Passband Communication
Passband Communication
Introduction
Beginning with this lecture, we will study wireless communication. The focus of these lectures will be on point to point communication. Communication on a wireless channel is inherently dierent from that on a wireline channel. The main dierence is that unlike wireline channel, wireless is a shared medium. The medium is considered as a federal resource and is federally regulated. The entire spectrum is split into many licensed and unlicensed bands. An example of the the point to point communication in the licensed band is the cellular phone communication, whereas wi-, cordless phones and blue tooth are some of the examples of communication in the unlicensed band. The transmission over a wireless channel is restricted to a range of frequencies (fc W , fc + W ) around the central carrier frequency fc . The wire is a low pass lter and hence 2 2 the carrier frequency for the wireline channel is fc = 0. This restriction immediately poses some questions about the design of the wireless communication systems. The foremost question being how is reliable communication related to the carrier frequency? Is the communication strategy and hence the transmitter-receiver design particular to the specic carrier frequency? Do we have to design the system based on fc ? It turns out that we can always work in with the baseband signal (i.e., the signal with fc = 0) even for the wireless communication and then convert the baseband signal to the passband signal (a signal that is centered around some nonzero carrier fraquency) with the desired carrier frequency. This makes the design of the transmitter and receiver transparent to the carrier frequency. Thus, only the front end of of the sytem needs to be changed if we change fc . Also since the bandwidth of the signal W (typically in KHz) is much smaller than the carrier frequency fc (typically in MHz), the design of DAC and ADC becomes much easier and modular. The focus of this lecture will be on the conversion of the baseband signal to the passband signal and vice-versa. Also, the actual wireless channel aects the passband signal. How do these eects translate in the baseband domain, i.e., is there a baseband equivalent of the wireless channel? We will also address this question.
|Xb (f )|
1
W 2
W 2
|X(f )|
1 2
fc
fc
Figure 1: Magnitude spectrum of the real baseband signal and its passband signal
baseband before processing. Therefore it is most important to have a baseband equivalent representation of signals. Lets begin with the real baseband signal xb (t) (of double sided bandwidth W ) that we want to transmit over the wireless channel in a band centered around fc . In wireline channel, xb (t) would be the signal at the output of the DAC. We know that we can up-convert this signal by multiplying it by cos 2fc t. x(t) = xb (t) 2 cos 2fc t (1) The resulting signal x(t) has spectrum centered around fc and fc . Figure 1 shows this transformation diagramatically. We scale the carrier by 2 as cos 2fc t has power 1 . Thus, 2 by scaling, we are keeping the power in xb (t) and x(t) same. Note that since xb (t) is real, the magnitude of its Fourier transform, Xb (f ) is symmetric in f and hence the magnitude of the spectrum of the RF signal, X(f ) is symmetric around fc and fc . We note that to get real x(t), we need not have X(f ) symmetric around fc and fc . This is a consequence of xb (t) being real. To get back the baseband signal, we multiply x(t) again by 2 cos 2fc t and then pass
xb (t)
the signal through a low pass lter with bandwidth W . x(t) 2 cos 2fc t = 2 cos2 (2fc t).xb (t) = (1 + cos 4fc t)xb (t)
(2) (3)
The low pass lter will discard the signal xb (t) cos 4fc t as it is the bandpass signal centered around 2fc . Figure 2 shows this transformation diagramatically. One can see that if we multiply x(t) by 2 sin 2fc t instead of 2 cos 2fc t, we get xb (t) sin 4fc t and low pass lter will discard this signal completely. There will be a similar outcome had we modulated the baseband signal on 2 sin 2fc t and try to recover it by using 2 cos 2fc t. Thus, 1. Since the only dierence in 2 cos 2fc t and 2 sin 2fc t is the phase lag of , syn2 chronization of carrier phase is crucial in up-conversion and down-conversion. 2. We also note that the signals modulated on 2 cos 2fc t and 2 sin 2fc t never get mixed up in the process of down-conversion. Though both the signals share same frequency band, they are orthogonal to each other. Thus, we could have transmitted two real baseband signals in the same frequency band and doubled the data rate. This is possible as now we are using total double sided bandwidth of 2W instead of W as in wireline channel. The resulting RF signal is still real. However, the magnitude of the spectrum of the RF signal need not be symmetric around fc and fc . Thus, we can now have the RF signal x(t) which is x(t) = xb1 (t) 2 cos 2fc t xb2 (t) 2 sin 2fc t
(4)
signals xb1 (t) and xb2 (t) are obtained at the receiver by multiplying x(t) by The baseband 2 cos 2fc t and 2 sin 2fc t separately and then passing both the outputs through the low pass lters. Here we are modulating the amplitude of the carrier by the baseband data. Such a scheme is called amplitude modulation. When we modulate both sin and cos parts of the carrier by two undependent baseband signals, the scheme is called Quadrature Amplitude Modulation (QAM). 1
Can we up-convert one more baseband signal and still be able to recover it at the receiver? The answer is negative. This is because, we can express cos(2fc t + ) as cos cos 2fc t sin sin 2fc t. Thus any phase change is uniquely determined by the amplitudes of cos 2fc t and sin 2fc t.
