K6OQK FMT Methodology
This is a picture of the off-air frequency measuring setup used at K6OQK. I also use this setup for the commercial broadcast off-air measurement service that I established about 35 years ago.
The
frequency standard is GPS referenced using two, DATUM 9390, Phase steered
Frequency and Time Code Generators with separate antennas, one as a backup,
which is continuously phase compared to the main GPS receiver.
For
the FMTs I use a GPS referenced HP-3586B as my receiver. One of the things you
quickly discover about the HP-3586 series SLM receivers is that the 1850 Hz
tone, while accurate enough for its original intended purpose, is far from the
accuracy required for our current level of precision frequency measurements.
This is because the 1850 Hz tone is derived from the Product Detector that uses
one of two crystals, dependent on the sideband selected, to generate its
carrier. These crystals are not
precise. They tend to be off by several tenths of a Hertz and they drift.
However,
the conversion oscillators within the HP-3586 are derived from the time base.
This means that for an incoming signal that is exactly on the
"tuned-to" frequency of the receiver, the resultant I.F. will be
exactly 15,625.000 Hz. I've added a pick-off point for the receiver's 15,625 Hz
I.F. that I can feed to one side of an oscilloscope and at the same time to
Spectrum Lab* that runs on my Laptop computer. Spectrum Lab has been calibrated
to this laptop. The other side of the oscilloscope is fed with a GPS referenced
HP-3336A precision signal generator, which coincidentally is the companion
generator for the HP-3586 series instruments. The signal generator is then set
to 15,625.00 Hz to produce a 1:1 Lissajou display on the oscilloscope.
Because
of the conversion scheme used in the HP-3586B, the I.F. error is in the opposite
direction of the incoming signal's error. By using this I.F. method I have
totally removed the steps and the need to calibrate around the Product Detector
and its 1850 Hz tone errors.
For
commercial measurements I use various techniques depending on the type of signal
I'm measuring and I usually have plenty of time to examine a signal.
But, since our FMTs only give us a key-down period of about 3 minutes or
less, my preferred method is as follows:
I start by presetting my receiver’s memories to the FMT’s published frequencies as well as the receiver parameters I want to start with. Since I’ve recently started using the time-averaged results from Spectrum Lab for measurements over long paths, I want to get as close to the correct frequency as soon as I can in order to give Spectrum Lab as much time as possible to time-average the signal without my tuning the receiver.
As
soon as I hear the FMT station’s CW Call-Up announcement I immediately begin
to tune the HP-3586B receiver, starting with the 3.1 kHz bandwidth, first in 10
Hz steps and then as I get closer to the actual frequency I reduce the frequency
steps to 1 Hz and reduce the I.F. bandwidth of the HP-3586B to 400 Hz and
eventually to the 20 Hz bandwidth position, all the time homing in to produce as
close to a 1:1 stationary Lissajous as possible. Using the oscilloscope and the
Lissajou I can usually get within 1 Hz or very close to 1 Hz during the CW
Call-Up announcement.
The
reason I personally prefer to use 1 Hz steps is that it is much easier for me to
quickly work with whole numbers than fractional or decimal numbers. Once the
receiver is tuned to the nearest whole cycle (Hz) I can move the HP-3336A’s
frequency in mHz steps to make the Lissajous pattern stand still - something that
can rarely be accomplished in a actual FMT over a long path due to Doppler
shift. The receiver’s 15,625.00 Hz I.F. is also monitored using the waterfall
display of Spectrum Lab set for +/- 5 Hz of the 15,625 Hz center frequency. This
allows me to time average the incoming signal's frequency for the duration of
the FMT's key down period.
Here’s
an example: If the HP-3586 receiver ends up tuned to 3,555,654.0 Hz to produce
the closest to a 1:1 stationary display then I know I am within 1 Hz. I can then
move the HP-3336A generator to try and make the Lissajous stand still. If the
HP-3336A ends up being 15,624.33, then I know the error is 0.67 Hz (15,625.00 Hz
-15,624.33 Hz = -0.67 Hz) (corrected to a +0.67 Hz because of the conversion
direction) as received at that moment. If Spectrum Lab and the Lissajous agree
over the three-minute key down period it's a miracle. Depending on the direction
of the error I will either add or subtract that I.F. error to the
"tuned-to" frequency of the HP-3586B. In this example the measured
frequency is 3,555,654.67 Hz.
Note
that with this method, tuning the 3336A does not affect the operation or
accuracy of Spectrum Lab’s time averaging of the frequency.
Once the HP-3586B is tuned and left alone, the HP-3336A and the
oscilloscope and Spectrum Lab are two separate and unrelated processes.
I’ve
experimented with injecting a precise low level signal on a nearby whole-Hertz
frequency and measuring the difference using Spectrum Lab. The results have been
very comparable to the method I described above.
My
antenna is an Inverted-V, 105' on a side running north to south with the apex
about 40' up in a tree. It's fed with balanced line all the way into my Shoppe.
For my ICOM radio I use a MFJ 974H balanced tuner and for the FMTs and the
HP-3586B I use a MFJ-902H tuner. There is a very basic DPDT knife switch that
switches the balanced antenna feed line between the two tuners.
The
results of all of this speak for themselves: Sometimes I'm close.
Burt, K6OQK
*Spectrum Lab is a free download and can be found at:
http://freenet-homepage.de/dl4yhf/spectra1.htmlCopyright 2007 by Connie Marshall and Burt Weiner