ESSA TR ERL 172-ITS 110 
 
 A UNITED STATES 
 DEPARTMENT OF 
 COMMERCE 
 PUBLICATION 
 
 
 "V,,*"** 
 
 ESSA Technical Report ERL 172-ITS 110 
 
 U.S. DEPARTMENT OF COMMERCE 
 Environmental Science Services Administration 
 Research Laboratories 
 
 Modulation Studies for IGOSS 
 
 IIROSHI AKIMA 
 
 BOULDER, COLO. 
 JUNE 1970 
 
ESSA RESEARCH LABORATORIES 
 
 The mission of the Research Laboratories is to study the oceans, inland waters, the 
 lower and upper atmosphere, the space environment, and the earth, in search of the under- 
 standing needed to provide more useful services in improving man's prospects for survival 
 as influenced by the physical environment. Laboratories contributing to these studies are: 
 
 Earth Sciences Laboratories: Geomagnetism, seismology, geodesy, and related earth 
 sciences; earthquake processes, internal structure and accurate figure of the Earth, and 
 distribution of the Earth's mass. 
 
 Atlantic Oceanographic and Meteorological Laboratories: Oceanography, with emphasis 
 on the geology and geophysics of ocean basins, oceanic processes, sea- air interactions, 
 hurricane research, and weather modification (Miami, Florida). 
 
 Pacific Oceanographic Laboratories: Oceanography; geology and geophysics of the 
 Pacific Basin and margins; oceanic processes and dynamics; tsunami generation, propaga- 
 tion, modification, detection, and monitoring (Seattle, Washington). 
 
 Atmospheric Physics and Chemistry Laboratory: Cloud physics and precipitation; chem- 
 ical composition and nucleating substances in the lower atmosphere; and laboratory and 
 field experiments toward developing feasible methods of weather modification. 
 
 Air Resources Laboratories: Diffusion, transport, and dissipation of atmospheric con- 
 taminants; development of methods for prediction and control of atmospheric pollution 
 (Silver Spring, Maryland). 
 
 Geophysical Fluid Dynamics Laboratory: Dynamics and physics of geophysical fluid 
 systems; development of a theoretical basis, through mathematical modeling and computer 
 simulation, for the behavior and properties of the atmosphere and the oceans (Princeton, 
 New Jersey). 
 
 National Severe Storms Laboratory: Tornadoes, squall lines, thunderstorms, and other 
 severe local convective phenomena toward achieving improved methods of forecasting, 
 detecting, and providing advance warnings (Norman, Oklahoma). 
 
 Space Disturbances Laboratory: Nature, behavior, and mechanisms of space disturb- 
 ances; development and use of techniques for continuous monitoring and early detection 
 and reporting of important disturbances. 
 
 Aeronomy Laboratory: Theoretical, laboratory, rocket, and satellite studies of the 
 physical and chemical processes controlling the ionosphere and exosphere of the earth and 
 other planets. 
 
 Wave Propagation Laboratory: Development of new methods for remote sensing of the 
 geophysical environment; special emphasis on propagation of sound waves, and electro- 
 magnetic waves at millimeter, infrared, and optical frequencies. 
 
 Institute for Telecommunication Sciences: Central federal agency for research and 
 services in propagation of radio waves, radio properties of the earth and its atmosphere, 
 nature of radio noise and interference, information transmission and antennas, and meth- 
 ods for the more effective use of the radio spectrum for telecommunications. 
 
 Research Flight Facility: Outfits and operates aircraft specially instrumented for re- 
 search; and meets needs of ESSA and other groups for environmental measurements for 
 aircraft (Miami, Florida). 
 
 ENVIRONMENTAL SCIENCE SERVICES ADMINISTRATION 
 
 BOULDER, COLORADO 80302 
 
<tAENTO f 
 
 Se, lHC[ siVf^ 
 
 U. S. DEPARTMENT OF COMMERCE 
 Maurice H. Stans, Secretary 
 
 ENVIRONMENTAL SCIENCE SERVICES ADMINISTRATION 
 Robert M. White, Administrator 
 RESEARCH LABORATORIES 
 Wilmot N. Hess, Director 
 
 ESSA TECHNICAL REPORT ERL 172-ITS 110 
 
 Modulation Studies for IGOSS 
 
 HIROSHI AKIMA 
 
 >. 
 a. 
 o 
 u 
 
 o 
 
 '8 
 
 Q. 
 <D 
 
 o 
 
 INSTITUTE FOR TELECOMMUNICATION SCIENCES 
 BOULDER, COLORADO 
 June 1970 
 
 For sale by the Superintendent of Documents, U. S. Government Printing Office, Washington, D. C. 20402 
 
 Price 55 cents 
 
FOREWORD 
 
 The work described in this report was done 
 at the request of Commandant (FS-1), U.S. Coast 
 Guard. It was funded under Military Interdepart- 
 mental Purchase Request No. Z-70099-9-91574 
 dated 16 September 1968, and monitored by 
 LCDR Marshall E. Gilbert, USCG. 
 
TABLE OF CONTENTS 
 
 Page 
 
 FOREWORD ii 
 
 ABSTRACT 1 
 
 1. INTRODUCTION 1 
 
 2. PRELIMINARY SELECTION OF MODULATION SYSTEMS 2 
 
 2.1. On -Off Keying System 3 
 
 2.2. Noncoherent Frequency-Shift-Keying (NCFSK) System 3 
 
 2.3. Time -Differential Phase -Shift -Keying (TDPSK) 
 
 System 4 
 
 2.4. Frequency-Differential Phase -Shift-Keying (FDPSK) 
 
 System 4 
 
 2.5. Other Techniques and Modems 5 
 
 3. EFFECTS OF NOISE AND INTERFERENCE ON THE 
 
 SYSTEM PERFORMANCES 5 
 
 3? 1. Nonfading Signal and Atmospheric Noise Without 
 
 Interference 7 
 
 3. 2. Fading Signal and Atmospheric Noise Without 
 
 Interference (Nondiversity Reception) 7 
 
 3.3. Fading Signal and Atmospheric Noise Without 
 
 Interference (Diversity Reception) 11 
 
 3.4. Nonfading Signal and Nonfading Interference With 
 
 Gaussian Noise 17 
 
 3.5. Nonfading Signal and Nonfading Interference With 
 
 Atmospheric Noise 18 
 
 3. 6. Fading Signal and Fading Interference Without 
 
 Noise (Nondiversity Reception) 23 
 
 3.7. Fading Signal and Fading Interference Without 
 
 Noise (Diversity Reception) 24 
 
 3.8. Fading Signal and Fading Interference With Noise 27 
 
 in 
 
4. EFFECTS OF ADVERSE PROPAGATION CONDITIONS 
 
 ON THE SYSTEM PERFORMANCES 30 
 
 4.1. Multipath Time -Delay Spread 31 
 
 4. 2. Frequency Fluctuations 32 
 4.3. Phase Fluctuations 33 
 
 5. FREQUENCY STABILITY 34 
 
 5. 1. Frequency Tolerance for NCFSK Systems 35 
 
 5.2. Frequency Tolerance for TDPSK Systems 38 
 
 5.3. Frequency Stabilization 39 
 
 6. SELECTIVITY OF RECEIVERS 42 
 
 6. 1. Requirements for the Selectivity of IGOSS 
 
 Shore -Station Receivers 43 
 
 6.2. Receiver Structure 46 
 
 7. CONCLUSIONS 48 
 
 8. ACKNOWLEDGMENTS 49 
 
 9. REFERENCES 50 
 
 IV 
 
MODULATION STUDIES FOR IGOSS 
 Hiroshi Akima 
 
 Modulation techniques for high frequency (HF) 
 data transmission for the Integrated Global Ocean 
 Station System (IGOSS) are discussed. Noncoherent 
 frequency -shift -keying (NCFSK) and time -differential 
 phase -shift -keying (TDPSK) systems are preliminarily 
 selected as preferred systems. Performances of these 
 preferred systems are discussed both in the presence 
 of noise and/ or interfering signal and under adverse 
 propagation conditions. A serious drawback inherent 
 in a TDPSK system under adverse propagation con- 
 ditions is pointed out. Frequency tolerance, frequency 
 stabilization techniques, and selectivity of receivers 
 are also discussed. 
 
 Key Words: Data transmission, frequency stability, 
 high frequency (HF), Integrated Global 
 Ocean Station System (IGOSS), modulation 
 technique, selectivity. 
 
 1. INTRODUCTION 
 
 This report describes certain aspects of modulation techniques 
 and related topics to be considered for the Integrated Global Ocean 
 Station System (IGOSS). The intention is not to recommend the use of 
 a specific modem (modulator and demodulator), but to provide 
 information necessary for selecting a modem and other related factors 
 that may affect the design of the IGOSS telecommunication network. 
 
 Many system parameters of the IGOSS are still undefined; how- 
 ever, the long range plan for the IGOSS is for several platforms (buoys 
 and ships) to transmit data simultaneously to a common shore station in 
 a common 3-kHz high frequency (HF) band. This is similar to multi- 
 
channel, frequency -division -multiplex (FDM) data modems in point- 
 to-point HF communications. FDM modems, which can tolerate the 
 multipath time -delay spreads associated with HF ionospheric radio 
 propagation, have been studied for more than a decade, and the results 
 were used as guidelines for this study. 
 
 Despite the similarity between the IGOSS and FDM point-to-point 
 communications, we must also be aware of the difference between them. 
 In a point-to-point system, signals in different channels are transmitted 
 from a common transmitter, but in the IGOSS they are transmitted from 
 different geographically dispersed platforms. An IGOSS shore station 
 must receive simultaneously, in a common 3 -kHz band, several signals 
 that may have different median values and fade independently of each 
 other. Therefore, interchannel interference is more severe than in 
 point-to-point systems, and special attention must be given to the 
 selectivity of the receiver and demodulator. 
 
 Independent channel transmitters also mean that synchronization 
 between channels cannot be expected, so that demodulation techniques 
 that rely on orthogonality between channel signals cannot be used. 
 
 Since service intervals of up to 1 year are desirable, high reli- 
 ability and small power consumption of the buoy transmitter are among 
 the criteria for selecting a specific modulation technique and modem. 
 
 2. PRELIMINARY SELECTION OF MODULATION SYSTEMS 
 
 Among the various modulation systems for digital data trans- 
 mission, described here in the order of their development, non- 
 coherent frequency -shift -keying (NCFSK) and time -differential phase ■ 
 shift -keying (TDPSK) systems appear to be the most attractive 
 candidates for the IGOSS. 
 
