HP 5071A Guide

29th Annual Precise Time and Time Interual (PTTI) Meeting
MAINTENANCE OF
HP 5071A PRIMARY
FREQUENCY
STANDARDS
USNO
AT
H. Chadsey (hc @planck.usno.navy.mil
A. Kubik ([e-mail]1)
Time Service Department
U.S. Naval Observatory
Washington, D.C. 20392
Abstract
The US. Naval Observatory (VSNO) has been operating Hewleti-Packard model 507lA
cesium-beamfraquenq standardr for over five years. During that period, there have been
a variety of failures and these devices have shownfrequency and phase changes.
The
HP 5071A model primary frequency standard oflers a very useful troubleshootingtool
by outputting the status of 22 digerenr operating parameters. This paper will present an
explanation of the parameters and show any correlation of them with the time,frequency,
and environmentalchanges. This paper will also
ofet some indicators to predictfuture
device problems.
INTRODUCTION
The Hewlett-Packard (HP) model 5071A Primary Frequency Standard is a quantum
leap forward compared to the HP model 5061B. Among the improvements is the
greater frequency stability through a broader temperature and humidity range. HP
also improved the device's operational parameter monitoring capabilities. No
longer must people enter the room housing the device (disrupting its environment)
and
physically make the measurements on the device. The new standards can
output their parameters via an RS-232 connection to a computer for permanent
filing and analysis.
49

The analysis of these permanent files can range from a snapshot picture of the
parameters to
a time series analysis with plotting. Having the extreme long-term
history
of the parameters has additional benefits, especially if that history extends
throughout the life
of the standard. USNO has 45 standards from the 100, 200,
300, 400, 500, 700, and 1000 production series. Each device has its own
"personality." Some characteristics have been correlated with production series.
This paper will present the general characteristics of the
USNO devices and note
where they may differ.
The parameters will be discussed in the order of their
output.
The more important parameters will be noted and discussed in greater
detail.
THE PARAMETERS
There are 22 parameter values available via the RS-232 connection. These
parameters are: frequency offset, oscillator control percentage,
rf amplitude 1
percentage, rf amplitude 2 percentage, Zeeman frequency, C field current, electron
multiplier voltage, signal gain percentage,tube oven voltage,
tube oven temperature
error, oscillator oven voltage, ion pump current, hot wire ionizer voltage, mass
spectrometer voltage, SAW tuning voltage, DRO tuning voltage, 87 MHz PLL
voltage, UP clock PLL voltage, +12 volt supply voltage, -12 volt supply voltage,
5 volt supply voltage, and internal temperature. USNO collects these data once per
hour. The information appears as:
MJD 48587 21:03:42
CBT ID: 6-temp
Status summary: Operating normally
Power source:
AC
Log status: Empty
Freq
Offset: Oe-15 Osc. control:
RF amplitude 1: 20.2 96 RF amplitude 2:
2emi Freq.
E-multiplier
:
CBT Oven:
Osc. Oven:
HW Ionizer:
SAW Tuning:
87MHz PLL:
+12v supply:
+5v supply:
39949 Hz
1870 V
6,2 V
-8.8 V
1.0 v
3.5 v
-0.8 v
12.3 V
5.3 v
c-tielicum:
Signal
Gain:
CBT Oven Err:
Ion Pump
Mass spec:
DRO Tuning
UP clock PILL:
-12v supply:
Thermometer:
-1.67 %
19.9 %
12.137 mA
28.8 %
0,OO c
0.2 uA
9.1 V
6.8 V
2.9 V
35.0 C
-12.4 V
50
7

USNO permanently files these values as:
50717.007569 CBT ID: 3128AOO110(H) /Status summary: Operating
normally Power source:
AC /Log status: Empty Oe-15 29.94 26.5 25.4
39949 12.166 2553 15.4 8.3 0.00 -8.7 0.0 0.8 12.8 1.8 5.8 1.7
2.6 12.3 -12.3 5.2 41.3
A computer program can sort and display the information in whatever format is
needed. One might want several devices listed on one page when looking at
snapshotpictures. On the
other hand, one might want several consecutive readings
for a particular device listed on one page to look for trends. Outputting the values
to
a plotting program is better for trend observation. Although the snapshot
pictures will tell if
a device is operating normally, it is the plot of the parameters'
history that will help most in detecting the onset of problems and in fixing the
device when
a failure occurs.
CLOCK PERFORMANCE
The frequency offset can be set by either the manufacturer or the user. Typically,
it is set to Oe-15.
Oscillator control percentage is a very important value to monitor. The value can
be positive or negative and like most of the other parameters, the value is very
characteristic
of the device, One standard may have an oscillator control value of
-8.6% and another standard may have a value of -43.1%. Neither of these values
done means that the clock is performing (or will perform) poorly. A variation in
the value may suggest a future failure, and that will be covered in the next section.
What is important about the oscillator control percentage
for performance is its
frequency
and amount of change. This is an example of when continuous, periodic
readings are needed. For example, USNO has had at least two standards that
exhibit a phase or rate change that was coincident with a change in its oscillator
control percentage. Such
an occurrence may be due to the standard experiencing
an environmental change (temperature andor humidity), a Cesium Beam Tube
(CBT) change, or a CBT controller board change, or it may be a precursor to a
failure.
The
rf amps have been found to have relatively constant values between 20.0 and
31.0 for each standard after initial burn-in. (The values may change somewhat
during the first
1 112 to 2 years of operation.) The values will differ slightly from
51

