Back to Bulletins index | Back to INAG-62 Contents page |
This article is an update on notes that were circulated to several people about two years ago.
Preliminary Comment: I am not an antenna expert and wish to avoid pretenses to that effect, but I have worked actively with a number of experts (mainly Dick Grubb, Walt Plywaski, Dick Kressman, and their own colleagues in this subject area) and I hope that what I have learned from them is not misrepresented here. I have tried to make some simple tests to reveal operational properties of antennas with which I have worked. These notes are intended to summarize these results for your information and critical comment. Some ideas for further work are included. Note added 17 Oct 1997: Having recently obtained some data from the San Diego dynasonde, I added its curve to Figure 5 together with a paragraph of discussion. For INAG, I have omitted personal names from the last section: 'Suggestions for Further Work' (you know who you are). Following the suggestions made, now two years ago when the document was distributed in hard copy to about 20 colleagues, no one has done anything except for some trials of the DATONG active antenna. I analyzed those results, with moderately discouraging conclusions, in a separate set of notes from which one paragraph is quoted later.
In contrast to most other radars, the ionosonde is dependent on "echoes of opportunity" by critical reflection from ionospheric density contours; these are often tilted and irregular, much like an ocean surface. We desire broad antenna patterns, so as to sample as much of these structures as geometry and refraction will permit.
Ionosondes require antennas for transmission (Tx) and reception (Rx). Since spaced receiving arrays are essential for digital ionosondes, the question of using one and the same antenna for Tx and Rx does not arise (but such schemes are ill advised anyway, because of the nature of the MF/HF noise environment). The data to be shown here compare (at present) the lumped Tx, Rx performance of several systems. Since the available Tx choices differ more significantly than the Rx choices, we begin with a discussion of ionosonde transmitting antennas.
Difficulties arise in Tx design for vertical incidence ionospheric sounding in the MF/HF bands. The difficulties are greater for the digital research ionosondes of interest here, since we want quantifiable performance and must minimize two obvious, likely, and undesirable properties of simple antenna designs. With mention of their consequences for research ionosondes, they are:
a) Uneven radiationpattern variations with frequency, which mask or bias ionospheric information bearing on
b) Uneven powergain variations with frequency (in consequence of (a), or because of impedance mismatch, ground losses, etc.), which mask or bias ionospheric effects bearing on:
Such measurements are the 'bread and butter' of the digital research ionosonde, and the choices of Tx and Rx are clearly critical to its success.
There are also some practical and engineering requirements that deserve attention:
c) The Tx must accommodate site constraints, aircraft (height) restrictions, climate hazards, and budget limits.
d) The Tx should minimize, and not provoke unnecessarily, RFI problems to civil and military telecommunications, nor to other experimental systems; in practice, this entails:
Additional considerations arise in considering the Rx options, of which the most important (not to be considered further here, however) is the need for accurate phase calibration.
There are two categories of practical Tx design: those characterized by a set of resonant (usually halfwave) lengths, and notional nonresonant structures. The latter (§2.1 and §2.2 below) manage to maintain some radiation at wavelengths small compared to the height and scale of the antenna ... but at the price of poor performance everywhere. They may be stabilized by a terminating resistor which consumes a significant part of the RF power.
The following are the Tx options known to me.
This is a cheap, dependable design, easy to erect on a single mast, long the standard for analog ionosondes at least in the USA program. I have used it (under protest) in an Equatorial dynasonde campaign, and therefore have some comparative data.
This is a classic design, often used for long haul broad band communications (then mounted more or less horizontally). I have no experience with this design. Mounted vertically, it might perform somewhat like the delta (which in fact is sometimes called a halfrhombic). It was long favored by German workers, and is still the choice of the Lowell digisonde group. It is suspended from a single mast. [A large example is installed with the digisonde at Ramfjordmoen, near the dynasonde's LPA, §2.5 below; this affords an excellent opportunity for objective comparisons].
This requires two supporting masts of moderate height, perhaps more masts if made long enough to perform well below about 2 MHz. A quasiresonant behavior is maintained by passive RC 'traps' at intervals within the dipole. It is used at USU's Bear Lake observatory and earlier, I believe, in their work at Siple Station Antarctica. I have very little experience with data from this antenna, but what I do have, is discussed below.