2. 1. On -Off Keying System 
 
 Digital information is transmitted by on -off keying of the unmodu- 
 lated carrier located at the center of the channel. For transmitters 
 limited in peak envelope power, the average power is half that of constant 
 amplitude systems, such as frequency -shift -keying (FSK) or phase-shift- 
 keying (PSK) systems. Since power is transmitted for only one of the 
 two states, a decision threshold must be set artificially. This can be 
 done easily if the signal wave is not fading and the noise is negligible 
 compared with the signal; however, the peak envelope power of 
 received radio wave signals propagated over HF ionospheric paths 
 generally fluctuates considerably, so that setting the decision threshold 
 is a difficult problem. For these reasons, on-off keying of a carrier is 
 thought unsuitable for automated data transmission; therefore, this 
 system is not considered further. 
 
 2.2. Noncoherent Frequency -Shift -Keying (NCFSK) System 
 
 A certain number of equally spaced frequencies f , f,, . . . , f 
 
 o 1 n-1 
 
 are chosen in each channel, where n is the number of chosen modulation 
 conditions. The channel subcarrier is frequency modulated (frequency- 
 shift -keyed) in accordance with the digital information signal, so that the 
 frequency of the subcarrier is equal to f. during an interval in which the 
 
 information content is equal to i (i=0, 1,2,..., n-1). Binary modems 
 (n=2) are most common. Quaternary modems (n=4), sometimes called 
 diplex or twinplex, are also used. NCFSK is one of the most widely 
 used systems for digital data transmission over HF ionospheric paths. 
 It is a strong candidate for the IGOSS. 
 
2. 3. Time -Differential Phase -Shift -Keying (TDPSK) System 
 
 In a PSK system, a subcarrier is phase modulated (phase -shift- 
 keyed) by the digital information signal in such a way that the difference 
 between the phase of the subcarrier cp and the reference phase cp is equal 
 to 
 
 cp - cp = (2i + 1 - n) n/n, (i = 0, 1, 2, . . . , n - 1), 
 
 where n is the number of modulation conditions. Both binary (n=2) and 
 quaternary (n=4) modems are widely used. In a TDPSK system, both 
 modulation and demodulation are referenced to the phase of the carrier 
 during the previous signalling element. TDPSK is also widely used for 
 digital data transmission over HF ionospheric paths. It is also a 
 strong candidate for the IGOSS. 
 
 2.4. Frequency -Differential Phase -Shift -Keying (FDPSK) System 
 
 In the FDPSK system, both modulation and demodulation are 
 referenced to the phase of another subcarrier (either modulated or un- 
 modulated), located near the modulated subcarrier (de Haas, 1965). 
 Although this technique is not widely used, perhaps it can yield the 
 highest data signalling rate in a given bandwidth among various 
 techniques of high-speed data transmission over HF ionospheric paths. 
 It has a serious drawback, however, when applied to single -channel 
 circuits. Since a reference -phase carrier must always accompany the 
 modulated carrier, a linear power amplifier, instead of a class -C 
 power amplifier, is required in the transmitter, resulting in a 
 considerably lower transmitter efficiency. Therefore, this system is 
 not considered further in this study. 
 
 4 
 
2. 5. Other Techniques and Modems 
 
 There are several other modems based on different techniques: 
 the Kathryn modem (Zimmerman and Kirsch, 19 65), the Adapticom 
 modem (Di Toro et al. , 1965), and one based on a swept-frequency 
 (chirp) modulation technique (Dayton, 1968). These modems are not 
 considered further here, because (a) a too high level of sophistication 
 requires complicated equipment or (b) available operating data are 
 insufficient to adequately predict actual performance characteristics. 
 
 3. EFFECTS OF NOISE AND INTERFERENCE 
 ON THE SYSTEM PERFORMANCE 
 
 When the desired signal is not fading or experiencing slow, 
 flat fading (i.e. , the fading is slow compared with the modulation rate, 
 and different spectrum components are fading simultaneously), the ele- 
 ment error probability P e in a binary NCFSK system is equal to one- 
 half the probability that the instantaneous amplitude of the unde sired 
 signal U exceeds the instantaneous amplitude of the desired signal D 
 (Montgomery, 1954-), i.e., 
 
 P e = \ Pr { U > D } 
 
 = \ f fp( D > U) dD dU, (1) 
 
 U>D 
 
 where Pr -| [ stands for "the probability that," and p(D, U) is the 
 probability density function of D and U. This applies to an NCFSK 
 system where the modulation index is not less than unity and no low- 
 pass filter is used before the decision-making circuit. For proper 
 
application of (1), the statistics of the undesired signal must be 
 measured at the input of the demodulator, i. e. , at the input to the 
 limiter in a limiter -discriminator demodulator, and in a bandwidth 
 equivalent to the sum of the bandwidths of the two filters in a dual- 
 filter demodulator. Except for a few simple cases, such as a non- 
 fading signal or a Ray leigh -fading signal in the presence of Gaussian 
 noise, this integral cannot be evaluated analytically but must be 
 computed numerically. 
 
 The element error probability in a binary TDPSK system is 
 closely approximated by (1); it is exactly equal to the result computed 
 by (1), when the undesired signal consists of Gaussian noise (Cahn, 
 1959)- (Since the bandwidth for a TDPSK system is usually considered 
 one -half that for an NCFSK system, its required ratio of signal energy 
 per bit to noise power density can be 3 dB lower than that for NCFSK 
 to give an equal error probability. ) 
 
 Amplitudes of signals propagated by the ionosphere vary con- 
 siderably with time. For short intervals (3 to 7 min), amplitude 
 distribution functions close to the Rayleigh distribution predominate. 
 Over longer intervals (30 to 60 min), on the other hand, amplitude 
 distributions more often follow a log -normal law. Although the form 
 of the measured distribution may differ from the Rayleigh distribution, 
 the observed fading range, defined as the ratio of the upper and lower 
 deciles, is the same order as 13.4 dB expected for the Rayleigh dis- 
 tribution (CCIR, 1967g). 
 
 In HF channels noise is mostly atmospheric. The amplitude - 
 probability distribution (APD) of atmospheric noise can be represented, 
 accurately enough for most applications, by an appropriate curve 
 chosen from a family of idealized curves. The choice can be made by 
 specifying a single parameter, defined as the ratio of the rms to the 
 
average of the envelope voltage and denoted by V in dB (CCIR, 1964). 
 
 For simplicity we assume that the interfering signal is either 
 a continuous wave (CW) or an FSK or PSK signal. 
 
 The element error probability in each particular case is 
 generally given as a function of signal-to-noise ratio (SNR) and /or 
 signal-to-interference ratio (SIR). The SNR is defined as the ratio 
 of desired signal power to average noise power, and SIR is the ratio 
 of desired signal power to interfering signal power, where both the 
 noise power and interfering signal power are measured at the demod- 
 ulator input. For a fading signal, either desired or interfering, 
 its median power is used in defining the SNR and SIR. 
 
 The integral in (1) has not been computed for very general 
 cases, such as a log-normal-fading signal in the presence of atmo- 
 spheric noise and a log -normal- fading interfering signal with dual 
 diversity reception. Some particular cases, however, have been 
 studied and described below. 
 
 3. 1. Nonfading Signal and Atmospheric Noise Without Interference 
 
 The element error probability is one -half the value of the APD 
 of the atmospheric noise corresponding to the signal amplitude. The 
 relation between SNR and error probability, parametric in noise param- 
 eter V, (Akima et al. , 19 69), is shown in figure 1. 
 d 
 
 3.2. Fading Signal and Atmospheric Noise Without Interference 
 
 (Nondiversity Reception) 
 
 The relation between SNR and element error probability, 
 
 parametric in noise parameter V.» computed by numerical integration 
 
 d 
 
 (Akima et al. , 1969), are given in figure 2 for Rayleigh-fading signals 
 and in figure 3 for log-normal-fading signals (fading range = 13. 4 dB). 
 
10 
 
 10 
 
 r2 
 
 o 
 
 w 
 i_ 
 
 UJ 
 
 c 
 E 
 
 UJ 
 
 O 
 
 o .4 
 
 -§ io 4 
 
 0_ 
 
 10 
 
 r5 
 
 10 
 
 r6 
 
 1 
 
 1 1 1 
 
 1 1 1 1 
 
 1 1 1 
 
 1 
 
 1 1 1 1 | 1 1 1 1 
 
 Nonfading Signal 
 
 Atmospheric Noise 
 
 No Interference 
 
 1 1 
 
 1 
 
 — 
 
 
 
 
 
 — 
 
 
 \ 
 
 
 
 
 : 
 
 — 
 
 
 
 
 
 
 S^- IO dB 
 
 
 — 
 
 
 V d =1.049 dB » 
 
 I 2 dB ^n 
 
 , 4 
 
 dB-A \ ^ 
 
 
 
 
 
 
 
 
 
 
 \X-8 dB 
 
 
 
 1 
 
 1 1 
 
 II 1 1 
 
 II I I 
 
 l\ 
 
 1 1 l\ 
 
 VAX 
 
 _6dB 
 
 1 1 1 
 
 — 
 
 \\\ 
 
 •10 
 
 10 20 30 
 
 SNR (dB) 
 
 40 
 
 50 
 
 Figure 1. Element error -probability in a single «- channel 
 NCFSK system under stable (nonfading) conditions with 
 atmospheric noise and no interference . 
 
20 
 SNR (dB) 
 
 Figure 2. Element error -probability in single- channel NCFSK 
 system under Ray leigh- fading conditions with atmospheric 
 noise and no inter ference > nondiversity reception. 
 
10" 
 
 fc 10" 
 
 Itoi 
 
 c 
 
 CD 
 
 E 
 
 10" 
 
 o 
 
 -O 
 O 
 
 -O 
 
 o 
 
 10" 
 
 10" 
 
 10" 
 
 1 1 1 1 
 
 I I I l 
 
 
 Log- Normal - 
 
 i i i i 
 
 Fading Signal 
 
 1 1 1 1 
 
 — ^^^ 
 
 
 
 ( Fading Range = l3.4dB) 
 
 Atmospheric Noise 
 
 No Interference 
 
 Nondiversity 
 
 — 
 
 — 
 
 
 
 
 
 — 
 
 — 
 
 
 
 
 v ^-~-Vd = 10 dB 
 
 — 
 
 
 4dB 
 
 2dB-— 
 V d =1.049 dB- 
 
 
 \\V^^-— - 8 
 
 dB _ 
 — 6dB _ 
 
 i i i r 
 
 i i i i 
 
 1 1 1 1 
 
 i i Ai 
 
 
 \\ 1 1 
 
 •10 
 
 10 
 
 20 
 SNR (dB 
 
 30 
 
 40 
 
 50 
 
 Figure 3. Element error probability in a single- channel NCFSK 
 system under log-normal-fading conditions with atmospheric 
 noise and no interference, nondiversity reception. (The 
 fading range is assumed to be 13.4 dB, which is the same as 
 expected for Rayleigh fading.) 
 
