WHICH ANTENNA IS THE BEST R. P. Haviland, W4MB I have always made a practice of log- ging the information which comes across in DX QSOs. Typically, this in- dudes antenna type and height, and rig type or power. Some years ago, a question of an- tenna usage prompted me to do a short analysis of the logs, to develop antenna type and height data. Very recently, a similar analysis by ZL1OI, but of U.S. stations, was reported. Also very recently, thoughts of a new antenna system for W4MB prompted an extensive review of log data. The idea was to try to get an answer, or at least an indication of answers, to typi- cal questions, such as: Which antenna is the best? How important is height? How important is power? This is a report of these studies. Relative Antenna Usage The analysis started with a review of usage, of the two studies previously reported, plus the data from the latest logs. The results are summarized in Table 1. Results of the three analyses are very similar, showing exactly the same relative order of usage and sur- prisingly small differences between U.S. and DX stations. Very nearly three-fourths of the stations use beams, and one-fourth use simple an- tennas. The Yagi dominates the beam usage; almost three times as many as its closest competitor, the quad. Verti- cal antennas are somewhat more popular than dipoles. The data at W4MB is primarily for 15-meter contacts, and is markedly influenced by European practices, since a majority of QSOs are with these stations. Comments received indicate that a major factor in the use of verticals is space for installation. This seems to be true for the suburbs, where many of the verticals are at ground level, and for the cities, where antennas are at roof height. Other comments indicate that the quad would be more popular if space were available, and that ready availability of commercial triband Yagis is a factor in their popularity. It will be noted that the difference between any of the three usage fig- ures and the average of the three is small, at most just over 3 per cent. This difference may not be significant. For example, for the W4MB data, 3 per cent would amount to a ten- to fif- teen-station change. However, there is some reason to believe that some of the indicated difference is real. The last data are for a period of relatively high solar activity, and reflects a change in QSO area. For example, the recent data include more USSR stations, which are often users of quads. As will be seen, the improved conditions can be responsible for the increased percentage of simple anten- nas. No reason for the decrease in the use of "other" antenna types has appeared; these others include rhom- bics, long wires, log periodics, and SO on. Signal Strength as a Measure of Performance The only performance measure available from the logs at W4MB is the signal-strength report given to the DX station. For the last five years, I have taken some effort to make these reports meaningful. The report is the S-meter reading on my TS-820, with enough attention being given to en- sure that it was the maximum, inter- ference-free signal. The meter calibra- tion of the 820 was checked just after purchase and found to be almost ex- actly 5 dB per S-unit. The relative and absolute values have been checked at intervals since, most often by check- ing the crystal calibrator signal, but also by using a signal generator. Sig- nal-strength-indication stability seems to be excellent. In most radio circuits, signal strength is a statistical variant follow- ing a Rayleigh distribution.' How- ever, this analysis is dealing with many circuits and a combination of factors, including transmitter power, antenna design, and solar activity. Thus, it seemed that the signal strengths should follow the Gaussian distribution fairly closely.t Also, it seemed that the distribution data could be used to check the results for bias, since a systematic change would change the distribution in some respect . Accordingly, a section of the log was examined for signal-strength probability. The results are shown in Fig. 1, the curve indicated by the Xs. On this type of graph paper, a Gaus- sian, or "normal," random distribu- tion plots as a straight line. Since the curve is not straight, the distribution departs from Gaussian. However, it should be noted that there are upper limits, due to legal restrictions on radi- ated power, and practical limits on antenna height. These have the effect of bending the upper part of the curve downward, as observed. Neglecting 'Rayleigh distribution -- A mathematical state- ment of a natural distribution of random variables. Gaussian distribution -- A distribution of ran- dom variables comparable to that found in nature, characterized by a symmetrical and continuous distribution decreasing gradually to zero on either side of the most probable value. this effect, the distribution below about S7 is very nearly Gaussian. Also shown are points for two other types of data, for contacts where height was given, and where power was given. (In QSOs, often only power or type of transmitter is given). These samples fall very close to the upper curve. Overall, the data indi- cates that the height and power data are associated with about a one S-unit stronger signal, as compared with an average contact. The difference be- comes smaller