December 2009 Soapbox

Name Call Comments
Mike Naruta AA8K Open HPSDR, Trimble Thunderbolt, Spectrum Lab 80M -110 dBm 40M -76 dBm 20M no signal
Edward G Neubauer KB3SZZ I was able to use the frequency measuring test to assist the poor. First, I set up W8KSE's FMT in my empty basement, just playing its lonely tone in sync with the 800 Hz sidetone CW pitch set on my Yaesu FT-450. I was just listening and relaxing, reading the book One Flew Over the Cuckoo's Nest. The sound drew the attention of my wife, who came downstairs to see what was going on. I told her that the W8KSE beacon was playing my favorite song. She just shook her head - after all, as she says 'Ham Radio is a nutty hobby.' The sound must have propagated through the back door because our dog started howling in the backyard at exactly....800 Hz. I turned up the volume, pumping the monotone into my stereo speakers, which were by now, set up on the back patio. Neighbors began to show up from all around, each with an uncontrollable dog at the end of a leash. They had migrated to the magnetic (a term used here only metaphorically) stimulus of the tone coming from far away in Ohio to our little townhouse in suburban Baltimore County. I determined that the magic of these waves from the heavens just could not be allowed to pass through history without doing some good.

Now, as it turns out, this was the day of the big neighborhood yard sale. Furniture, toys, clothing, and the general junk of all the good citizens was neatly lined up on tables throughout the community. By now, those citizens had lost all interest in making money. Hands over their ears, they were screaming, "Make it go away, please." Some were angry about the dogs. Others were upset about the blaring speakers. But all had a headache and were writhing in agony.

On her own, my daughter walked around to talk with them. "Would you like me to make it go away?" she asked. "Yes, yes – please! We'll do anything little girl," they responded. So she made them promise to donate their items to the poor. They readily agreed, and she promptly dispensed two capsules of ibuprofen to each of them.

Quickly, I drove throughout the 'hood, collecting furniture into the back of a moving van I had rented for a different reason, but that's another story. Then, the beacon broadcast suddenly ended, like the sun reappearing after a total eclipse. People stopped screaming, and the dogs went home. (It was a hell of a clean-up, though.) With my son's help, we delivered the contents of the truck to the local charity organization, which distributed the goods to the needy.

Some Testimonials:

From Johnny (not his real name) in the woods at the corner of Rossville Blvd. and Pulaski Highway: "Thank goodness for W8KSE's beacon. I have never had such a fine coffee table in my tar paper shack before. And the Weber grill keeps it nice and warm, if a bit smoky."

From Kitzy, somewhere in Dundalk: "I would just like to thank W8KSE from the bottom of my shoes - please excuse the gum - to the top of my beehive hairdo! Because of their tone, my world has changed for the better. My children now have the toys they need to keep them busy and out of my hair."

And from Bart, dog owner: "The best thing about the W8KSE transmission was when it ended. Spot (not his real name) will never be the same. Every time I walk him and we pass by that yard, he just looks longingly and gives the 800 Hz howl."

Robin Hood or Anti-Grinch? You be the judge.... The bottom line is, because of W8KSE, the poor people of Baltimore are just a little less poor than before.

