Path of 500 foot Beverage On Ground. Feedpoint is at upper right.
In the early part of this year I installed a 500 foot long Beverage antenna on the ground (BOG), and I have been pleased with its performance. With this antenna, I have been able to receive WSPR spots from Europe, Hawaii, Australia, and throughout the Continental USA and the Caribbean.
During the late Spring, I installed an Inverted L for use as a transmit antenna. Its vertical element is roughly 53 feet tall and the horizontal tophat extends for 210 feet.
The BOG runs along an axis of 230-250 degrees from its feedpoint to its termination. The Inverted-L runs along an axis of 12-25 degrees from its feedpoint.
I was curious how reception from the Inverted-L would compare with that of the BOG, so on the night of October 17, 2018 I set up two identical receive stations, with one station being fed by the BOG and the other being fed by the Inverted-L, and then I monitored both WSPR and JT9 on both stations from 0000 to 1200 UTC on October 18, 2018 and compared the results using the WSPRNet reports for each station (one station used the callsign W3SZ and the other station used the callsign W3SZ/IL).
I used identical openHPSDR Hermes transceivers for this test. Each Hermes was running OpenHPSDR-PowerSDR mRx PS v 3.4.9 with two instances of WSJTX 1.9, one instance of WSJTX on each Hermes receiving WSPR transmissions and the other instance on each radio receiving JT9 transmissions.
The results were a bit surprising to me, which may just reflect my inexperience. The BOG DID do better than the Inverted-L, as I expected, but the performance difference between the BOG and the Inverted-L, while highly statistically significant, was not as large as I would have expected. I will give a summary of results first, and then show some graphs. This report will be confined to the WSPR results.
On WSPR, the BOG spotted a total of 875 receptions. The Inverted-L spotted a total of 788 receptions during the same time frame. There were 780 simultaneous receptions by the two stations that allowed for direct comparison of received signal strengths between the two antennas. Receptions were obtained of a total of 23 unique callsigns, with one of those 23 callsigns being a one-time-only reception of WB0DBQ at -30 dB by the Inverted-L that was not detected by the BOG. The other 22 unique callsigns were detected by both antennas. There were 90 Receptions that were made by the BOG that were not detected by the Inverted-L, and 12 receptions (including the one of WB0DBQ just mentioned) that were made by the inverted-L but not by the BOG. Most (73/90) of the receptions by the BOG that were missed by the Inverted-L were at distance of greater than 1100 km, whereas most (7/12) of the receptions by the Inverted-L that were missed by the BOG were at distances of less than 600 km. The successful decodes by the BOG that were missed by the Inverted-L were all at signal levels of -24 dB or weaker, while half of the decodes by the Inverted-L that were missed by the BOG were -20 dB or stronger. So it seems, qualitatively, that the BOG did a better job on the weaker, easier to miss signals than did the Inverted-L.
The graphs below will show signal strength (or difference in signal strength) in WSJT units (dB) on the Y axis, and either callsign or azimuth on the X axis. For each X value all data points for that X value will be shown, along with the mean and standard deviation error bars for those values. In addition, to make interpretation easier, the Azimuth, Callsign, and Distance to the received station from FN20ag (W3SZ location) will be shown in red near the upper error bar. Where that text data is not given on the graphs of signal strength vs azimuth, the omission is because two or more stations shared a single azimuth value. A horizontal green line will mark the 0 dB signal level on each graph.
The first graph below compares the signal strengths of simultaneously received signals on the BOG and the Inverted-L by displaying the value (BOG SNR minus Inverted-L SNR) vs Callsign. A positive value (above the green line) indicates that the BOG heard better than the Inverted-L, and a negative value (below the green line) indicates that the Inverted-L heard better than the BOG. As indicated above, the mean Y value for each X value is shown in red, as are the error bars above and below the mean value.