one device to another. The Zeeman frequency is set at the factory and should be
39949.0. The C field current should be between 12.0 and 12.2 milliamps and
remain fairly constant. Again, the exact value will differ between standards.
The electron multiplier voltage is another very important parameter to watch.
There
is a limit at 2553 volts. USNO does have a standard performing quite well
below the
lo00 volt specificationfound in the 5071A owner's manual. (Its startup
voltage was about
440 volts,) Typically, our standards have a startup voltage
between 1150
and 1500. The voltage will decrease slightly and then begin to rise
slowly during the first two years of operation. The total change will usually range
from 150to
200 volts. The voltage should continue to increase slowly after this
initial startup
period.
The absolute value is characteristic of the device. The
standard should be performing well if the value is changing slowly and
CONSISTENTLY during the entire life time of the CBT.
The signal gain should be constant at 14.4 percent as long as the standard is below
its maximum electron multiplier voltage. When the electron multiplier voltage
reaches its
limit, the signal gain will increase to keep the overall gain constant.
The cesium-beam oven voltage depends on ambient temperature. The
USNO
standards values are consistent between 7.5 and 8.0 volts. Actual values differ
between devices. The cesium-beam tube error must be small (+I- 0.1 volts) for the
standard to be. operating properly.
The oscillator oven voltage typically has a value
of -8.7 to -8.8 volts during normal
operation. It should change by no more than 0.1 volts throughout the life of the
device.
The ion
pump current should have a low startup value (typically near 0.0) and
remain constant. Some devices, however, are working quite well with constant
values
of 10,20, or higher. (USNO has one device working well with an ion pump
current of
36.0 microamps.) This value is another one of those "personality"
characteristics
of each device. A high current value can indicate a vacuum or
electrical leak, that the tube has been off for a long period, or that the tube is
contaminated. A current greater than 50 microamps will cause shutdown.
The hot wire ionizer voltage references the voltage across the ionizer ribbon.
A
value of 1.0 is ideal and it should be between 0.9 and 1.1 volts.
The
mass spectrometer value will range from 10.0 to 14.0 volts. It should remain
the same value during normal performance due to environmental changes.
52

The SAW and DRO tuning, 87 MHz PLL, UP clock PLL, +/-12 volt supply, and
5 volt supply values should be relatively constant from initial startup. All may
vary by 4- 0.1 volts during normal operation,
The
temperature value is the internal temperature of the standard. The value should
typically range from
35 to 45 degrees Celsius and should remain constant. It will
change
as the standards environment changes,
CLOCK FAILURES
The recording of the parameters once per hour, every hour, and retaining the
information
in permanent data files has a great benefit for detecting and analyzing
failures.
It also allows for the prediction of some future failures.
Usually, analysis
of the historical parameters may only indicate that a failure will occur, but not
when.
USNO has experienceda few problems with the standardsand often looking
at the parameters has pointed directly to the needed repair. By taking hourly
readings of
the parameters, one can watch a standard fail and prevent a related
system disaster.
An example of a minor problem that can be seen in the parameters is a significant
change in the standard’s environmental temperature. A change in the environmental
temperature can show up as an unusually large change in the electron multiplier
voltage (see Figures
1, 2, and 3),
More serious problems are seen when an electronic card inside the standard fails.
USNO has had some problems with the A2 and A6 cards failing and has
documented the parameters showing the indications of
a failure.
A failure in the A2 CBT controllercard shows up as a significantchange in only
one parameter.
The electron multiplier voltage willjump more than 10 volts. This
greater than 10-voltchange will occur over a 12- to %hour period (see Figure 4).
Although a step in this parameter is not always an indication of an A2 card failure,
the change
is an indication that it is probable. No parameter changes seem to
foretell
this problem.
A failure in the A6 servo card causes the clock to lose frequency lock and shows
up
as a change in most of the parameters. The failure is most notable in the
oscillator control,
rf amplitudes, and CBT parameters. The oscillator control will
jump more than 5 percent when the failure occurs (see Figure 5). The rf
53

