CHAPTER

HYBRID MQAM-LFSK (HQFM) SIGNALING

Quadrature multiplexing techniques have been used for a long time both in analog and digital communications. One digital modulation scheme of this type is the quadrature amplitude modulation (QAM), where both in phase {I} and quadrature {Q} components, are modulated by an independent series of M1-ary and M2-ary rectangular pulses respectively. The sum of the both I- and Q- components yields M = M1M2 rectangular signal constellation. As M increases, the bandwidth efficiency is increased at the expense of power efficiency (measure of Eb/No). In contrast, with multi-dimensional, L-ary orthogonal frequency shift keying (FSK) signal sets, power efficiency increases with L at the expense of the bandwidth efficiency. The hybrid of both of the modulation schemes, referred to as hybrid MQAM/ LFSK modulation (HQFM), can offer the advantages of both techniques. The results obtained are quite favorable. This hybrid technique applies the principle of quadrature-carrier multiplexing to L-dimensional orthogonal constellations in order to enhance the bandwidth efficiency and at the same time retain the power efficiency of orthogonal signaling.

CHAPTER 3

HYBRID MQAM-LFSK (HQFM) SIGNALING

It is appropriate to compare the proposed HQFM scheme with QAM. Like QAM, HQFM yields a signal set of size Q (= ML) with a non-constant envelope and energy. HQFM is simultaneously power and bandwidth efficient modulation scheme where its error performance is always better than that of unfiltered QAM. Interestingly enough, however, under certain conditions (e.g., in terms of 99%-power containment bandwidth) the bandwidth efficiency of HQFM is approximately the same as that of QQAM. In this chapter, only the signaling properties of the proposed HQFM are discussed. Its error rate performance will be discussed in chapter 5 and 6. Before discussing the signaling properties, brief review of general digital modulation scheme (MPSK, MQAM and MFSK) in term of signal waveforms, energy characteristics, correlation or Euclidean distance properties is given. A literature review of hybrid modulation based on general modulation schemes is also given. Power spectral density, null-to-null bandwidth occupancy and bandwidth efficiency in terms of 90% and 99% power containment of HQFM is compared with conventional L- FSK, M-, Q- (Q = ML) QAMs in this chapter.

3.1

Brief Review of MPSK, MQAM and MFSK
In general, any M-ary signaling system, the waveform used to transmit the

information will be encoded as {sm(t); 1£m£M; 0£t£Ts} where M = 2n, n being number of bits per encoded symbol [Proakis: 89]. There bandpass generic representation, with initial phase {q0} and centre frequency {fc} is
xm (t ) = Re um (t )e j (2p fct +q0 )

{

}

(3.1)

The equivalent lowpass representation of equation (3.1) is
j f ( t ) +q ( t ) um ( t ) = Am (t )e ( m m ) g ( t )

(3.2)

where Im(t) = Am(t)ej(fm(t)+qm(t)) and g(t) are information bearing signal and pulse shaping filter respectively. For the sake of convenience, q0 = 0 and g(t) is assumed to be rectangular pulse shaping filter with period Ts = Tblog2M where Tb being bit duration in

then sm(t) corresponds to phase shift keying (PSK).umN] is equivalent complex valued lowpass N-dimensional signal vector. E = 1 and minimum Euclidean distance of 2Eg (1 .2) shows that Im(t) in the pth transmission interval. These M signals.1: Signal Space diagram of different MPSK If um(t) = ej(qm)g(t) = (cosqm+jsinqm)g(t). are also characterized by their energy 2 Em = ò sm (t )dt = 1 2 um (t ) . (p-1)Ts £ t £ pTs. 2 0 Ts (3. HYBRID MQAM-LFSK (HQFM) SIGNALING g (t ) = 2Eg Ts (3.6) (a)BPSK (b)QPSK (c)16-PSK FIGURE 3.cos 2p / M ) . which is defined as 2 d ij = si . may differ either in amplitude {Am(t)}.5) where um = [um1.e.um2.CHAPTER 3 seconds i. or phase {qm(t)} or frequency {fm(t)} or combination of all or some of these parameters and is chosen according to the designer’s need and requirement.3) Equation (3. 2 Eg . These signal waveforms.e.4) and mutually by their complex cross correlation coefficients 1 rij = Ei E j ò Ts 0 si (t ) s (t )dt = * j ui g u*j ui u j (3.s j = Ei + E j . 1£m£M. having equal energies i. qm(t) = qm = 2p(m-1)/M are the possible M phases of the carrier that convey the transmitted information.…. Another parameter measuring the similarity or dissimilarity of the set of M signal waveforms is Euclidean distance.2 Ei E j Re { rij } 2 (3.

2: Signal Space diagram of different MQAM If um(t) = (AmI+jAmQ)g(t) then like MPSK a two dimensional signal space is produced where AmI and AmQ defines the I.14dB more power than 16QAM. the signal space diagram is rectangular as shown in figure 3. In this case each n bit symbol is divided into (n1. This type of signaling is called quadrature amplitude modulation (QAM). M = M1M2}. Thus. where Am2 = AmI2 + AmQ2 and q m = tan -1 ( AmI / AmQ ) [Proakis: 89]. which justifies the use of 16QAM over 16PSK. n2) bits and each n1 = log2(M1) and n2 = log2(M2) bits are ASK-encoded separately and combined. For example. the probability of detecting correct MQAM symbol is 10log[3M2/2(M-1)p2] dB more than MPSK for same number of signaling waveforms. therefore are less power efficient [Proakis: 89]. From this. One drawback of MPSK is that increasing M brings the signal points closer to each other so the probability of detecting correct MPSK signals decreases with increasing M.e. . M [Proakis: 89].2. The bandwidth efficiency of MQAM is same as that of MPSK but are more power efficient than MPSK i. QAM can be easily utilized in high data rate applications giving more power efficiency for the same bandwidth efficiency as compared to MPSK. These signal 2 waveforms have energies Em = 1 2 Am Eg with minimum Euclidean distance of 2 Eg between adjacent points. An alternative way of representing the amplitudes is |Am|ejqm.1. 1£m2£M2. 16PSK requires at least 4. {1£m1£M1. In special case where the signal amplitudes take the set of discrete values (2m1-1-M1) + j(2m2-1-M2).CHAPTER 3 HYBRID MQAM-LFSK (HQFM) SIGNALING produces a circular signal space diagram as shown in figure 3. it is apparent that QAM signals can be viewed as combined amplitude {|Am|} and phase {qm} modulation. (a)4-QAM (b)16-QAM (c)64-QAM FIGURE 3.and Q-axis of the signal space diagram.