Received Callsigns are listed alphabetically, left-to-right. Nineteen of the 22 stations were on average heard better on the BOG than on the Inverted-L, and 3 stations were heard better on the Inverted-L. You can see that the greatest difference between the two antennas was for AA1A, which was on average received 12 dB better on the BOG than on the Inverted-L. The next-greatest spread between the antennas was for AA3GZ, with the BOG being better by 7 dB for this station, but there were much wider error bars for AA3GZ’s results due to fewer simultaneous data points being acquired. Next were AA8HS and K1BZ, both stations being received about 3 dB better on the BOG. Of the remaining stations, all except for VA3TX, VE7BDQ, and K2BLA were on average better heard on the BOG than on the Inverted-L, although in some cases the difference in received signal strength was very small. 13 of the 22 stations had the error bars completely above the green 0 dB line, and only one, VA3TX, had the error bars completely below the green 0 dB line. A T test of all of the simultaneously received BOG and Inverted-L signal strength results gave a p value of 2 x 10-11 for the comparison. In general, a p value of less than 10-2 is considered to indicate a significant difference, so this results indicates a highly significant difference between the received signal strengths of the BOG and Inverted-L antennas, in favor of the BOG.
The graph below shows the same parameter displayed for the Y values, (BOG SNR minus Inverted-L SNR), and the X axis parameter is azimuth of the received station from W3SZ.
In the graph above, it looks like the BOG is most superior to the Inverted-L at azimuths of less than 90 degrees (specifically, based on the available data, at azimuths of 67-87 degrees), and it does not have an obvious peak in its performance relative to the Inverted-L in its preferred direction of 230-250 degrees. Of course a 500 ft BOG would not be expected to have much directivity at 630M, for its wavelength assuming a velocity factor of 55% would be only (500/0.55 * 12/39.37)/630 = 277/630 = 0.44 wavelengths. So it would be even less directional than the elevated 89m-long 160M (0.56 wavelength) Beverage in this illustration taken from ON4UN’s text where the -3 dB points are at approximately +/- 70 degrees:
And my Inverted-L, if I am interpreting L.B. Cebik’s work correctly, would have minimal if any directivity on 630M; see the plot below for a 50 x 50 Inverted-L on 3.55 MHz. With my Inverted-L dimensions, at 630M my Inverted-L would be expected to be even less directional than the plot for 3.55 MHz in the illustration below, and the azimuth plot of the 3.55 MHz antenna is nearly a perfect circle:
What is really surprising to me is that the station where the biggest advantage is seen, AA1A, is at azimuth 67 degrees, which is almost exactly in the 180 degree null of the BOG, which would be expected to be between 50 and 70 degrees. And we can’t explain this anomaly by invoking directivity of the Inverted-L, because that should have almost no directivity.
I guess other explanations must be in play here; perhaps including the signal’s angle of arrival at each antenna, distortions of one or both antenna’s expected patterns by local terrain and adjacent structures, the unavoidably asymmetric ground system of the Inverted-L (due to adjacent buildings), and other factors.
To sort this out, I thought that it might be helpful to look at the difference in signal strengths between the two antennas for each callsign/azimuth received as a function of time, so I produced the graph below. It shows a small graph of (BOG SNR minus Inverted-L SNR) vs time for each simultaneous reception for each callsign received. I set this up so that the X axis will autoscale for each individual graph to give the best display of datapoints for that graph. As a result, the X axes differ from graph to graph, so stay alert as you peruse this image. You may find it helpful to right-click on the image and then open it in a separate tab so that you can enlarge it to better see the details. A few of the graphs (e.g. VK4YB, K9FD, AA3GZ) where the station was only received over a very short interval, show minutes instead of hours on the X axis:
Unfortunately, this analysis provided me with no additional insights, although there was a suggestion that the difference between the two antennas tended to diminish after approximately 1000 UTC in some cases (e.g N1HO, WD8DAS), as the band “closed” for the day.
Below are, without comment, some graphs of the raw data for each antenna with means and error bars shown, with x axis factors as above being either Callsign or Azimuth from W3SZ.
I will take a look at the JT9 data next, and if it offers any new insights, will report it in similar fashion to the above.
73,
Roger Rehr
W3SZ