The Fading Multipath Channel
I. Introduction:
The multipath is the typical model used to describe
wireless or mobile communication channels. The
received signal is a sum of multiple reflected copies of
the transmitted one, each of a different attenuation or
phase shift. At the receiver side, the multipath appears
as multiplicative distortion to the transmitted signal.
Time variation in the communication channel causes
Fading. Usually, it happens due to motion or climate
change. Figure 1: Typical urban propagation scenario
II. Fading Types:
Assuming 𝑠(𝑡) is the transmitted signal, and there are N propagation paths leading to the
receiver. The received signal can be written as
𝑁
𝑟(𝑡) = ∑ 𝛼𝑛 (𝑡)𝑠(𝑡 − 𝜏𝑛 ) (1)
𝑛=1
Where 𝛼𝑛 (𝑡) is an attenuation factor and 𝜏𝑛 is the relative time delay for the path
numbered n. For some channels, such as the tropospheric scatter channel, the received
signal is a continuum of multipath components. Thus, we can write (1) in integral form:
∞
𝑟(𝑡) = ∫ 𝛼(𝜏, 𝑡)𝑠(𝑡 − 𝜏)𝑑𝜏 (2)
−∞
Hence, the multipath channel can be represented by a time-varying filter whose impulse
response is 𝛼(𝜏, 𝑡).
For a fixed time 𝑡 = 𝑡0, if the channel filter impulse response 𝛼(𝜏, 𝑡0 ) does not distort the
transmitted signal 𝑠(𝑡); the multipath channel is said to be frequency-nonselective
channel. Otherwise, the channel would be frequency-selective.
Fading occurs when the channel filter 𝛼(𝜏, 𝑡) varies with time 𝑡.If the channel filter
𝛼(𝜏0 , 𝑡) at time delay 𝜏 = 𝜏0 varies slowly with 𝑡 in relative to the transmitted symbol
time 𝑇𝑠 ; the multipath channel is a slowly fading channel. Otherwise, the channel would
be a fast fading one.
Hence, each multipath channel is characterized by two parameters: the coherence time
(∆𝑡)𝑐 and the coherence bandwidth (∆𝑓)𝑐 .
When the transmitted signal bandwidth 𝑊 does not exceed the coherence bandwidth,
the channel is described as frequency-nonselective channel.
When the transmitted symbol time 𝑇𝑠 is smaller than the coherence time, the channel has
slow fading. Otherwise, it has fast fading. Usually, fading is characterized by the Doppler
Spread 𝐵𝑑 = 1⁄(∆𝑡)𝑐 .
I. Statistical Models for Fading Channels:
Rayleigh Fading. This model is valid when there are a lot of paths, each with a different
received power or phase shift, and there is no dominant path. According to the central
limit theorem, the sum of paths’ contributions appears as random, zero-mean and
Gaussian variable on both the real and the imaginary part of the channel response.
Assuming 𝑋 and 𝑌 are two i.i.d gaussian variables, the received signal amplitude and
phase are described by the random variables 𝑅 and Φ as following
𝑌
𝑅 = √𝑋 2 + 𝑌 2 , Φ = tan−1 (3)
𝑋
Where 𝑅 is has a Rayleigh probability distribution –see Figure2.a-, and its PDF is given by
(4). While Φ has a uniform distribution on the interval[– 𝜋, 𝜋].
𝑟 −𝑟22
𝑃𝑅 (𝑟|σ) = 2 𝑒 2σ 𝑟≥0 (4)
σ
𝜋 4−𝜋 2
𝐸[𝑅] = √ 𝜎 , 𝐸[𝑅 2 ] = 𝜎 (5)
2 2
(a) Rayleigh Fading (b) Rician Fading
Figure2 : Received power variation in cases of (a) Rayleigh fading and (b) Rician fading
Rician Fading. When the received power comes mainly from one path, the distribution
of the received signal’s amplitude becomes Rician. Usually, this is the case when there is
a LOS between the transmitter and the receiver.
The same statistical model for the Rayleigh channel remains valid, but with 𝑋 or 𝑌 having
non-zero means, due to the effect of the specular component (the LOS path). The Rician
PDF is given by
2 2
𝑟 −(𝑟 +𝑣 ) 𝑥𝑣
𝑃𝑅 (𝑟|𝜎, 𝑣) = 2 𝑒 2σ2 𝐼0 ( 2 ) 𝑟≥0 (6)
𝜎 𝜎
Equivalently,
𝑅 = √𝑋 2 + 𝑌 2 , 𝑋~𝑁(𝑣𝑐𝑜𝑠𝜃, 𝜎 2 ) 𝑎𝑛𝑑 𝑌~𝑁(𝑣𝑠𝑖𝑛𝜃, 𝜎 2 ) (7)
𝜎 is the standard deviation of the local power,
𝑣 peak amplitude of the dominant signal and
𝐼0 (. . .) modified Bessel Function of the first kind and zero order.
Note:
The Bessel functions are more frequently defined as solutions to the differential equation
The 𝐾-factor is the ratio of the specular-to-diffuse power for the LOS path.
𝑣2
𝐾= (8)
2𝜎 2
Its value is typically between 1 and 10. A 𝐾-factor of 0 corresponds to Rayleigh fading.
The rician fading is identified by two parameters: they could be either (𝜎, 𝑣) as written in
(6), or(𝑖𝑡𝑠 𝑝𝑜𝑤𝑒𝑟 𝐸[𝑅 2 ], the 𝐾 − factor).
II. Lab Work:
For the following, we assume frequency-nonselective slowly fading channel, a matched
filter receiver, perfect timing and coherent demodulation with exact phase estimation
for the channel filter.
A. Plot the histogram of a:
1. Rayleigh distributed random variable of unity power. Estimate its mean and
variance and compare with what you calculate using (5).
Recall that, If 𝑋 and 𝑌 are i.i.d gaussian random variables with zero mean, then
𝑅 = 𝑎𝑏𝑠(𝑋 + 𝑖 ∗ 𝑌) has Rayleigh distribution.
2. Rician distributed random variable of unity power and a 𝐾-factor of 0, 2, 5, 10 and
20 (these values are not in dB!). Use just ONE graph and COMMENT.
Recall that, if 𝑋 and 𝑌 are i.i.d gaussian random variables with zero mean, then
𝑅 = 𝑎𝑏𝑠((𝑋 + 𝑣𝑐𝑜𝑠𝜃) + 𝑖 ∗ (𝑌 + 𝑣𝑠𝑖𝑛𝜃)) has Rician distribution -𝜃 is additional
degree of freedom-.
B. Estimate the performance of BPSK modulation assuming Rayleigh fading when
Eb/N0=0:3:33 dB. Compare with the performance in AWGN without fading.
Estimate the performance of BPSK modulation assuming Rician fading for a 𝐾-factor of 0,
2, 5, 10 and 20. Compare with the results you have got in B.