8) Schemes discussed so far are not bandwidth and power efficient simultaneously. the channel bandwidth required for each MFSK signal is fD. more symbols can be accommodated in a given bandwidth as number of M signals increases. total transmission bandwidth required is MfD. these waveforms are no more 2-D. To meet the condition of orthogonality. In contrast to MQAM or MPSK. there is no over crowding in the signal space. On the other hand. Hence.5 h = f DTs = í î 1 non-coherent (3. are highly bandwidth efficient i.pTs ò 0 g (t )dt pTs £ t £ ( p + 1)Ts (3. I ) = 2p h å I i i =-¥ p t . rather are M-dimensional [Proakis: 89]. While MQAM or MPSK.CHAPTER 3 HYBRID MQAM-LFSK (HQFM) SIGNALING Now consider um(t) = ejfm(t)g(t). these signals are represented in terms of symbol denoting the accumulation of all symbols until time (p-1)Ts. the probability of detecting correct symbols increases with increasing M and approaches Shannon’s channel capacity as M ® ¥. For the case of orthogonal MFSK signals and increase in signal dimensions. {fD}. This type of FSK is called continuous phase FSK (CP-FSK) [Proakis: 89]. the probability of detecting correct symbol decreases with increasing M. But. which decreases the bandwidth efficiency. the minimum frequency separation. fm(t) takes the form f m (t ) = j (t . This type of signaling is referred as frequency shift keying (FSK). then M = 2n signals are generated by shifting the carrier by an amount fm(t) = p(2m-1-M)fDt. Note that in interval pTs £ t £ (p+1)Ts. 1£m£M} and h =fDTs is known as the modulation index. therefore. must satisfy coherent ì0. Being orthogonal to one another. To corporate smooth transition from one frequency to another.e. MFSK signaling is a function of time. cannot be represented as discrete points in a signal space diagram. so it is termed as non-linear modulation [Lathi: 98]. 1£m£M. where fD is the minimum frequency spacing between two adjacent signals or tones.7) where Ii is the sequence of M-ary information symbols selected from the set {2m-1-M. due to overcrowding of discrete points in a signal space diagram. hence. .e. MFSK is power efficient i. in contrast to MQAM or MPSK.

2 Hybrid Modulation: Literature Review Many hybrid modulation techniques and their applications are discussed in literature. A study of coded modulation format. these systems cannot be readily generalized to differential detection of differential QFPM for a general case of L. The concept of making simultaneous use of phase and frequency was first introduced by Reed and Scholtz in 1966 as N-Orthogonal Phase modulated codes [Ref: Ghareeb. Fleisher: 92]. . [Chung: 99] proposed and analyzed the maximum-likelihood differential-detection (MLDD) algorithm for pilot symbol assisted differentially-encoded QFPM (PSA-DQFPM) signaling format in AWGN channel for a general case of N. The constant envelope QFPM (CEQFPM).CHAPTER 3 HYBRID MQAM-LFSK (HQFM) SIGNALING 3. [Ghareeb: 95] generalized the concept of [Fleisher. Symbol-by symbol differential detection of LFSK/ MDPSK and differential Q2PSK for various satellite mobile channels were studied in [Wei. Neither of these schemes considers the combination of orthogonal FSK with MQAM. he proposed a non-coherent detection of LFSK frequencies (h = 1) and differential detection of MPSK (DPSK). In his classical paper.5). These signals were of type LFSK/2PSK and have focused on the performance of the coherent systems only. Fleisher: 92] [Fleisher. Qu: 95]. referred as Quadrature Frequency Phase Modulation (QFPM). are of particular interest for satellite communication channels. Although these studies have demonstrated the possibility of differential detection. This section briefly reviews these available hybrid modulation techniques. one is cosinusoidal with frequency either of (fc ± 1/4Ts). Quadrature-Quadrature Phase Shift Keying (Q2PSK) [Saha. These are derived from quadrature-sum of two biorthogonal L-dimensional NFSK/2PSK signal sets [Ref: Periyalwar. Birdsall: 89] uses two parallel MSK signals (similar to binary CP-FSK with modulation index h = 0. the other is sinusoidal with frequency either of (fc ± 1/4Ts). Qu: 95] and extendend the idea to LFSK/MPSK modulation schemes and named it as Joint Frequency Phase Modulation (JFPM). Generally all these schemes make use of hybridization of orthogonal FSK and MPSK. Korn: 95]. which is in fact 4L-ary LFSK/QPSK. based on quadrature multiplexing of two L-dimensional biorthogonal set with modulation index h = 1 was carried out by [Periyalwar. Yongaçoğlu: 94].

so that a better channel utilization and a reliable communication link is provided. Depending on the channel conditions the system employs either MFSK (M = 2. Recently a variable-rate. 16. employs both MFSK and MDPSK simultaneously. The authors presented an idea of using adaptive MFSK when the MQAM system declares an outage. It is to be noted that the OMOAM format. Gohar: 06 c-a & -b]. satisfies the condition of orthogonality stated in 1 2 The whole bits in a data (log2M = n) is either FSK symbol or QAM alphabet. taking discrete amplitude levels. 4. Alouini: 03a & b]. Alouini: 06] showing a switching mechanism between the two modulation schemes . 32.8. 4. Since MQAM is spectrally as efficient as MDPSK but shows better error rate performance for M > 8. two new modulation families were introduced by substituting the on–off-keyed signaling for the orthogonal group signaling adopted in OMOAM and OMOPM [Chung. fD. 3. each n bit symbol is divided into n-k = log2L and k = log2M bits. The minimum frequency separation.3 HQFM Signals In L/M HQFM [Latif . Later on.: 05]. See figure in Ref [Digham. so replacing QAM with PSK is a better replacement to all the above mentioned hybrid modulation schemes. 64}1. Alouini: 06]. 8 and 16} or MQAM {M = 2.CHAPTER 3 HYBRID MQAM-LFSK (HQFM) SIGNALING BER and spectral performance characteristics were explored for the orthogonally multiplexed orthogonal amplitude modulation (OMOAM) [Chung: 02] and orthogonally multiplexed orthogonal phase modulation (OMOPM) [Chung: 03] signals constructed from various time-limited rectangular and sinusoidal pulsed basis sets and bandlimited root-raised-cosine pulsed basis sets. A power loading function in terms of total channel power gain is defined which is used to determine when to switch from FSK to QAM2 [Digham. The arbitrarily chosen n-k bits are used to select the frequency fl from non-coherent LFSK according to fl = (2l-1-L)fD . 1£l£L. explicitly unify several hybrid modulations discussed above. The hybrid schemes discussed so far. Liaw. variable-power scheme is proposed which employs both non-coherent MFSK and MQAM [Digham. which is the main idea presented in this dissertation.