as signal strength in- creases. Inspection of the raw data showed a tendency for the QSOs where power and height were given to be longer than QSOs where they were not. They tended to be more "rag chews" than "hello-goodbye" contacts. Part- ly, these occurred under conditions of better signal or less QRM. Also, the longer period gave a greater chance of observing a short-term increase in signal strength. Thus, the observed difference in the three sets of samples seems to be an observable fact, rather than some fault of the method of analysis. Performance of Simple Antennas The performance of simple anten- nas -- dipoles and verticals -- was considered first. For each value of sig- nal strength, the percentage of sta- tions using these antennas was deter- mined. The results are shown in Fig. 2. The performance character is strik- ing, and the trend is definite. Simple antennas do get signals through, but the signals tend to be on the low side; markedly lower than the signals from beams. The measured data can be approxi- mated reasonably well by a simple re- lation: for each unit increase of signal strength, there will be about a five per cent reduction in the number of sig- nals from dipoles, and about another five per cent reduction in the signals from verticals; there will be a corre- sponding increase in the number of signals coming from beams. While these antennas tend to pro- duce weaker signals, the performance is not really that bad. Verticals, for ex ample, account for about 18 per cent of all antennas. Comparing this value with the curve in Fig. 2 shows that the average expected signal from a verti- cal is nearly S6. For a dipole the usage is 13 per cent, which intersects the dipole curve at (18 + 13) or 31 per cent, again nearly S6. These are perfectly respectable signals, some 25 dB above normal noise levels; not broadcast quality, but excellent com- munications quality. To the limits of analysis accuracy, verticals and di- poles, as used, give identical results. Performance of Beam Antennas For this analysis, the amount of data restricted beam consideration to only two types, Yagis and quads. The percentage of these at each signal level is shown in Fig. 3. This seems to indicate a striking, indeed startling, re- suit: as a family, Yagis seem to out- perform quads. Over the range of S4 to S9 signals, those from quads are nearly constant in percentage, while over the same range, those from Yagis increase in percentage, by about 5 per cent use for each unit of signal strength. Obviously, this is an important find- ing. If it is really true, it could settle the long arguments of quad os Yagi. Ac- cordingly, some additional tests were made. The first of these was to re- ex- amine the source data for antenna size. Since the usual quad has two elements, any antenna with more than two elements was considered a "large quad." Also, since a two-ele- ment quad is usually considered to be equal to a three-element Yagi, de- signs with four or more elements were considered "large Yagis." While some of the antenna notes recorded the specific design, TA33, Th-6, and so on, there were not enough recorded as these or as mono-banders to per- mit detailed evaluation. The results of this size analysis is shown in Fig. 4. Here, the percent- age of the smaller Yagis is plotted first. The trend to increasing percent- age with increasing signal still ap- pears, but the larger Yagis seem to be increasing at a faster rate. This sug- gests that the apparent better per- formance of Yagis as compared with quads is partly due to larger antennas. A different way of looking at the data is shown in Fig. 5. This is de veloped from the number of antennas of a given type producing a given sig- nal, plotted as a percentage of the total number of that type. It seems clear that there are two main group- ings, one for the simple antennas -- verticals and dipoles, and the other for the beams -- the quads and Yagis . On the average, the beam group produces about 11/2 S-units, about 71/2 dB greater signal level than do the simple antennas. Interestingly, this is very nearly the theoretical gain of three-element Yagis and two-ele- ment quads. Also interestingly, the beam group seems to produce a more consistent signal: at low signal levels the beams are about two S-units bet- ter than the simple antennas. Data for the big antennas, the four or more element Yagis and three or more element quads, are also shown. The number of such antennas is rather small, so the points are rather scattered. It seems to indicate about one-half S-unit improvement in signal as compared with the smaller beams. Again, this is interesting in that it rep- resents about the expected gain in- crease. Plotted in this way, the type/signal- strength data indicates that there is no difference between quads and Yagis, or between dipoles and verticals. In- stead, the data suggests strongly that the significant factor is the amount of gain, rather than the type of antenna which produces the gain, at least for the types surveyed. However, before considering this relative comparison, factors common to all antenna use should be looked at. Effect of Antenna Height The early W4MB study included evaluation of antenna height. This was done again for this second study. Results are tabulated in Table 2, and plotted in distribution