1
The baseband signal xb (t) is now dened in terms of the pair (xb1 (t), xb2 (t)). In literature, this pair is denoted as (xI (t), xQ (t)), where I stands for in phase signal and Q stands for b b quadrature phase signal. To make the notation compact we can think of xb (t) as a complex signal dened as follows: def xb (t) = xI (t) + jxQ (t) (5) b b We will follow this notation hereafter. If the wireless channel is just the AWGN channel, then we know how to recover the baseband signal from the RF signal at the receiver and we are done. However, wireless channel is not AWGN channel. If h(t) denote the impulse response of the (time-invarient) wireless channel, the received RF signal is y(t) = h(t) x(t) + w(t) (6)
where w(t) is the RF noise. We will ignore the noise for the time being. Then, y(t) = h(t) x(t). x(t) is obtained by up-converting the baseband signal xb (t). We obtain the baseband signal yb (t) at the receiver by down-converting the received RF signal y(t). The question we want to address now is How does the channel impulse response manifests itself in baseband? How are the the baseband signals yb (t) and xb (t) related? It turns out that there is an baseband equivalent lter hb (t) of the channel lter h(t). The transmitted baseband signal xb (t) is ltered through the baseband channel lter hb (t) to give the received baseband signal yb (t). 2 yb (t) = hb (t) xb (t) To understand the relation between h(t) and hb (t), lets consider a few examples. 1. Lets take the simple case when h(t) = (t). Then, y(t) = x(t) and hence yb (t) = xb (t). Hence, hb (t) = (t). 2. Lets consider h(t) = (t t0 ). In this case, y(t) = x(t t0 ) = xI (t t0 ) 2 cos 2fc (t t0 ) xQ (t t0 ) 2 sin 2fc (t t0 ) b b
2
(7)
(8)
The intuition behind this relation can be obtained by representing the signals in frequency domain. We know that in frequency domain, Y (f ) = H(f )X(f ). We have also seen from Figure 1 that the X(f ) and Xb (f ) are related by translation in frequency domain and so are Y (f ) and Yb (f ). Thus, if we translate H(f ) appropriately, we will have Hb (f ) so that Yb (f ) = Hb (f )Xb (f ), i.e., in time domain, yb (t) = hb (t) xb (t).
We obtain the baseband signal yb (t) as I yb (t) = LPF y(t) 2 cos 2fc t = LPF (2 cos 2fc (t t0 ) cos 2fc t)xI (t t0 ) b (2 sin 2fc (t t0 ) cos 2fc t)xQ (t t0 ) b = LPF (cos 2fc (2t t0 ) + cos 2fc t0 )xI (t t0 ) b (sin 2fc (2t t0 ) sin 2fc t0 )xQ (t t0 ) b = xI (t t0 ) cos 2fc t0 + xQ (t t0 ) sin 2fc t0 b b = R xb (t t0 )ej2fc t0
Q Similarly, we obtain yb (t) as
(9)
(10)
Thus, yb (t) = xb (t t0 )ej2fc t0 Hence, hb (t) = ej2fc t0 (t t0 ) Thus, the baseband signal also gets delayed by the same amount as the passband signal. However, its phase also changes. This phase lag depends on the delay t0 as well as on the carrier frequency fc . We can generalize the second example to obtain the baseband equivalent representation of a generalized channel. Suppose the wireless channel is given by
L1
h(t) =
l=0
al (t tl )
(18)
hb (t) =
l=0
al ej2fc tl (t tl )
(19)
We will see in the next lecture that the wireless channel can actually be modeled as Equation 18.
Looking Ahead
In this lecture, we have seen the baseband representation of the RF signal. Writing baseband signal as a complex number simplies the notation a lot. In next few lectures, we will see that this notation will enable us to use many of the results from the wireline channel. We have also obtained the baseband equivalent of the channel given by Equation 18. In the next lecture we will see that this is a reasonable model for the wireless channel. We will also see that how the wireless cannels are dierent from the wireline channels. 5