In equation(3.10) can also be rewritten as xml . The complex form of HQFM signal in the qth signaling interval can then be expressed as uml .l (t )u * j . M1 = M2.q = xml .q (t ) = Am .l = 1 2 Am .15) . Also. During the same symbol period.lk | f DTs ) jp |ll -lk | fDTs e p | ll . k (t ) dt = Ai .q +fl . the value of fq depends not only on the current data.q ) fq . It is a hard-cored hybrid system defining no switching threshold. h = fDTs is the modulation index of the CP-FSK employed. Unlike the adaptive hybrid modulator described in [Digham.12) (3. Alouini: 03a & b.13) 2 2 Am.and Q-axis of the signal space diagram.q e j (Jm .lk | f DTs (3.8). q = AmI . M = M1M2} respectively.11) (3.13). the remaining k = log2M bits are then mapped using ordinary MQAM.3) and xml .14) 2 These waveforms have energies Em .l = p hI q . Therefore. but also on the past data.q ) (3.q = tan -1 ( AmI .q +p hll .q = Am. 06].l Eg and cross correlation coefficient ril . k (t ) dt = 1 Ts ò xi . q e j (Jm . {1£m1£M1. Also Il’s can take up discrete values ll = 2l-1-L {1£l£L}.l (t ) x* j . it is noted that at any time index t. L/M HQFM employs simultaneously FSK and MQAM.l A* j .CHAPTER 3 HYBRID MQAM-LFSK (HQFM) SIGNALING equation(3.l -q j .q / AmQ . equation (3. q Jm.l + p hå I s . Ts.k ) sin(p | ll .q g (t ) (3.k e j (qi . 1£m2£M2.9) where g(t) is rectangular pulse shaping filter defined in equation(3. jk = 1 2 Eil E jk Ts 0 ò Ts 0 ui . q + AmQ .q t / Ts ) (3.10) (3. For square MQAM. defining the I.l s =0 q -1 where AmI’s and AmQ’s can take up discrete values (2m1-1-M1) and (2m2-1-M2).

Q/8… QAMs with frequencies taken from 2.CHAPTER 3 HYBRID MQAM-LFSK (HQFM) SIGNALING To meet the condition of orthogonality. 8… FSK respectively.jk| = 0. This is the same as required by non-coherent CP-FSK (or merely FSK) defined in equation (3.e. the phase reference for FSK is absent i.lk | f DTs ) = sin(2p | l . sin(p | ll . k £ L Since adjacent frequency slot |l-k| = 1. we can say that equivalent QAM1 signal set of size Q can be split into small Q/2.8).3). therefore. where each member in one subset is orthogonal to every member in other subsets. For non-coherent L/M HQFM signal set. where each plane is distinguished by its corresponding FSK frequency (figure 3. Or in other words. it can be observed that L/M HQFM signal set constitute of L subsets. the cross correlation coefficient in equation (3.15). each with MQAM modulated symbols. Q/4. 4/4 HQFM 2/8 HQFM FIGURE 3.3: Decomposition of 16QAM into 4/4 and 2/8 HQFM From equation(3. it implies that fD =1/Ts for l ≠ k is the required minimum frequency separation to satisfy the condition of orthogonality.15) must be zero. . |ril.k | f DTs ) = 0 1 £ l. In other words. 1 Equivalency in the sense of same number of bits per symbol when HQFM is compared with pure QAM. The points lying in the HQFM signal space can be viewed as points lying in a QAM with less order on L different orthogonal planes. fD must be the integral multiple of 1/Ts. 4. In other words.

The symbol time of FSK must then be reduced to Ts¢ = Ts (n-k) / n. A generic modulator representing the generation of HQFM signals is shown in figure 3. Consider again equation(3. It is worth noting that QAM uses 2D constellations while HQFM uses 2L+1 dimensional signaling. we have L = 1. Birdsall: 89] which is a member of a general class of modulation formats known as Joint Phase Frequency Modulation (JPFM) [Ghareeb: 95].4 where LFSK and MQAM signals are produced conventionally described in [Proakis:89] [Couch: 02]. HQFM reduces to a special modulation format known as Q2PSK [Saha. however. Power spectral density (PSD) is one of the measures of channel bandwidth required to transmit an information bearing data signals.4 Power Spectral Density (PSD) The available channel bandwidth is limited in most of the digital communication systems. in order to maintain the same information bit rate. so one has to consider the spectral contents of the digital modulation under consideration. for ordinary MQAM. For L = 2 and M = 4. The random process representing the equivalent low-pass waveform of the HQFM has the form . by LFSK having the same number of tones L = 2n-k.4: HQFM (Hybrid Quadrature Frequency Modulation) 3.14). during each symbol time Ts¢ . The minimum frequency separation between two adjacent tones in this equivalent FSK system is 1/ Ts¢ . {n = (n-k) + k} bits of information are transmitted. Also. Only (n-k) bits are transmitted.CHAPTER 3 HYBRID MQAM-LFSK (HQFM) SIGNALING During each symbol time Ts. and is (n-k)/n wider than the HQFM tone separation of 1/ Ts¢ . FIGURE 3.

