2003 ARRL Frequency Measurment Test at KD5TFD


The ARRL runs an annual Frequency Measurement Test (FMT). It sounded like an intriguing idea, so I tried my hand at 2003 FMT and got some decent results.


The Test

The test is basically W1AW, the ARRL station, sending signals on 80, 40, 20, and 15 meters for about 15 minutes. Receiving stations try to measure the frequency of the transmitted signal as accurately as they can.


The Setup

My setup for the test consisted of a Kenwood TS 2000 HF receiver and simple inverted V wire antennas. I'd left the receiver on for about 2 days before the test to make sure it was waremd up and termperature stable. (Two days is probably overkill, near as I can tell it comes to a stable frequency in an hour or less). The receiver was tuned to 1 khz or so below the received signal in USB mode. This will produce a 1khz audio tone, if there are no errors in the receiver, or soundcard frequency detrmining elements. The audio output of the receiver was fed to two computers, one running Argo and the other running a wave file recorder sampling at 44.1 khz. Argo displays a waterfall and was used simply to get a rough estimate of the frequency of the received audio tone. As the test ran, I recorded three 1 minute wave files for 80, 40 and 20 meters. Propagation being what it was at the time, I did not hear the 15 meter test signal.

Shortly before and after the test I also performed the same measurements/recordings on WWV (2.5, 5.0, 10.0 and 15.0 MHz) and CHU (3.330, 7.335, 14.670 MHz) time stations. This set of measurements is used to determine the frequency error in my radio and computer sound card, since the frequncy of these time beacons can be considered to be pretty accurate (at least more accurate than anything I have here!).

Once I recorded all of the wave files I used Spectrum Lab (a really great program) to determine the peak of the recorded signals. Then averaged these with some homebrew scripts to throw out values that looked like outliers. Using the WWV and CHU wave files I was then able to produce an error correction chart for my setup.

I performed essentially the same analysis on the W1AW wave files, with the exception of the 20 meter signal. The 20 meter signal was so weak, there were substantial sections of the wave files that did not have enough signal to work with, so I subsetted the 20 meter wave files down to sections that had good signal strength. The 20 meter singal was not audible, it could only be detected with Argo's waterfall, so I was not terribly confident in the results. Once I had results from Spectrum Lab processing and averging I applied corrections based on my error chart and submitted my results.


The Results

Band ARRL Frequency My Measurement Absolute Error PPM error
80 3,585.373.7 Hz 3,585.373.6 Hz 0.10 Hz 0.03 ppm
40 7,050,409.9 Hz 7,050,409.8 Hz 0.10 Hz 0.01 ppm
20 14,050,075.7 Hz 14,050,076.2 Hz -0.50 Hz -0.04 ppm

I was pretty happy with the results, especially the 20 meter result as that signal was so weak. Was also a bit surprised as to how accurate the results came out, amazing what one can do with a modern PC, free software, and a little basic math.

I hope the ARRL does another test. It was a bit of fun to come up wiht the setup and data analyis procedure for this and would be interested to see if I can get more accurate results.


Comments to: Bill Tracey (kd5tfd@ewjt.com)

Last Updated: 11 September 2004

Copyright © Bill Tracey 2004