Log Periodic Antennas comprise a class which aims at frequency independence by the device of selfsimilarity in 'frequency periods' defined by a scaling factor < 1. LPAs can be directed toward the ground, or upward. Provided the ground is electrically 'good', performance is generally better when the antenna is directed downward, so that each radiating element is at a constant height above its ground image, relative to its resonant length.
This design resembles a plane triangle usually suspended between two masts. For the "Kinesonde in France" experiments, we used a single central mast instead, and declined the dipole elements toward ground at about 30 degrees below horizontal. We made extensive scale modeling and computer simulations, which suggested broad (roughly circular) patterns in both E and H planes, about 3 dB down at 60º zenith angle, and good frequencyindependence within its 2 to 20 MHz design range. Below 2 MHz radiated power decreased smoothly at about 14 dB/octave. There was some evidence (in the scale model work) of selfinterference toward the upper end of the frequency range. I do not have comparative data from this antenna at present, but the design is of some interest because commercial versions are available.
This is the Troms antenna, with two planes. It is suspended from 4 masts at the corners of a square, within ropes that descend from the top of the masts to the center of the square at the ground. Two opposite, or all 4 triangular planes thus defined, may be filled with the zigzag wire structure. We have some, but limited, experience with 4 plane structures. Except for certain special applications (see below), there seems to be no particular advantage to the 4-plane design. A detailed design drawing similar to the structure at Ramfjordmoen appears in Figure 1. This shows an antenna for the range 2.186 30.0 MHz with = 0.92 suspended among masts of height 32m. (The Tromsø antenna, as built, may differ somewhat from these values). Unfortunately, figure 1 is the best small size figure that could be produced for publication. A larger, more readable version is available from the author for people seriously interested in constructing the antenna. For others, here is a description of the construction.
The antenna is strung on ropes attached (by pulleys for hoisting, repair and tensioning) to the tops of 4 poles 106' (21.3m) in height; the ropes meet at the center of the square defined by the poles, 450' (137.16m) on a side. Starting at two diagonally-opposite pole-top rope points, #14 AWG copperweld wire is punched through the ropes and wrapped tight. The wires then run in a clockwise direction to the next ropes to measured points, where a hairpin bend in the wires is again punched through the ropes and wrapped tight. The wires then run back again to the first ropes, zigzag fashion, and to the next (down) measured points. The measured points are expressed as the distance from the center vertex of this structure (near the ground) up along each rope. The first measured tie point (where the wires are first attached) is at 330.0' (100.58m) from the vertex. Subsequent tie points are at 0.92 of the preceding distance. For example, the 2nd tie points (on the other ropes clockwise), are at 303.6' (92.54m), and the 3rd tie points (back on the first ropes) are at 279.3' (85.13m). The whole thing is truncated at the 49th tie points at 6.03' (1.84m) on the first ropes, near where the zag length is half-wave at 25 or 30 MHz. We never use it up there.
In each of the two opposite tetrahedron faces thus filled by zigzags, a 3-wire feedline extends from near the ground out to the highest wires. The center wires of the feed lines go up the center to the center of the highest wires. The two other wires diverge linearly, to end at 15' (4.57m) each side of the center feed wire on the highest radiating wire. All wires are #14 copperweld. All 3 feed wires in each radiating plane are wrapped and soldered to the radiating wires where they cross.
The antenna is fed from a big 50-ohm coax line into a 50-ohm unbalanced-to-balanced 300-ohm transformer.
All of the distances cited are with the antenna tensioned (1000 lbs. in the diagonal ropes, plus some small tensioning ropes at the tie points). The ropes were chosen to have 5% maximum extension under tension.
The ground is good for RF, as far as I know, but there is no metallic ground plane. It is almost flat and horizontal, except for a 1.5m rise just outside the square on one side. If there is a standard way to quantify good or fair ground, it might be useful to compare with perfect ground, to learn what a metallic ground screen would accomplish.
The masts are the only expensive part of the construction. And metal masts, even if they are the most readily available and traditional for antenna work, are not really desirable: they tend to radiate parasitic energy, and they do so with vertical polarisation, the worst situation for RFI to nearby listeners. Also, they may damage the antenna performance. It needs to be modelled. A much better choice, if tall 'tree-masts' are unavailable or too expensive, are laminated wooden beams, as factory-made for auditorium rafters and similar applications. With (say) 25x25cm cross-section, and rope guys at top and mid-height; they serve very well (and have been in use at Tromsø for 17 years).
Another point to notice is that the antenna can be scaled to other low-frequency cuttoffs, which will change the required mast-height and ground area. The LPA might (with advance planning of its azimuth) serve for medium-distance oblique-incidence sounding between two similar ionosondes. For this purpose, one of the two radiating planes could be fed unbalanced to ground. This idea has not been modelled, and is conjectural; we will, however, do the modelling.
It is worth mentioning one option available with this design which to my knowledge has not been attempted, but which may well argue its consideration in certain situations (e.g., the prospective pair of dynasondes at Tromsø and Svalbard). For bistatic sounding between instruments spaced at distances of 500 to 1500 km, some power gain is necessary at the appropriate azimuth and zenith angle. Provided the LPA orientation is planned for this, a selected single plane of the structure may be driven unbalanced against ground to achieve this. The 4plane structure provides, in addition to as many azimuth options, the feature that two planes might be reserved for vertical incidence work.
We can consider only Rx designs suitable in small arrays of 3 to 8 antennas within an area comparable to a favored radio wavelength, usually 100m or less. Radiation isolation between the antennas should be kept large, > 40 dB; this argues antennas small compared to their spacing. Four choices are currently of interest.
This was the choice advocated originally for the dynasonde. They are in use at the USU Bear Lake site, but were replaced at Tromsø in 1985 because of poor performance at frequencies below about 3 MHz.
These have not yet been tried, but are of interest because one version is commercially available, and because they may overcome the limitations of the small passive dipole.
These are used at Tromsø at present in a 'fat' version in which the elements are made of thick (15cm) tubing. Similar dipoles were used with the Huancayo dynasonde, but made of 5cm 'ladder wire'.
These are favored by the Lowell Digisondes.
We are in position to make some comparisons among the following antenna systems, in each case part of a NOAA dynasonde installation:
Some idea of the performance of such systems, in the absence of absolute power and pattern calibrations, is given by analysis of ionospheric measurements made using them. I compare here some results using programs which read any number of ionosonde recordings, and which form the arithmetic means and standard deviations of selected echo variables in uniform (0.1 MHz) bins of radio frequency. Of particular interest, and generality, are the frequency profiles of echodecibel powers, also further sorted by echolocations. Although a wide variety of natural processes affect these parameters, we can reasonably expect their large sample means to reveal some properties of the antenna systems in use. Of course, such results do not distinguish properties of the transmitting and receiving systems. The results also depend on the sample size, and (in part) on whether system calibrations exist. Comments about the role of such factors will be made in context.
Echo amplitude is computed as the arithmetic mean of the modulus of the complex amplitude components, among all receivingantenna and deltaf 'channels' of a pulse set In our work at present, there are usually 8 such channels. For the work described here, the DSP system selects one complex amplitude for each channel nearest the mean echo peak. The linear amplitudes (A, initially expressed simply in ADC counts) are expanded according to the dynasonde decibel step attenuator ("AGC") settings, so that a very wide dynamic range is available in principle. The decibel power equivalent is . These steps are common to all of the data discussed here. However, a difference between the Huancayo data and the other examples must be noted. In the 1982 Huancayo case, the dynasondeoriginal ADCs were in use; a raw ADC count in the range 2048 was a voltage in the range 8 volts. A recent hardware upgrade applicable to the Troms and USU Bear Lake data now expresses echo complex amplitude components in the range 5 volts. Thus 4.08 dB must be subtracted from the Huancayo decibel powers for comparison with the other data.
The 8 measuring channels are usually divided equally between the dynasonde's two 'identical' receivers, and often unequally among the several receiving antennas. Practices in this respect vary with the experiment, the experimenters' preferences, site constraints, etc. Although the individual receiving antennas could be studied, in the present work a single amplitude is first obtained as the average among the 8 channels.
To form the mean and standard deviation of echo decibel powers shown below, each echo was taken in sequence and assigned to its radiofrequency bin (0.1 MHz equal intervals). The mean and variance of the linear amplitude was updated in that bin by an algorithm designed for this purpose. The decibel equivalent of the mean and standard deviation were then obtained after all echoes were subsumed in this way. In another analysis routine, the echoes were further sorted into bins according to their zenith and azimuth angles of arrival. The mean, sigma variations of dB(f), without regard to echolocation, are shown first.
Figure 2 contains results from 243 Bmode recordings during the May 1994 UKDYNA campaign, and (analyzed separately) 1304 recordings from January 1993. Sigma 'error bars' are shown only for the May data. The two months differ significantly in solar illumination and consequent Dregion absorption. Between 1.8 MHz and 5 MHz (May data), or up to 7 MHz (January data) the mean echo power is constant 2.5 dB, oscillating distinctly with a 'period' of 0.8. This oscillation is the logperiodicity designed into the antenna. The apparent performance degradations above 5.5 MHz (May) and 7 MHz (January) may be identified with two causes:
(a) Insufficient sampling as the respective Fregion penetration frequencies are reached. This is clear from the two lower curves, which show the (log) echo counts. Clearly, consistent mean powers and smooth frequency variations are obtained when a few100's of echoes are available in each bin; this condition is satisfied for all of the Jan93 bins (1,294,623 echoes subsumed, in all), but not uniformly so above 5.8 MHz in May94.
(b) Range dilation. Echo amplitude A decreases linearly with echo range; an echo from 600 km group range (R') suffers 15.6 dB more attenuation than an echo from 100 km for this reason. On average, above about 5.0 MHz in May, and above about 7 MHz in January, R' and dR'/df are large and increasing (with f).
Since fxF2 seldom exceeds about 8 MHz at Tromsø, it is difficult to appraise the system power performance at greater frequencies.
The January mean values exceed the May values by about 5 dB near 2 MHz, the difference decreasing to about zero by 5 MHz. This is qualitatively as expected by greater Dregion absorption in May.
Below 1.8 MHz (May data only; a starting frequency of 1.7 MHz was used in January), the decibel amplitudes decrease by 18 dB per octave, as will be shown by a function (and its curve) later, in Figure 6. A decrement of 14 dB per octave below cutoff was obtained in modeling studies (1971) of the Log Periodic Dipole Antenna (§2.4). The decrement by the 2plane antenna may well be less than that calculated for the LPDA, but the present data contain a contribution to this decrement from Dregion absorption. The agreement with modeling therefore seems more than qualitatively satisfactory. It is also worth emphasizing that the LPA remains effective for data acquisition over a substantial frequency range below its design cutoff.
The decibel standard deviations are remarkably constant with frequency, about 12 dB, and as we shall see, they do not differ much among the other sites and antennas studied. The largest amplitude fluctuations are probably caused by geometric focusing (and, to a lesser extent, defocusing) by ionospheric reflectionlevel curvatures, so it is not surprising that little variation of sigma occurs with frequency, antenna design, and observing site.
This is a much smaller sample of data (10 recordings); below 2 MHz and above 7.8 MHz the dB variations are not meaningful for this reason. The effect of R' dilation above about 7.6 MHz is also evident. The 'spikes' of large count at 4 discrete frequencies near 3 MHz arises from the inclusion of two multifixed frequency "Kinesonde Mode" recordings in the sample. It may be noted that the mean, sigma in 3 of these frequency bins are not greatly different from their Bmode neighbors, by the factor 10 increase of echo count.
These are daytime recordings, and part of the general decrement of echo power with radio frequency < 5 MHz is the result of Dregion absorption. However, the irregular fluctuations with frequency, and in particular the odd behavior near 3 MHz, are likely to be characteristic of the antenna system. I do not know the radio frequencies corresponding to the resonant LC traps in the TWD, so cannot associate them with features in the figure.
Over most of the graph, the data count is adequate to define the mean and sigma. The recordings span the transition from day to night, so Dregion absorption again contributes to the decreasing power below, say, 6 MHz. However, the broad scale shape of the curve is probably characteristic of the Delta antenna. Isolated peak counts near 2.9, 4.5 and 6.8 MHz result from multifixedfrequency recordings in Kmode. Count minima at 4.8, 6.15 and 9.7 MHz are harder to explain. The latter two minima are associated with wider decibel peaks, but no decibel feature is seen at 4.8 MHz. We shall see below that an explanation is perhaps found in the antenna radiation pattern.
The dimensions of this "small delta" are not definitely known to me, but a characteristic scale is probably about 25m in height and base length; if 50m approximates its natural "resonance" wavelength, (although the delta is intended not to be a resonant structure), this may explain its general peak of performance near 6 MHz, as well as the sharp peak near there.
Recall from the discussion above, that 4.08 dB should be subtracted from the entire Huancayo graph of Figure 4, for consistency with Figures 2 and 3.
This comparison is made explicit in Figure 5, where all 3 systems are represented on the same decibel scale and without their sigma data. Since the transmitter power is about equal among these systems, and the receiving antennas at Tromsø and Huancayo are fairly similar, the LPA vs Delta comparison is more or less direct. However, a further 3 dB might be subtracted from the Huancayo curve since that system was laid out so as to put all of the radiated energy into the (equatorial linear) Omode polarization. The other systems enable O and X (circular) polarizations, and the power is divided about equally between them.
To compare the USU system with the others, we need some idea of the role played by the 30 MHz receiving dipoles at Bear Lake. In July 1985, while similar 30 MHz receiving dipoles were still in service at Tromsø, a prototype of the present fat dipole was constructed. Here are the decibel differences for the same ionospheric echoes received by each type:
Radio Frequency | 0.7 | 1.0 | 1.6 | 2.0 | 2.6 | 3.0 | 3.6 | 4.6 | MHz |
FatD30MHz dipole | 15 | 13 | 10 | 10 | 10 | 10 | 10 | 10 | dB |
With these points in mind, we may conclude from Figure 5 that the TWD performs somewhat better (in terms of received echo power only) than the Delta near and below 2 MHz, but both are at least 15 dB inferior to the LPA. The TWD and Delta then become roughly comparable toward higher frequencies, although the TWD is less smooth. By about 6 MHz, all three antennas provide comparable echo power.
There are too few echoes at Tromsø and Bear Lake at higher frequencies to permit good comparisons. However, because of its self-similarity up through 20 MHz, there are few reasons to expect that the LPA performance changes markedly, whereas the Delta almost surely continues the declining performance already evident above 10 MHz. If the Delta were made larger to improve its performance at low frequencies, the highfrequency rolloff would commence proportionately lower too. Presumably, the bandwidth of the TWD can be extended upward with additional LC traps, but perhaps at the price of further degradation of its performance at low frequencies.
Figure 5 now (Oct 1997) includes results from the San Diego dynasonde. Unfortunately, I do not know the details of its Tx antenna, but it was probably a delta built to larger scale than the Huancayo delta. The curve is about 10 dB above the Huancayo curve, which can probably be attributed in part to the San Diego receiving antennas: they were "fat dipoles" built to conform rather closely to the Tromso Rx antennas. The table above compares these dipoles to the USU-type short dipoles, not to the ladder wire Huancayo dipoles of length similar to Tromso and San Diego. It is very likely that if a delta was used, it included 'broadbanding' improvements (a diverging 3-wire structure) which were not used at Huancayo. These factors probably account for the 10 dB difference. The San Diego curve includes many more ionograms that for Huancayo, accounting for its smoother variation.
Results in the preceding paragraphs are "allsky" averages. It is of interest to know how these systems differ in their spatial directivity. Where we see variations of performance with radio frequency, it is possible that variations of radiation pattern are occurring as well. The desired information is therefore threedimensional (decibel dependence on azimuth sector (AZ), zenith zone (ZN) and radio frequency, f), which considerably aggravates the problems of binning sufficient data and of graphical representation. (Actually, the methods used here do not require largememory arrays, and the concise results can be updated with new data as they become available, utterly without limit. But the methods are at present 'datastarved').