19) can be written in the form ¡( f1 . f 2 ) = E í å xq G ( f1 .CHAPTER 3 HYBRID MQAM-LFSK (HQFM) SIGNALING u (t ) = q =-¥ å ¥ g (t . ll )e ç å x p G ( f 2 .qTs .qejqm.f2 ) qTs r =-¥ å f ( r )e ¥ (3. l )}e ¥ q =-¥ .)* in equation(3. f 2 ) .q = ll = 2l-1-L.f 2 ) qTs e j 2p f2 rTs j 2p f 2 rTs = E {G ( f1 . ll ) ý îq =-¥ þ = q =-¥ å ¥ xq Á{ g (t .qT .17) where G(f. l q s ¥ l . M = M1M2} and ll.18) denotes statistical expectation and complex conjugation respectively.q is a random variable which can take up the value (2m1 . xq .j 2p ( f1 .qTs .f 2 ) qTs e j 2p f 2 rTs q =-¥ r =-¥ å å f (r )E {G ( f .j 2p fqTs (3. f 2 ) = E {U ( f1 )U * ( f 2 )} (3. ll) is the Fourier transform of g(t.qTs .} and (. {1£m1£M1. ll )} = q =-¥ å x G ( f . f 2 ) defined by ¡( f1 . l )e q l ¥ . Taking the Fourier transform of equation(3. F( f ) can be obtained by finding the line masses of the function ¡( f1 .j 2p ( f1 . q ) = q =-¥ å x g (t . ll )G * ( f 2 . The PSD of u(t) . Since the random process u(t) is non-stationary process so its Fourier transform U( f ) = Á{u(t)}exists [Lathi: 98]. 1£m2£M2. l )G ( f .1 . The operator E{.16) ì ¥ ü U ( f ) = Á í å xq g (t .19) Assuming {xq} and {ll}.M 1 ) + j (2m2 .j 2p ( f1 . ll . l )}e * q q+r * 1 l 2 l * 1 l 2 l ¥ ¥ . ll )e ÷ ý è p =-¥ ø ï ï î q =-¥ þ (3. * ¥ ì ü ï ¥ .M 2 ) .16) where xq = Am.1 . both as stationary and statistically independent random variables and letting p = q + r. l )G ( f . f 2 ) = = q =-¥ r =-¥ ¥ ¥ å å E { x x } E {G ( f .j 2p f 2 pTs ö ï ¡( f1 .q ) (3. ll )} å e.18) which is located on the bisector of the plane ( f1 .20) .j 2p f1qTs æ . equation(3. ll).

equation (3. f 2 ) located on the bisector of the plane ( f1 . ll ) = G ( f .23) Equation (3.22) again. l ) l =1 l L 2 (3. Since the elements of the sequence {xq} are equally likely and symmetrically positioned on complex plane.23) shows that the function ¡( f1 .CHAPTER 3 HYBRID MQAM-LFSK (HQFM) SIGNALING where f(r) = E{xqxq+r} is the autocorrelation function of QAM symbols drawn from {xq}. The PSD F( f ) . f 2 ) = E {G ( f1 .20) can be rewritten as ¡( f1 .21) Substituting f(r) from equation (3.26) . comprised of the line masses of the function ¡( f1 . For equal probable symbols Pl = 1/L for all q. ll )} å e . ll ) Ts Ts l =1 { } (3. is given by F( f ) = s2 s2 L 2 E G ( f .22) equation (3. Equation (3. f 2 ) contains both continuous and discrete spectrum.1 2 f D ll ) . ll )G * ( f 2 .24) Equation (3.24) is obtained by using identity (3.21) and using identity [Lathi: 98] r =-¥ å e j 2p rfTs = ¥ 1 Ts r =-¥ å d ç f -T è ¥ æ r ö ÷ s ø (3.25) where Pl is the probability of occurrence of G ( f . then [Praokis: 89] ìs 2 + m 2 f (r ) = í 2 î m r =0 r¹0 (3. q =-¥ ¥ æ 2 m2 s + ç ç Ts è æ r d ç f2 å Ts r =-¥ è ¥ öö ÷÷ ÷ øø (3.25) becomes F( f ) = s2 LTs å G( f .23) then becomes ¥ æ s2 r ö ¡( f1 . f 2 ) = E {G ( f1 .f 2 .÷ Ts Ts ø r =-¥ è (3.j 2p ( f1 .f2 ) qTs . ll )} å d ç f1 . ll ) = å Pl G( f . f 2 ) . therefore mean m = 0. ll )G * ( f 2 . Assuming {xq} as identically distributed uncorrelated random variables with mean m and variance s2.

1 2 f D ll ) Ts p(f -1 2 f D ll ) Ts e .30) Our observation on (3. the power spectral density of a HQFM signals is reduced to the power spectrum density of an MQAM [Couch: 02] i. the spectrum of HQFM does not contain a discrete spectral component but consists of a continuous spectrum whose shape depends only on the spectral characteristic of the signal pulse.[2l . 07] s 2Ts F( f ) = L æ sin p ( fTs .CHAPTER 3 HYBRID MQAM-LFSK (HQFM) SIGNALING Thus.1 ö 2 ll h) ç ÷ å 1 l =1 è p ( fTs .31) .27) Putting fDTs = h G ( f .1 .28) Hence equation (3.ll h / 2 ) p ( fTs .29) For non-coherent HQFM system. the spectra of HQFM signals fall off at a rate proportional to the inverse of the frequency difference squared. ll ) with unit amplitude A = 1 and width Ts [Couch: 02] i. The first factor is the shape of the complex envelope waveform s(t).e.26) becomes s 2Ts F( f ) = L æ sin p ( fTs . Gohar: 06 c-a & -b.ll h / 2 ) e .1 2 f D ll ) = Ts sin p ( f .L ] / 2) ø L 2 (3.1 . we get [Latif. æ sin p fTs ö F( f ) = s Ts ç ÷ è p fTs ø 2 2 (3.jp ( fTs .e. ll ) = Ts sin p ( fTs .29).30) is that. G ( f .[2l . when L = 1. The second factor is the correlation of the sequence {eq}.ll h / 2 ) (3. The above analysis can be concluded by observing two factors which affect the shape of the signal spectrum independently. Now consider the G ( f . h = fDTs = 1 and by substituting the value of ll = 2l-1-L into(3. When no spectral shaping is employed. and for frequencies far from the center frequency.1 f l T 2 D l) s (3. ll ) as the Fourier transform of a rectangular pulse g (t .jp ( f .L] / 2) ö ç ÷ å l =1 è p ( fTs .2 ll h ) ø L 2 (3. ll ) = G ( f .

L ) h ) -y cos p ( i + j . F( f )/Tb for Square MQAM M = {4. 64 and 256.32) where U i = U ( f . nulls occurs at integral multiple of fTb = 0. the PSD obtained is F( f ) = L L ü 1 ìL 2 L + 2 B G + 4 Bij Gi G j ý ( ) í åå ii i 2 å Ts L î i =1 i =1 j > i þ (3. {fc}. For orthogonal LFSK.2y cos 2p fTs (3.25. Hence the null-to-null bandwidth for MQAM is 1/Tblog2M. therefore increases by increasing M = 2n i.31) for square QAM with different values of M is plotted in figure 3.1 . 0. the number of bits per signal.1 . The origin refers to centre frequency. 16. Similarly. li ) is the pulse shape used and Bij ( f ) = cos p ( 2 fTs .5 3 Normalized Frequency.e. Equation (3.5: Normalized Power Spectral Density.5 for comparison only.1667.33) .CHAPTER 3 Normalized Power Spectral Density. 256} For MPSK.5. in case of M = 16. 64.5 2 2. 0.125 respectively.5 1 1. Only the right half is shown because the power spectral density is symmetrical around fc or origin. The bandwidth efficiency hB = 1/BTb = log2M. F( f )/Tb.L ) h 1 + y 2 . fTb FIGURE 3. The figure clearly shows that the nulls for M = 4 occurs at integral multiple of fTb = 0. s2 = 1.( i + j . dB 10 0 -10 -20 -30 -40 -50 -60 HYBRID MQAM-LFSK (HQFM) SIGNALING M=4 M = 16 M = 64 M = 256 0 0.

