[time-nuts] Brooks Shera's GPS locking circuit
Brian Kirby
kirbybq at bellsouth.net
Thu Dec 14 02:43:14 EST 2006
I would like to share my experience using the Brooks Shera controller
with a rubidium oscillator. 3 Years ago, I decided to discipline a
Efratom FRS-C rubidium. For a 1PPS timing reference, I am using a
Motorola M12+ GPS receiver.
Starting with the rubidium, which was new old stock on EBay and part of
an AT&T timing sub-system, I burned it in for about 3 months to
stabilize it and to find its center control voltage point, with respect
to UTC. When centered on UTC (we'll it was somewhere like -8.8x10-13),
I found that dialing the EFC up one volt and down one volt (reference
the UTC setting) gave a plus and minus change of 5x10-10. UTC in my
case was around 2.1 volts on the EFC line. So I made the following
assumptions on EFC: 5x10-10 for a 1 volt change, 5x10-11 for a 100mv
change, 5x10-12 for a 10mv change, 5x10-13 for a 1mv change, 5x10-14 for
a 100uV change, etc.
Using the original Shera controller (because of its voltage sensitivity
design) , I knew that I would need an amplifier, an offset voltage
generator, and a summing amplifier and a buffer.
I built an variable voltage high stability reference, designed to give
me a voltage from 1.5 to 2.5 volts. I built a variable gain amplifier
that gave voltage gain from x9 to x20. This unit was constructed using
the LM299 voltage reference and OP07 op amps.
After some early data, I found the 30 second update rate for the DAC
output in the controller, was too short for the rubidium. I contacted
Brook's and we worked on some details, and he was able to change the
gain of the controller to a voltage sensitivity of 1x10-9. We also
programmed the PIC for a 120 second update rate for the DAC output, and
to average 120 input samples.
I was able to remove the voltage gain amplifiers from the circuits. The
circuit was further refined to have a fixed voltage referenced of
2.000x... volts and a single summing amplifier (one input for the Shera
controller, the other input for the 2 volt reference. I used a LM399
voltage reference and LTC1002 op amps for the interface circuitry.
These circuits use 0.1 percent resistors and a lot of attention was paid
to, stability, low thermal EMF/Kelvin type problems, grounding and
induced noise, etc. It was quite a learning curve.
I found early during the test, that the power supply had to be very
stiff and quiet. My final units start with +/- 24 volt regulators, then
some noise filters, then some precision +/- 12 volt regulators (using
precision references), and they feed Shera's board which uses some
common LM2940/2990 regulators. My notes show the voltage drift was
maintained to around 1 mV.
I also learned the limitations of my test equipment, trying to test
these precision oscillators. For references, I have a FRK-L that is on
UPS power and its in a temperature controlled environment, I also have
two HP Z3801A GPSDOs, 1 Motorola M12+ GPS receiver, and 4 Motorola
Oncore VP GPS receivers. I use a HP53131A time interval counter, with
TAC32 Plus software, as the measurement system. What I found, is the
53131A noise floor dominates the measurements below 100 seconds Allan
deviations when trying to measure the precision oscillators. As this is
a serious hobby and I have financial obligations, I cannot afford a dual
mixer test set and a real primary frequency standard. So I have to
test with what I can work with, and realize that I cannot achieve
certain gains. So, if you plan to become a "time nut", baseline your
time interval system, and make sure you can work in and report real
world results (within your realized limitations).
And I also had fun baselining all the GPS receivers, which are above the
noise baseline of the 53131A. You may find one that stands out and I
found some minor quirks between same units.
Because I did not have a dual mixer test set and I realized that I was
needing instrumentation that would handle below 100 second variances, I
decided on using a simple phase detector, by using balanced mixers, lo
pass filters and a strip chart recorder. As a visual guide, one channel
of the recorder was feed from the disciplined rubidium and the other
channel was feed from another Z3801A and both were referenced to another
Z3801A. So I could see perturbatations relative to each other. I found
the 120 second PIC seemed to have a temperature problem, and the
disciplined rubidium was much more noisy compared to an undisciplined
rubidium and the Z3801s in the short term. At that time, I had no way
to digitally record any data from the phase measurement system. I now
have a Fluke 45 multimeter, that has RS232 outputs and I can record the
phase detector voltages and convert the voltage phase data back to
meaningful results. Characterizing the Fluke 45 is also another story....