Gordon Batey WA4FJC HP 3586B using a trap dipole antenna. Nothing heard on 20 meters. Next time I will try to use a sound card program and some long time data analysis. 73 Gordon
Bill Riches WA2DVU Reading done at 1300Z on 12/29. 40 M sig s7-s9 on Icom Pro3, Sig on 3586B was -72-76 dBm. Antenna is a Mosley Pro54 at 60 feet. WWV error at 5mhz was abt 240uHZ high. I did not figure WWV error into my entry. 73, Bill WA2DVU
Stuart Cole N5LBZ I really did not expect to hear anything with the transmitted signal at only 15W. but I was surprised to be able to hear the signal on all three bands. 20M was the weakest signal. I monitored the signal on each band for about an hour or until the signal dropped out. I connected my HP 3325A Synthesizer [locked to GPS] to a small dipole in my shack and adjusted the synthesizer freq until the audio trace on spectrum lab matched the trace of the unknown signal as close as I could read it. I noticed the freq of the received signal on all bands would vary slightly over time so I might be off a little if I measured it during a drift. All in all a very good FMT. Good job to all who put on the test Stuart
Bradley D. Miskimen N5LUL Glad to be a part of this event! I was using a Yaesu FT-920 and nothing else. Would zero beat (by ear) on USB, switch to LSB and do the same. Then took the average of the two findings. On 20M, I have found in past FMT's, that the receiver reads a little high. Took the average of the two readings and then subtracted a 'fudge-factor' of 32.43hz. We'll see how it goes this time. Brad- N5LUL
Michael Statom KB0OLA
Robert Snyder W6CP HW: Yaesu FT-2000, 32-bit PC with Ensoniq AudioPCI: ES1371 SW: fldigi in Freq Analysis mode Multiple runs switching between test signal and WWV at 15 MHz to measure error. Subtracted WWV error corrected for frequency (15MHz/Fo) from measured frequencies for each run. Threw out bad runs and outliers. Averaged each of the remaining runs for final result for each band. Note: Not a big deal, but your Latitude and Longitude input formats ('DD MM.SSS' and 'DDD MM.SSS') seem a bit confusing to me. Seconds after the dot? Then why three digits? Or did you want DD MM.MMM (i.e. decimal minutes)? I entered SSS as seconds where the digits are tens,units,tenths. Many thanks, Bob W6CP
Richard Dabney K6BZZ Thanks for running the test. Attic antenna here so couldn't get real strong signals. I think the ones I got were yours. '73 de Richard in Arizona
Bill Eaton K9AYA Wavetek 2407 sig gen HP 5345A counter both locked to HP Z3801A Yaesu FT-857 rcvr took about 20 100 second average counts of several different beat freqs. set sig gen about 1000 hz above then below then did 2000 hz above and below. ran the audio to the counter. The last digit could be + or - 1 or 2
Burt Weiner K6OQK I was able to hear the three HF signals fine at my location, so the FMT transmitter power was not a problem for me. My receiver is a HP-3586B fed by an Inverted-V, 105 feet on a side favoring east and west. I have a MFJ 920H tuner between the antenna and the 3586B receiver. A detailed description of my FMT methodology is described at: There was a great deal of Doppler on all frequencies on both days even at a time of day when conditions should've been stable. I did look at all bands at different times of the day and night hoping for a miracle. At times Spectrum Lab showed dithering as much as 2 Hz wide, so my entered readings are almost tongue-in-cheek. Tomorrow morning my wife is taking me to the Eye Doctor to get my eyes un-crossed, the result of staring at a dithering Lissajou scope trace too long. My coordinates shown above use NAD-27 DATUM. This has been a fun and interesting FMT. Thanks to all who worked so hard to make it a successful FMT.
James M. Galm W8WTS
Richard Davis W9ZB For HF, used home-brew copy of SDR1000 receiver w/100 MHz TCXO ref. Sigs on 20M were too weak for good reading, others just ok. For 144 MHz, used Icom706MKIIG and took reading of Ch 9 DTV pilot for Calibration, then eye-balled the reading from the Icom dial, adding cal correction and drift amount (kentucky windage).
Frank Getz N3FG I used an IC-7200 with Spectran. Checked equipment error against CHU at 7850 and 3330.
John Moody W5RVT Never heard signal on 80M. Ionosphere shift made this difficult. Equipment - HP3586C with Z3801 timebase ,Argo S/W ,HP8656B signal gen phase locked to Z3801 to check the setup.