M = 8}.e. On the other hand the main lobe of 4FSK is approximately 50% wider than 4/4 HQFM. h = fDTs = 1. is also shown for comparison. by increasing the number of bits per symbol and QAM size {M}. It can be shown that sidelobes in the PSD of LFSK falls with the rate of fourth power of frequency difference from centre frequency (see appendix A for proof).7(b) clearly shows a spectral narrowing of the main lobe with increasing M.32) respectively. M = 8. the spectral occupancy of HQFM decreases. 256} in figure 3. Keeping {L} constant i. The possibility of 8/2 HQFM {L = 8.625 which is greater than both M = 4 {fTb = 0. Also. The resulting PSDs are also compared with ML-QAM. M = 2} is not considered due to the non availability of MQAM with M < 4. M = 16. Thus. the bandwidth occupancy is the inverse function of M. Therefore for fixed L. therefore occupying more bandwidth.7(a) and 4/L HQFM {L = 4. .6(b) compares the PSD of 2/8 HQFM with 16QAM.30) are shown. Thus. it is shown that the main lobe of the PSD widens by increasing the number of keying frequencies. only the use of noncoherent FSK is considered i. 128} in figure 3. Figure 3.25}. Figure 3.e. This in turn.CHAPTER 3 HYBRID MQAM-LFSK (HQFM) SIGNALING Here y º y(jh) is the characteristic function of the random information sequence. This is also true for ordinary QAM for which L = 1. The PSDs of individual 4QAM and 4FSK using equation (3. for 2/M HQFM {L = 2. Figure 3. Comparing figure 3. In the following paragraphs different power spectral densities for different HQFM formats will be discussed. {L}. 32. M = 4} compared by 16QAM {ML = Q = 16}. increases the bandwidth efficiency for fixed number of keying frequencies {L}. therefore increasing the bandwidth occupancy.7 compare PSD plots when number of bits per HQFM symbol increases.5} and ML = 16QAMs {fTb = 0. 64. Different plots employing equation (3. For rectangular pulse shape it is defined by equation(3.6 (a) & (b).6(a) shows the power spectral density of 4/4 HQFM {L = 4. Other possible HQFM format for {log2ML = 4} bits symbol is 2/8 {L = 2.28). the bandwidth occupancy of HQFM is less than the bandwidth occupancy of FSK. The figure clearly illustrates that the main lobe of 4/4 HQFM occurs at fTb = 0.31) and(3.

fTb FIGURE 3.6: Normalized Power Spectral Densities. 4 .5 1 1.5 1 1. dB -20 -25 -30 -35 -40 0 0. F( f )/Tb for 4 bit HQFM formats (a) 4/4 HQFM compared with M = 4.CHAPTER 3 HYBRID MQAM-LFSK (HQFM) SIGNALING 10 5 0 -5 -10 -15 4/4HQFM 16QAM 4QAM 4FSK Normalized Power Spectral Density.5 2 2. F( f )/Tb.5 3 2/8HQFM 4/4HQFM 16QAM Normalized Frequency.5 3 10 5 0 -5 -10 -15 -20 -25 -30 -35 -40 0 0. 16 QAM and 4FSK (b) Behavior of PSD with increasing L = 2.5 2 2.

64.7: Normalized Power Spectral Densities.5 3 10 5 0 -5 -10 -15 -20 -25 -30 -35 -40 0 0. L (a) 2/M HQFM {M = 8. 128} (b) 4/M HQFM {M = 16.5 2 2.5 2 2. F( f )/Tb. fTb FIGURE 3. 32. dB -20 -25 -30 -35 -40 0 0. 256} .5 1 1. F( f )/Tb for HQFM formats with fixed number of frequencies.CHAPTER 3 HYBRID MQAM-LFSK (HQFM) SIGNALING 10 5 0 -5 -10 -15 2/8HQFM 2/32HQFM 2/128HQFM Normalized Power Spectral Density.5 3 4/16HQFM 4/64HQFM 4/256HQFM Normalized Frequency.5 1 1.

M = 16} By comparing PSDs in figure 3. Rb = 1/Tb. F( f )/Tb for L/16 HQFM {L = 2.7 and 3. In all cases (figure 3. Gohar: 06 c-a & -b. dB 5 0 -5 -10 -15 -20 -25 -30 -35 -40 0 0. this can be approximated by B = L/Ts. for non-coherent HQFM. it is evident that for fixed {M}. Hence.5) i. For orthogonal HQFM with modulation index h = 1.5 2 2/16HQFM 4/16HQFM 8/16HQFM Normalized Frequency.34) . the required transmission bandwidth will almost always be determined by its frequency components.8. The bit rate.5 1 1. F( f )/Tb. 3. 07] hB = Rb 1 log 2 L + log 2 M = = B BTb L (3.5 Bandwidth Efficiency In L/M HQFM.e.8) the rate of fall of sidelobes decreases at the same rate as QAM (figure 3. the bandwidth efficiency is [Latif. the main spectral lobe of L/M HQFM becomes wider with increasing {L}.8: Normalized Power Spectral Densities.CHAPTER 3 10 HYBRID MQAM-LFSK (HQFM) SIGNALING Normalized Power Spectral Density. 3. is defined as Rb = n/Ts = log2ML/Ts. squared frequency offset from centre frequency fc (in our case origin).6. 8. 4. fTb 2.5 3 FIGURE 3. although a reduction in the out-of-band spectral tails is observed.