I decided to change to a 60 second controller - and at this point I
really started keeping some real detailed data and records. I ran
variance test for weeks at a time using one second data and started
running variances curves above 100 seconds to 86400 seconds. Some of
these test took 12 weeks to get meaningful data on the "daily" variance
- and I have to wonder about little introduced biases, such as the
weather. Then we learn that our 1 PPS reference dividers also have
temperature drifts.... Did I tell you the basement has it own air
conditioner and heating system, all this equipment is running on AC
voltage regulators and UPS systems ? And with the recent energy cost, I
really need to shut this project down....
OK, here is some TIC data from the project, using 128, 1024, 8192, and
65536 second variances: And the baseline noise floor. All are modified
Allan variances.
Noise floor 3x10-12, 4x10-13, 5x10-14, 7x10-15
FRK-L rubidium vs M12 GPS receiver 1x10-10, 1x10-11, 2x10-12, 1x10-12
FRK-L vs FRS-C (undisciplined) 3x10-12, 6x10-13, 7x10-13, 6x10-13
FRK-L vs FRS-C disciplined 4x10-12, 7x10-13, 9x10-13, 9x10-13
FRK-L vs Z3801(A) 4x10-12, 3x10-12, 1x10-12, 1x10-12
FRK-L vs Z3801(B) 4x10-12, 4x10-12, 1x10-12, 2x10-12
Noise floor 3x10-12, 4x10-13, 5x10-14, 7x10-15
Z3801(A) vs M12 1x10-10, 1x10-11, 2x10-12, 2x10-13
Z3801(A) vs FRK-L 4x10-12, 3x10-12, 1x10-12,3x10-12
Z3801(A) vs FRS-C 3x10-12, 2x10-12, 1x10-12, 2x10-12
Z3801(A) vs FRS disciplined 4x10-12, 3x10-12, 9x10-13, 1x10-13
Z3801(A) vs Z3801(B) 4x10-12, 6x10-12, 2x10-123x10-13
Noise floor 3x10-12, 4x10-13, 5x10-14, 7x10-15
FRS-disciplined vs M12 1x10-10, 1x10-11, 2x10-12, 2x10-13
FRS-disciplined vs FRK-L 4x10-12, 7x10-13, 1x10-12, 1x10-12
FRS-disciplined vs Z3801(A) 4x10-12, 4x10-12, 2x10-12, 3x10-13
FRS-disciplined vs Z3801(B) 4x10-12, 2x10-12, 1x10-12, 2x10-13
It appears my limitation is the references, mainly the Z3801s.
I am in the process of recording 1, 10, and 100 second phase data with
the balanced mixer and voltmeter system. The 1 second data is
completed, but the 10 second and 100 second test are in progress.
FRS-disciplined VS Z3801(A) is 1.6x10-10 for 1 second.
Its been fun, but I have also learned to be patient, and I consult with
several friends who use to be in the test equipment and instrumentation
calibration business. Let me point out, that a lot of us time nuts are
"amateurs" and in my case, I cobbled up what I could afford and find. I
was lucky that I got a head start in precision timing as I got to use
cesium beam clocks when I was a younger man in the Air Force - and at
that time, getting real knowledge seemed to be limited - it was books,
reports and who you knew - the internet has changed our world, as we
knew it ! My later experiences came from the telecommunications
industry, which is very dependent on precision timing. I use to head up
a chronic problem team on high speed data circuits and one of the first
things you check is the timing configurations of the circuits. I later
worked with NASA at a mission control facility and again, precision
timing for the instrumentation and communications.
I have somewhat retired and because of the 401K rapes and corporate
failures, I have went back to work and I maintain radio systems - about
the only timing I run into nowadays is a GPSDO for base station
references. I sure miss having the precision gear to test with.
It boils down too now having precision frequency references that I
thought I would never be able to afford in my lifetime. I attribute it
to the availability of the GPS system, and the constant improvement and
reliability of electronics. What do I use the systems for ? - I drive
several frequency synthesizers and frequency counters on the bench. I
believe that I can warrant my frequency generation and measurements to
reflect the x10-10 level minimum, right now, instantly, and x10-12 level
per 24 hours. And this is reference to UTC for real. Not too long
ago, you would be lucky, if your counters time base could hold x10-9
performance over a year and you had everything slaved to it. UTC
reference - well, when it was sent off and calibrated....
I look forward to our discussions on the Time Nuts list server - we have
got a great crowd here and its small, tight and right.
Brian - N4FMN
>
More information about the time-nuts
mailing list