Wireless
Digital Communications
Topic 2
Capacity of Fading Channels
Classification of Wireless Channels
多径性 时变性
Flat Fading Time-Invariant
(Narrow-Band Transmission)
Frequency-Selective Fading Time-Variant
(Broad-Band Transmission) (Fast, Slow)
Outline
Capacity of Wireless Channels
Pioneered by Claude Shannon in 1948
Defined as the maximum MI of channel
Maximum error-free data rate a channel
can support.
Theoretical limit (not achievable)
Channel characteristic
Not dependent on design techniques
Basic Principles
Channel Capacity
Example 4.1
Power and Bandwidth Limited Regions
Capacity of Flat-Fading Channels
Fading Value Known at Receiver Only (Receiver CSI):
The channel gain is known to the receiver, and both the
transmitter and receiver know the distribution.
Fading Value Known at Transmitter and Receiver
(Transmitter and Receiver CSI): The channel gain is
known to the transmitter and receiver, and both the
transmitter and receiver know the distribution.
Capacity of Flat-Fading Channels
Receiver CSI Only
P g[i ]
N0 B
Capacity of Flat-Fading Channels
Receiver CSI Only
Example 4.2
Capacity of Flat-Fading Channels
Receiver CSI Only and Constant SNR over a Large
Number of Transmissions
Capacity with Outage
C B log 2 (1 min )
Pout P( min )
Average Rate Correctly Received
Cout (1 Pout ) B log 2 (1 min )
Capacity of Flat-Fading Channels
Transmitter and Receiver CSI
Transmit power P() can be adapted
Leads to optimization problem
max P( )
C
P( ) : E[ P( )] P 0 B log 2 1
P
p ( )d
P g[i ]
N0 B P ( ) g[i ]
N0 B
Capacity of Flat-Fading Channels
Transmitter and Receiver CSI
Optimal Power Allocation:
Water-Filling Scheme
Capacity Comparisons
Capacity Comparisons-Low SNR
Example 4.4
Example 4.4
Example 4.4
Channel Inversion
Fading inverted to maintain constant SNR
P( )
Simplifies design (fixed rate) P
Greatly reduces capacity
Capacity is zero in Rayleigh fading
Truncated inversion
Invert channel above cutoff fade depth
Constant SNR (fixed rate) above cutoff
Cutoff greatly increases capacity
Close to optimal
Channel Inversion
in More Details
Channel inversion
If target SNR is σ,transmit at P( ) P /
Expected power constraint gives 1 / E[1 / ]
What if E[1/γ]=∞?
Truncted channel inversion
Do not use the channel if γ<γ1(outage)
Expected power constraint gives 1 / E [1 / ]
1
Example 4.5
Example 4.6
Capacity of Frequency Selective
Fading Channels（Time-Invariant ）
Capacity of Frequency Selective
Fading Channels（Time-Invariant ）
Frequency selective fading channel is assumed to be a
set of AWGN channels in parallel, power of jth channel
is Pj 2
H j Pj 2
SNR of jth channel: Hj P
N0 B j
Capacity N0 B
Pj
C max B log 2 (1 j )
Pj : Pj P P
j
j
Optimal power allocation, obtained through Lagrangian
Technique leads to water-filling power allocation in
frequency
Capacity of Frequency Selective
Fading Channels（Time-Invariant ）
Water Filling in Frequency Domain
Example 4.7
Capacity of Frequency Selective
Fading Channels-Continuous Fading
Capacity of Frequency Selective
Fading Channels-Continuous Fading
Main Points
Capacity depends on degree of channel knowledge , CSI at receiver
only, CSI at both receiver and transmitter
Capacity with Transceiver CSI knowledge: variable-rate variable-
power transmission (water filling) optimal
▲ (Time-variant)Flat fading channels water filling in instantaneous
SNR
▲ (Time-invariant)Frequency selective fading water filling in
frequency
Channel inversion practical, but should truncate
Water-filling provides significant improvement at low SNR
Ex. 4.11