amplitudes will jump and remain at a constant value (see Figure. 6). The CBT oven
voltage will go to 0.0 volts (see Figure 7). The CBT temperature error will step
more than 1 degree Celsius and remain constant (see Figure 8). These jumps occur
within a few hours and can occur between two consecutive hourly readings. Again,
although a step in these parameters is not always an indication of an A6 card
failure,
the change suggests that it is probable. The occurrence of this failure is
sometimes predictable. One cannot, however, foretell when it will occur.
The
signs are in the CBT parameters. The CBT oven voltage will vary more than
4-0.4 volts, rather than the normal less than +/- 0.1 volts (see Figure 7).
The
CBT temperature error values will also show an abnormally large variation. They
will vary more than 4- 0.15 degree Celsius rather than the normal less than 4- 0.1
degree Celsius (see Figure 8). Another type of failure in the A6 card normally
occurs only at startup. A short can cause a fuse to fail on the A6 card. Notice that
after repair, the
rf amplitudes have returned to their prefailure characteristics (see
Figure 6) and the CBT parameters show a "normal" amount of variation (see
Figures
7 and 8), It is also worth noting that the oscillator control and electron
multiplier have assumed new initial states after repair (see Figures 5 and 9).
The standards at USNO are now getting old enough so that we are starting to see
end-of-life CBT failures. This is probably the only failure that every organization
will
see if they keep the standard for a long time or receive an old standard. The
Hewlett-Packard manual says that the standard
CBT should last about 5 years. The
high performance is warranted for 3 years. USNOs experience is that the high
performance
CBT lasts on average more than 5 years. Two parameters that say the
CBT has failed are the electron multiplier voltage and the signal gain percentage.
The electron multiplier voltage will
rise quickly during the failure. It will read
2553 (the limit) when the failurehas occurred (see Figure 10). The signal gain will
very quickly rise to
100% after the electron multiplier voltage has hit its limit (see
Figure
11). The alarm light should but will NOT always be lit after the CBT
failure has occurred.
The oscillator control percentage will slowly increase its amount of variation during
the failure. The size of the variation will increase as time passes after the CBT
failure (see Figure 12).
The failure will also be observed in the rf amplitude 1 and 2 parameters. They will
both shaw a large increase in variation either at the time of CBT failure or shortly
thereafter.
There
seem to be three parameters that suggest a CBT failure will occur. These are
the electron multiplier and rf amplitudes 1 and 2. (Although the signal gain does
show that the CBT has failed, it does not show any sign of when the failure will
occur.)
The first and most obvious foretelling event is a rapid increase in the
electron multiplier voltage (see Figure 10, MJD 50712 to 50716). USNO found
54
1

that the clock is sometimes already outside our performance criteria during this
period. The step in the electron multiplier voltage (see Figure 10 at MJD 50708)
indicates that the failure is about to begin. This step will usually occur between
2 weeks and 2 days before the electron multiplier voltage reaches its limiting value
of
2553 volts and indicates that end-of-life is inevitable. This is, however, not much
advance warning.
Advance
warning is provided by the rf amplitude 1 and 2 parameters. The
variation of the rf amplitude values is very large (see Figures 13 and 14 after MJD
50716). However, notice how the percentage values decrease just before the
failure.
This dlows for some lead time before the failure. More lead
time is
provided by the
rf amplitudes starting to show a tendency to change. For rf
amplitude 1, this can occur about a week before the failure (see Figure 13 at MJD
50708), RF amplitude 2 can show this tendency as much as 25 days before the
failure (see Figure 14 at MJD 50689). Both rf amplitudes can be either a more
positive or negative variation. Either way, they will show
an average decreasejust
before
the CBT failure occurs. A ward of caution, though: just because the rf
amplitudes are showing these early signs does not mean that the CBT will fail in
less then 30 days. It should be taken as a warning that the device needs to be
watched mare closely.
More concern should be used when the tendency becomes
more pronounced (see Figure 14 after MJD 50706). Although this is not a great
amount of notice before the inevitable failure, some notice is better than none.
The
notice will hopefully be enough to avoid a system crash or other operational
failures*
CONCLUSIONS
The Hewlett-Packard model 5071A primary frequency standard is a great
operational improvement over the model 5061B. This operational improvement is
enhanced further by the device’s ability to output its operating parameters
to a
computer for analysis. The analysis can help in the diagnosis of problems (e.g.,
cards gone bad).
Close analysis
of the parameters and their historical values will in some cases allow
the foretelling
of failures. These clues can be as simple as seeing extreme
environmental effects on
the device andor a change in a cards operation.
If the
parameters are watched closely enough, the inevitable CBT failure can be
anticipated
and correctivemeasures taken.
55

Most of the time, the important parameters to be watched are: electron multiplier
voltage, ion pump current, oscillator control percentage, signal gain percentage,
internal temperature, and readings of rf amplitude t and 2 percentages.
These
Seven should enable a person to predict the performance and life of a frequency
Standard.
ACKNOWLEDGMENTS
The authors would like to thank Randolph Clarke for clock performance analysis
assistance,
and Edward Powers for providing historical clock parameter data. A
special thanks to Wendy King for the parameter information collection
programming.
REFERENCE
Hewlett-Packad Company, HP 5071A primary Frequency Standard Operating and
Pmg&ng Manual, 1992.
1
56

3