Gene Hinkle K5PA The tools used here include: 90 Foot dipole, ICOM756PRO-III, Spectran V2, S-Meter Lite, WWV, and Excel. Spent way too much time making measurements and understanding what I was seeing. Found the IC756 PBT (TWIN Pass Band Tuning) DOES introduce spectral spreading of signal. After a lot of time learning, finally found a simple method using the spread sheet to account for frequency offset errors making data entry easier. I AM READY FOR NEXT TIME! Take my measured and calculated frequencies this time with a grain of salt. I also plotted S-meter readings, 4 times per second, to have record of signal strength for the frequencies measured. FUTURE: I would like to use Spectrum Lab software to record time versus spectral peak and calculate standard deviations and compare these to the standard deviations of the S-meter and absolute signal strength. My hypothesis that is probably obvious, but not demonstrated, is the higher the signal to noise levels the more accurate the frequency measurement. Question is whe
Jim Hatley K7TT The 20M signal was very readable here into my attic trap dipole when the band was open which was about 1600Z to 2100Z. The signal, however, was somewhat distorted due to propagation effects I assume. WWV 10Mhz and 15Mhz were also quite distorted making readings difficult. My method depends on a measurement of WWV to minimize errors in the receiver and PC sound card. I made 24 separate measurements which resulted in a spread of about +/- 0.4 hertz. I submitted an simple arithmetic average. I'll be amazed if it is even close as I usually make some ridiculous math error... During the test I never at any time heard a recognizable FMT signal on 80M or 40M and I spent considerable time trying to find them. I did hear some signals but they never identified when the 20M station did so I assumed they were spurs and ignored them. The FMT was a fun, challenging event and hope it will be repeated again in the future. Thanks to all the contributors there for letting us participate. Jim - K7TT
Scott Burkhardt W�KU SWL report only. Checking my frequency. 10 mHz reception of WWV is not strong here. The signal from Ft. Collins CO skips over Denver! Rig is a Ten Tec Omni 6+ with TCXO zero beat to WWV on 10 mHz. I was unable to hear 40M night or day. 80M + 20M were 599 2 ele quad @ 50' for 20M, sloper off tower for 80M + 40M Strong signals to my mobile also. 579-599 Thanks for the effort, Happy New Year. Scott W�KU
Doug Nelson K6HGF The QRP 20 meter signal was very good here in Southern California. Wish I could have said the same for the 40 & 80 meter signals. I'm probably way off on those. Thanks for the FMT.
Bruce Beford N1RX Location is in west central New Hampshire. I used a 40M extended double-zepp (176ft) doublet at 75 ft high, fed with 450 ohm windowed feedline to a 4:1 balun. This fed an HPSDR 'Mercury' receiver. The reciver's master oscillator was locked to a Trimble Thunderbolt GPSDO. The audio output of the receiver was fed to the shack PC, running Spectrum Lab software. The receiver was used in LSB mode with narrow filtering, and tuned to produce a nominal 600 Hz output tone. Reference runs were made, using WWV at 10 MHz and 15 MHz. This was used to guage the overall accuracy of the system. The receiver was used in LSB mode, and tuned to produce a nominal 600 Hz output tone. The output tone was monitored in Spectrum Lab, displaying a range of 598-602 Hz, using an FFT of 512K sample length. The cumulative spectrum was 'eyeballed' to determine the deviation from the nominal 600 Hz tone. The final calculated frequency results were determined algebraically. This was a lot of fun! Thanks for running this extended mode t
Marvin Collins W6OQI My FMT setup uses a HP Z3801B GPS disciplined oscillator. The 10 MHz output from the HP Z3801B is connected to the external reference input on a HP 3586B tuneable signal level meter and a HP 3336B signal generator. The 15625 Hz I. F. output from the HP 3586B is fed to the X input of a X-Y scope. The HP 3336B signal generator is tuned to 15625 Hz and fed to the Y input of the X-Y scope. The HP 3586B is tuned to the signal being measured. If the HP 3586B is tuned precisely to the incoming signal under test, the I. F. will be exactly 15625 Hz. The 15625 Hz I. F. is also fed to a computer running Spectrum Lab. It is usually not possible to tune the HP 3586B exactly to the incoming signal. There will almost always be a fraction of a Hz error. This error will manifest itself as the 15625 Hz I. F. signal being slightly off frequency. Spectrum Lab is used to indicate the freqeuency of the 15625 I. F. The tuning error shown by Spectrum Lab is then mathematically added or subtracted from the freque
Doug Nelson K6HGF This is a re-submission. Dyslexia got me on the first 20 meter submission. Please use these results. Thanks, Doug K6HGF
John King WA1ABI Tnx for the FMT.
