ECE 461

Fall 2006 August 31, 2006

Complex Baseband Representation

Channel Model for Point-to-Point Communications Point-to-point communications systems are well modeled using a bandpass additive noise channel model of the form shown in Figure 1. |S(f )| fc n(t) Figure 1: Real bandpass channel model for point-to-point communications The message bearing signal s(t) is a real-valued bandpass signal whose spectrum is concentrated in the vicinity of some carrier frequency fc . Distortions introduced by the channel are characterized by a linear time invariant system with impulse response h(t), and frequency response H(f ) concentrated around fc . The channel re sponse h(t) may or may not be known at the receiver. In the simplest case, the response h(t) corresponds to a an ideal bandpass lter with bandwidth corresponding to that of the signal s(t). The additive noise process n(t) is usually idealized by White Gaussian Noise (WGN). The received signal r(t) is a real-valued bandpass process as well. In the following we convert the bandpass channel model into a more convenient and equivalent complex baseband channel model. Complex baseband representation for signal Since the signal s(t) is real, its spectrum S(f ) is symmetric about f = 0. Hence all of the information about the signal s(t) is contained in the positive half of the spectrum S(f ), which we dene to be 1 (1) S+ (f ) = 2S(f )1 {f 0} . where u() is the unit step function. The factor of 2 in the above equation makes the signal s+ (t) have the same energy as the signal s(t). The inverse Fourier transform of the spectrum S+ (f ) is easily shown to be the complex signal 1 s+ (t) = [(t) + j s(t)] , s 2 c V.V. Veeravalli, 2006 (2) 1 fc f

s(t)

h(t)

y (t)

r (t)

 where the signal s(t) is the Hilbert transform of s(t) (i.e., the Fourier transform of s(t) is jsgn(f )S(f )). + (f ) down to the The signal s+ (t) is called the pre-envelope of s(t). If we shift the spectrum of S origin, we get the baseband signal s(t) with S(f ) = S+ (f + fc ), and s(t) = s+ (t)ej2fc t . (3)

Note that since S(f ) is not necessarily symmetric around the origin, the signal s(t) is in general complex-valued. The signal s(t) is called the complex envelope or the complex baseband representation of the real signal s(t). From (2) and (3), we get (4) s(t) = Re[ 2 s+ (t)] = Re[ 2 s(t)ej2fc t ] . The complex envelope s(t) can be written in terms of its real and imaginary parts as s(t) = sI (t) + jsQ (t) . From this and (4) we get s(t) = 2[sI (t) cos 2fc t sQ (t) sin 2fc t] = 2 a(t) cos[2fc t + (t)] , where a(t) = s2 (t) + s2 (t), I Q and (t) = tan1 sQ (t) . sI (t) (5)

(6)

(7)

The signal a(t) is called the envelope of s(t), and (t) is called the phase of s(t). It is to be noted that every bandpass signal can be written in the forms given in (6). Equation (6) also suggests a practical way to generate the (components of) complex envelope from the passband signal. It is easy to see that if we multiply s(t) by 2 cos(2fc t) and low-pass lter (LPF) the output, we produce sI (t). Similarly, if we multiply by 2 sin(2fc t) and LPF the output, we get sQ (t). The conversion from passband to baseband and vice-versa is illustrated below in Figure 2 LPF s(t) 2 cos 2fc t LPF 2 sin 2fc t sQ (t) 2 sin 2fc t sI (t) 2 cos 2fc t s(t)

Figure 2: Conversion from passband to baseband and vice-versa.

Complex baseband representation of channel response Referring to Figure 1, since the output of the channel y (t) is a bandpass signal, it has the complex baseband representation y(t) = y+ (t) ej2fc t . The signal y (t) is related to s(t) through the convo lution integral, i.e., y (t) = h s(t). The question that we ask now is whether the complex envelopes c V.V. Veeravalli, 2006 2

s(t) and y(t) are related in a similar fashion, and if so, what is the corresponding complex impulse response? You will show in HW#2 that this in indeed the case and that the corresponding complex baseband channel response is given by: 1 H(f ) = H+ (f + fc ) 2 Note the additional factor of Note that y(t) = h s(t) = (hI + jhQ ) (sI + jsQ )(t) implies that the I and Q components of y(t) can be computed separately as yI (t) = hI sI (t) hQ sQ (t) , and yQ (t) = hI sQ (t) + hQ sI (t) . This suggests a way to implement the passband lter h using real baseband operations. (10) (9) = 1 h(t) = h+ (t)ej2fc t 2 and h(t) = 2Re[h(t)ej2fc t ] . (8)

2 in the equation relating h(t) and h(t).

c V.V. Veeravalli, 2006