 10 
 
3.3. Fading Signal and Atmospheric Noise Without Interference 
 
 (Diversity Reception) 
 
 When dual diversity with a selection-switching combiner is 
 used, the probability of element error in an NCFSK system is given 
 by 
 
 P g = - Pr { U a > D x , V a > V 2 } 
 
 + \ Pr { U 2 > D ? , V 2 > V x }, (2) 
 
 where 
 
 D. = voltage of desired signal from antenna i 
 or in channel i (i = 1 , 2), 
 
 U. = voltage of undesired signal (noise) from 
 antenna i or channel i (i = 1 , 2), 
 
 V. = vector sum of D. and U. 
 
 l 11 
 
 = voltage of signal plus noise from 
 antenna i or in channel i (i = 1 , 2). 
 
 Since D 8 and U 2 follow the same probability distributions as 
 D-l and \J 1 , respectively, the second term in (2) is equal to the first 
 term. Therefore, -we have 
 
 P e = Pr { U x > D lf V a > V s } (3) 
 
 as an expression of element error probability in an NCFSK system 
 with dual selection-switching diversity reception. 
 
 This result can easily be extended to cover higher order diver 
 sity with selection-switching combining. For example, 
 
 11 
 
Pe = 2Pr {u x > D 1 , V 1 > V 2 , V, > V 3 , V x > V 4 | (4) 
 
 is the expression of el ment error probability for quadruple diversity- 
 reception. 
 
 Evaluating P from (3) or (4) is the evaluation of a definite 
 integral of a joint probability density function and can best be done 
 by the Monte Carlo method or numerical simulation. One big advantage 
 of the Monte Carlo method is that it does not become much more 
 complicated when the order of diversity is increased. When we use 
 this method, sequences of random numbers are chosen to follow the 
 probability distribution functions of the signal, the noise, and the 
 phase difference between the signal and the noise. The necessary 
 condition for the occurrence of an error, as given by (3) or (4), is 
 then tested with successive sets of elements; each element is picked 
 from the corresponding sequence of random numbers, and the number 
 of times an error occurs is counted. 
 
 We assumed statistical independence among the signal and noise 
 amplitudes and a random phase relationship between signal and noise. 
 We then tested one million sets of elements for each combination of 
 two types of signal fading, dual and quadruple diversity, six values of 
 V , and several values of SNR in 2-dB steps. Results of these compu- 
 tations for dual diversity are shown in figures 4 and 5 for Rayleigh 
 fading signal and log -normal -fa ding signal, respectively, and for 
 quadruple diversity in figures 6 and 7. Values of the required SNR for 
 element error probabilities of 10 and 10 are read from figures 
 2 through 7 and shown in table 1. This table indicates that the gain in 
 required SNR obtained either by the use of dual diversity instead of 
 nondiversity or by the use of quadruple diversity instead of dual diversity 
 depends on the type of fading, the value of V , and the allowable error 
 
 12 
 
io- 
 
 5 I0" 2 
 
 c 
 
 E 
 <v 
 
 ^lO" 3 
 o 
 
 >> 
 
 X> 
 
 o 
 
 XI 
 
 o 
 
 10" 
 
 10" 
 
 1 1 1 1 
 
 1 1 1 1 
 
 1 1 1 1 
 
 MM MM 
 
 Rayleigh - Fading Signal 
 
 — 
 
 — v 
 
 
 
 Atmospheric Noise 
 No Interference 
 
 Dual Diversity 
 (Selection Switching) 
 
 — 
 
 — 
 
 
 
 
 
 — 
 
 — 
 
 
 2dB^ 
 
 
 ^V d = 10 dB 
 
 — 
 
 Vd 
 
 = l.049dB^ 
 
 
 
 5dB _ 
 -6dB 
 
 1 1 1 1 
 
 1 1 1 1 
 
 1 1 1 1 
 
 \ \ 
 
 \ 
 
 \ 
 \ 
 
 1 1 1 1 
 
 \ \\\ 
 
 \ \ \ V 
 
 \ \ \N 
 \ N N 
 
 \ 
 W 
 
 \N 1 1 
 
 ■10 
 
 10 
 
 20 
 SNR (dB 
 
 30 
 
 40 
 
 50 
 
 Figure 4. Element error probability in a single-channel NCFSK 
 system under Rayleigh -fading conditions with atmospheric 
 noise and no interference, dual (selection-switching ) diver- 
 sity reception. 
 
 13 
 
20 30 
 
 SNR (dB) 
 
 Figure 5. Element error probability in a single-channel NCFSK 
 system under log-normal-fading oonditons with atmospheric 
 noise and no interference* dual (selection-switching) diver- 
 sity reception. (The fading range is assumed to be 13.4 dB . ) 
 
 14 
 
10" 
 
 o I0' 2 
 
 c 
 
 V 
 
 E 
 
 u 10- 3 
 
 O 
 -O 
 O 
 
 £ 10" 
 
 10 
 
 -5 
 
 10" 
 
 1 1 1 1 
 
 1 1 1 1 
 
 1 1 1 1 
 
 1 1 1 1 MM 
 
 Rayleigh - Fading Signal 
 
 Mil 
 
 — 
 
 
 
 Atmospheric Noise 
 No Interference 
 
 Quadruple Diversity 
 (Selection Switching) 
 
 — 
 
 I 
 
 \ 
 
 
 
 
 — 
 
 — 
 
 4dB' 
 
 \ "*\ 
 
 
 -v d = io de 
 
 - 8dB 
 
 — 
 
 — 
 
 2dE 
 
 V d = 1.049 
 
 IB — "*\ * 
 
 \ 
 
 \ \ 
 
 V— 6dB 
 
 — 
 
 1 1 1 1 
 
 1 1 1 1 
 
 \ 
 \ 
 \ 
 
 1 1 1 l\ 
 
 \ \ 
 \ \ 
 \ \ 
 
 1 \ 1 1 ^ 
 
 \ \ X 
 
 1 \\\ 
 
 \l 1 
 
 1 i 
 
 ■10 
 
 10 
 
 20 30 
 
 SNR (dB) 
 
 40 
 
 50 
 
 Figure 6. Element error probability in a single- channel NCFSK 
 system under Ray leigh- fading conditions with atmospheric 
 noise and no interferences quadruple (selection-switching) 
 diversity reception. 
 
 15 
 
10" 
 
 o I0" 2 
 
 UJ 
 
 c 
 E 
 
 LxJ 
 
 10" 
 
 JO 
 
 o 
 
 O 
 
 CL 10" 
 
 10 
 
 -5 
 
 10" 
 
 1 1 1 1 
 
 1 1 1 1 
 
 1 
 
 1 1 
 
 
 1 1 II 
 
 1 1 1 1 
 
 1 1 1 1 
 
 "^^^^^^^v 
 
 
 
 
 
 Log- Normal - 
 
 Fading Signal 
 
 — 
 
 
 
 
 
 
 (Fading Range = 13.4 dB ) 
 
 
 
 
 
 
 
 
 Atmospheric Noise 
 No Interference 
 
 — 
 
 
 
 
 
 
 Quadruple Diversity 
 
 
 — 
 
 
 
 ( Selection 
 
 Switching ) 
 
 — 
 
 — 
 
 
 
 
 ^V d = 10 dB 
 
 — 
 
 
 4dB- 
 
 
 
 
 
 ^^8dB 
 
 
 
 2dB—"" 
 
 
 
 
 YV^^-6dB 
 
 
 
 V d = 1.049 <JB^^ 
 
 \ 
 
 \ 
 
 \ 
 
 \ XX 
 
 
 1 1 1 1 
 
 1 1 1 1 
 
 1 
 
 \ 
 I 
 \ 
 
 ill 
 
 \ 
 
 \ 
 
 | \ 
 
 .MM ' 
 
 \ U 
 
 i 1 \ M\ 
 
 1 1 1 1 
 
 ■10 
 
 10 
 
 20 
 SNR (dB) 
 
 30 
 
 40 
 
 50 
 
 Figure 7. Element error -probability in a single- channel NCFSK 
 system under log-normal-fading conditions with atmospheric 
 noise and no interference, quadruple (selection-switching ) 
 diversity reception. (The fading range is assumed to be 
 13.4 dB.) 
 
 16 
 
Table 1. Required SNR for a Single -Channel NCFSK System 
 Under Fading Conditions With Atmospheric Noise and No 
 Interference. (P is the element error probability and 
 V£ is the CCIR nozse parameter in dB . Selection-switch- 
 ing diversity is assumed. ) 
 
 Signal 
 Fading 
 
 
 
 
 Requi 
 
 red SNR in 
 
 dB 
 
 
 P e 
 
 Diversity 
 
 v d 
 
 — 
 
 
 
 
 
 
 
 
 1.049 2 
 
 4 
 
 6 
 
 8 
 
 10 
 
 
 
 None 
 
 25 
 
 25 
 
 25 
 
 25 
 
 24 
 
 24 
 
 
 10" 3 
 
 Dual 
 
 15 
 
 17 
 
 19 
 
 20 
 
 21 
 
 22 
 
 Rayleigh 
 
 
 Quadruple 
 
 11 
 
 14 
 
 18 
 
 20 
 
 21 
 
 22 
 
 
 
 
 
 
 
 
 
 
 
 None 
 
 35 
 
 35 
 
 35 
 
 35 
 
 35 
 
 35 
 
 
 10- 4 
 
 Dual 
 
 21 
 
 23 
 
 26 
 
 28 
 
 29 
 
 30 
 
 
 
 Quadruple 
 
 14 
 
 18 
 
 23 
 
 26 
 
 28 
 
 30 
 
 
 
 None 
 
 17 
 
 18 
 
 20 
 
 21 
 
 21 
 
 21 
 
 
 10" 3 
 
 Dual 
 
 14 
 
 16 
 
 18 
 
 20 
 
 21 
 
 21 
 
 Log-Normal 
 
 
 Quadruple 
 
 11 
 
 14 
 
 18 
 
 19 
 
 21 
 
 21 
 
 
 
 
 
 
 
 
 
 
 
 None 
 
 21 
 
 23 
 
 27 
 
 29 
 
 30 
 
 31 
 
 
 10" 4 
 
 Dual 
 
 17 
 
 20 
 
 24 
 
 27 
 
 29 
 
 30 
 
 
 
 Quadruple 
 
 14 
 
 18 
 
 23 
 
 26 
 
 28 
 
 29 
 
 probability. The gain is generally less for a log-normal-fading signal 
 than for a Rayleigh -fading signal. The higher the V, of the noise, the 
 less the gain from diversity. The lower the allowable error probability, 
 the greater the gain from diversity. 
 
 3.4. Nonfading Signal and Nonfading Interference With Gaussian Noise 
 
 Since the distribution of an undesired signal consisting of a non- 
 fading interfering signal and Gaussian noise follows the Nakagami-Rice 
 distribution, the error probability P can be expressed as 
 
 17 
 
P e =jQ(a,b), (5) 
 
 where Q(a, b) is a Marcum 1 s Q function (Marcum, 1950), a is the ratio 
 
 of the amplitude of the interfering signal to the rms noise voltage, and 
 
 b is the ratio of the amplitude of the desired signal to the rms noise 
 
 voltage. Both a and b are measured at the demodulator input. The 
 
 relation between SNR and error probability, parametric in SIR, is 
 
 given in figure 8. 
 
