Marc Torres, Sept,2002
For VLBI the excellent phase stability characteristic of the
Hydrogen
Maser needs to be transferred to the first LO of the receiver with
little
or no degradation, so as the overall VLBI experiment can be effectively
maser-limited.
On the PdB, this is performed by sending a signal to the antennas
in the 2 GHz range, which is locally multiplied by ~100 to reach 200
GHz.
In the building, a frequency synthesizer gets the 5 MHz from the maser
multiplied by 400 to generate the 2 GHz. Those tremendous
multiplication
factors place very stringent requirements on all the phase-locked loops
involved.
Commercial synthesizers are specified in terms of spectral purity, but manufacturers give little or no information for frequency offsets below 10 Hz from the carrier. This is probably because the telecom market does not care about stability in this range, or because the use of a non-proprietary external 10 MHz reference might be conflictuous. Depending on the cleanliness of their respective designs, synthesizers exhibit phase stability records that can vary by a large amount.
To fill this range, which is not covered by data sheets, we have
designed
a DC-coupled analog phase meter, which is stable enough for doubtlessly
comparing various synthesizers from the market.
1. The phase meter
The 10 MHz reference frequency is converted to ECL differential,
then
drives a push-push power amplifier. This one makes use of fast discrete
transistors, loaded by a moderate Q, very stable tank circuit
tuned
at 20 MHz. It delivers 200mW to the sampling mixer, a device from
Metelics
which includes a step recovery diode (SRD) and two very fast Schottkys.
The output of the mixer is lowpass filtered to 20 Hz , and a few op
amps
raise the level to several volts.
The pictures above show the principle and the implementation of the
DC phasemeter, which is due to student Florent Boulon, from IUT de
Grenoble
technical school. When an RF signal is applied, the output would issue
a sinewave at (RF-N*20MHz) frequency.
For example, with a synthesizer set at 920.000001 MHz , the
output
is exactly 1 cycle (360 degreees) per second, as shown below:
The optimal value of N depends on many things and needs to be found
experimentally betwen 50 and 80, as usual. If we assume that the
portion
of the sinewave between -30 and + 30 degrees is linear, we can
say
that one quarter of the peak-to-peak voltage corresponds to 30 degrees,
and this allows to calibrate the sensitivity of the phasemeter.
2. Tests on a few commercial synthesizers
Five commercial devices have been tested under the same conditions:
- Anritsu MG3690A lent by the local representative
- HP 8644B from the lab
- IFR 2025 from the lab
- Racal-Dana lent by Iram-Granada
- Rohde & Schwartz SMT06 lent by the local representative
- June,2005 : test on a CW-only
Anritsu MG3692B
- March, 2007 : test on an Agilent E8663B
-Jan, 2009 : test on a Rohde&Schwarz SMA100A
- Jan, 2009 : test on a Schomandl ND1000D
The synthesizers and the DC phasemeter were fed by an external high quality 10 MHz , via semirigid cables. Unfortunately the Racal only accepts 5 MHz (it has been modified for RA), and for its measurement, its internal 10 MHz source was sent to the phasemeter. So its record is not strictly comparable with the 3 others, but is still a good indication.
Records were taken over 100 seconds after a long stabilization period for the setup. The results are presented below and are self-explanatory.
Left:Anritsu, scale:2deg/div ; Right:
hp, scale :0.47 deg/div
Rohde&Schwartz , scale : 0.55 deg/div
Left:IFR, scale: 0.07deg/div ; Right:
Racal, scale :0.07 deg/div
The two first machines are definitely not usable for VLBI. The Racal
has less deviation than the IFR in the 1-10Hz range, but its drift over
100 seconds is similar. The Pk-Pk deviation of ~0.3 deg @900 MHz
observed on those graphs translates to 30 degrees @ 90 GHz, which
is acceptable for VLBI.
Direct spectra were simultaneously recorded. Click
here.
3. Experimental issues
These measurements require a very stable environmement. The physical parameters that degrade phase stability are mainly of thermical and mechanical nature.
The heaviest instruments are less sensitive to fast ambient temperature changes, and to microphonics. This instantaneous phase measurement technique clearly evidences this common-sense statement.
The electrical length of the connections used also need to be carefully stabilized. The connectors must be firmly tightened and the cables secured. The use of double-braided coaxial cable, or semirigid, is mandatory. From this point of view, it is a pity that all the commercial synthesizers make use of a BNC connector for their external 10 MHz input. BNC is spring-loaded and does not allow for good control of the electrical length.
The stability of the DC phasemeter itself presently matches or
exceeds
the stability of the synthesizers it is intended to test, essentially
because
it is a far simpler device. The measurements it delivers are better
taken
as comparative information than as absolute values. It could be
improved
by increasing its thermal and mechanical mass, and by adding some
layers
of thermal and electrical regulation.