Long-term characteristics of impulsive atmospheric noise

Abstract

The most important source of interference to the reception of radio signals in the frequency range 1.5�0 MHz at tropical latitudes is atmospheric noise. It appears either as a form of continuous noise or as distinct impulses. Such impulses get superposed on the continuous noise which is invariably present. When atmospheric noise appears in the form of continuous noise only, its value is of about the same order as that of the interference arising from other sources of noise-radiations from stations, etc. The interference arising from the other causes mentioned can be regarded as the background. Therefore, the continuous form of atmospheric noise can be regarded as showing a strong tendency to merge with the background. In view of this fact, a separate assessment of atmospheric noise in its continuous form is not of practical significance at tropical latitudes. Further, even if an assessment of the noise in this form is required to be made, the problem of actual measurement becomes extremely difficult because it is almost impossible to devise methods to separate the noise from the background. In the tropics, which are the world centres of thunderstorm activity, the number of local and/or near sources that are active at any given time is very much larger than at high latitudes. Therefore, the number of impulses in the continuous background becomes much larger. The electrical discharges associated with a flash in a thunderstorm give rise to impulses; for a large number of these impulses (noise level), the magnitude-as indicated by the quasi peak value-is mostly over 10 dB above the background; quite often it is much higher. Having regard to the fact that the impulses appear distinct and that their magnitude is large, it is obvious that these impulses can be studied free from contamination, particularly when measurements are accompanied by continuous monitoring. Further, the impulsive form of atmospheric noise is the one that is important for practical assessment of the interfering effect of the noise. A measure of this impulsive noise over a short period is the average value of the 50 highest impulses received in a five minute period. This parameter is called the noise level and is about 3 dB higher than the mean value of all the impulses received during the five minute period. It has been observed that the distribution of the magnitudes of the impulses during a period of five minutes is random, and this justifies the choice of five minutes. The noise level, therefore, is a suitable short term characteristic of the noise in terms of which the long term characteristics can be investigated. The thesis is concerned with the analysis of the data collected for atmospheric noise in its impulsive form in terms of the noise level at 1.6 MHz and 2.9 MHz. Actual experimental investigations had shown that the correlation between noise levels during successive five minute periods was generally so high that any evaluation of the half hourly noise level by averaging would not satisfactorily represent the noise level during the half hour. Therefore, a noise measurement carried out for a five minute period was taken as a measure of the noise level of an entire half hour. This led to a decision to measure noise at each frequency for a period of five minutes only during a half hour. Hence, the noise level data for evaluating the long term characteristics get reduced to a maximum of 48 per day (i.e., one per half hour per frequency; in practice, up to 24 per frequency). Noise level data were collected for each of the half hours during 12� h I.S.T., for one complete year. By taking a season as a satisfactory unit for examining the long term characteristics, noise level values for about 90 days is the maximum that can be expected theoretically. Having regard to measurement difficulties, in practice about 60 values only become available because some days are lost due to power or equipment failure, etc. This set of ~60 noise level observations provides the raw data for computing seasonal median and higher decile values for a half hour. When the median and higher decile values of noise level show hour to hour variations, the nature of the variation is examined. On this basis, it has been found that the noise level shows systematic variations during the hours 12� I.S.T. Further, the percentage of time for which impulsive noise is present (on a seasonal basis) has also been investigated. Even this is found to show systematic variations. Generally, whenever there is systematic variation, the half hourly noise level (median or higher decile) increases from half hour to half hour as the hour of day advances from 12 to about 20 h I.S.T. The exact nature of variation of the median and higher decile values is different; further, there are differences from season to season. During the hours 20� I.S.T. there was practically no systematic variation during the seasons March朚ay, June朅ugust and September朜ovember. There was a little systematic variation for these hours during December朏ebruary. Therefore, the noise levels could be considered random for the four hours 20� I.S.T. during three seasons of the year. Before subjecting the data for these seasons to statistical analysis, the correlation between the half hourly noise levels on any one day was examined and it was found that the correlation was very high. Effectively, therefore, the eight half hour units of the time block 20� I.S.T. get reduced to practically one. That is, the number of noise levels available for statistical analysis for the entire time block 20� h becomes practically equivalent to the number of observations available for just one half hour unit. Therefore, although a larger number of observations are available for a time block, the statistical significance of the analysis gets reduced to that obtained by just about 60 independent observations. An examination of the noise level data for the time block 20� I.S.T., on a seasonal basis, has shown that it consists of two log normal distributions. Assuming that one log normal distribution corresponds to one set of sources, the conclusion from the analysis would be that there are two sets of sources responsible for the noise. This is easy to visualise on the basis of existing thunderstorm data. For one set of sources, reception of the noise may be by the direct ray or by the direct cum ground ray. Reception by this mode of propagation can be expected for sources which may lie at distances up to over 100 km (or even 200 km). This can be expected because discharges within the cloud take place at a height of about 3 km or more above mean sea level and they are responsible for noise at the frequencies concerned. Hence, one could think of these sources as local sources (the word local is conventionally meant to imply distances much smaller than 100�0 km, e.g., 20� km; here it is used more broadly for 搉earby� sources). The second set of sources could be visualised as all near sources from which noise is received via the ionosphere. Such sources-even if they are spread over distances extending from a few hundred kilometres to over 1000 km-could be expected to contribute noise of about the same magnitude. The implications of the experimental results discussed above could be visualised to mean that, for noise estimation purposes for tropical latitudes, two sets of sources should be reckoned with. Information about the percentage of time in a season for a time block or a half hour for which each one of them is expected to contribute to the received noise has to be evaluated. This can be used along with estimates of mean value and standard deviation for each set of sources for computing seasonal noise levels. (An examination of the C.C.I.R. data appears to indicate that the local sources as described here are not considered in computing noise data.) For time blocks for which there are both random and systematic variations, viz., 12� and 16� I.S.T., both types of variations will have to be taken into account for computing noise data. This problem does not lend itself to any simple solution. Such a solution will require collection of more extensive data and that too over a wider area. The raw noise level data collected during a season at the two frequencies, viz., 1.6 MHz and 2.9 MHz, were examined to find out whether it was possible to predict noise at one frequency knowing that at the other frequency. The scatter diagram indicated that this should be possible. Even when the scatter diagram was drawn in different ways, viz., for each set of sources separately and for both sets of sources together, the trend appeared to be the same. During the time block 20� I.S.T., it had already been found that there were no systematic variations of noise levels at any one frequency. Therefore, for examining the correlation between noise levels at the two frequencies, all the raw data for the entire time block during a season were pooled together. The result showed very good correlation and the necessary regression analysis was carried out. The conclusion emerging from the analysis is that if the noise level is known at one frequency, that at the other can be predicted within an accuracy of about 3 dB. On examining this problem for a more extended range of frequencies by using data collected by others collaborating with us, it was found that as the frequency gap increases the correlation is poorer and the accuracy of prediction lower. A practical conclusion from this investigation is: for predicting noise in the frequency range 1.5�0 MHz, it is better to measure noise at about 3.0 MHz if measurements have to be carried out at one frequency only.