 As expected, interference necessitates an increase in required 
 
 SNR. At P = 10" 3 , for example, SIR of 3 dB, 6 dB, and 10 dB 
 e 
 
 necessitate increases of 9 dB, 5 dB, and 2 dB, respectively, in 
 required SNR. 
 
 3. 5. Nonfading Signal and Nonfading Interference With Atmospheric Noise 
 
 The element error probability is equal to one -half the value of 
 the APD, for the level corresponding to the signal amplitude, of the 
 composite waveform which consists of the interfering signal and 
 atmospheric noise. The APD of the composite waveform for a 
 specified level is equal to a two-dimensional integral of the probability 
 density function of the composite waveform outside a circle whose 
 center is the origin of the coordinate system and radius equals the 
 specified level. Assuming randomness between the phases of the 
 interfering signal and atmospheric noise, we can evaluate the integral 
 by numerical integration. The relations between SNR and error 
 probability, parametric in SIR, are given in figures 9, 10, and 11 for 
 the V values of 4, 6, and 10 dB, respectively. The effect of 
 interference is less severe in the presence of atmospheric noise than 
 
 in Gaussian noise. At P = 10" 3 and V =4 dB, SIRs of 3, 6, and 10 dB 
 
 e d 
 
 necessitate increases of 7, 3, and 1 dB, respectively, in required SNR, 
 
 18 
 
10 
 
 -I 
 
 2 10 
 ui 
 
 E 
 
 UJ 
 
 o 
 
 -Q 
 
 o 
 
 10 
 
 -3 
 
 10 
 
 10 
 
 10 
 
 -6 
 
 1 II 1 
 
 1 1 1 1 
 
 1 1 1 1 
 
 1 1 1 1 | 1 1 1 1 
 Nonfading Signal 
 Nonfading Interference 
 
 1 1 1 1 
 
 
 
 
 Gaussiar 
 
 i Noise 
 
 
 
 
 
 
 _ SIR- l dB 
 
 
 — 
 
 
 
 
 2dB 
 
 — 
 
 
 1 
 
 
 
 ___ 3dB 
 
 
 — 
 
 \ 
 
 
 
 €><i& 
 
 — 
 
 _ 
 
 in h n 1 
 
 
 
 
 
 
 IU d d 1 
 
 
 
 20dB M I I I 
 
 
 
 
 — 
 
 00 Jj 1 \ 
 
 
 
 — 
 
 1 1 1 1 
 
 1 1 1 1 
 
 111 ll ll 1 
 
 ll 1 
 
 1 1 1 1 
 
 1 1 1 1 
 
 -10 
 
 10 20 30 
 
 SNR (dB) 
 
 40 
 
 50 
 
 Figure 8. Element error probability in a single-channel NCFSK 
 system under stable conditions with Gaussian noise and non- 
 fading interference . 
 
 
 19 
 
o 
 
 l_ 
 LU 
 
 E 
 
 o 
 
 O 
 O 
 
 -I 
 
 10 
 
 ,-2 
 
 10 
 
 -3 
 
 10 
 
 10 
 
 «-6 
 
 1 1 1 1 
 
 1 II 1 
 
 1 1 1 1 
 
 1 1 1 1 
 
 1 1 1 1 
 
 1 1 1 1 
 
 
 
 
 Nonfading Signal 
 
 
 — -^^&s^^ 
 
 
 
 Nonfading Interference 
 Atmospheric Noise ( V d = 4 dB ) 
 
 — 
 
 — 
 
 
 
 
 
 — 
 
 — 
 
 
 
 
 -SIR = IdB 
 
 — 
 
 — 
 
 
 
 
 _— 2dB 
 
 — 
 
 
 
 
 
 
 \~— 3 dB 
 
 
 
 — 
 
 
 
 
 N^6dB 
 
 — 
 
 — 
 
 
 10 dB 
 
 "W \ 
 
 
 i — 
 
 
 
 — 
 
 
 00 
 
 
 
 \ — 
 
 
 
 1 1 1 1 
 
 1 1 1 1 
 
 1 1 1 1 
 
 1 1 l\\ 
 
 \l 1 \ l\ 
 
 1 \l 1 
 
 10 
 
 10 20 30 
 
 SNR (dB) 
 
 40 
 
 50 
 
 Figure 9. Element error probability in a single-channel NCFSK 
 system under stable conditions with atmospheric noise 
 CVrf = 4 dB] and nonfading interference. 
 
 20 
 
o 
 
 k_ 
 LxJ 
 
 c 
 E 
 
 o 
 
 O 
 O 
 
 10 20 30 
 SNR (dB) 
 
 Figure 10. Element error probability in a single- channel NCFSK 
 system under stable conditions with atmospheric noise 
 (\ '£ = 6 dB) and nonfading interference . 
 
 21 
 
10 
 
 -I 
 
 2 10 
 
 UJ 
 
 c 
 
 E 
 «> 
 
 UJ 
 
 o 
 
 jO 
 
 o 
 o 
 
 10 
 
 -3 
 
 10 
 
 H 
 
 10 
 
 -5 
 
 10 
 
 1 1 1 1 
 
 1 1 1 1 
 
 1 II 1 
 
 [III fill 
 
 Nonfading Signal 
 Nonfading Interference 
 Atmospheric Noise (V d = 10 dB) 
 
 1 1 1 1 
 
 — ^^s^ 
 
 
 
 
 
 — 
 
 — 
 
 
 
 
 .SIR = 1 dB 
 
 — 
 
 — 
 
 
 
 
 ^____ 2dB 
 
 — 
 
 — 
 
 
 
 
 3dB 
 
 — 
 
 — 
 
 
 
 
 \^6dB 
 
 — 
 
 — 
 
 
 
 10 dB y>\ 
 
 
 \ ~~ 
 
 — 
 
 
 
 OCK 
 
 
 \ \ — 
 
 1 1 II 
 
 1 1 1 1 
 
 1 1 1 1 
 
 1 1 1 1 
 
 1 1 l\\ 
 
 \ l\ \ 
 
 -10 
 
 10 
 
 20 
 SNR (dB) 
 
 30 
 
 40 
 
 50 
 
 Figure 11. Element error probability in a single- channel NCFSK 
 system under stable conditions with atmospheric noise 
 (V^ = 10 dB) and nonfading interference. 
 
 22 
 
compared with 9, 5, and 2 dB for Gaussian noise. The corresponding 
 
 values are 6, 3, and 1 dB for V , = 6 dB, and 6, 2, and 1 dB for 
 
 d 
 
 V J = 10 dB. 
 d 
 
 3. 6. Fading Signal and Fading Interference Without Noise 
 
 (Nondiversity Reception) 
 
 When both the desired signal and the interference are Rayleigh- 
 fading the element error probability is the same as that for a Rayleigh' 
 fading desired signal in the presence of Gaussian noise and no 
 interference, because the APD of the Rayleigh -fading interference 
 is the same as that of Gaussian noise. For a Rayleigh -fading signal 
 and Gaussian noise, Montgomery (1954) showed that the element error 
 probability is given by 
 
 P e-I ■ TTR- ' < 6 » 
 
 where R is the ratio of average signal power to average noise power. 
 This relation applies to our case, if R is interpreted as the SIR. 
 (Although the SIR is defined here as the ratio of median desired signal 
 power to median interfering signal power, it is equal to the ratio of 
 average desired signal power to average interfering signal power 
 when both signals follow an identical distribution law. ) This relation 
 for nondiversity is plotted in figure 12, section 3.7. 
 
 For log -normal -fading signals, desired and interfering, a 
 linear transformation of the coordinate system can reduce the double 
 integral in (1) to a single integral of the form 
 
 P e = I . Ff __2 LI . (7) 
 
 23 
 
where 
 
 00 
 
 1 f t 2 
 
 V2TT J 
 
 F(x) = — / exp(--)dt, (8) 
 
 X 
 
 m and m. are the median levels in dB of the desired and interfering 
 si 
 
 signals, respectively, and a and o. are the respective standard devi- 
 ations, also in dB. The o . is related to the fading range R by 
 
 S f X X 
 
 R = 2.563o. (9) 
 
 The relation between SIR and element error probability for log-normal 
 fading (fading range = 13.4 dB) and nondiversity is shown in figure 13, 
 section 3.7. 
 
 3.7. Fading Signal and Fading Interference Without Noise 
 
 (Diversity Reception) 
 
 We start with quadruple space and frequency diversity with a 
 selection-switching combiner: two modulated signals (both correspond- 
 ing to an identical digital signal) are transmitted simultaneously in two 
 channels separated in frequency, channel 1 and channel 2, and are 
 received on two antennas separated in space, antenna 1 and antenna 2; 
 the strongest signal among the four received signals is selected at the 
 diversity combiner. The probability of element error in an NCFSK 
 system is then given by 
 
 P =-Pr<fu >D , V = max (V ) } 
 
 e ci \ a^c^ a^c^ a^c^ a^c-j ' 
 
 + -Pr{u >D , V = max(V . .) } 
 
 c, \ a^Cg a i c s 3-i c s a i c j ' 
 
 24 
 
+ ^-Priu >D , V = max (V ) > 
 
 + ^-Pr|u >D , V = max(V ) > , 
 
 2 { &2 C 2 a.2Cg &2 C 2 a i c j 
 
 (10) 
 
 where 
 
 D_ = voltage of desired signal from antenna i in channel j 
 (i = 1, 2; j = 1, 2), 
 
 a i c J 
 
 U = voltage of unde sired signal (interference) from 
 
 antenna i in channel j (i = 1, 2; j = 1, 2), 
 
 a i c j 
 
 V = vector sum of D and U 
 
 a^c-i ii \ 1 
 
 = voltage of signal plus interference from antenna i 
 in channel j (i = 1, 2; j = 1, 2), 
 
 and max (V ) is a voltage having a maximum amplitude among the 
 
 a i c j 
 four voltages, V ., V , V , and V . From this general 
 
 a l c l 3-1 c S a 2 c l 3-aCg 
 
 form we shall derive three useful relations. 
 
 First, we assume quadruple space and frequency diversity- 
 reception with an interfering signal in channel 1 only. Since no inter- 
 fering signal exists in channel 2, both U and U are zero and 
 
 a i c a &-z c 2 
 
 cannot exceed D and D , respectively; therefore, the second 
 
 a i c a a 2 c 2 
 
 and the fourth terms of (10) are zero. Since, in general, D and 
 
 a 2 c l 
 
 U follow the same probability distributions as D and U , 
 
 a 2 c 1 a 1 c l a^ 
 
 respectively, the third term of (10) should be equal to the first term. 
 Therefore, we have 
 
 25 
 
P = Pr < U >D , V = max (V ) > 
 
 G \ cl^C^ ^*1^1 ^1^1 1^1 
 
 = Pr < U >D , V > D , 
 
 V a^C]_ cL-j^C^ cL^C^ &]_Cg 
 
 V > V , V > D > (11) 
 
 a^C^ 3-3^1 3-1^1 3-2^2 ' 
 
 as an expression of element error probability in this case. 
 
