Dec 2008
I purchased a Used Efratom Rubidium Oscillator. It was manafactured in the 35th week of 1991, but I think is was refurbished in 11th week 1996
which still makes it pretty old.
The part number is 100323-001. It looks physically similar to an Efratom FRS rubidium oscillator, and as I have a manula for that, I'll presume its
roughly the same.
An FRS-C has bord parts with numbers 100339-xxx
The unit runs off 24V. requires about 1.5A at startup and 0.4A once its reached operating temperature. From the FRS specs, the stability is:
< 5E-11/Month, +/-5E-10/Year, 1E-10/Day
After 7 day continuous operation, one day stability is 4E-11
At 1PPS 1E-10 = 0.1nS phase change per pulse.
01/12/2008
I have put the oscillator in a 2U rack case, along with a 16C84 PPS divider with software from
ppsdiv.asm
Note , I had to add pseudo directives to the assembler code:
#include__CONFIG _CP_OFF & _WDT_OFF & _PWRTE_ON & _XT_OSC
As by default the microchip complier tries to use the RC oscillator (which adds a divide by 16 to the clock) and enables the watchdog timer (which
resets the code every few 100s of milliseconds).
The power supply is a linear regulated 3A 24V supply. I have a diode 'or' in the power supply output so I can power everything off a DC supply is I
need to remove 240V power for any reason. I don't know howm much use that will be , but it only costs 2 diodes and some binding posts.
The PPS output goes to the NTP box hazchem.smoke.com.au where I can compare the offset to the GPS input. Another split of the 1PPS will go to my HP
5315A frequency meter.
I am now monitoring the phase shift of the Rubidum oscillator as compared to GPS. I seem to have about +2uS over 28 hours, but its early days on the oscillator settling down. Its only probably run <72 hours total since I got it. I'll wait at least a week before trying to tweak...
Oscilator start time is ~ 04:30 01/12/2008 UTC
03/12/2008
From specs, frequency trim range = +/-1E-9
That means a total range of 2E-9, if the trimmer (R14) is a 25 turn pot then thats 8E-11 per turn, 1E-11 per 1/8th turn.
04/12/2008 18:18 EDST, 07:18 UTC
Adjusted trimmer 1/2 turn anti-clockwise
Best-fit was 5.39556e-11 * x
Adjusting PPS offset on the fly :
$ ntpdc ntpdc> fudge 127.127.22.0 time1 .665232 Keyid: xxxxxx MD5 Password: yyyyyy done! ntpdc> q $
to read current fudge value:
ntpdc> clockstat 127.127.22.0 clock address: 127.127.22.0 clock type: PPS Clock Discipline (22) last event: 0 current status: 0 number of polls: 112 no response to poll: 0 bad format responses: 0 bad data responses: 0 running time: 3308 fudge time 1: 0.665232 fudge time 2: 0.000000 stratum: 16 reference ID: ATOM fudge flags: 0x0
06/12/2008 12:35 EDST, 01:35 UTC
Best fit = 2.01146e-11 *x + 0.382428
Adjusted trimmer 1/4 turn anti clockwise.
07/12/2008 15:34 EDST, 04:34 UTC
Best fit = 1.06206e-11 *x + 0.394092
Adjusted trimmer 1/4 turn anti clockwise.
12/12/2008 23:05 EDST
At about 0200UTC on 10/12 the oscillator changed from < 1E-12 to about -20 E-12.
I had been previously messing around with the GPS in the same box, so this could have contributed
to this somehow.
14/12/2008 21:45 EDST
I have added internal temperature and Rubidium module Xtal osc VCO control voltage monitoring using the picaxe-08m recovered from the old XTAl osc. I am plotting this on the offset chart which is getting a bit busy
16/12/2008 22:00 EDST
This tables shows how much time difference a specific offset will build up over different periods
| Offset | sec | minute | hour | day | week | 28days | 365days |
| 1E-9 | 1.0nS | 60nS | 3.6uS | 86.4uS | 604.8uS | 2.4192mS | 31.536mS |
| 1E-10 | 0.1nS | 6.0nS | 360nS | 8.64uS | 60.48uS | 241.92uS | 3.1536mS |
| 1E-11 | 0.01nS | 0.6nS | 36nS | .864uS | 6.048uS | 24.192uS | 315.36uS |
| 1E-12 | 0.001nS | 0.06nS | 3.6nS | 86.4ns | .6048uS | 2.4192uS | 31.536uS |
| 1E-13 | 0.0001nS | 0.006nS | 0.36nS | 8.64nS | 60.48nS | 241.92nS | 3.1536uS |