2188 0.125 3 2 1.25 0.625 0.34). Similarly 2/128 has same bandwidth efficiency as 16 QAM but it supports higher data rate (double) i.e.375 0. 4/64 HQFM has same efficiency as 4QAM.5 0. TABLE 3.5 0.25 1.I: BANDWIDTH EFFICIENCIES (hB) FOR DIFFERENT HQFM FORMATS L/M HQFM L 11 2 4 8 16 32 64 FSK L/4 --0. Table 3.I relates the bandwidth efficiencies of different HQFM formats. this measure does not fully describe where the majority of signal’s power is contained.375 0.5 1 0. From a practical 1 L = 1 corresponds to the bandwidth efficiency of pure MQAM . 2/8.34) shows that hB is directly proportional to QAM level {M} and is an inverse function of number of FSK tones {L}.156 0.094 2 1. Although most of the signal’s energy lies in its main lobe.1719 L/32 6 3. In fact. defined in equation(3.g.25 0.875 0.6875 0.25 0. It is also evident that this equation reduces to the bandwidth efficiencies of pure QAM for L = 1 or pure FSK when M = 1. 8 bits per symbol (2/128 HQFM) as compared to 4bits per symbol (16QAM).25 0.5 2 1. null-to-null bandwidth efficiency.3125 0. The first null-to-null bandwidth is defined as the width of main lobe of the signal’s PSD in frequency domain.4375 0.75 1 0.125 0. Therefore. Also.5 0.3438 0.1875 L/64 L/128 7 4 2.75 0.5 1.625 0. It can be observed that similar bandwidth efficiencies can be obtained for higher data rates e.5625 0. cannot be used. it depends entirely upon the pulse shaping filter employed. the bandwidth efficiency of HQFM lies between the bandwidth efficiency of FSK and QAM.5 0.375 0.2031 3.CHAPTER 3 HYBRID MQAM-LFSK (HQFM) SIGNALING Equation (3.1563 L/16 5 3 1.2813 0.6 Fractional Out-of-Band Power (FOBP) There are several definitions of bandwidth that are popular in literature.1406 L/8 4 2.

36) becomes ¥ Pt = -¥ ò F( f )df =s2 (3. To calculate the total power Pt.9 or 0.cos 2 z =ò dz z2 0 ¥ ¥ ¥ (3.38) . and let the total power be Pt.CHAPTER 3 HYBRID MQAM-LFSK (HQFM) SIGNALING point of view. it is stated in US FCC regulations. the bandwidth efficiency of a modulated signal is calculated in terms of fractional out-of-band power (FOBP) containment and is fixed to some value typically 0.35) where Pt = ò-¥ F ( f )df is the total power confined and PB = ò.1 2 ll h ) . consider equation (3. Also.2 ll h ) ø L ¥ 2 s 2Ts = L s 2 L ¥ sin 2 z = å ò z 2 df p L l =1 -¥ where z = p ( fTs .1 ö 2 ll h) ç ÷ df å ò 1 l =1 -¥ è p ( fTs .99[Couch: 02].cos pz pp dz = [Spiegel: 68]. In general.The total power defined by 2 z 2 equation (3. the power in a certain band should be 99% of the total power. Let the bandwidth B be defined as the percent fractional power bandwidth. The fractional out-of-band power is defined as mB = 1 ¥ PB Pt B /2 (3.B /2 F ( f ) df is the power confined by the modulator over a bandwidth B.29) again ¥ Pt = -¥ ò F( f )df æ sin p ( fTs .36) (3. the main lobe contains more than 90% of the total energy. In general.37) As we know that ò ¥ 0 1 . sin 2 z sin 2 z dz = 2 dz ò z2 ò z2 -¥ 0 1 .

p ( BT s l è 2 s l 2 ø p ( BT .43) .ll h) ) + Si (p ( BTs + ll h) ) pè (1 + p cos p BTs cos pll h) BTs + pll h sin p BTs sin pll h ö + ÷ 2 2 ÷ (p BTs ) . 2 p ( BTs .1 2 ll h ) .1 ö Pt 2 ll h ) ç ÷ df = å ò 1 L l =1 .Si(2b) ) + ( b cos 2a . The integrand (3.l h ) s l 2 p ( BTs .1 2 B.CHAPTER 3 HYBRID MQAM-LFSK (HQFM) SIGNALING Power confined by the modulator over a bandwidth B is defined as B /2 PB = .2 ll h ) ø L B /2 2 s 2Ts = L åP l =1 L Bl (3.a cos 2b ) ab where Si(x) is x known as sine integral and is defined as Si( x) = ò0 sin t tdt [Spiegel: 68].(pll h )2 .B /2 è p ( fTs .p ( BTs + ll h ) z 2p 2 p ( BTs .39) Therefore.p ( BTò +l h) z 2 dz ÷ ÷ 2p ç .41) Again.ll h )} . Equation (3. 2 B} to z = { 2 p ( BTs + ll h ) .( BTs + ll h ) p ( BT .41) becomes z = p ( fTs .B /2 è B /2 2 (3.cos 2 z dz 2 z p .(pll h ) ø (3.l h ) s l 2 ò 2 = 1 æ dz cos 2 z ö ç 2 ò +l h ) z .p ( BT s l è 2 ø (3. change 1 1 1 f = {.35) reduces to mB = 1 where 1 L å PB L l =1 l (3.ll h ) 1 . the FOBP of L/M HQFM defined in equation (3.40) æ sin p ( fTs .ll h ) 2 4 BTs 1 æ cos 2 z ö ç =+ ò +l h ) z 2 dz ÷ ÷ 2p ç (p BTs ) 2 .B /2 ò F ( f )df æ sin p ( fTs .1 ö 2 ll h ) PBl = Ts ò ç ÷ df 1 p ( fT l h ) s 2 l ø .42) can now be rewritten as PBl = 2æ ç Si (p ( BTs . the limits in PBl becomes from 1 2 sin 2 z 1 PBl = dz = 2 ò p .42) Since òb cos 2 z a z 2 dz = 4 ( Si(2a ) .l h ) s l 2 p ( BT .

