66 2.26. 2.27. 2.28. 2.29, 2.30, 2.31, 2.32. 2.33. Spread Spectrum and CDMA. Amplitude 4 of the signal s(t; 4) = s(t) needs to be measured. The reference signal s(¢) for three cases is given in Figure 2.33. In which of these cases will the accuracy of measuring A be highest? vu (0) uh s(t) UR a) ce (a) (b) (c) Figure 2.33 Three forms of a reference signal Amplitude 4 of the signal s(; 4) = As(0) is measured. Someone is dissatisfied with the accuracy of the amplitude estimation. How many times should the duration ar the reference signal s() be increased in order to halve the standard deviation of the estimation of A, all the other parameters of s(¢) remaining the same? Amplitude 4 of the signal s(1;4) = As(2) is measured. The amplitude of the reference signal s() is doubled while its duration is halved. What happens to the standard deviation of the estimation of 4? The initial phase of the bandpass sign: (a) Carrier frequency of the signal is doubled? (b) Signal duration is doubled? (©) Signal amplitude is halved? (@) Signal amplitude is doubled and duration is reduced by four times? ‘The initial phase of the bandpass signal needs to be ‘measured. Three variants of the Signal envelope are given in Figure 2.33. In which ofthese cases wll the accuracy of phase measurement be highest? The initial phase of the LFM signal of Problem 2.22 is measured. Variation of which of the parameters A,fo, Wa,T,7 and in which direction will affect the Precision of the phase estimation? What happens to the standard deviation of the phase estimation when A,/o, W are all increased by V3 times while T and + are halved? Sketch the autocorrelation functions ofthe three signals shown in Figure 2.34 jie bandpass BPSK signal consists of three consecutive retangular pulses cach being of duration A. The phases of the first two ‘equal zero while the third phase is r. Sketch the autocorrelation function of the signal. ‘The matched filter for a rectangular baseband pulse of duration A is given. What Sort of circuitry should be added to it in order to obtain the matched filter for the signals of Figure 2.34? Sketch the response of the filter matched to signal (c) when this very signal inputs the filter. “Ney 8Merits of spread spectrum 107 3.3. In some system power SIR at the matched filter output degrades 101 times as compared to power SNR, while the system can preserve its operation capability with only two-times degrading SIR. What should be changed in the signal, SNR remaining constant, if (a) Only plain signals are allowed? (b) Signal peak power cannot be raised (what should the signal processing gain be in this case)? 3.4, In some system a band-elimination filter neutralizes the narrowband jammer. Due to this, the matched filter SNR degrades by 3B. (a) What should be done to the plain signal duration and amplitude if only 2% degradation in power SNR is tolerable (SNR in the absence of band-elimination is fixed)? (b) Is it possible to push degradation below 2% without increasing signal power? If ‘so, what should the signal processing gain be? 3.5. A system can operate with SNR no smaller than 104B, Due to a barrage jammer ‘SNR drops to —3dB. How could the signal parameters be changed to neutralize the jammer, if: (a) Only plain signal of the fixed energy is allowed? (b) Only plain signal of the same peak power can be used? (©) Peak power and energy of the signal are fixed with no other constraints? (d) Peak power of the signal is fixed and its bandwidth can be increased only 10 times? Find the processing gain in cases (c) and (d). 3.6. In conflict with a barrage jammer transmitter, a system increases signal duration by 4 times with simultaneous halving of signal power and widening of the bandwidth by 50 times. The jammer transmitter is capable of increasing its power no more than 13dB. Who will be the winner in this game? 3.7. A signal occupies two separate sub-bands of identical width Wo. Total signal energy is distributed between them in the proportion 9:16, In the receiver two matched filters process both sub-band parts and their outputs are combined optimally to maximize a resultant SNR. There is a barrage jammer transmitter. ‘What is the most harmful distribution of its total power between the signal sub-bands? Matched filter SNR for an intended receiver equals 14dB. The processing gain of the signal WT = 400. Find: (a) Ratio of power spectrum densities of the signal and background AWGN. (b) SNR at the integrator output of an interceptor radiometer. 3.9. In the BPSK data transmission system error probability per bit Pr = 1.5 x 10° is required. A system designer wants the SNR of the interceptor radiometer to be no greater than —10dB per one transmitted bit duration. What processing gain per one bit would be satisfactory?Discrete spread spectrum signals 145, where {Fi,,/=0,1, .....N ~ 1} is the frequency code sequence of the kth signal. It is obvious that computation of CCF is a just tallying up a number of coincident frequencies in the pair of signals time-shifted by m chip positions. 