Phil Walker W1PW I really liked this new 48-hour FMT format. it was fascinating to see how well the 15 watt transmitters could be heard. At times, both the 80 meter and 20 meter signals were considerably over S9 here in Arizona. The 40 meter signal was somewhat more challenging, but Spectrum Lab was able to hear it OK. During night-time hours, the 20 meter signal would completely disappear, but the 80 meter signal made up for it! Equipment: Yaesu FT-817nd calibrated periodically using the 10 MHz output of a Trimble Thunderbolt GPS Disciplined frequency standard. A 600 Hz heterodyne was fed to Spectrum Laboratory with outputs into a text file every 10 seconds. The rest of the math was done with an Excel worksheet for calibration corrections. Thanks very much for this unique test and the unique opportunity it gave us to hear the longer-term propagation effects on the signals. Sorry I couldn't hear the 2-meter signal -- guess I need a much higher vertical antenna for that, or a remote antenna and a 2,000 mile feedlin
D. Daniel McGlothin KB3MUN This was the FMT I have participated in. I received, via the LinuxHam Yahoo Group courtesy of W6CP, a notice of this FMT shortly after it started. Since I had just made my first HF QSO from my home station, I thought 'Why not, just jump right in, and do an FMT next.' My equipment chain is: K3MT Grasswire (pointing on a bearing of about 20 degrees) to LDG Z100 Autotuner to Kenwood TS-140S to BuxComm Rascal GLX digital mode interface to cheap USB sound card to Ubuntu v9.10 PC running FLdigi v3.13AE I started listening around 9:00 EST, and quickly picked out the 80M and the 40M signals. Quick measurements using FLdigi's Frequency Analysis tool I measured the frequency of those signals. Not really expecting to find the 20m signal given that its beam was pointed directly away from me, I none the less found a 'ghost' of a trace on the waterfall--I supposed I was catching it off the back of the beam. The faint trace in the waterfall seemed to break into CW in synchronisation with t
Bill Bradrick WB0LXZ Thanks for another chance to practice using Spectrum Lab. I used 80M dipole, HP3586B, and HP3335A with timebase locked to GPS with a PC and QBASIC program. I had plenty of time to zero in on signals using SpecLab in QRSS 60 or 120 second dot mode. I was surprised to receive enough 80 meter signal to make my measurement around 1:30pm CST on the29th. My 20 meter measurement was from 7 to 8 am CST on the 30th. -bill
Ted wa0eir Used a very basic set up. Icom IC-746Pro and my psk31 program, TWPSK Hope I got within a cycle or two :-) 73, Ted - wa0eir
Hisami Dejima 7L4IOU Dear OM Happy New Year! Thank you very much for the chance. I submit 20M freq first. still looking 40M but no found yet. I can upload WAV file to briefcase if you need. 73 Hisami 7L4IOU
SHINTARO SHIBUYA JF1VRU It was too hard for receiving a signal on both 40m and 80m in Japan though long time transmission of a signal.
Philip Scoggins KD5WBW The FMT was monitored first on frequency 1405435 at 1949 Z on 12-28-09. Later in the day and evening, the signal was not detected. On 12-29-09, copy was taken from approximately 1600 to 2040 Z on frequency 1405450. Again later in the day and in the evening, the signal could not be located. The signal was also not detected at all on 40 meters. Copy was taken with CW Decoder XP (my CW needs improvement and the software could copy without me being present.) Had I been better at copying CW, I likely could have copied better with the R in RST being a 4 or 5. The signal fluctuated down to zero at times. However, the S and T were generally 6 and 9 respectively. Here is the copy on frequency 1405435 at 1950 Z on 12-28-09. tm u r ib e t v eehe a h 6aedeh i ee AR t eve HI r AS ettet u e n cetett tt e a s u r i n g t e sit eue tk eeti l o itt s f at t er m t e erett t tee e eitt ttti k s e f m t f m a er m t aeie tteisee ie ei 5 a er l e e u e n c y m e a s u r i n g t e s t ste o l l o w s
Lyle Kraft AA6LK comments: Same setup as for 2009 ARRL FMT (FT-847, 5BTV, HP Z3801A, Marconi 2019, Laptop w/ Spectrum Lab). 20M signal generally between S5-S9; 80M signal around S6; 40M signal briefly heard, then only S3. Interesting to see how signal changes with changes in propagation. At about the same hour (10:30 PST) on both 12/28 and 12/29, the 20M signal appeared to split into two paths; one continued on steady frequency, the other began raising the frequency up to about 0.4Hz, dwell for a few minutes, then slowly come back down to the original frequency, over about a 15 minute period. There were other times when signals would shift up or down by ~0.5Hz for short periods of time, on the order of several minutes. I noticed this mostly on 80M and 20M. Was only briefly able to hear the 40M signal on 12/29 ~17:00 PST. On one measurement I displayed both the incoming signal and my nearby generator signal. This showed that the variations were truly in the incoming signals since the variations in generator signal were mostly due to receiver drift and independent of incoming signal. I saved off the images if anyone's interested. Definitely pays to know when signals are stable when making a measurement. Thanks to the MVUS gang for setting up the event.
Ron Smith G3SVW I listened to the signals from W8KSE using my HF transceiver, a Yaesu FT1000MP, with a 40-metre dipole antenna and 500 ohm open-wire feeder.