form in Fig. 6. A small difference, up to about 10 per cent deviation from the average, is found between the two data peri- ods. This may be real, or it may be due to "sampling noise." There are some differences between the two periods. The second covers a period of improved propagation, which would tend to increase the percentage of marginal signal contacts, as ob- served. The second period also in- cludes a larger fraction of 10-meter contacts, where height is known to be less important. More important than the matter of usage is the effect of antenna height. To study this, signal reports were tab- ulated for height blocks of 0-15 feet, 15-30 feet, and up to 120-240 feet. The reason for this grouping choice was the sometimes-used approxima- tion that doubling the height will im- prove the signal by one-half to one S- unit. The results of this analysis are shown in Fig. 7. While five height-groups are plot- ted, only three curves are shown. The data for the 0-15 height, shown by Xs, follows a Gaussian distribution al- most exactly. Data for the groups. 15-30 feet and 30-60 feet follow the distribution fairly closely, but with a different slope, and with a small in- crease in average signal strength. The change in slope indicates that the sig- nal for these groups is more consistent than that for the low group. The dope change is also apparent for the 60-120 foot group, which shows a further increase in average signal strength . The 120-240 foot group seems to fall along a line which is parallel to, but above, the 0-15 foot group. A check of the source data did indicate that this high antenna group included many simple antennas mounted on apartment roofs. An attempt was made to evaluate the effect of anten- na type, but there were not enough such contacts to be meaningful. Overall, it appears that high anten- nas behave according to reputation. There is a definite increase in average signal strength, in the range of 1-2 S- units. There is also good indication of greater signal consistency; a high an- tenna will make contacts, where a low one may not. The first W4MB study did not eval- uate power, but an evaluation was done for this study. These data are tabulated in Table 3, together with the ZL101 data for comparison. The influence of the one-package trans- ceiver, of about 180 watts input, seems clear. Comparison of the two sets indicates a greater percentage of high-power stations in the U.S.A., certainly in line with our reputation. There seems to be increasing use of powers in the range of 1 to 30 watts. It is most pronounced on 10 meters, partly due to new commercial trans- ceivers of this class, partly due to CB conversion, and partly due to power/band restrictions of some license classes in some countries. However, QRP operation is not re- stricted to 10 meters; growth seems tc be occurring on all bands. The results of the signal strength/ power study are shown in Fig. 8. With one exception, the trend is deal signals, and more consistent signals. The average increase is almost exactly in accord with power, i.e., two S- units for 10 dB increase. The exception is for the power class 10-30 watts, which appears to be bet- ter than the 30-100 watt class. This may be real -- several QRP operators have indicated that they took special care with the antenna installation. Many more samples would be needed to separate the effects of power and antenna size. Some time ago, W4MB ran a two- month check of power benefit by operating at 180 W PEP rather than the usual 1500-1600 watt level. The number of contacts made in a month did not change greatly, but a differ- ence in operating practice was neces- sary to do this. Whereas, with the linear, a short CQ had a high proba bility of one or more replies, the CQ replies dropped way down when oper- ating "barefoot," and it was necessary to start replying to DX stations CQs to keep the number of contacts up. Power does make a difference. Re-evaluation of Beam Antenna Type Several methods of checking the possible difference between quads and Yagi beams were considered. The two finally adopted are covered here. If one type of beam is truly better, it should perform better under poor conditions. This would be reflected as s an increase in the percentage of con- tacts using that type of antenna. The measure of conditions adopted was the A index, the second numeri- cal value transmitted by WWV. It was found that a reasonable sample size could be obtained by selecting con- tacts where A was ten or less for good conditions, and A of fifteen or greater for poor conditions. Contacts for in- termediate values, or for days when the A index was not recorded, were ignored. For comparison, the per- centage of antenna use for all values of A was also tabulated. The results are shown in Table 4. First, it is noted that the variations between the three sets of data is small. The largest is for dipoles, changing by 4.6 per cent from the all- A-value column to the poor column. For small Yagis, the percentages show a small increase from poor to good conditions, but almost exactly a compensating decrease for large Yagis. For quads, there is a decrease of 4.2 per cent between good and poor conditions; the small increase in large quads does not compensate, so there is a net decrease in the number of quad contacts under poor condi- tions. A