 Second, simplifying the reception scheme, we assume dual 
 frequency diversity reception with an interfering signal in channel 1 
 only. Since we do not have antenna 2, the third and the fourth terms 
 of (10), based on voltages from antenna 2, are zero. In addition, 
 since no interference is present in channel 2, the second term of (10) 
 is zero. Therefore, we obtain 
 
 P = - Pr {u > D , V = max(V ) > 
 
 e Li ' a^Ci a^c^ a^c^ a;c^ " 
 
 = - Pr { U > D , V > D > 
 
 for this case. 
 
 (12) 
 
 Finally, we consider dual space diversity with an interfering 
 signal. Since we do not have channel 2, the second and the fourth 
 terms of (10), based on voltages in channel 2, are zero. In addition, 
 as in quadruple diversity reception, the third term of (10) is equal 
 to the first term; therefore, we have 
 
 P = Pr < U > D , V = max(V ) [ 
 
 = Pr fu >D , V > V } 
 
 \ a^Ci a^c^ a^c^ agC^ * 
 
 (13) 
 
 for this case. 
 
 26 
 
Using (11), (12), and (13), we evaluated the probability of element 
 error for quadruple space and frequency diversity, dual frequency 
 diversity, and dual space diversity, respectively. A Monte Carlo 
 method was used for the evaluation, which simulated error performance 
 tests with a total of 10 6 bits for each type of diversity. Results of 
 these evaluations are shown for Rayleigh fading in figure 12 and for 
 log -normal fading in figure 13. They are also compared with the 
 previously derived results for nondiversity reception. Note that the 
 gain in required SIR expected from the use of diversity depends on the 
 type of fading of signal and interference. Only a small gain can be 
 expected when both desired and interfering signals are fading with a 
 log -normal distribution; for example the gain expected from quadruple 
 space and frequency diversity instead of nondiversity is 6 dB at an 
 element error probability of 10" . 
 
 3. 8. Fading Signal and Fading Interference With Noise 
 
 Except for some special cases, the element error probability 
 must be computed by numerically integrating the joint probability den- 
 sity function of the desired signal, interfering signal, and noise. Per- 
 haps, the Monte Carlo method would be useful for this and might be 
 mandatory for diversity reception. Even with the Monte Carlo method, 
 the computation is expensive and time consuming, if we try to obtain 
 data for numerous combinations of input parameters, such as types of 
 fading, diversity techniques, noise parameter, SNR, and SIR. 
 
 For Rayleigh -fading desired and interfering signals and Gaussian 
 noise with nondiversity reception, the element error probability can be 
 calculated analytically. Since the probability distribution function of 
 the composite wave of a Rayleigh -fading interference and Gaussian 
 noise is the same as that of Gaussian noise, the element error 
 probability can be computed by (6). In this case, however, we must use, 
 
 27 
 
UJ 
 
 c 
 <u 
 
 E 
 u 
 o 
 
 -Q 
 o 
 
 -O 
 
 o 
 
 10 20 30 
 SIR (dB) 
 
 Figure 12. Element error probability in a single- channel NCFSK 
 system under Ray leigh- fading conditions with Ray leigh- fading 
 interference and no noise. 
 
 28 
 
10 20 30 
 SIR (dB) 
 
 Figure 13. Element error probability in a single-channel NCFSK 
 system under log-normal fading conditions with log-normal 
 fading interference and no noise. 
 
 29 
 
as the value of R, the ratio of the average desired signal power to the 
 average power of the composite wave of the interference and noise. 
 
 For a Rayleigh -fading desired signal with nondiversity reception, 
 the element error probability is almost independent of the probability 
 distribution function of undesired signals in the lower error probability 
 region. It depends only on the ratio of the desired signal power to 
 the total power of undesired signals. Therefore, the element error 
 probability in the lower error probability region can be read from figure 
 2, if we use, as the SNR in the abscissa of this figure, the ratio of the 
 median desired signal power to the average power of the composite 
 signal of the interference and noise. 
 
 Except for a Rayleigh -fading desired signal with nondiversity 
 reception, the element error probability is higher with a higher V 
 value in the range of error probability of our interest, say 10 " ,i or less. 
 Since a composite wave of interference and noise is generally less 
 impulsive than the noise alone, each curve in figures 3 to 7 gives an upper 
 bound of the element error probability for a particular value of V , if 
 the SNR used is the same as used in the preceding paragraph. 
 
 4. EFFECTS OF ADVERSE PROPAGATION CONDITIONS 
 ON THE SYSTEM PERFORMANCES 
 
 HF signals propagated by the ionosphere suffer a number of 
 peculiar distortions that affect the selection of a modem and the design 
 of the telecommunication network. For some types of distortion both 
 NCFSK and TDPSK systems deteriorate similarly; other types of 
 distortion affect these systems in different ways. The effects of 
 ionospheric distortions on NCFSK and TDPSK systems have been 
 analyzed in detail by Bello and Nelin (1962, 1963, 1964) and compre- 
 hensively surveyed from the standpoint of FDM data transmission 
 
 30 
 
systems design (Akima, 1967). A discussion of these effects in 
 relation to the IGOSS, and of a serious drawback inherent in a TDPSK 
 system, follows. 
 
 4. 1. Multipath Time -Delay Spread 
 
 Complex paths with both low and high rays having a different 
 number of ionospheric and ground reflections are a major source of 
 multipath long-time delay spread. Multipath spread is a function of the 
 distance between the transmitter and the receiver and of the multipath 
 reduction factor, defined as the ratio of the operating frequency to the 
 highest of the classical MUFs (maximum useful frequencies) for simple 
 E, F l} and F 2 modes (Salaman, 1962). A spread of 1 to 2 ms is 
 fairly common. The spread becomes smaller and smaller as the multi- 
 path reduction factor approaches unity. However, in a scheduled 
 operation with a limited choice of operating frequencies, as for the 
 IGOSS, it is virtually impossible always to achieve operation with a 
 multipath reduction factor close to unity. 
 
 To minimize the deleterious effect of multipath time -delay 
 spread in both NCFSK and TDPSK systems, we must make the element 
 length relatively long compared with the multipath time -delay spread. 
 For this reason the minimum element length is usually about 10 ms, 
 resulting in a maximum modulation rate of approximately 100 bauds. 
 For the IGOSS, the final recommended choice of element length will 
 depend on the above factors and on considerations of overall system 
 efficiency, i. e. , maximizing the total amount of usable information that 
 can be obtained in a given time and with a given number of operating 
 frequencies and channels. 
 
 31 
 
4. 2. Frequency Fluctuations 
 
 The instantaneous frequency of a received signal fluctuates due 
 to propagation medium characteristics (Doppler effects) regardless of 
 the stability of the transmitter frequency. Long-term changes in 
 frequency (10 min or longer) are relatively small, usually less than a 
 fraction of 1 Hz, but short-term changes (a few minutes or shorter) 
 can be much greater and may be as much as tens of Hz under very 
 severe conditions (Davies and Baker, 1966). 
 
 Frequency changes in signals in adjacent channels (originating 
 from different locations) are not necessarily in the same direction, 
 because the propagation paths for the two signals are different. Thus, 
 in either an NCFSK system or a TDPSK system, channel spacings 
 wider than those commonly used in FDM modems are recommended 
 for circuits where such severe conditions may be expected (e.g., 
 auroral and trans equatorial circuits). 
 
 Another logical solution is that propagation paths exhibiting 
 severe frequency fluctuations should be avoided as much as possible 
 when the configuration of the IGOSS is designed. In practice, this is 
 limited to choosing shore station locations and to considering these 
 effects in preparing operational schedules. 
 
 To combat the effect of frequency fluctuations on an NCFSK 
 system, a wider frequency shift and, thus, a wider channel bandwidth 
 than those required by the modulation rate are necessary. Although 
 they require a higher signal level for maintaining the same SNR, these 
 requirements do not conflict with other constraints, such as the longer 
 element length dictated by multipath time -delay spread. Thus, for an 
 NCFSK system, adverse propagation condition can be accommodated at 
 the cost of a decreased data signalling capacity in the 3 -kHz bandwidth 
 
 32 
 
and increased transmitter power. For this system, one possibility- 
 would be for severe circuits to use twice the values of deviation and 
 channel bandwidth used for less severe circuits. 
 
 Frequency fluctuation created by transmitter and receiver 
 frequency instabilities is discussed in section 5. 
 
 4. 3. Phase Fluctuations 
 
 Several observations of phase fluctuation of signals propagated 
 by the ionosphere have been reported (Koch and Beery, 1962; Porter 
 et al. , 1963; David et al. , 1965). They indicate an increase in phase 
 fluctuations with an increase in the fading rate for a given sampling 
 interval (e. g. , 5 or 10 ms) and with an increase in the sampling 
 interval for a given fading rate. For example, the probabilities that 
 phase fluctuations exceeded 90 for sampling intervals of 8 ms and 
 20 ms over a transauroral path were 1 percent and 5 percent, 
 respectively, when the fading rate was 7 Hz (Koch and Beery, 1962). 
 
 Phase fluctuations limit the maximum element length that can 
 be used in TDPSK systems, because in these systems decisions are 
 made by comparing the phases of two successive elements. To avoid 
 the harmful effect of phase fluctuations in TDPSK systems, element 
 length must be kept as short as possible. The minimum length, how- 
 ever, is limited by the multipath time-delay spread in data trans- 
 mission channels over HF ionospheric paths, as noted in section 4. 2. 
 The fact that adverse propagation conditions impose these basically 
 conflicting requirements is perhaps the most serious drawback of 
 TDPSK systems for data transmission over HF ionospheric paths. 
 It is very difficult, if even possible, to find an element length suitable 
 for the range of propagation conditions that may be encountered in the 
 IGOSS. Furthermore, we cannot escape from these two contradictory 
 
 33 
 
requirements, even if we are willing to reduce the data signalling rate 
 in the 3 -kHz bandwidth. 
 
 Frequency fluctuation or offset caused by unstable or inaccurate 
 transmitter and receiver frequencies can be translated into phase 
 fluctuation and phase offset. This is discussed in section 5. 
 
 5. FREQUENCY STABILITY 
 
 Recent advances in frequency stabilization techniques allow us 
 to obtain, at an increased cost for a higher frequency stability, 
 oscillators that are as stable as desired for most communication 
 applications. Therefore, frequency stability to be specified for a 
 transmitter or a receiver should be determined as a compromise 
 between the loss in bandwidth use (or system performance) due to 
 increased frequency tolerances and the increased cost for higher 
 stability of the oscillator frequency. 
 