To measure the signal frequency I used my Hewlett-Packard 3336A frequency synthesizer. The instrument is several years old and has not been calibrated for a long time, although using measurements I have made of known broadcast signals in England, I am confident its accuracy is as good as 1 in 106. For interest, I can tell you the synthesizer was recovered from a West African scrap tip in 2002! It was a lucky find for me.

I listened for W8KSE between 08:00 and 09:30 UTC on Tuesday 29th December and identified the callsign clearly on the 7 MHz band. Signal strength was good, starting at S3 on the RST scale, and rising to S7 by 08:45 UTC. Fade was taking the signal down by 09:30 UTC. At that early time I could not positively identify the FMT carrier on 14 MHz, although I did hear it later in the day, but not as strong as the 7 MHz signal. Unfortunately I was unaware of the change to the 3.5 MHz carrier frequency and did not hear the FMT on that band at all.

I was not in a position to do any frequency measurement on the first day, so I listened again at 08:00 UTC on Wednesday December 30th. The 7 MHz signal was not as strong as on the previous day, only reaching about S4 at best. Fade on the signal was quite rapid and turbulent at times. To make matters worse, there was considerable interference from continental SSB signals.

To measure the frequency of the 7 MHz signal, I tuned to W8KSE, with a 200 Hz bandwidth to minimise interference, and with an audio pitch of about 600 Hz. I loosely coupled the H-P 3336A synthesizer to my receiver input and tuned it to zero-beat with W8KSE to get an initial estimation of the signal frequency. The turbulent nature of the fade didn't make it possible to gauge the zero-beat very precisely, so I tuned first to one side of the signal to get an approximate 1 Hz beat and read off the synthesizer output frequency, then tuned to the other side of the signal for a 1 Hz beat and read off the new synthesizer setting. Listening to a 1 Hz beat was easier than trying to get an accurate zero-beat.

If my estimation of the low-side and high-side 1 Hz beats was accurate, then my two synthesizer read-outs would differ by 2 Hz, and the signal frequency would lie between these two readings.

I used the frequency synthesizer with a 9-digit display to give a resolution of 0.01 Hz. I'm probably being a bit too optimistic with that last digit, so I don't have a lot of confidence in it, but recorded it anyway.

The frequency of W8KSE on 7 MHz I measured and calculated to be 7, 054, 684 . 52 Hz.

This is the first time I have experienced an FMT and enjoyed it. Of particular interest to me here is the effect of ionospheric Doppler - I am on the Propagation Studies Committee of the RSGB and iono-Doppler was discussed five weeks ago. In a future FMT I would try to get frequency measurements at different times of the day to see if any deviations could be attributed to Doppler at sunrise and sunset.

I would like to hear more news of future tests and suggest they are run maybe twice per year. If there is another FMT I am thinking of ways to make the measurement easier next time. Additionally, I would be interested to see a report of how others compared in the test and what methods they use.

Thanks to all concerned for running the FMT.

