The MVUS FMT:
October 2007 N8UR Monitoring Site Description
N8UR Monitoring SiteTo cross-check the calculated frequencies from the W8KSE transmitter site, John, N8UR, monitored the transmitted signals off the air. John's station is about 3.7 miles from the site (at 84 10' 20" W, 39 42' 49" N)
Here is a block diagram of the setup described below.
And here is a photo gallery showing the equipment at both sites.
Measurement HardwareI used a ground-mounted vertical antenna to receive the FMT signals. The antenna was fed, through a high-pass filter that cut down the very strong broadcast stations in the area, into a six-way splitter. Three of the outputs were used to drive the measurement system. Each antenna signal went to one of the inputs of a two-way splitter/combiner.
The 3586C selective voltmeter that I use as an HF test receiver has a tracking generator output that makes available a 0 dBm signal on the frequency to which the receiver is tuned. That output is fed to a variable attenuator (up to 100 dB attenuation may be needed) and from the attenuator into the other input of the combiner. The combiner's output was fed to the 3586C antenna input. The audio output of the 3586C was sent to one of the inputs of an M-Audio Delta 44 sound card.
Here is a diagram of how the 3586C receiver is configured for this measurement:
This configuration was repeated three times to allow monitoring all three W8KSE signals simultaneously.
An HP 5065A Rubidium frequency standard provided the frequency reference for the three receivers. It was monitored against GPS so an accurate frequency offset could be established. Here are graphs showing the 5065A's offset over a long time period, and during the weekend of the FMT:
During the test period, the offset was about -4.75x10-12. Note that the short-term plot scale is an order of magnitude greater than the long-term plot. That's because the GPS PPS signal is very noisy for short-term averages, which this plot uses. The important thing is the green line which shows the frequency offset.
An 1850 Hertz tone generated by an HP 3325B function generator (which was locked to a Z3801A GPS disciplined oscillator) was fed into the fourth channel of the Delta 44 sound card.
Measurement SoftwareThe Delta 44 was installed in an Athlon X2 64 bit computer running Linux. The jackd audio routing system fed the four input streams to two programs.
The magic of this system is the 3586C and its tracking generator. The tracking generator signal, when looped to the antenna input, generates a tone at nominally 1850 Hertz no matter what frequency the receiver is tuned to (the receiver's passband is centered around 1850 Hz). So, as you tune an unknown frequency through the passband, you will see it move relative to the "marker" created by the tracking generator input. If the receiver is in upper sideband mode, and you position the unknown signal slightly above the 1850 Hertz marker, its frequency is simply the 3586C dial setting plus the difference between the marker and the unknown.
I use Baudline for tuning the signal, and measuring the delta between the two. It has a function that automatically measures the delta between the two largest peaks in the display, and the measurement resolution extends down to the microHertz.
To get maximum accuracy, I average the signals over the transmission period, and use the delta frequency function to measure the peaks in the average display.
One issue with this method is the accuracy of the clock driving the sound card's sampling. While one of my long-term plans is to replace the sound card crystal with a synthesizer I can lock to an external reference, I haven't gotten around to that yet.
So, the 1850 Hertz pilot tone (mentioned above) that is fed into the fourth channel of the sound card is used to measure the sample rate error. The pilot frequency is derived from a GPSDO, so it is accurate to at least parts in 10-12. Baudline's frequency measurement tool provides its idea of the pilot tone, which can be compared to the known 1850 Hertz frequency to determine the error.
This calculation works backwards from the way fractional frequency offset
is usually determined, because a sample rate slower than nominal
will result in a frequency measurement that is higher than it
should be. So, the formula is
To determine the actual tuned frequency, the 3586C's indicated
dial frequency is adjusted by the HP 5065A's measured offset (in this
case, about -4.8x10-12, and the delta frequency by the sample
rate error (about 4.9x10-5) and added.
In other words, assuming that the receiver is used in upper sideband
mode and the unknown signal is tuned above the 1850 Hertz marker tone: