Spaced antenna diversity in temperate latitude meteor burst systems operating near 40 MHz : Variation of signal goss-correlation coefficient with time

The potential for meteor burst (MB) spaced antenna diversity is studied using crosscorrelation techniques at a temperate latitude, using 37-MHz signals scattered over an 800km path. Approximately 9 hours of data were analyzed from a 2-day period in February 1990, when the antenna separation was 10k. The correlation variations with time of signals with durations > 0.75 s, which were categorized as "underdense," overdense," or "not known" (NK), were investigated. The cross-correlation coefficients of signals from underdense and overdense trails are high when correlated over their total signal envelopes, but this is not true when the signal envelope is segmented in time. NK signals are observed to be more uncorrelated than underdense or overdense signals and are, therefore, likely to be the most advantageous to a diversity system. Analysis of data from all three categories combined shows that 40% of signals with duration > 0.75 s have correlation values of less than 0.6 after the first 0.25 s, The correlation-time dependency observed for NK and underdense signals is not identified for the overdense signal category. It is proposed that for underdense and NK signals the correlation-time dependency is due to the vector addition of other weak signal modes. These weak signal modes, however, have little effect on overdense signals which, typically, exhibit higher signal powers in the early stages of the signal envelope.


of obtaining uncorrelated signals is by reception on two appropriately spaced antennas (space diversity).
An investigation of MB space diversity on long-duration (-4 s) echoes was performed by Ladd [1961].The experiment utilized a 1-kW transmitter with an operating frequency of 49 MHz, and signals were received on two antennas separated by either 22)• or 60)• (where)• is the signal wavelength).Signals were recorded and digitized using a sampling interval of 100 ms.The correlation coefficients were computed for underdense, specular overdense and nonspecular overdense signals.High correlations were observed (0.9946, 0.9819, and 0.7970, respectively), and Ladd [1961] concluded that little advantage could be gained using space diversity.However, fading periods on signals in this frequency band range typically between 100 ms and 1 s [Greenhow, 1950] and the 100 ms sampling period adopted would, therefore, have been too infrequent.The crosscorrelation coefficients were also computed over complete signal durations.Similar analyses performed by Cannon et al. [1988] showed the correlation coefficient to be a function of signal decay time.In this latter work, high-correlation coefficients were reported at the start of the trail decay [Manning, 1959] with the coefficients decreasing towards the end of the signal lifetime.The measurement technique of Ladd [ 1961] was, therefore, flawed.
A cross-correlation coefficient of 0.6, or below, was adopted by Ladd [1961] as a "practical" indication of useful signal decorrelation [Staras, 1956].Such a condition is, however, only applicable to Rayleigh fading conditions.Our studies of signal envelopes have convinced us that fading cannot always be described as Rayleigh.Interference between two, or a small number of glints (<5), resulting in a deep periodic fading envelope which approximates a rectified sine wave, is more likely than Rayleigh fading.Greenhow [1952] suggested that approximately 400 ms are required for the formation of the first glint.The applicability of the Rayleigh fading model, requiring the formation of multiple (>6) glints, is the.n restricted to long-enduring meteor trails.Rayleigh fading is, therefore, not generally applicable to MB diversity models.The deep periodic fading patterns resulting from two or three glints [Manning, 1959] also relies on the formation of glints each of a similar dimension.Superposition of signals from smaller scattering regions and signals from the main scattering body will cause an amplitude fluctuation as opposed to deep periodic fading.
In contrast to Ladd [1961], Bartholomd [1962] suggested that MB space diversity is a useful technique.Bartholomd [1962] performed experiments in the 38 to 41 MHz band with 400-watt transmissions using the receiver automatic gain control (AGC) voltage as a measure of the signal amplitude.The attack and decay times for these AGC measurements were 10 ms and 40 ms, respectively.An antenna separation of 4)• in a line perpendicular to the great circle path from the transmitter to the receiver was adopted.For specular reflections the diversity effect reported was small both for underdense and overdense wails.When the trails were long (>2 s), however, and wind distortion had caused several glints, quasi-independent, time-varying signals were obtained at each of the separated antennas.
The only fully developed MB communications system incorporating diversity is COMET [Bartholomd and Vogt, 1968].The COMET system implemented space and height diversity, and signal combination was achieved using a multidetector.COMET showed a remarkably high duty cycle when compared to other MB communications systems [Cannon and Reed, 1987].Bartholomd and Vogt [1968] believed that diversity contributed significantly to this high throughput.
In summary, the efficacy of space diversity in MB systems is unclear.MB communications systems generally provide a low average data rate, with average rates varying from a few, to generally only tens of bits per second [Cannon and Reed, 1987].Uncorrelated signals on spaced antennas could be exploited to enhance this throughput using an appropriate diversity combiner.which output the logarithm of the detected input signal level.

Although this paper addresses meteor scatter signals, but it
Each output was recorded on a FM channel of a tape recorder in a 1.25-kHz bandwidth.The audio output from one receiver was also recorded.
Each data tape was calibrated at 28-min intervals with signals ranging in power levels from -133 dBm to -76 dBm in 3-dBm steps.These calibration levels enabled the absolute values of the signal strength to be derived.The receiving and recording equipment was located in a temperature-controlled environment to reduce thermal drift.Cross-correlation values for each segment were grouped together and analyzed.Implicit in this analysis technique is the assumption of an average ionization height and location.This averaging ignores the differing geometrical factors which affect the decay of meteor trails occurring within the antenna common volume.

Trail categorization
Signals were assigned to one of three signal categories: underdense, overdense or "not known" (NK) (Figure 1 A specular overdense signal (Figure lb) is also characterized by a fast rising increase in signal strength until the Fresnel zone is formed.This is followed by a further period of slowly increasing signal strength, resulting in a rounded top to the received envelope, before the signal decays.Overdense decay durations and received signal strengths are usually greater than those from underdense trails.In the overdense signal example (Figure lb) a diversity antenna switch would not increase the effective signal duration and would, therefore, not be advantageous.
Signals not of the above format were categorized as "not known" (NK).Figure l c shows a highly uncorrelated waveform which was categorized as a NK signal.In the example we note that during the period included by the dashed vertical lines, the received power on channel 1 is above the threshold but that on channel 2 is below the threshold.A diversity antenna switch would be advantageous to a MB system for this NK signal example.
Signals which suffer from echo-overlap decorrelation [Berry et al., 1961]  It is usual to attribute received signal amplitude fluctuations to perturbations of the trail by mesospheric winds.Consequently, both overdense and underdense trails would be affected identically after trail formation.The two trail types would then be expected to decorrelate within similar time scales.The contrasting time-dependent correlation results between the underdense and overdense categories suggest, however, that the amplitude fluctuation mechanism is traildependent.We suggest that the differences can be attributed to the typically greater received signal strengths observed from overdense trails.Received signal power analysis (Figure 7   We also draw the reader's attention to the work of Staras [1956], which contributed to the conclusions of Ladd [ 1961], concerning the degree of decorrelation required before the signals are considered to be useful in a diversity system.The theoretical basis of the calculations is the joint probability density function of two Rayleigh fading channels which is often not applicable to the MB channel.
In conclusion, the results presented in this paper suggest that space diversity may be advantageous, especially when signals propagate via NK trails.A diversity contribution can, however, also be exlx?cted from overdense and underdense trails, particularly towards the end of their trail lifetimes when significant decorrelation occurs.
is also to be noted that three other sources of signal exist in the MB frequency band; namely, ionoscatter, sporadic E and propagation during periods of high sunspot activity via the F2 layer.The magnitude and occurrence of signals propagated via these modes depend on mechanisms with various temporal scales and upon the radio frequency used.These other signals may add vectorially with MB signals providing the possibility of diversity gain (similar conclusions were reached by Bartholornd [1962]).Correlation investigations of signals received on two spatially separated antennas will reveal if decorrelated signals exist which might then be exploited by MB communications systems incorporating diversity.The received signals are categorized as underdense, overdense or "not known" (NK), and the advantages of each category are examined.The "NK" category contains signals of an undefined format which may be associated with scattering from sporadic E, ionoscatter, or nonspecular meteor trails.Our results and conclusions apply to an 800-km, temperate latitude path at operating of sporadic E signals which might confuse the study of MB signals, data were collected in February 1990 when the expected occurrence frequency of these modes was at a minimum.Morning and afternoon data were collected using two antennas separated by 10)• (
).A similar signal categorization was performed byOstergaard et al. [1985].Signals scattered from an underdense trail are characterized by a fast (-0.1 s) rising increase in signal strength followed by a slower linear decay (in decibels) which starts immediately after the envelope has reached its maximum amplitude.An underdense signal is illustrated in Figurelawhere the decay duration on channel 1 is greater than channel 2 by an amount A. In this example a simple diversity antenna switch would effectively increase the signal duration by the amount At, an increase in duration which may be advantageous in a MB system.

Fig. 6 .
Fig. 2.An example of echo overlap signal.Figures3a and 3billustrate the deep periodic fading due to[Manning, et al., 1952] two (or more) Doppler-shifted components beating together to produce a Isin tl modulation pattern.The total decay duration of the underdense signal (Figure3a) is -2.6 s.In a conventional communications system, however, this single trail might be interpreted as three shorter (<1 s) trails because of the three deep (-10 riB) fades present during the signal decay.The potential data throughput from such a signal is then not fully exploited because of the time lost reestablishing the link after each signal fade.In this particular example of a fading underdense signal, diversity

Fig. 7 .
Fig. 7. Average power present in the three signal categories for signals of duration > 0.75 s.Antenna separation is 10)•.

Fig
Fig. 9. Cumulative correlation probability curves applicable to all received signals of duration >_ 0.75 s Antenna separation is 10•,.