 In this study we will consider only frequency stability of a 
 platform transmitter, because the frequency of a shore -station receiver 
 can be stabilized to the desired extent at a small cost compared with 
 the total cost of the shore station facilities. 
 
 The following discussions suggest that the frequency stability 
 to be specified for the platform transmitters lies between approximately 
 one part in 10 7 and one part in 10 s . The oscillator for 10~ 7 stability 
 is expensive but achieves high efficiency in bandwidth use. The 
 oscillator for 10" 6 stability is less expensive but can only be used with 
 a penalty in bandwidth and, therefore, in system transmission time. 
 A final decision should be based on many factors, some of which are 
 beyond the scope of this study. 
 
 34 
 
5. 1. Frequency Tolerance for NCFSK Systems 
 
 The CCIR (1967d) gives a value of 3 Hz as the frequency 
 tolerance required for reception without automatic frequency control 
 of A7 emissions (multichannel voice -frequency telegraphy) in the 4 to 
 29.7 MHz band. This tolerance corresponds to a frequency stability 
 of 1.4 x 10~ 7 at 22 MHz, which is the highest frequency allocated for 
 the IGOSS, by the World Administrative Radio Conference (WARC) held 
 in Geneva in 1967. A frequency stability of 10" can be achieved within 
 state-of-the-art techniques; therefore, this is one of the prospective 
 values for the IGOSS. 
 
 Even though it can be achieved, a frequency stability of 10" 
 requires a high quality frequency source. We shall discuss whether 
 it is possible, and if so to what extent, to relax the tolerance beyond 
 3 Hz for the IGOSS. 
 
 Montgomery (1954) showed that, as long as the instantaneous 
 
 signal amplitude exceeds the instantaneous noise amplitude, the 
 discriminator input frequency cannot deviate from the signal frequency 
 by more than 0. 5 R Hz, where R is the modulation rate in bauds. The 
 CCIR (1967b) has reported the optimum deviation for a modulation rate 
 of R bauds to be ± 0. 4 R Hz for HF radio circuits, with a required 
 minimum bandwidth (at the -3 dB points) of R Hz. This indicates that 
 the probability that the discriminator input frequency deviates from 
 the signal frequency by more than 0.4 R Hz is negligibly small, though 
 not zero, as long as the instantaneous signal amplitude exceeds the 
 instantaneous noise amplitude. Using numerical examples we now 
 discuss the possibility of relaxing the tolerance. 
 
 The first example we assume is an NCFSK system with a fre- 
 quency deviation of ±42. 5 Hz, a channel separation of 170 Hz, and a 
 channel filter bandwidth of 120 Hz at the -3 dB points. These values of 
 
 35 
 
frequency deviation and channel separation are the same as in FDM 
 NCFSK systems, as recommended by the CCIR (1967b), operating at a 
 modulation rate of approximately 100 bauds over HF radio circuits. 
 In the assumed system with a deviation of ± 42. 5 Hz and a bandwidth 
 of 120 Hz, the noise bandwidth is slightly wider than that of the 
 optimum system, but since the difference is small, we may assume, to 
 a first approximation, that a frequency error of 42. 5 - 0. 4 R = 
 42,5 - 40 = 2. 5 Hz would result in negligible degradation of the per- 
 formance of our system. This value is close to the value of 3 Hz given 
 by the CCIR (1967d). 
 
 To relax the frequency tolerance of this system without increas- 
 ing the error probability, we must reduce the modulation rate while 
 keeping the bandwidth the same. Since the modulation index is now very 
 different from the optimum system it is better to rely on the criterion 
 derived by Montgomery (1954). In this case, the frequency tolerance is 
 equal to 42. 5 - 0. 5 R Hz, i. e. , 42. 5 - 37. 5 = 5 Hz and 42. 5 - 25 = 17. 5 Hz, 
 when the modulation rate is reduced to 75 and 50 bauds, respectively. 
 
 These values of the frequency tolerance apply when the signal 
 frequency deviates toward the center of the channel. However, when the 
 signal frequency deviates toward the edge of the channel filter, the 
 signal also degrades. The tolerance against the frequency error in 
 this direction can be estimated as follows. As quoted earlier, the 
 required minimum bandwidth (at the -3 dB points) is R Hz for the 
 optimum deviation of ± 0.4 R Hz. Therefore, the difference between 
 the -3 dB point of the filter and the signal frequency must be at least 
 R/ 2 - 0.4 R = 0. 1 R Hz. In the assumed system with a 120 -Hz band- 
 width, the frequency error cannot exceed 120/ 2 - 0. 1 R - 42. 5 = 
 17. 5 - 0. 1 R Hz; therefore, we have 7. 5, 10, and 12. 5 Hz as the values 
 of the frequency tolerance for R = 100, 7 5, and 50 bauds, respectively. 
 
 36 
 
Combining these values with those obtained in the preceding paragraph 
 and taking the smaller one, we have 2. 5, 5, and 12. 5 Hz as the values 
 of frequency tolerance for the modulation rates of 100, 7 5, and 50 bauds, 
 respectively. These tolerances roughly correspond to frequency 
 stabilities of 10~ 7 , 2 x 10~ 7 , and 5 x 10~ 7 at the highest transmitting 
 
 frequency of 22 MHz . 
 
 The second example we assume is an NCFSK system with a 
 frequency deviation of ± 85 Hz, a channel separation of 340 Hz (twice 
 the values given in CCIR, 1967b), and a channel filter bandwidth of 
 240 Hz at the -3 dB point. In this system the bandwidth use is halved 
 and thus the total system transmission time is doubled, as compared 
 with the system having a ± 42. 5 Hz frequency deviation. Since the 
 channel bandwidth is doubled, the transmitter power should be 
 approximately 3 dB higher to keep the error probability the same for 
 an equal modulation rate. Applying the same argument as in the 
 first example, we have 85 - 0. 5 R Hz as the tolerance for the frequency 
 error toward the center of the channel filter and 240/ 2 - 0. 1 R - 85 = 
 35 - 0. 1 R Hz for the error toward the edge of the channel filter. The 
 first equals 35, 47. 5, and 60 Hz for the modulation rates of 100, 7 5, 
 and 50 bauds, respectively, and the other equals 25, 27. 5, and 30 Hz 
 for the same modulation rates. Thus, the frequency tolerance in this 
 example is determined only by the latter and is equal to 25, 27. 5, and 
 30 Hz for the modulation rates of 100, 7 5, and 50 bauds, respectively. 
 These values of frequency tolerance roughly correspond to frequency 
 stabilities of 1. 1 x 10 " 6 , 1.2 xlO -6 , and 1. 4 x 10 " 6 at the highest 
 transmitter frequency of 22 MHz. 
 
 In summary, the frequency tolerance of 3 Hz can be relaxed to 
 12. 5 Hz at the cost of halving the modulation rate and thus doubling 
 the transmission time in the first system, and relaxed to 25 Hz at 
 
 37 
 
the cost of doubling the bandwidth in the second system, -which results 
 in doubling the total system transmission time and requires twice the 
 required transmitter power. 
 
 5. 2. Frequency Tolerance for TDPSK Systems 
 
 In a binary TDPSK system the difference between the received 
 frequency and the allocated frequency results in a loss in the signal 
 amplitude by a factor of cos (2n» Af • T) and, accordingly, a loss in the 
 required SNR by the same amount, where Af is the frequency difference 
 and T is the element length of the digital information signal. If we 
 assume that a 1-dB increase in the required SNR is allowed, Af • T can 
 be as large as 0. 075, and from this we obtain 7. 5, 5. 6, and 3. 8 Hz as 
 frequency tolerances for modulation rates of 100, 75, and 50 bauds, 
 respectively. Thus, in a TDPSK system the frequency tolerance 
 becomes more stringent when the modulation rate is reduced. When a 
 3-dB increase in the required SNR is allowed, Af • T can be 0. 125, and 
 the frequency tolerances can be relaxed to 12. 5, 9.4, and 6.25 Hz for 
 modulation rates of 100, 75, and 50 bauds, respectively. However, 
 Af » T increases as only a square root of the increase in the required 
 SNR in dB; thus frequency tolerance cannot be relaxed greatly by allow- 
 ing a large increase in the required SNR. 
 
 In a quaternary TDPSK system the frequency difference not only 
 causes a loss in the signal amplitude of the in-phase component by a 
 factor of cos (2n • Af • T), but also produces a quadrature -component 
 response of magnitude sin (2TT • Af • T). An error occurs in this system 
 when the amplitude of the quadrature component of signal plus noise 
 exceeds the amplitude of the in-phase component of signal plus noise. 
 The error in this system can be compared to the error in an NCFSK 
 system with a dual -filter demodulator, -where an error occurs when the 
 amplitude of an interference plus noise in a filter exceeds the amplitude 
 
 38 
 
of the desired signal plus noise in the other filter. Therefore, the 
 error probability in a quaternary TDPSK system can be discussed in 
 approximately the same manner as in an NCFSK system in the presence 
 of interference and noise. At Af • T = 0. 05 the loss in the signal 
 amplitude of the in-phase component is 0.4 dB, and the equivalent SIR 
 (that is the ratio of the in-phase component of the signal to the quadra- 
 ture component) is about 10 dB. Since an increase of 1 to 2 dB is 
 necessary in the required SNR by an interfering signal at the level of 
 SIR = 10 dB (see figs. 8 through 11), this value of Af * T is about the 
 maximum tolerance allowable in a quaternary TDPSK system. This 
 corresponds to frequency tolerances of 5, 3.75, and 2. 5 Hz at 
 modulation rates of 100, 7 5, and 50 bauds, respectively. 
 
 In summary, the frequency tolerance for a TDPSK system 
 operating at a modulation rate of 50 to 100 bauds lies in a range 
 between 2. 5 and 12. 5 Hz, depending on the modulation rate, the loss 
 in the required SNR that can be tolerated, and the number of phases 
 used in modulation. These values correspond to frequency stabilities 
 of 10~ 7 to 5 x 10" 7 for a transmitter frequency of 22 MHz. 
 
 5. 3 Frequency Stabilization 
 
 Quartz crystal oscillators are most widely used for stabilizing 
 transmitter frequencies. Rubidium -beam or cesium-beam oscillators 
 are also used when higher stabilities are required. The CCIR (1967c) 
 gives data on the present status of frequency stabilization techniques 
 that rely on quartz crystals. 
 
 There are two ways of generating a necessary frequency, one 
 by using a crystal cut for that frequency and the other by using a 
 frequency synthesizer. A frequency synthesizer has a master 
 oscillator of high stability and synthesizes the desired frequency from 
 
 39 
 
the single frequency source by multiplication, division, and addition 
 processes; the stability of the synthesized frequency thus equals that 
 of the source. 
 
