The ARRL runs a neato contest twice a year - the Frequency Measuring Test (FMT). The objective is to measure a transmission's unknown frequency as accurately as possible. There are 4 transmissions: 2 different geographical locations on 2 different bands (40m and 80m). The tests are run in the evening when the ionosphere is changing which adds an unpredictable element (doppler shift) to the results. Like any good “contest”, it's part skill and part luck!
The results are reported in accuracy windows:
Greater than 10 Hz doesn't get reported in a window but does get reported in your results listed by callsign.
You'll notice there's a method/soapbox with each callsign. If the contestant chooses, they can report their setup. Some of the setups consist of some amazing gear. My setup is basically my radio station with the addition of a GPS disciplined oscillator, namely the leo bodnar.
Here I'll note how I did my measurements so I can remember for future contests. I'll update this as I learn new things.
My hardware setup has used a KX3 or a Kenwood TS-690S with the SO-2 TCXO option. The Kenwood with the TCXO option is quite accurate at +/- 0.5 ppm. I suspect it could come close to a 1 Hz result without an external reference. At 7 MHz the spec would be +/- 3.5 Hz however my experience is that it's within 1 Hz. The KX3 seems to be within around 4 Hz so perhaps better than 1 ppm. I'll need to look up its spec.
As mentioned above I use the leo bodnar GPSDO. It is programmable so it can be set to a frequency close to the test frequency, namely to be within the receiver passband so that both the reference and the test signal can be measured at the same time.
I combined the GPSDO with my antenna through a simple tee. Of course this would affect the SWR but that doesn't really matter for receive. Do be very careful not to transmit while the setup is connected. I turn the power level all the way down and ensure the radio is set to PTT, not VOX. It's a good idea to disconnect keys and microphones, anything that could trigger PTT.
The GPSDO emits around 0 dBm. The signal needs to be reduced to a reasonable receiver level. -73 dBm is about S9. The GPSDO signal should be about the same strength as the received signal. We want fldigi to see both signals about the same strength. I use fixed attenuation to get the GPSDO down to S9+10 and then a switched attenuator so I can quickly change the strength to match the received test signal as it varies. I found the PX3 with the KX3 to be convenient to see the relative strengths and make adjustments.
Note that all hardware should warm up for at least 30 minutes. I turn everything on a few hours before the test.
There are multiple different software choices. I've been using fldigi since I started so I continue to use it. I suspect changing software won't change my results since the accuracy should be physical, not analytical. However different software might be easier to use.
I use fldigi in the FMT mode. It's a good idea to look up the instructions. Also note the recommendation to calibrate the audio codec. This compensates for frequency drift of the sound card sampling clock. It seems you could process this out in the final data but it's definitely easier when your reference appears stable during the measurement.
In the FMT configuration, I don't think the frequency correction is important to the final result. It doesn't actually correct your result. It is handy for getting the trace to 0 so you can zoom in. The display only zooms into 0, not an arbitrary offset.
I set the FIR filter to be the maximum of 10 Hz. I don't know if the measurement could be improved with a smaller value. I find that I want to be sure the test and reference signals are captured for the duration of the test. If the signals go outside the measurement window, fldigi still reports numbers that seem ok but are just an average of the noise inside the filter. A clue is that the reported signal strength drops to the filter's noise level. As you would expect, I've gotten bad results if my signal isn't inside this passband.
Use WWV to practice making a measurement. Set the radio to DATA mode (for example the KX3) or USB with sufficient bandwidth (for example the TS-690S). Set the carrier to 9.999 MHz for the 10 MHz WWV or similarly 1 kHz below a different WWV frequency. The AM carrier will show up at 1 kHz in the audio passband. There will be other artifacts associated with the WWV tone, ticks, speaking, etc. Setting the GSPDO to 2 kHz in the audio passband should generally keep it out of the audio artifacts.
The most important things when gathering the data are to remember to start recording (I've messed this up more than once), to note the key down and key up times as close to the second as possible and to note your configured GPSDO frequency. In addition, monitor the relative signal strengths of the test and reference signals, making adjustments before key down.
The FMT tells you the approximate frequency to find the test stations. They'll be within 10's of Hz of the published frequency, on their test frequency. So you can be sure to be able to find them but they'll be on the “unknown” frequency to be measured. They do the call up and the test on the same frequency so you can get all your parameters set during the 5 minutes of call up.
You can start recording at the beginning of call up but the file gets kind of large. I suggest waiting until they send their series of dits just prior to the key down to start recording. As soon as key down starts, make a note of the seconds on your clock so you can find the row corresponding with the start of the test. Similarly, make a note of the seconds on your clock right at the end of the test. This will enable you to find the rows of data to average and find your result. The key down duration is approximately 1 minute.
In one case I could not hear the signal very well but was able to get a measurement. I believe my results were taken on the CW call-up portion rather than key down. Nevertheless I got within 1 Hz results. So recording the call up portion certainly isn't bad. But you do want to know where the key down portion of the data is in your file.
Be sure to keep original copies of the CSV files. If you save over the CSV file the formulas will be lost. You could, for example, open the CSV file in Excel and then immediately save it as an XLSX.
I immediately freeze the top row of data and add digits to the right of the decimal point, namely in the calculated columns like G, L and O.
You'll notice the reference frequency you selected in fldigi is at audio frequency in column E. Column C and E are supposed to add up to your GSPDO frequency. However they will not since its going to be off by the amount your receiver is off.
In order to keep the data unmodified, I add a new column T. Also in V1-W1,V2-W2,V3-W3 I add rows for 1) my final result (to be calculated), 2) the GSPDO frequency I programmed into my GPSDO, and 3) the GPSDO correction that will be an adjustment to fix where Column E is off.
We'll assume columns C and E are constant throughout the test. If you made any changes, find the row you want to use. I'll use row 4 in the case of constant columns. With the GPSDO frequency in W2, set W3 to '=W2-(C4+E4)'. This is a factor that you'll add to fldigi's calculated Column O. Therefore Column T with the final results will be '=O4+W$3'. Note the use of '$' in W3. The idea is that when you populate column T with this formula, you want all cells to point to W3.
Finally, you'll need to decide which rows you want to use. W1 will contain the AVERAGE() of the range of values in Column T that you want to use. At most use all cells between the key up and key down times you recorded during the test. You might also take a look at power levels in columns H and M. Within the key down time, as long as they only vary a few dB, the data is probably good. If they change a lot, something may have happened - perhaps QSB or perhaps drift outside the fldigi filter. In this case, use the range of times with reasonable signal strengths.
W1 is the result you'll report to the FMT website.