Using a VNA for rapid frequency standard alignment & accuracy assessment:
Return to OH2AUE homepage
Yet further misuse of my HP8753B !
- ok, let's start with a disclaimer: any ideas from my website you apply at your own risk !
time and frequency accuracy are the basis for pretty much everything in todays society
Finnish hams had great fun in 2008 with our (RATS) national HF Frequency Accuracy Measurement Competion (in Finnish, but check images)
being the happy owner of an HP8405A, I often use this for HF antenna alignment, microwave device group delay measurement and frequency stability measurements
using an A/D converter with this Vector Voltmeter, the phase output can be handled by a PC, otherwise a regular pen recorder is just fine too
I finally wanted to make the frequency error measurement even more flexible, so I turned to my old faithful HP8753B, that I have many a time misused before too...
with the firmware installed on my particular model is v.02.01, June 27th. 1988 there seems to be no problem with direct phase analysis
there are several real advantages of using a VNA:
- variable detection bandwidth
- arbitrary center frequency
- simple frequency alignment
- direct display
- electronic documentation
- instability issues detected easily
so far I have used the VNA for comparing stability of 10 MHz and 300 MHz sources, but it will work at any frequency within the analyser coverage
initially I used my Trimble Thunderbolt as the "trusted reference", after at least a couple of weeks ontime
the KO4BB Trimble Thunderbolt GPS Disciplined Oscillator Display does a fine job of showing critical data without a PC, mine is built on a piece of veroboard
first I did some stability analysis of the system over periods of 10 seconds to 1 hour as a manner of calibration (the analyser allows 24 hour logging)
phase variation is less than 40 millidegrees over 60 seconds without any particular filtering, smoothing or averageing
with the stabilized GPS secondary reference I made a test with my Ball Efratom Rubidium reference against the Trimble
here is a 60 minute plot of the phase drift after the rubidium standard has been powered up for 4 hours (it took 41 minutes and 42 secods for a full 360 degree swing)
this measurement shows a Δf/f of 2.8*10-11, but the same ballpark figure can be achieved in just a few minutes as the short term stability of the Rubidium reference is not too bad
alignment of an oscillator under test is fun, as you can see the results immediately on the VNA screen, but this is an iterative process taking a while due to restabilisation effects
here is a plot of a typical OCXO after alignment: the short-term stability is excellent and the phase drift is -154.4 degrees in 50 seconds yielding a Δf/f of 8.6*10-10 (downwards)
an example of alignment of the above oscillator as a 60 second plot: initial error in divisions 7-8, and after a couple of tweaks the error is visible in divisions 3 to 6
the OCXO will be very accurate for short periods of time, but will drift back and forth, remaining however in the order of 10-9
so the ghastly bit of this story is the horrendously poor short term stability of my Meinberg 166 with 10 MHz reference option LQ as in Low Quality :-(
what can I say ?
OK, so the above measured QK OCXO is going to replace the personal computer grade 10 MHz reference in the Meinberg, pretty damn soon too
some fun shots: warmup of the Rubidium reference, warmup of another OCXO that is 300 Hz too high, typical frequency change in the OCXO a minute or two after alignment
Here is a simplified process for the measurement:
- make a coarse comparison between your references using a regular dual channel oscilloscope
- you can either use the Y/T Mode or the X/Y Mode on your scope for Lissajous diagrams
- if necessary, make coarse alingment with the scope (this VNA method is for high accuracy calibration !)
- disconnect your directional device(s) from the HP8753
- set the VNA to CW mode and set the center frequency to be the same as your reference and DUT
- connect your reference (e.g. 10 MHz) directly to the Reference input channel (observe damage levels etc.)
- the RF input maximum level on my unit is +20 dBm, 25 VDC, much less RF is actually necessary
- your reference drive level should be appropriate to extinguish the Phase Lock Alarm
- connect your DUT to channel A (similarly observing damage levels etc.)
- set the analyser to A/R measurement
- set the display format to Phase
- set the scale/div to 36 degrees/div
- set the scale reference position to 5 DIV and the reference value to 0 degrees
- I prefer to use 1601 sample points
- set Trigger to Continuous
- start with a Sweep Time of 60 s (6 seconds/div)
- you may use Averaging/Smoothing/IF BW reduction to deal with noise and/or short term instability, but do it smartly !
- using the above may hide potential problems if you are not careful !
- if you want to use really narrow IF bandwidths, you may need to synchronise the VNA to your (e.g. 10 MHz) source in case the VNA timebase is off calibration...
- this system is sensitive enough to detect changes in your reference frequencies due to e.g. magnetic and graviational orientation, barometric pressure...
- it is possible to analyse down to the millidegree level with some care, but at least I do not have any MASER's, yet anyway ;-)
OM Ramppa, OH2LIY has independently verified this method by using it for calibration of his Rubidium reference source
after 1 hour warmup and about half an hour of alignment, 1 hour of measurement on the VNA gave: plot 1, plot 2
for comparison he used his Trimble Thunderbolt GPS secondary reference at 10 MHz
My colleague Juha Kiili, OH2LKV, also tested the Rohde & Schwarz ZVA vector network analyser and this also seems to work perfectly for time correlation analysis.
- on the ZVA, the maximum sweep time is 127500 secods, so it should be possible to analyse really small difference in time/phase.
- OM Hubert, DB7ME confirmed this method works fine also on the HP8753C ! (15.10.2014)
Return to OH2AUE
Established 12.06.2009, updated 15.10.2014