Premium
Development of a minicomputer atmospheric noise model
Author(s) -
Sailors D. B.,
Brown R. P.
Publication year - 1983
Publication title -
radio science
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.371
H-Index - 84
eISSN - 1944-799X
pISSN - 0048-6604
DOI - 10.1029/rs018i004p00625
Subject(s) - atmospheric noise , noise (video) , algorithm , mathematics , residual , computer science , statistics , telecommunications , artificial intelligence , very low frequency , image (mathematics)
A practical but simplified atmospheric noise model in universal time for 1‐MHz F am usable in estimating system signal‐to‐noise ratios in microminicomputer‐based HF propagation prediction systems is presented. Numerical mapping techniques were used to represent worldwide atmospheric noise in 24 numerical maps: one for each of six 4‐hour UT time blocks and each of four 3‐month periods. This representation results in a minimum reduction by one fifth in the number of coefficients necessary in computer memory over that of an earlier model in which the map represented an entire 24‐hour day. The simpler expression for the median noise levels ( F am ) results in significant savings in computer code and time. Several versions were produced, each with a different number of coefficients (number of harmonics) in the Fourier representation for each time block. The number of coefficients ranged from 192 coefficients per time block for the most accurate versions to 35 coefficients for the least accurate. The accuracies were determined by calculating the rms residual between values of F am used to develop the model and the corresponding values computed from each model. The most accurate model had an average rms error for all times and seasons of 1.3 dB, whereas the least accurate had a rms error of 4.5 dB. During summer, the worst season, the most accurate model had a rms error of 1.4 dB, whereas the least accurate had a rms error of 5.6 dB. The International Radio Consultative Committee noise data used to generate the models had an average standard deviation of the error at 1 MHz of 5.4 dB and an error of 6.7 dB during summer; the numerical map in use in current HF prediction programs has an average rms error of 1.7 dB and an error of 2.0 dB during summer. The locations of the maximum absolute deviations of the models were confined to a small region in central Africa.