 In services that require changes over a number of transmitter 
 frequencies, all frequencies can be (1) generated with individual 
 crystals, (2) synthesized from a master oscillator frequency by a 
 common variable -frequency synthesizing circuit, or (3) synthesized 
 from a master oscillator frequency by individual fixed -frequency 
 synthesizing circuits. The first method has been exclusively used 
 and is still widely used even now but becomes progressively more 
 expensive if a high stability is required. For a frequency stability 
 of 10 " 7 per year the estimated current cost is approximately $800 to 
 $1, 000 per required frequency. The second method costs $7, 000 to 
 $11, 000 including $2, 000 for the master oscillator; additional circuitry 
 for remote control of the frequency is generally required when this 
 method is used in an unmanned station. The third method costs 
 approximately $2, 400 + $200N including $2, 000 for the master 
 oscillator and $400 for the power supply for the synthesizing circuits, 
 where N is the number of frequencies: a remote control circuit for 
 this method is much simpler than that for the second method. The third 
 method, i. e. , a master oscillator and fixed -frequency synthesizing 
 circuits, seems to be the most promising for our application. 
 
 In an FSK transmitter, frequency stability also depends on the 
 way of modulation. If two stabilized frequencies are generated and one 
 of them is selected according to the binary information signal, 
 frequency stability of the FSK signal is equal to that of the stabilized 
 frequencies. But, if a variable frequency oscillator (VFO) output, 
 frequency -shift -keyed by the binary signal, is heterodyned with a 
 stabilized oscillator output, the stability of the FSK signal may be 
 
 40 
 
worse than that of a stabilized oscillator. Because of a possible 
 service interval of 1 year, this last method does not look very 
 attractive for the IGOSS. 
 
 Assuming that an NCFSK signal is to be transmitted in one of 
 15 channels (such as recommended by the CCIR, 1967b) in one of the 
 six HF bands allocated by the WARC in 1967, we shall briefly discuss 
 several possible schemes of generating the platform transmitter 
 frequencies. Generating an NCFSK signal is equivalent to generating 
 one of 6x15x2 = 180 possible frequencies f ( i = 1, 2, . . . , 6; 
 j = 1, 2, . . . , 15; k = 1,2) according to the band-selecting signal i, the 
 channel-selecting signal j, and the binary data signal k. 
 
 One extreme scheme is to generate all possible frequencies 
 simultaneously and to select one of them. All the frequencies could 
 be generated separately from individual crystals or synthesized by 
 appropriate fixed -frequency synthesizing networks. But, in any case, 
 this scheme has redundant circuit construction and does not seem to 
 be practical. 
 
 The other extreme scheme is to use a single variable -frequency 
 synthesizer. Although only one master oscillator is required in this 
 scheme, a variable -frequency synthesizing network is rather expensive 
 ($7, 000 - $11, 000, including the master oscillator), and the control 
 circuitry is also fairly complicated. 
 
 Another scheme can be as follows: one frequency is selected 
 by a band -selecting switch out of six that correspond to the six HF 
 bands; one by a channel -selecting switch out of 15 frequencies 
 corresponding to the 15 channels; one by a data signalling switch 
 (frequency -shift -keyer) out of two frequencies corresponding to "0" 
 and "1" of the binary data signal. The output signals from the three 
 switches are then mixed so that the frequency of the output signal is 
 
 41 
 
the sum of the frequencies of the three signals. The necessary 
 6 + 15 + 2 = 23 frequencies for this scheme can be synthesized from a 
 master oscillator frequency by 23 fixed -frequency synthesizing net- 
 works. The cost is estimated at $7, 000, which includes $2, 000 for 
 a master oscillator, $200 for each synthesizing network, and $400 
 for a power supply. 
 
 The scheme described in the preceding paragraph is not the only 
 possible one; it can be modified in several ways. Since the 15 
 channel frequencies are separated by equal intervals, the number of 
 necessary frequencies can be reduced by generating the 15 channel 
 frequencies in two steps. For example, we can generate a subgroup 
 of five frequencies and heterodyne the selected frequency with one of 
 three frequencies. The number of frequencies required is then 
 6 + 5 + 3 + 2= 16, reducing the estimated cost to approximately 
 $5, 600, a substantial saving over $7, 000. 
 
 6. SELECTIVITY OF RECEIVERS 
 
 The selectivity of a receiver is a measure of its ability to 
 discriminate between wanted and unwanted signals. Criteria for 
 establishing the selectivity of a receiver and definitions to be used for 
 the purpose of studying the selectivity are recommended by the CCIR 
 (1967a). Some data for the selectivity characteristics of various 
 classes of receivers and several methods of measuring selectivity 
 are also described by the CCIR (1967a, e, f). 
 
 This selectivity is expressed in terms of the single -signal 
 selectivity and the effective selectivity. The single -signal selectivity 
 represents the selectivity in the linear region; it can be studied by 
 measuring, with one signal, the passband, attenuation- slope, image- 
 
 42 
 
rejection ratio, intermediate -frequency rejection ratio, and other 
 spurious -response rejection ratios. The effective selectivity includes 
 the effects of amplitude nonlinear ity; it can be investigated by 
 measuring, with the wanted signal and an unwanted signal both present, 
 blocking and adjacent -signal selectivity (adjacent -channel selectivity, 
 if there is regular channelling); and by measuring, with two unwanted 
 signals present, radio -frequency intermodulation distortion. 
 
 In this section we discuss requirements for the selectivity of 
 IGOSS shore -station receivers and design factors necessary to meet 
 those requirements. 
 
 6. 1. Requirements for the Selectivity of IGOSS 
 Shore -Station Receivers 
 
 Since the HF bands allocated to the IGOSS have a 2 50 -Hz guard 
 band on each side of them, it is easy to protect the IGOSS signals 
 against unwanted signals of other services. 
 
 In the IGOSS, however, there is a peculiar problem of inter - 
 channel interference. An IGOSS shore station must receive simul- 
 taneously, in a common 3-kHz bandwidth, as many signals as there 
 are channels in the band. Since these signals come from geographically 
 dispersed platforms, they fade independently and also may differ 
 considerably in their median levels. Therefore, the requirements for 
 selectivity are more severe in the IGOSS than in a frequency -division - 
 multiplexed (FDM) data transmission system, where the signals in the 
 different channels have an equal median level. 
 
 The difference in the median level of signals received in a 
 common 3-kHz HF band is thus an important factor for specifying the 
 selectivity of IGOSS shore -station receivers. As an example we have 
 the following result computed by Hatfield and Adams (private 
 
 43 
 
communication). They assume a system operation plan in which 
 a total of 5501 platforms are randomly located all over the world and 
 each platform transmits observed data to one of 18 shore stations, 
 in one of 13 channels preassigned to each platform, in one of the six 
 HF bands. Using the method described by Barghausen et al. (1969), 
 they computed the reliability for each possible combination of platform, 
 shore station, and HF band; on the basis of this, they selected a 
 systems assignment plan that would provide maximum total system 
 reliability. Circuits using the same shore station, frequency band, 
 and channel were assigned sequential time slots as the assignments 
 occurred. With the system operation plan thus determined, statistics 
 were taken of the median level difference between pairs of adjacent 
 channels. The result was that the probabilities that the level difference 
 exceeds 6, 9, 12, and 15 dB are approximately 15, 5, 2, and 1 percent, 
 respectively. 
 
 This result indicates that significant level differences can 
 occur if time slots are assigned sequentially without any precautions. 
 However, we can limit large level differences by reassigning time 
 slots, with a possible penalty of increased number of time slots and, 
 accordingly, increased total transmission time. Since the probability 
 of occurrence of large level differences, such as 15 dB or greater, 
 is relatively small, we can expect that the possible increase in the 
 number of time slots resulting from the elimination of level differences 
 greater than 15 dB is also small. Therefore, we assume that the 
 maximum difference in the median signal levels between a pair of 
 adjacent channels will not exceed 20 dB, including 5 dB for the 
 uncertainty in predicting median signal levels. 
 
 Another important factor to be considered in specifying the 
 selectivity of a receiver is s the maximum tolerable effect of interference. 
 
 44 
 
Although the CCIR (1967a) uses the output power from the receiver 
 to represent the effect of interference in the definition of the 
 selectivity of a receiver, this is valid only for a receiver used for 
 analog -signal transmission. For a receiver used for digital data 
 transmission, the power at a point just before the decision -making 
 circuit could be used. But here we use the power at the demodulator 
 input to represent the effect of interference, because, from a practical 
 standpoint, it is better suited for this purpose. Consequently the 
 maximum tolerable amount of interference is in terms of the minimum 
 SIR required at the demodulator input. 
 
 The minimum required SIR is a function of the allowable error 
 probability and can be determined from figures 12 and 13. These 
 figures indicate that the SIR at the demodulator input should be 
 approximately 20 and 25 dB for element error probabilities of 10"' 
 and 10 ~ 4 , respectively, when dual space diversity is used. 
 
 If all the wanted and unwanted signal levels were always in the 
 linear region of the receiver, 45 -dB attenuation of signals in other 
 channels would be sufficient for specifying the selectivity of an IGOSS 
 shore -station receiver, because it would secure an SIR of 25 dB 
 against an unwanted signal 20 dB stronger than the wanted signal. 
 Since very strong signals may sometimes occur, this is insufficient, 
 and the selectivity should be specified in terms of effective 
 selectivity. 
 
 To specify effective selectivity, we must specify the level of 
 the wanted signal at which to make the measurements. In our case 
 the level of the wanted signal should be 20 dB less than the maximum 
 signal level expected. This, in turn, depends on the overall plan 
 of the IGOSS, and no reliable data are now available. 
 
 45 
 
Adjacent -channel selectivity and intermodulation distortion are 
 the most important factors governing systems performance, and as a 
 result of the above discussion the following two requirements are 
 recommended to specify the selectivity of IGOSS shore -station 
 receivers: 
 
 (1) The SIR at the demodulator input shall be 25 dB or higher, 
 when a wanted signal is applied to one channel and an un- 
 wanted signal is applied to another channel at a level of 
 20 dB stronger than that of the wanted signal. 
 
 (2) The SIR at the demodulator input shall be 25 dB or higher 
 when a wanted signal is applied to one channel and two un- 
 wanted signals are applied, one to the adjacent channel on 
 one side and the other next to the adjacent channel on the 
 same side, both 20 dB stronger than the wanted signal. 
 
 The wanted signal level in both cases will be specified as soon as the 
 overall plan of the IGOSS has been established in sufficient detail. 
 