Instead of pursuing the full 3dimensional problem, I have therefore elected to bin and display two 2dimensional representations: AZ(f), averaged over all zenith angles; and ZN(f), averaged over all azimuths. The "SKYSORT" analysis routine permits choice of the bin resolution in each dimension. It would be desirable to use at least 8 azimuth sectors, and perhaps the same number of zenith zones. There are not enough data (except at Tromsø) to support this at present, so for the sake of uniformity just 4 AZ sectors (90°) and 6 ZN zones (10°, all echoes >60° lumped into the outer zone) have been specified for each site. The count of echoes occurring in each bin is subject to a number of geophysical factors (differing among the 3 sites presently available), in addition to antenna effects, but in most cases the geophysical effects should have little influence on the mean echo power.
Only the May 1994 data were analyzed for the present purpose; the January 1993 data (included in Figures 2, 5) were obtained using only 4 of the available 6 antennas at Tromsø, and as a result they suffer excessive echolocation aliasing. In Figure 6 the mean dB(f) is shown for the four cardinal azimuth sectors, superimposed. Except at a few frequencies, there is no significant dependence on azimuth. There is a substantial dependence on zenith angle, as shown in Figure 7. The logperiodicity of the LPA remains evident in both figures, and there is no evident tendency for the curves to converge or diverge in concert with the ± 2.5 dB oscillation identified with this design parameter.
The zenithangle variation is fairly well described by a cos12 (ZN) law (e.g., 8 dB at 30°, 14 dB at 40°): a cos3 amplitude law is often assumed for a horizontal dipole near ground, so cos6 applies to power, and this is squared again if both Tx and Rx behave as dipoles.
The dynasonde receiving array at Tromsø is very near the eastern edge of a large 2.75 MHz dipole array of the 'PRE' experiment; in fact, the dynasonde 'W' antenna is literally beneath the outer dipole elements of PRE. (This very undesirable situation is a result of crowded antenna siting constraints there). The effect of this arrangement is not known apriori, but we have found in other work that echo amplitudes from the W antenna are lower, and some phase shift is introduced near 2.75 MHz. It is likely that features in Figures 6, 7 between 2 and 3 MHz give further evidence of these coupling effects.
It should be noted that even the most extreme zenith angles (i.e., 60) are represented in Figure 7, throughout the frequency range. Some workers might question the usefulness of this broad skyview; I am confident, however, that this is a useful feature: the most compelling reason arises in the estimation of vector velocities, where the horizontal components must be determined by their projection on the lineofsight Doppler measurement. The more remote echoes contribute the most definitive information in this way.
Although the point is peripheral to our present purpose, it is also worth noting that the lowest amplitudes in Figure 7 are systematically at the largest zenith angles. This is quite compelling evidence that the 6ANT array scheme, together with the full 'UNIPHASE' analysis developed for it, accomplishes its intended echolocation antialiasing purpose. As mentioned above, I could not use the large Jan93 database for this analysis because only the outer 4 antennas of the 6ANT array were used in that campaign. When the present SKYSORT analysis is nevertheless applied to this dataset, very characteristic and systematic aliasing effects are immediately apparent. More specifically, out of the 1.31x106 echoes, none are observed at a limiting zenith angle ZNL, which increases systematically with radio frequency:
ZNL | 6 | 12 | 18 | 24 | 30 | 36 | 42 | 48 | degrees | |
radio freq. | 1.95 | 2.15 | 2.35 | 2.65 | 3.15 | 3.75 | 4.95 | 6.95 | MHz |
This means that, for a given zenith angle ZN > ZNL, all such echoes alias into a smaller ZN. It also means that at any radio frequency > 1.95 MHz using 4ANT, there is a substantial risk (increasing with frequency) that any chosen echo may have returned from a larger zenith angle than the value estimated.
There are other aliasing risks to consider, and 6ANT is not free of them; nor is it completely free of echolocation aliasing at the highest frequencies. New work on these questions is underway.
The results of the present section, together with those of §4.1, show that the Tromsø antenna system performs rather faithfully in accord with LPA and dipole theory (apart from the PRE effects). The results suggest that an absolute calibration of the system power (permitting corrections to individual data for Tx rolloff <1.8 MHz, for the ±2.5 dB logperiodic factor, and for the cos12 (ZN) law) may be within reach. I believe that the results also justify taking the Tromso system as a 'benchmark' for comparison with the other two systems available at present; this tone which will be perceived in the remainder of these notes.