Connie Marshall K5CM Signals could always be detected with SL, even though at times the signals could not be copied by ear. Will send files to the email address. Thanks and 73, Connie K5CM
Jim Keeth AF9A 80m signals were pretty good here in Indy during the daylight hours. Signals in the afternoons were up to 30dB above the noise with QSB. Didn't hear anything on 40m or 20m at the times I was able to look for signals. The receiver used was a home brew version of the SDR1000 with an ovenized LO calibrated with WWV. Thanks for the test and 73, Jim AF9A
Mike Kowalsky AC8Y Thanks to all involved for running the test. Before going to work in the morning, I set up to log data on 40 meters, and then switched to 80 meters when I got back home. The 20 meter signal was too weak to mess with most of the time. A normal FMT is too short to allow much averaging. This test was better. Plots of measured frequency variation ( reveal the inherent variability of ionospheric path delay. Even with 96,000 samples, picking a center frequency with millihertz resolution was still a guessing game. It took a few days to reduce all the data. Maybe next time (?) the signal could be time tagged so path delay variation could be determined. A 1 pps phase shift could be added to the carriers. It would be fun to add a time interval counter to the measurement set-up. Equipment: Watkins Johnson 8718 receiver; HP-58503A GPS receiver; HP5334A counter; HP3324A synthesizer. 73 Mike AC8Y
Ed Moxon K1GGI Method is to beat the signal with a known heterodyne oscillator and measure the beat note with SpectrumLab. 80m reading came from about an hour of averaging around midnight local time on the 28th. 40m reading came from about a half hour of averaging around midnight the 29th; at that time the spectral line was quite stable and sharp. At other times, 40 wandered by hundreds of mHz. 20m is a guess from a short recording made around 1500 local time on the 28th. Thanks for a fun fmt.
Jeff Walker W3JW Propagation software tools were used to determine time of day when the path from my QTH to Dayton was most favorable for each of the HF bands. 80 M was available the most, 40 meters somewhat less and 20 M the least. However, with both Spectran and Spectrum Lab I could make useable measurements for most of my waking hours. Data was captured (SL) for a total of approximately 600 minutes (all HF bands) over the test period. Screen captures, wave files and exported data files were colected. All data was reduced manually (UGH!)
Dana Whitlow K8YUM NOTES ON W8KSE FMT EXPERIMENT I am privileged to work in the electronics department at the Arecibo Observatory in Puerto Rico, and one of the perks of my position is access to certain exotic resources. Among these is a hydrogen maser clock, whose derived reference frequencies are widely distributed about the observatory. Also available is a 20M vertical antenna on a nearby hilltop, used mainly for occasional ham radio 'events' held in the interest of good public relations.

I had been thinking for some time that participation in a frequency measuring test might be fun, given these resouces. The W8KSE test particularly struck my fancy because the minutes-long periods without modulation or keying seemed to offer the potential for a no-brains application of extremely narrow bandwidth to obtain decent signal to noise ratios on weak signals. But life was not to be that simple, as you'll see below.

Over about the last 18 months I had been developing a series of PC programs aimed at taking time domain series from either a DSO or my SDR-14, then deriving phase and envelope amplitude histories of the signal contained within a post-acquisition- selected band of frequencies. It had long since occurred to me that this capability ought to be handy for an FMT. The basic idea is to measure the slope of the phase history versus time, which defines the frequency. With decent SNR the phase slope can be measured to a small fraction of one cycle over any given time interval, putting me well ahead of the game compared to simply using a frequency counter or finding the peak bin of a spectrum calculation based on the same time interval.

My initial thinking had revolved around building a front end and baseband mixer, then capturing the IQ outputs with a TEK DSO at a low sample rare. The L.O. signal was to come from a synthesizer locked to the observatory standard. But time flies (at my age it seems to do so whether I'm having fun or not!), and the weekend before the test found me with none of hardware yet even started. So what was I to do? I decided on the following alternate strategy since it did not require building any new hardware, but at the cost of shifting the burden to software work later on.

THE SCHEME: (embellished from a writeup I made the day before the test in order to collect my thoughts)

Use my R-8500 receiver and SDR-14 to receive the signal and capture the IQ data. But drift has to be accounted for, as neither the R-8500 nor the SDR have external reference inputs.