35) becomes mB = 1 - 4 L ì å íSi (p ( BTs . This expression can also be obtained by replacing ll = 0 in equation (3.(pll h ) ï þ (3. This figure also plots the FOBP of LFSK by numerically solving equation (3. . This figure clearly shows that LFSK is spectrally less efficient as compared to MQAM.5 mB = 1 - æ BT + (-1)( ll -1)/ 2 pl sin p BT 4 L ì ï l s 1 Si p ( BT l ) + 2 ç s ( ) í å s 2 l 2 2 ç p L l =1 ï ( 2p BTs ) .46) When L = 1.ll h) ) p L l =1 î (1 + p cos p BTs cos pll h) BTs + pll h sin p BTs sin pll h ü ï + ý 2 2 (p BTs ) .(pll ) ï î þ (3.46) reduces to mB = 1 - (1 + p cos p BTs ) ü 4ì íSi(p BTs ) ý pî p 2 BTs þ (3.44) is plotted against the bandwidth normalized by the bit rate (BTb). bandwidth efficiency can be obtained directly for a given fractional power bandwidth.CHAPTER 3 HYBRID MQAM-LFSK (HQFM) SIGNALING Now FOBP of L/M HQFM defined by (3.40).47) are plotted versus the normalized bandwidth BTb in figure 3. Special Cases h = 0.45) h=1 mB = 1 - (1 + p cos p BTs ) BTs ü 4 L ì ï ï Si p ( BT l ) ( ) í å s l 2 2 ý p L l =1 ï (p BTs ) .9.(pll ) è î öü ï ÷ý ÷ ï øþ (3.47) This is the fractional out-of band power expression for ordinary MQAM with symbol period Ts = Tblog2M. The FOBP containments of MQAM obtained by equation (3.32).35) using equation(3.43) and substituting the resulting expression in equation(3. equation (3.44) If FOBP defined in (3.

dB -5 L/4 HQFM L/16 HQFM L/64 HQFM (a) L = 2 -10 -15 -20 (b) L = 4 -25 0 0. M = {4. 16. L = {2. dB -5 -10 M=4 M = 16 M = 64 M = 256 L=2 L=4 L=8 -15 -20 -25 0 1 2 3 Normalized Bandwidth. mB. 4. and 64} .5 1 1.CHAPTER 3 HYBRID MQAM-LFSK (HQFM) SIGNALING 0 Fractional Out-of Band Power. 64. BTb 2. and 256} and LFSK.10: Fractional Out-of-Band Power for L/M HQFM with L = {2. mB.5 3 FIGURE 3. BTb 4 5 FIGURE 3. 16. and 8} 0 Fractional Out-of Band Power.9: Comparison of FOBP as a function of Normalized Bandwidth BTb for square QAM. 4} and M = {4.5 2 Normalized Bandwidth.

which make HQFM with fixed M.246 0. 90% and 99% power bandwidth of LFSK using numerical techniques is also given in these tables for comparison. one can observe that.180 1.415 4.1) and 99% (mB = 0.890 5.456 4.800 0.765 0.850 1.II and 3. we see that MQAM signals have higher outof-band spectral tails.636 2. From these plots. by using L/M HQFM.46) are plotted versus the normalized bandwidth BTb in figure 3.456 In terms of 90% power bandwidth efficiency. respectively.II and 3.490 2. the bandwidth efficiency of HQFM is increased and the out-of-band spectral tails are reduced.10 for L/M HQFM with L= {2.158 8.246 2.208 5.478 2.568 0.800 2.830 1.104 3.790 L/32 0. FOBP for different formats of L/M HQFM using equation (3. the bandwidth efficiency of HQFM is greater than that of the CPFSK when the size M of the signal set is greater than 4.350 0. taking any combination of L/M HQFM. 16. TABLE 3.835 3.375 0. we conclude that. Also.III is that.571 0.988 1.638 6.434 L/16 0.745 1.786 1. whereas HQFM tends to reduce the out-of-band spectral tails.065 1.281 0. 4} and M = {4.567 0. Our observation of the data in Tables 3.650 1.CHAPTER 3 HYBRID MQAM-LFSK (HQFM) SIGNALING The FOBP of non-coherent L/M HQFM obtained by equation(3. LFSK AND DIFFERENT L/M HQFM L/M HQFM L 1 2 4 8 16 32 64 FSK L/4 -0. as order of QAM {M} increases. for any fixed value of Q = ML.400 0.265 L/64 0.490 0. there are as many possibilities as could be desired for choosing the bandwidth efficiency (value of 1/BTb at which mB = -10 dB or -20dB respectively) of the desired system (as allowed in the design).322 0. 64} respectively.338 2. more bandwidth efficient than MQAM.848 0.01) power containment. the bandwidth efficiency of HQFM is smaller than . for any fixed {L}.435 0.III for 90% (mB = 0.821 L/128 0.272 L/8 0.47) are given in Table 3.647 4.930 1.300 0. It can also be observed that.46) and MQAM (case when L=1) using equation (3.80 0.II: COMPARISON OF 90% (-10dB) POWER BANDWIDTH (BTb) FOR MQAM.890 4.458 0.

it increases as the signal size increases.800 5.030 3.48) where T = NTs is OFDM symbol duration and fq = (2q-1-N)/NTs. TABLE 3.137 5.529 2.165 8. Gohar: 07]. in both systems the bandwidth efficiencies are very close [Latif.540 2.685 3.450 3. the null-to-null bandwidth efficiency of HQFM is always greater that 90% power bandwidth.48) clearly states that PSD of OFDM . we see that the bandwidth efficiency of CPFSK decreases whereas in the case of HQFM.900 2.200 2.360 2.040 4.400 2.200 5.CHAPTER 3 HYBRID MQAM-LFSK (HQFM) SIGNALING that of Q-ary QAM.f q )T N -1 ö ÷ ÷ ø 2 (3.828 6.950 3.860 2. Equation (3.876 4.160 2. modulated by rectangular pulse is of the form of sinx/x pulses.050 3. Comparing table 3.926 In terms of 99% power bandwidth efficiency.30 6.630 7.415 3. PSD of each orthogonal subcarrier. n being number of bits per subcarrier.380 3.085 3. we have to say that for OFDM.700 2.070 4. 3.I and 3.720 5.881 2.380 1.II.f q )T F( f ) = å ç ç q = 0 è p ( f .408 L/32 4.266 1.252 2. Ts = nTb.005 L/8 6.600 3. Firstly the bandwidth efficiency of both increases as the size Q = ML of the signal set increase and secondly. Compared to CPFSK.115 L/16 5.7 Spectral Properties of HQFM-OFDM Lastly.445 5.020 4. LFSK AND DIFFERENT L/M HQFM L/M HQFM L 1 2 4 8 16 32 64 FSK L/4 -1. we observe that there are strong analogies between L/M HQFM and Q-ary QAM.600 3.III: COMPARISON OF 99% (-20dB) POWER BANDWIDTH (BTb) FOR MQAM.200 1.400 2.300 10.125 3.320 4.818 L/64 3.100 4. thus the overall PSD for the complex envelope can be evaluated as [Couch: 02] æ sin p ( f .335 L/128 2.

