Capacity

© Ammar Abu-Hudrouss
١
Islamic University Gaza

Shannon Capacity

 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

C  B log 2 1   
  P / N 0 B 
The channel bandwidth B
The received power P
N0 is the power spectral density of the noise

Wireless Communications
Slide 2

١
 Shannon Capacity

 Example
 Consider a wireless channel where power falloff with distance
follows the formula Pr(d) = Pt (d0/d )3 for d0 = 10 m.
Assume the channel has bandwidth B = 30 KHz and AWGN with
noise power spectral density of N0 = 10−9 W/Hz. For a transmit
power of 1 W, find the capacity of this channel for a transmit -
receive distance of 100 m and 1 Km.

 γ = 33 for d = 100 and C = 152.6 Kbps

 γ = 0.033 for d = 1 km and C = 1.4 Kbps

Wireless Communications
Slide 3

Capacity of Flat-Fading Channels

 We will consider three different scenarios

The channel distribution information is known at transmitter
and receiver
Receiver CSI
Transmitter and receiver CSI

Wireless Communications
Slide 4

٢
 Channel side information at the receiver

 Capacity defines theoretical rate limit

Maximum error free rate a channel can support
 Depends on what is known about channel
 Fading Statistics Known
Hard to find capacity
 Shannon capacity when fading Known at Receiver Only

C   B log 2 1    p( )d  B log 2 (1   )
0

Wireless Communications
Slide 5

Channel side information at the receiver

 Consider a flat-fading channel with i.i.d. channel gain g[i] which

can take on three possible values:
 g1 = .05 with probability p1 = .1, g2 = .5 with probability p2 = .5,
and g3 = 1 with probability p3 = .4.
 The transmit power is 10 mW, the noise spectral density is N0 =
10−9 W/Hz, and the channel bandwidth is 30 KHz. Assume the
receiver has knowledge of the instantaneous value of g[i] but
the transmitter does not. Find the Shannon capacity of this
channel and compare with the capacity of an AWGN channel with
the same average SNR.

γ1 = 0.8333, γ2 = 83.333 and γ2 = 333.3

C = 199.26 Kbps
C for AWGN = 223.8 Kbps

Wireless Communications
Slide 6

٣
 Outage capacity

 In this case, a minimum value of receive SNR is specified (γmin)

 The data rate is given by

R  B log 2 1   min 
 The data is correctly received as long as γ > γmin
 Outage probability Pout = p (γ < γmin)
 The outage capacity is defined as
C0  1  pout B log 2 1   min 
 The Pout that maximize capacity can maximize the average rate
correctly received

Wireless Communications
Slide 7

Outage Capacity

Normalized data rate versus outage probability

Wireless Communications
Slide 8

٤
  Example:
 Assume the same channel as in the previous example, with a
bandwidth of 30 KHz and three possible received SNRs: γ1 =
.8333 with p(γ1) = .1, γ2 = 83.33 with p(γ2) = .5, and γ3 = 333.33
with p(γ3) = .4.
 Find the capacity versus outage for this channel, and find the
average rate correctly received for outage probabilities
pout < .1, pout = .1 and pout = .6.
 solution
 For pout < 0.1 , we have to decode correctly for all the channel
states (including the weakest)
 C = B log2(1 + γmin )= 26.23 Kpbs
 For 0.1 < pout < 0.6
 C = B log2(1 + γmin )= 191.94 Kpbs
 And finally C = 251.55 Kpbs for 0.6 ≤ pout < 1

Wireless Communications
Slide 9

 Where as the outage capacity for each case

 Co = (1 - pout) C
 For p < 0.1
 C = 26.23 Kpbs
 For 0.1 < pout < 0.6
 C = = (1 - 0.1)* 191.94 = 171 Kpbs
 for 0.6 ≤ pout < 1
 And finally C = (1-0.6)* 251.55 = 172.75 Kpbs