Inject a reference signal from a stable source (locked to the observatory frequency standard) at a frequency very near that of the test signal. In software, isolate the two signals in the frequency domain, calculate the separate time domain signals, then take their complex ratio. This will be a low- frequency beat signal containing the desired information (amplitude and phase histories of the test signal), with receiver and sampling system frequency drifts virtually cancelled. Also amplitude and phase changes through the receiver (due to AGC action etc) will be cancelled. Then measuring the slope of the phase history wouldtell me the precise frequency offset between the reference and test signals.

Note that for strict cancellation of sampling rate drifts the beat frequency would have to be zero, but this would preclude the possibility of separating the test and reference signals in software. Therefore the non-zero beat frequency that appears will be mis-scaled by the sampling rate error, but if the beat frequency is kept down to a few Hz, results should be good enough for this purpose. Example: 10 Hz beat frequency, 1 PPM sampling rate drift after warmup- leads to 10 uHz frequency error out of 14 MHz, less than 1 part in 1012. I think propagation path effects will strongly dominate over this.

Now in an ideal world, this looks superb- I could capture a 24-hour span, yielding a bandwidth as little as ~12 microHz. But now reality sets in. For one thing I realized that the ionosphere would broaden the signal to a much greater extent, and receiver drift would require that the initial software pre-filtering also be much wider than this. So at this point I realized that there was no need to process the captured signal in a single chunk, which was much nicer for several reasons. High among these is that I would be doing the processing on my teeny netbook (my good PC is dead) with only 1 GB of RAM. In the end I settled on processing in 8 mega- sample chunks (about 1030 seconds each at 8138 Hz), for which worst case receiver drift was about 0.5 Hz. Also ionospheric signal broadening turned out to be in the range of about 0.25 to 0.5 Hz at my location.

Fortunately I realized in time that this would require me to set the strength of the reference signal high enough that it would have decent signal-to-noise in a bandwidth wide enough to accommodate receiver drift.

I should mention that Puerto Rico was not the best place to be for this test since the transmit power was low and the signal was directed towards the west from the Dayton area. Perhaps I was lucky to hear it at all!


So on the morning of Dec 28 I dragged my apparatus onto an empty spot on a bench of the control room and made peace with the telescope operator, assuring him that this setup would in no way interfere with ongoing observations. To further this aim I inserted an external 100-MHz LPF in the line from the antenna to catch whatever the R-8500 might be leaking out its antenna port. The control room area is already well-enough shielded to bottle up emissions from the dozen or so off-the-shelf PCs and other similar equipment there, so I was not too concerned about leakage from the cabinets and cables of my equipment.

A few minutes after 1445 UTC I found the code announcement of the test and began tuning up the local synthesizer which was to provide the stable reference. Using the SDR-14's own spectrum analyzer function (implemented by the Moetronix 'Spectra-Vue' software) I worked the synthesizer down to a little under 10 Hz below the signal frequency and was able to start data acquisition at roughly 1510 UTC. I acquired data at 8138 samples per second, the lowest supported by the software.

During this setup I came to appreciate the unwitting wisdom of my decision to use this arrangement rather than the home- built baseband mixer and DSO. With that setup I would probably have had much much more difficulty estabishing the reference signal's frequency where I needed it to be since that setup would provide no real-time view of the signal. I still would have needed at least a receiver in the setup.

I collected data from then until early evening, then stopped to restart another file. At that time I had collected a bit over 800MB and did not want to risk exceeding 2GB file size over the course of the night. I left the site for dinner, then returned about 2330 UTC to check on things and found that the signal had faded into nothingness. So I discontinued my "observation" rather early into the 2nd file. I was feeling bad about too much disruption to the work I'm being paid for and decided not to pursue any further data collection beyond this point. The boss wasn't even around, so doesn't this show real dedication on my part?

During the day I had checked from time to time to see how things were going and can report that the signal sometimes reached 20-25 dB over the noise in 0.5 Hz bandwidth, but was usually quite a bit weaker. I think I was lucky to find and copy the code ID just before the start of the test.


Having seen enough clues to be confident I didn't need to prrocess the data in one chunk, I first divided the long file into a series of 8 megasample (8388608 if you want to get technical about it) exactly contiguous files. I wanted the contiguous files because at this point I still had the silly notion that I might be able to somehow link the phase histories from file to file and get somthing of a continuous phase history covering several hours (dream on, kid!).