USU TWD Antenna and passive 30 MHz dipoles, (Figures 8 (AZ) and 9 (ZN) )
There is a tendency, although not at all frequencies, for stronger echoes to return from the West and North in Figure 8. The antenna system is laid out on a hillside (orientation?), which might account for this effect. There are perhaps too few echoes per f, AZ bin to be confident of any conclusions, but there seems to be some evidence of southward 'lobeing' (periodic minima). The decibel range and variability with azimuth is certainly larger than for the LPA.
The zenithangle results, Figure 9, suggest that the system changes 'character' at 3 MHz and again at 6 MHz. Between these limits, the zenith (10°) zone returns echoes some 15 dB stronger than those from the outer zones, even from the 20° zone; this seems to argue a very directive antenna system in this range. With their Rx baseline (D = 30m), the USU receiving array cannot alias echolocations below 5 MHz. Even at 7 MHz, aliasing is only possible beyond , so this is not likely to account for the observations.
Here we encounter important physical effects which may bias the echolocations. The nearly horizontal magnetic field (dip 0.5°), together with fieldalignment of most irregularities, reduces the count of echoes from the N/S plane, except from near the zenith (as an aspect of measurement noise, if nothing else). This probably accounts for much of the tight clustering without regard to azimuth in figure 10. There is, however, consistent evidence for stronger echoes from W than from E, above about 7 MHz. This is perhaps a natural (plasmawave, or 'equatorial bubble') scattering effect.
The same D = 30m as at USU was used here, but a socalled 'NOPOL' Rx scheme (all antennas pointing northward) effectively doubled the smallest possible aliasing angle frequency, to 10 MHz. There is, accordingly, progressive absence of larger zenith angles with frequency > 10 MHz in Figure 11. Allowing for this, the decibel spread between small and large zenith angles increases fairly rapidly and steadily with frequency in Figure 11, although the power is always largest from the zenith. The implication is clearly of a broad polar diagram (comparable to the LPA or dipole) at low frequencies (< 3 MHz), becoming much more narrow at high frequencies.
I realize that this topic is not exciting physics, but I think it may well be material to doing good physics with digital ionosondes. I would therefore like to see it pushed further. Simply getting a bigger database for the sites studied here is both possible and desirable. Adding sites, especially where they involve different antenna choices, different ionospheric (and ground) conditions, etc., will help clarify the whole picture and permit more objective choices in establishing new sites or upgrading existing ones. I wish to end these notes with some specific suggestions. I will be grateful for comments, corrections and suggestions. Where more data analysis is indicated, I shall be happy to undertake it, provided I can receive the data (and read it !), on floppy, CDROM, QIC80 tape, or by ftp. Alternatively, I can provide copies of the two short Fortran subroutines that produced the results discussed here.
NOAA Boot Lake, Colo. | LPA, small dipoles, good ground |
EISCAT, Tromsø | LPA, fat dipoles fair ground |
Halley Bay; BAS | LPD, small dipoles, no ground |
USU Bear Lake;) | TWD, small dipoles |
Los Alamos; Huancayo | LPA, fat dipoles, poor ground |
University of Tromsø | Immense Rhombic; mag. loops |
La Trobe University | Small Rhombic? mag. loops? |
RAL, UK | ditto ? |
Anywhere | any other ? |
Canadian digital ionosonde | ? |
Are there others ? | ? |
[Note added, 7 Nov 1997: Some tests were done of two DATONG active antennas at Tromsø. They were not sufficiently extensive to answer all of the questions above, but they do suggest that the DATONG is not the active antenna of choice for improving ionosonde data. Here is a paragraph quoted from my notes of 14 March 1996:
"We could almost certainly NOT simply exchange
the existing NESWes 6ANT dipoles for Datongs, without unacceptable
penalties in lost or weak echoes and poorer S/N. On the other
hand, they are probably much better at low frequencies than the
non-active short dipoles (such as at Bear Lake). So a comparison
with those would still be informative".]
Back to Bulletins index | Back to INAG-62 Contents page |