pp: 546 . –S. Vol. No. so bandwidth occupancy of HQFM-OFDM. D.28 [CHUNG. Zander: 06]: B» N +1 1 » NTs Ts (3.. N. Vol..: 31430-01. 1999. LIAW..40 [CHUNG. IEEE Trans. pp: 428 .: 99]: “Differential Detection of Quadrature Frequency/Phase Modulated Signals”.: 02]: DIGITAL AND ANALOG COMMUNICATION SYSTEMS. the side lobes of the OFDM power spectrum falls faster with increasing frequency. 2005. SLIMANE. Mar. 2003. ZANDER. Art..CHAPTER 3 HYBRID MQAM-LFSK (HQFM) SIGNALING depends on frequency separation of each subcarrier (Df = 1/T = 1/NTs). ISBN: 9789144030807 [CHUNG. Commn.49) REFERENCES [AHLIN. [COUCH. pp: 415 . L. Studentlitteratur Publishing Group (Sweden). Odder (Denmark). the bandwidth of OFDM signals can be approximated as [Ahlin.. B. 50(3). It is observed that by increasing the number of subcarriers. ISBN: 81-7808-328-0 . Observing FOBP of OFDM signals for different number of subcarriers. C. 51(3). Jan. it can be shown that for large N.: 05]: “Orthogonally multiplexed On-Off Keyed Amplitude and Phase Modulation Families”.multiplexed orthogonal amplitude modulation family”. D. C. India 2002. 2002. irrespective of the modulation format used.99. As N goes to infinity. Commn. C. Commn. J. Vol. IEEE Trans. Vol.: 03]: “Coherent and Differentially coherent detections of Orthogonallymultiplexed orthogonal phase modulated signals”. C. W. IEEE Trans. 4(1). Apr. pp: 288 .. Mar. Gohar: 07]. D. in terms of PSD/ FOBP remains unchanged [Latif.: 02]: “Orthogonally. Prentice Hall PTR. Commn..: 06]: PRINCIPLES OF WIRELESS COMMUNICATIONS. 6th Edition. 47(4). IEEE Trans. Narayana Press.56 [CHUNG. D. H. the PSD of OFDM signals converges to a rectangular function with duration 1/Ts.

Oxford Univ.F.. Proc. Feb 1995. Proc. pp: 1512 – 16 [DIGHAM.F. Conf... ALOUINI. 2003). N. B. Vol. F. IEEE Trans. pp: 1513 . Conf. on Emerging Technologies (ICET 2006). ALOUINI. 2nd IEEE Int.. S. pp: 53 – 9 . F. Aug. D.47 [GHAREEB. GOHAR. 2nd Int. Proc.: 98]: MODERN DIGITAL AND ANALOG COMMUNICATION SYSTEMS. 2006. 2006 China [LATIF.: 06 c-a]: “A Hybrid MQAM-LFSK OFDM Transceiver with Low PAPR”. N. Sel. IEEE J. pp:964 . Vol. 11. 13(2). Vol. IEE Proc. M. F. S.-S.86 [LATHI. 2. ALOUINI. IEEE 58th Vehicular Technology Conference (VTC ‘03(Fall.: 03b]: “Variable-rate variable-power hybrid M-FSK M-QAM for fading channels”. pp: 276 . 1994. Computers and Signal Processing. 2003.CHAPTER 3 HYBRID MQAM-LFSK (HQFM) SIGNALING [DIGHAM.: 95]: “Quadrature Frequency/Phase Modulation”..: 06 c-b]: “BER Performance Evaluation and PSD Analysis of Non-Coherent Hybrid MQAM-LFSK OFDM Transmission System”. Feb/Mar/Apr 1995.. Press Inc. A. IEEE International Conference on Communications (ICC’06). Commn.24 [GHAREEB. on Wireless Communications. P. I. 22-24. YONGAÇOĞLU." Proc. A.: 06]: "Spectrally Efficient Hybrid FSK/QAM with Optimum Bit and Power Loading. pp: 5022 . ISBN10: 0195110099.: 94]: “Performance of Joint Frequency Phase Modulation over Rayleigh Fading Channels”. pp: 241 . [LATIF.-S. 6-9 Oct. IEEE Pacific Rim Conference on Communications.. 2003. Commn. Networking and Mobile Computing (WiCom06). Vol. London 1998. Vol. June 2006. GOHAR.F.27 [FLEISHER. M. Sept. 13 – 14.. ’03)..-S.: 95]: “Bit error Rate Performance and Power Spectral density of NonCoherent Hybrid Frequency-Phase Modulation system”. Vol. I. Areas Commn. 141(4). Nov.. Peshawar (Pakistan)..7 [DIGHAM. 43(2/3/4). 28-30 Aug.: 03a]: “Adaptive hybrid M-FSK M-QAM modulation”. (PACRIM. D. A. 3rd Edition. QU. M. 3.

2nd Edition. Sci. Vol.. S. Mc-Graw Hill Book Company.76 [PERIYALWAR.: 68]: MATHEMATICAL HANDBOOK OF FORMULAS AND TABLES. S.:07]: “Signal Properties of Hybrid LFSK Modulated MQAM (HQFM) OFDM Transceiver”. Veh. Vol..: 92]: “Trellis Coding of Quadrature Frequency/ Phase Modulated Signals”. Tech.CHAPTER 3 HYBRID MQAM-LFSK (HQFM) SIGNALING [LATIF. IEEE. D. M. A. 10(8). I. J. Jour. T. L. Sec. November 2007.: 95]: “Combined Frequency and Differential Phase Shift Keying with Non-Coherent Detection in Stallite Mobile Channel”. Int. Trans.448 [SPIEGEL. FLEISHER. ISBN: 007-050937-9 [SAHA. 1992. Oct. pp: 72.: 89]: DIGITAL COMMUNICATIONS.63 [PROAKIS. 1995. Aug. pp: 437 . 1968. J. S. Comp.. 37(5). KORN.. (IJCSNS) Vol. 7(11). GOHAR.. pp: 603 – 11 . IEEE Trans. pp: 1254 . International Series in Electrical Engineering (Communications and Signal Processing). G. Sel.. Areas Commn. Commn. Netw. N. G. May 1989. IEEE. 1989. New York.. Schaum’s Outline Series. D. New York. McGraw-Hill Book Company.. R. Vol.: 89]: “Quadrature-Quadrature Phase Shift Keying”. BIRDSALL. 44(3). [WEI.

CHAPTER 3 HYBRID MQAM-LFSK (HQFM) SIGNALING .