 1.93782496452 delta s 7 10000000.0004 Hz 13.5647189617 s 1.94401693344 delta s 8 10000000.0 Hz 15.5021290779 s 1.9374101162 delta s 9 10000000.0003 Hz 17.4395709038 s 1.93744182587 delta s 10 10000000.0002 Hz 19.3772699833 s 1.93769907951 delta s 11 10000000.0 Hz 21.3132119179 s 1.93594193459 delta s 12 9999999.9997 Hz 23.2543890476 s 1.94117712975 delta s 151.119617939 Sum s 2440.26380663 Sum s^2 120000000.0 Sum y 1.20000000001e+15 Sum y^2 23.2543890476 Sum ds 45.0639478477 Sum ds^2 1511196179.4 Sum ys 10000000.0 Hz mean 0.150755672289 Hz std dev 12 readings in 23.2543890476 seconds, should be 21.516 1.93786575397 delta t mean, 0.144865753969 dead time, 0.144865753969 also, 8.07951778966 % 0.00240610675374 delta s std dev 1.50755672288e-08 stability? 15.0755672 ppb 4.16655598912e-12 accuracy? 4.1665560 ppt second denominator 2.0 0.000215087265483 correlation 3.78864265201e-06 slope 9999999.99999 intercept 10000000.0 Hz hat in the middle 10000000.0001 Hz hat # 17 at about 32.9437178175 s bash-4.3$ -- Jeff Woolsey {{woolsey,jlw}@jlw,first.last@{gmail,jlw}}.com Nature abhors straight antennas, clean lenses, and empty storage. "Delete! Delete! OK!" -Dr. Bronner on disk space management Card-sorting, Joel. -Crow on solitaire _______________________________________________ time-nuts mailing list -- time-nuts@lists.febo.com To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com and follow the instructions there.