5.6 Processing gain of discrete signals Let us revisit the general model (5.1) to discuss the issue of the processing gain of a discrete signal. Suppose that all F; belong to the frequency alphabet of size M. in which consecutive frequencies are separated by a chip bandwidth providing orthogonality of chips with different frequencies. Then at cach of M available frequencies we have a signal subspace whose dimension is N, since we are fully free in selection of the amplitude- Phase code sequence, ic. N-dimensional vector a= (dp, 4),...,ay-1). Orthogonality of these subspaces means that the dimension of the whole signal space covering all Mf frequencies is MN. In Section 2.5 it was shown that the dimension of bandpass signal space coincides with the total time-frequency resource allocated to signals. Being interested only in spread spectrum signals, each occupying the whole available resource, we may predict that the time-frequency product of a discrete signal, ie. processing gain, equals MN. Tet us confirm this statement by a straightforward computation, setting A, = A. Estimating chip bandwidth as 1/A and taking into account that available bandwidth and time resources are then W = M/A and T = NA, we arrive at the result WT = MN Obviously, for APSK signals Mf = 1, and processing gain WT = N. Problems 5:1. A discrete signal of length WV = 5 has complex amplitudes ap = 1 +j, a, = —1 +j, a2 = 1 +J.a3 = —1—J, ag = 1~j and frequencies F; =0, i = 0, 1,2,3,4. Evalu- ate the phases and amplitudes of its chips and classify the signal by its modulation —~, mode. (5.2) A discrete signal has amplitude-phase and frequency codes a; = exp [jmi(i + 1)/2], ~~ Fi=0, ~1,0,1,.... Calculate the amplitudes and phases of its chips, Clas- sify the signal by its modulation mode. Is this signal periodic? If so, specily its period. 5.3. Prove the evenness of the periodic and aperiodic autocorrelation functions of code sequences of APSK signals. $4. An APSK signal is built of rectangular chips with A, =A and has the code sequence set by the vector a = (I, 1,0, 1,0,0, ~1). Calculate and sketch its aperiodic and periodic ACF. Do the same for the case A. = A/2, 5.5. What happens to the periodic and aperiodic ACF of an APSK signal under the following transformations of a code sequence: (a) Cyclic shift of elements? (b) Changing the signs of all elements? (©) Changing the signs of only the elements with even numbers? (@) Multiplying all elements by the same constant? (©) Mirror-like rearranging (ie, reading from right to left)?194 Spread Spectrum and CDMA than of radar systems, where the terminology stems from. For instance, a great number of regular algorithms for constructing Costas arrays [54], i.e. square (M = N) sonar arrays or FSK sequences having equal length and number of fre- quencies, are known. Problems 6.1. The frequency of a rectangular pulse drops linearly throughout its duration T = 10s from 110 to 90 MHz. Calculate the processir-g gain of the signal. What is the approximate duration of the signal at the matched filter output? Sketch the ambiguity function and ambiguity diagram. 6.2. The frequency of a rectangular pulse drops linearly over the first half of its duration T = 10ps from 110 to 90MHz and then grows linearly from 90 to 110 MHz over the second half, Calculate the processing gain of the signal. Sketch the ambiguity function and its low-level and high-level horizontal sections. 63. Calculate aperiodic and periodic ACF for a binary Barker code of length N = 11 ‘Try to do it the most economical way. 64, Take a periodic sequence {a;} of period N and form a new sequence {b,} picking each dth element of {a ia» Where multiplication in the index is modulo N. Such a transform is called the decimation of {a;} with index d. Prove that if {a)} has perfect periodic ACF and dis co-prime to N, {b;} alsa has perfect periodic ACF. 6.5. The binary {+1} code of length N = S has periodic ACF Ry(m) = 1,m # Omods. Values of its aperiodic ACF are Rq(1) = 0, Ra(2) = —1. Find R,(3) and R,(4). 6.6. The binary {+1} code of length V = 5 has constant component ay = +3 and Ral) Find Ry(m) and the rest of the values of Ra(m). 6.7. Cana binary {+1} code of an odd length N > 5 have R,(5) = -1? Can a binary code of an even length N > 6 have R,(6) = —1? For an arbitrary binary code formulate and prove the relation between parities of the three values: length N, shift m and level of Ra(m). 6.8. Isit possible for a binary {+1} sequence that Ra(m) = 1, Re(m + 1) = 3 for some m? What about parities of Ra(om) and Ra(om + 1)? 6.9. Is it possible for a binary {+1} sequence that R,(2) = 2, R,(2) = -1? 6.10. Suppose someone has found that each of the PSK sequences of length NV = 100 at his disposal has non-normalized periodic ACF taking on values +12 at some shifts m€ {1,2,...,N ~ 1}. Can a code with pg,max < 0.05 be present among them? 6.11. Construct a matched filter for a BPSK signal manipulated by a sequence {+-++} and show waveforms at its characteristic points, when the filter is ~. input by aperiodic and periodic versions of the signal. (G12) Construct a matched filter for a BPSK signal manipulated by a sequence {+—++———} and show waveforms at its characteristic points when aperiodic and periodic versions of the signal are applied to its input. 6.13, A student has calculated periodic ACF of the binary {+1} sequence of length 21 and obtained the following values: —Rj(m:) =9, Rp(m2) = -3, Ry(mmg) = ~5, Ry(ms) = —7, Ry(me) =7. Can all of these results be correct? Which (if any) are definitely incorrect?