 6. 2. Receiver Structure 
 
 In general, the selectivity of a receiver is not only limited by 
 the characteristics of the filter but also limited by unavoidable 
 amplitude nonlinearities, e. g. , cross -modulation of the wanted signal 
 by strong unwanted signals. Since the signal level generally increases 
 as it comes closer to the receiver output, there is more chance of a 
 loss of effective selectivity caused by amplitude nonlinearities in the 
 later stages of a receiver. As a general rule, the CCIR (1967a) 
 recommends that the filters that determine the selectivity shall be 
 included as near as possible to the receiver input, and the amplifying 
 stages preceding the filter shall be sufficiently linear, to avoid 
 significant loss of selectivity. For IGOSS shore -station receivers, 
 
 46 
 
this principle implies that a signal in each channel should be 
 separated from signals in other channels as early as possible. This 
 also implies that the number of common circuits shared by the signals 
 in the different channels should be restricted as much as possible. 
 
 There are several choices; in principle, a signal in each 
 channel can be separated at any stage in the receiver. In practice, 
 however, we have two practical schemes from which we must choose 
 one. One choice would be to separate each channel immediately 
 after the first frequency converter, i. e. , at the input of the inter- 
 mediate-frequency amplifier (IFA); the other choice would be to 
 separate the channels at audio frequency, as done in most FDM data- 
 transmission receivers. The first choice may be called the multiple 
 IFA scheme, because each channel has its own IFA; the second may 
 be called the common IFA scheme, because all channels share a 
 common IFA. 
 
 Although it would be difficult to prove the necessity of the 
 multiple IFA scheme for the IGOSS shore-station receivers, there is 
 no question that better effective selectivity can be achieved with this 
 scheme than with the common IFA scheme. Moreover, adopting the 
 multiple IFA scheme will allow the design of automatic gain control 
 (AGC) circuitry providing independent control of each channel. 
 Therefore, we recommend adopting the multiple IFA scheme. 
 
 The difference in cost between the two schemes is mainly 
 the difference in the type of filters and the number of IFAs required. 
 Since the center frequency of the filters for the multiple IFA scheme 
 is about 1 MHz and the bandwidth is about 100 Hz, a crystal filter 
 or a mechanical filter is required. The required selectivity suggests 
 using a five -pole or six -pole crystal filter; its cost is roughly 
 estimated at $150 to $200, compared with the cost of an audio -frequency 
 
 47 
 
filter for the common IFA scheme roughly estimated at $70 to $100. 
 The second factor is that a multiple IFA receiver must have the same 
 number of IFAs as the number of channels. This extra cost is 
 roughly estimated at $50 to $70 per channel. 
 
 Besides the protection against possible interchannel inter- 
 ference within the IGOSS, we must protect the IGOSS signals against 
 other unwanted signals. Therefore, we recommend that crystal 
 filters of approximately 3- to 4 -kHz bandwidth be used at the input 
 of the receiver for each of the six HF bands, regardless of the 
 schemes used for the receiver. Modern techniques enable designing 
 such filters to have an insertion loss of only 1 to 2 dB. 
 
 The discussion given in this section applies equally to NCFSK 
 or TDPSK. 
 
 7. CONCLUSIONS 
 
 To provide basic data relevant to selecting a modem and the 
 design of the IGOSS telecommunication network, we have discussed 
 modulation techniques for digital data transmission over HF 
 ionospheric paths and some related topics. 
 
 On a preliminary basis, NCFSK and TDPSK systems were 
 preferred; the performance of these systems in the presence of noise 
 and/ or interfering signal under normal propagation conditions has 
 been analyzed. These analyses were made on a fairly general basis: 
 nonfading and fading signals (either desired or undesired), Gaussian 
 and atmospheric noise, nondiversity and diversity reception. 
 
 The effects of adverse propagation conditions, peculiar to HF 
 ionospheric paths, on these preferred systems were also considered. 
 The results indicate that a TDPSK system has a serious drawback when 
 
 48 
 
applied to the IGOSS, where propagation conditions may be widely- 
 different and adverse conditions are sometimes likely to occur. 
 
 Frequency tolerances allowed for the preferred modulation 
 systems were studied, and some frequency stabilization techniques 
 that can be used for the IGOSS platform transmitters are given. 
 Requirements for the selectivity of the IGOSS shore -station receivers 
 were determined, and the multiple IFA scheme is recommended as 
 a receiver structure that satisfies these requirements. 
 
 8. ACKNOWLEDGMENTS 
 
 The author expresses his deep appreciation to many colleagues 
 in the ESSA Research Laboratories: T. deHaas for directing the 
 study; J. E. Adams and D. N. Hatfield for their valuable data 
 concerning the IGOSS; G. G. Ax for his helpful discussions on the 
 whole area of this study; W. D. Bensema for his information on a 
 computer program; E. C. Bolton, R. M. Coon, L. J. Demmer, 
 and C. C. Watterson for their valuable data concerning the state-of- 
 the-art electronics; and A. F. Barghausen and A. D. Spaulding for 
 their constructive criticisms on this report. 
 
 49 
 
9. REFERENCES 
 
 Akima, Hiroshi (1967), Speeding-up of HF radiotransmissions and 
 
 broadcasts (World Weather Watch Study T. 21), ESSA Tech. 
 Rept. IER 34-ITSA 34 (Clearinghouse for Federal Scientific 
 and Technical Information, Springfield, Va. 22151). 
 
 Akima, Hiroshi, Gene G. Ax, and Wesley M. Beery (1969), Required 
 signal-to-noise ratios for HF communication systems, ESSA 
 Tech. Rept. ERL 131 -ITS 92 (U. S. Government Printing 
 Office, Washington, D. C. 20402). 
 
 Barghausen, Alfred F. , James W. Finney, L. Lee Proctor, and 
 
 Larry D. Schultz (1969), Predicting long-term operational 
 parameters of high-frequency sky-wave telecommunication 
 systems, ESSA Tech. Rept. ERL 110-ITS 78 (U.S. Govern- 
 ment Printing Office, Washington, D. C. 20402). 
 
 Bello, P. A. , and B. D. Nelin (1962), The influence of fading spectrum 
 on the binary error probabilities of incoherent and differen- 
 tially coherent matched filter receivers, IRE Trans. Commun. 
 Systems CS-10 , 160-168. 
 
 Bello, P. A., and B. D. Nelin (1963), The effect of frequency selective 
 fading on the binary error probabilities of incoherent and 
 differentially coherent matched filter receivers, IEEE Trans. 
 Commun. Systems CS-1 1, 170-186. 
 
 Bello, P. A., and B. D. Nelin (1964), Optimization of subchannel data 
 rate in FDM-SSB transmission over selectively fading media, 
 IEEE Trans. Commun. Systems CS-12, 46-53. 
 
 Cahn, Charles R. (1959), Performance of digital phase -modulation 
 communication systems, IRE Trans. Commun. Systems 
 CS-7 , 3-6. 
 
 CCIR (International Radio Consultative Committee) (1964), World 
 
 distribution and characteristics of atmospheric radio noise, 
 Rept. 322, CClR Green Book, a separate volume (International 
 Telecommunication Union (ITU), Geneva, Switzerland). 
 
 50 
 
CCIR (1967 a), Selectivity of receivers, Recommendation 332-1, 
 
 CCIR Green Book I_, 158-186 (ITU, Geneva, Switzerland). 
 
 CCIR (1967 b), Arrangement of voice -frequency telegraph channels 
 working at a modulation rate of about 100 bauds over HF 
 radio circuits, Recommendation 436, CCIR Green Book III, 
 60-61 (ITU, Geneva, Switzerland). 
 
 CCIR (1967 c), Frequency stabilization of transmitters, Rept. 180-1, 
 CCIR Green Book I_, 91-93 (ITU, Geneva, Switzerland). 
 
 CCIR (1967 d), Frequency tolerance of transmitters, Rept. 181-1, 
 CCIR Green Book I_, 93-96 (ITU, Geneva, Switzerland). 
 
 CCIR(1967e), Selectivity of receivers, Rept. 185-1, CCIR Green 
 Book I_, 209-215 (ITU, Geneva, Switzerland). 
 
 CCIR (1967 f), Multiple -signal methods of measuring selectivity, 
 
 Rept. 186-1, CCIR Green Book I_, 216-217 (ITU, Geneva, 
 Switzerland). 
 
 CCIR (19 67 g), Fading of signals propagated by the ionosphere, 
 
 Rept. 266-1, CCIR Green Book II, 325-336 (ITU, Geneva, 
 Switzerland). 
 
 David, F. , A. P. Baker, P. W. Rudmann, H. Sherman, and R. Zach 
 (1965), Analytical and experimental study of correlation func- 
 tions over HF circuits, Final Rept. , 1 July 1964 to 31 October 
 1965, Signal Corps Contract No. DA 28-043-AMC-00145 (E) 
 (Communication Systems, Inc. , Paramus, N.J. ). 
 
 Da vies, Kenneth, and Donald M. Baker (1966), On frequency variations 
 of ionospherically propagated HF radio signals, Radio Science 
 1_ (New Series), 545-556. 
 
 Dayton, David S. (1968), FM chirp communications: Multiple access 
 to dispersive channels, IEEE Trans. Electromagnetic 
 Compatibility EMC -10 , 296-297. 
 
 de Haas, T. (1965), A frequency-differential PSK modem for HF data 
 communications, First IEEE Annual Commun. Conv. Record, 
 135-136 (IEEE, New York, N. Y. ). 
 
 51 
 
Di Toro, M. J. , J. Hanulec, and B. Goldberg (1965), Design and per- 
 formance of a new adaptive serial data modem on simulated 
 time -variable multipath HF link, First IEEE Annual Commun. 
 Conv. Record, 769-773 (IEEE, New York, N. Y. ). 
 
 Koch, J. "W. , and W. M. Beery (1962), Observations of radio wave 
 phase characteristics on a high-frequency auroral path, 
 J. Res. NBS 66D , 291-296. 
 
 Marcum, J. I. (1950), Tables of the Q -functions, Rept. RM-753 
 (Rand Corp. , Santa Monica, Calif. ). 
 
 Montgomery, G. Franklin (1954), A comparison of amplitude and angle 
 modulation for narrow-band communication of binary-coded 
 messages in fluctuation noise, Proc. IRE 42, 447-454. 
 
 Porter, G. C. , C. J. Zarcone, and W. F. Walker (1963), Study of 
 fine grain fading and phase stability of multiple CW signals, 
 Rept. #3 for task 3, Signal Corps Contract DA 36-039, SC- 
 88943 (General Dynamics /Electronics, Rochester, N. Y. ). 
 
 Salaman, Roger K. (1962), A new ionospheric multipath reduction 
 
 factor (MRF), IRE Trans. Commun. Systems CS-10 , 221-222. 
 
 Zimmerman, M. S. , and A. L. Kirsch (1965), The AN/GSC-10 
 
 (KATHRYN): A variable rate digital data terminal for HF 
 radio communication, Eleventh NATCOM Sympo. , 95 
 (IEEE, New York, N. Y. ). 
 
 52 
 
 GPO 859-846 
 
PENN STATE UNIVERSITY LIBRARIES 
 
 AQQ0Q7202 
 
 DAI