I added to one of my existing analysis programs to implement the processing steps described above. Next I checked for the proper behavior with a test signal from two separate signal synthesizers set a few Hz apart and both synchronized to the station frequency standard. The software appeared to be doing what was expected.

Next I started taking a look at the signals acquired on Dec 28. Wow! This is where my worst fears about the signal quality were confirmed. The drift-canceled test signal still exhibited typically about 0.25 Hz width and looked like so much noise in the spectrum view.

This observation caused me to pause and rethink the whole issue of what the phase history slope concept could tell me about the frequency of such a crummy signal. I did some heavy thinking and some experiments with noise away from the signal, and finally arrived at the notion that the phase slope is probably telling me the location of the center of mass of the energy in the selected bandwidth. This made me feel a little better because I had enough data to just average down to perhaps 100-200 milliHz uncertainty.

But then I decided that I could perhaps do better by using the following expedient: confine phase slope estimations to time regions wherein the fading seemed relatively less "evil". So I worked my way through the entire data set, manually choosing chunks wherein the fading nulls were not very deep, the signal strength was relatively high, and the structure seemed not overly "complicated". This was done by eyeballing dB plots of the signal amplitude versus time. This effort yielded several dozen individual frequency estimations, each taken on a data chunk of typically a few tens of seconds duration.

I don't know if this approach to winnowing the data is right or wrong but it's what I did, so there! I'd appreciate any comments that readers might have on the probable validity and/or effectiveness of this trick for improving accuracy of the frequency estimations.

While doing this I noted a distinct trend over time and thus chose to make a plot of the measurements rather than merely take their average. The PDF plot shows the result- it looks kind of ugly because of the irregularity of the decent- looking time regions I found, but the trend is readily seen.

The frequency looks pretty steady at about 14.055260910 MHz at first, then starts falling off about 1820 UTC, reaching a minimum about 0.3 Hz lower at 2200 UTC, then appears to rapidly recover to near its initial value by about 2240 UTC.

I have taken the rough average value (14.055260910 MHz) of the early flat portion of the curve as my official FMT entry.

I also calculated and plotted path length variation during the run to see if the size of the change looked plausible, which it does to my untrained eye. I also showed the plot to one of the aeronomers at the observatory, who indicated that the magnitude of the shift was believeable and that the trend was in the right direction. For this calculation I needd to make an assumption about the transmitted frequency; to the extent that the assumption is in error there will be a slope error to the plot. For now I have applied the same value that I have submitted as my FMT entry; once the actual transmitted frequency is released I will correct the plot to incorporate that value.

End of tale.

I will email my soapbox story and a couple of attached PDF files. BTW, the coordinates given above are for the receive antenna used, within perhaps 20 feet of correct. Sorry about the wrong format- I found myself without a convenient method for conversion when making this submission. 73, Dana Whitlow K8YUM
Jerry Flanders W4UK Thanks for putting on the test. Enjoyed the 24 hr format - more relaxed with the actual measurements, plenty of time for calibrations/recalibrations. Used rx, spectrun lab, wwv, pencil and paper. Jerry
Bill Lynch AB2UW Hi All - Excellent idea to run the FMT for 48 hours. It gave me a much better appreciation of how 'bad' an off-the-air radio signal can be for a frequency reference. I used a minimal setup for the FMT. Icom IC-7000 transceiver, SignaLink USB (which has a built-in sound card) and a netbook computer running Spectrum Lab V2.74b12 I tuned the radio to WWV (CW mode, 100Hz bandwidth) and after warmup, adjusted the IC-7000 'REF adjust' (internal frequency reference) until the sidetone was within 1 Hz of 800Hz (the sidetone frequency was set to '800Hz' in the radio's 'CW Pitch' submenu). Then, warmup of everything overnight. Spectrum lab FFT set to 524288 'input size' which give a bin width of 0.0210285 Hz. Tuned to closest WWV or CHU, (15MHz for 14055 kHz FMT signal, 7850 kHz for 7055 kHz FMT signal and 3330 kHz for 3583 kHz FMT signal), then allowed several minutes for FFT calculation, then tuned to FMT signal (and maybe back to WWV or CHU). Then did screen captures of spectru