The development of a stable 10 MHz reference
oscillator
Introduction
A microwave contact is influenced by a couple of variables. Some of these
variables can be eliminated or reduced to an acceptable error level. One
of these variables is frequency accurency. Following is a description of
an attempt to build a frequency reference locked to an external higher
standard.
The design
When you have an idea to make a design you first have to find out if someone
else already has done what you trying to develop. Why invent the wheel
let others do that. The results of a quick inventarisation are listed:
-
DCF77 receiver DJ3RV, UKW berichte 24 (1984) Heft 1.
-
DCF77 gelockte frequentiestandard PA0JBB, Electron april 1993.
-
MSF reference G4JNT, RadCom April/May 1994.
-
10 MHz Referentiebron, Elektuur september 1989.
After some discussion the decision was made to take the idea from the 153
kHz design. Studying the design in detail made clear that the idea was
good. Some improvements in the circuit had to be made. The following block
diagram gives an overview of the realized project.
Detailed design
Antenna & preamp
The antenna is a LW ferroxcube staff tuned to 153 kHz. The antenna is followed
by a preamp fed via the coax cable.
Receiver
The incoming 153 kHz signal is buffered by a transistor and filtered by
a double-tuned bandpass filter. After the filter the 153 kHz signal is
fed into a CA3089. This IC clips the signal and delivers a signal indicating
the strength of the input signal. The output signal of the CA3089 is mixed
with150 kHz in a SO42P. The 3 kHz SO42 output signal is filtered out by
a LC bandpass filter. An active band filter completes the receiver.
PLL
In the PLL the 3 kHz from the receiver and the 3 kHz from the 10 MHz TCXO
are frequency/phase compared in a 4046. Due to the long integration time
of the filter following the phase comparator there is not a direct phase
relationship between the two input signals. Changing the phase of one of
the input signals by 180 degrees takes about 15 minutes to regulate this
error out.
A long integration time is nescesarry because the received signal has
a very bad short term stability.
The only information left in the signal after the filter is an error
voltage which consist of a part due to aging and temperature change of
the 10 MHz TCXO and a part due to drift of the 153 kHz transmitter. (if
one calls 1E-10 drift) Some noise is also part of this signal. But by letting
the adjustment range of the 10 MHz TCXO small (0.1 Hz/V) the influence
of the noise is minimized.
10 MHz oscillator
A 10 MHz TCXO was chosen as oscillator. The first test showed that the
temperature compensation of the TCXO was not enough to achieve a stable
"lock" of the PLL.
A new circuit was designed where the TCXO was placed in a temperature
controlled oven. This oven was placed in a 3cm thick isolation material.
This was an big improvement but still temperature changes in the room are
visible as a slow drift of the loop voltage.
Divider
The 10 MHz from the TCXO is divided to achieve the 150 kHz and 3 kHz signals.
This is done by first multiplying the 10 MHz to 30 MHz. 30 MHz can easily
divided to the required frequencys.
Experience and performance
Two receivers were build.
Due to instability in the propagation path 180 deg phase shifts combined
with fading were observed. This happens mostly one or two hours after sunset.
This caused the phase comparator to generate an error voltage to adjust
this phase error. Due to the long integration time this take about 15 min.
in this period a couple of new 180 phase shifts were observed. This resulted
in an instable loop and the 10 MHz output signal is no longer accurate
to the wanted level.
After a couple of years of inactivity and the move to a new place (bigger
town --> more man made noise) the project was once again started up. When
the receiver was in its new place no lock conditions could be observed.
Watching the field strength it looks if this was lower as it was at the
previous place. But one of the biggest problems was the man made noise.
Clicks from light switches and heating equipment made it impossible to
have a stable lock situation.
This problem made clear that the 153 kHz wasn't no longer a good choice.
The results of the follow up are in part 2.
Back to PE1CQQ 
Page.
(c) PE1CQQ Last update 5-10-2000