check of the logs was made to see if this decrease were real. One factor noted was the almost complete absence of USSR contacts under poor conditions. Since a high percentage of these stations use quads, the ob- served change could be real. How- ever, it must be remembered that one percentage point is only two contacts for the poor column. Accordingly, the sampling noise is large. Because of this, a second method of checking was sought. The logs were again reviewed. It was found that a fair amount of QSO data gave height, power, and type of antenna. Since height and power are shown to be important, this data was checked for two groups; one included all an- tenna types, but only those stations running linears or over 400 watts (assumed to be the lower limit of lin- ears). The second group included on- ly Yagis, for stations operating at the same power level. The results of this analysis are shown in Fig. 9. The curve shown is copied from Fig. 1, the contacts giv- ing height, the other remaining vari- able. The o-points are for linear- power level contacts in general, com- pared with the x-points for contacts using Yagis. As seen, there is essen- tially no difference in signal perform- ance, and no difference when com- pared with contacts stating height. The number of contacts used in this compilation is not great, sixty-five for the all-antenna group, and thirty-six for the Yagi group. As a result, there is appreciable chance for error. How- ever, the indication that there is no real difference seems clear. Comparing the four tests of beam type makes it appear that there is little or no difference between Yagis and quads. The important factor, as noted before, seems to be antenna gain. Summary and Conclusions While there is some possibility of error due to sampling noise, this sta- tistical analysis indicates the following: Beam antennas produce better sig- nals than simple antennas on the average, by almost exactly the anten- na gain. For the common beams this amounts to about two S-units. There is good indication that the beams pro- duce a more consistent signal. Higher antennas produce better signals, by approximately one-half S- unit for doubled antenna height. An- tennas above thirty feet appear to produce somewhat more consistent signals than lower ones. Increasing the power improves sig- nal strength, almost exactly a 10 dB increase of signal for each 10 dB in- crease of power. Higher power sig- nals also appear to be appreciably more consistent than low-power ones. Although some tests appear to indi- cate that the Yagi beam produces stronger signals than the quad beam, other tests indicate absolutely no dif- ference. It appears that the gain of the particular antenna is much more im- portant than the type, and it may be the only significant factor. The most important single variable in received signal strength is propaga- tion variability, typically a range of 35 dB over the 5-95 per cent limits. (The range of signals in this log period is from SB to S9 + 40 dB, a total range of 85 dB). The range is so great that a combination of legal power limit, and maximum practical antenna height and gain cannot compensate for it. However, the super station is appre ciably more likely to get some signal through under poor condition. There is one overall conclusion: The single most important factor in getting good signal reports is choice of frequency band and time, as needed to catch optimum conditions. Using this variable, any station can be a good DX station. Notes on Extending the Study The solution to reducing the sam- pling noise mentioned several times is to increase the sample size -- the number of contacts studied. For ex- ample, a tenfold increase in the num- ber of samples will reduce the noise to about one-third. This may not be easy. The data segment of the W4MB logs is limited by rig and power changes at one end, and antenna changes at the other. It seemed best to avoid these added variables. Even so, some 1800 contacts were ex- amined -- not all QSOs give useful data. Since an improvement in analysis will require looking at some ten to twenty thousand QSOs, an extended study might make a good club project -- especially if the club has a comput- er available. The extra number of QSOs would allow some additional studies, such as variations in usage between continents or even countries. Just for example, it appears that UA antennas are much more likely to be quads, and DLs are much more likely to use linear amplifiers. If this club approach is used, some checks of the data will be needed. Dif- ferent operators have different report- ing practices, and a "station correc- tion factor" may be required. Devel- oping this from the data will take extra work . It is probably best to avoid analysis of contest and pile-up data. Contest reports are too stylized (5 x 9), and pile-up QSOs are apt to be influenced by the well known "DX-report" factor. Such an extended study would be interesting. Well done, it should pro- vide a definitive answer as to whether it is gain that is important, or whether antenna type also enters the picture. And perhaps it could look at the less common antenna types for usage and performance. Oh yes: W4MB's new antenna? It's a quad, except on 10 meters, where it is a Quagi. Why? It seemed a good idea at the time. It still seems so.