Wireless Communications
Slide 10

٥
 Fading Known at
Transmitter and Receiver
 For fixed transmit power, same as with only receiver knowledge
of fading
 Transmit power S() can also be adapted
 Leads to optimization problem


max  S ( ) 
C  B log 1 
2  p ( )d
S ( ) : E[ S ( )]  S 0 S 

Wireless Communications
Slide 11

Optimal Adaptive Scheme

 Power Adaptation Waterfilling

1

S( ) 10  1   0 0


S  0 else 1

0 
 Capacity

C  
  log2   p( )d .
B 0  0 

Wireless Communications
Slide 12

٦
 Optimal Adaptive Scheme

 To find the value of γ0 we use

 S( )
 p  1
0 S

 Or
  1 1
0  0   p  1
 This formula has to be found numerically for continuous pdf.

Wireless Communications
Slide 13

Wireless Communications
Slide 14

٧
  Example: Assume the same channel as in the previous example,
with a bandwidth of 30 KHz and three possible
 received SNRs: γ1 = .8333 with p(γ1) = .1, γ2 = 83.33 with p(γ2) =
.5, and γ3 = 333.33 with p(γ3) = .4.
 Find the ergodic capacity of this channel assuming both
transmitter and receiver have instantaneous CSI.

 First, we need to find the cutoff value γ0


 1 1
    p  1
0  0 

This yields that γ0 = 0.89 > 0.8333

This value is inconsistence with γ1 . So we have to eliminate and

recalculate for γ0 .

Wireless Communications
Slide 15

This yields that γ0 = 0.89. so by assuming that is not used, the

capacity of the channel becomes

3
C   B log2  i /  0  p i   200.82 Kpbs
i 2

The capacity is slightly higher than receiver CSI only but it is still
significantly below that off an AWGN channel. (why?!)

Wireless Communications
Slide 16

٨
 Channel Inversion
 Fading inverted to maintain constant SNR
 Simplifies design (fixed rate)
 The zero-outage capacity is given by

 1 
C  B log 2 1     B log 2 1  
 E 1 /   
 Greatly reduces capacity
Capacity is zero in Rayleigh fading

Wireless Communications
Slide 17

 Example Assume the same channel as in the previous example,

with a bandwidth of 30 KHz and three possible
 received SNRs: γ1 = .8333 with p(γ1) = .1, γ2 = 83.33 with p(γ2) =
.5, and γ3 = 333.33 with p(γ3) = .4.
 Assuming transmitter and receiver CSI, find the zero-outage
capacity of this channel

0.1 0 .5 0.4
E 1 /       0.1272
0.8333 83.33 333.33

C  B log 2 1     9443 Kbps

 This is less than half of the Shannon capacity with optimal

water-filling adaptation

Wireless Communications
Slide 18

٩
 Truncated Inversion
 Truncated inversion
Invert channel above cutoff fade depth
S( )     0

S 0 else
Constant SNR (fixed rate) above cutoff
  1
  E 0 1 /     p  d
0 

Cutoff greatly increases capacity (close to optimal)

 1 
C  max B log 2 1   p   0 
0  E 1 /   
 0 

Wireless Communications
Slide 19

 Example: Assume the same channel as in the previous example,

with a bandwidth of 30 KHz and three possible
 received SNRs: γ1 = .8333 with p(γ1) = .1, γ2 = 83.33 with p(γ2) =
.5, and γ3 = 333.33 with p(γ3) = .4.
 Find the outage capacity of this channel and associated outage
probabilities for cutoff values γ0 = .84 and γ0 = 83.4.
 Which of these cutoff values yields a larger outage capacity?
 For γ0 = .84
3
0 .5 0.4
E 1 /     p i  /  i    0.0072
i2 83.33 333.33
 1 
C  B log 2 1   p   0   192.45 Kbps
 E 1 /   
 0 

Wireless Communications
Slide 20

١٠
  For γ0 = 83.3
0.4
E 1 /    p 3  /  3   .0012
333.33
 1 
C  B log 2 1   p   0   116.45 Kbps
 E 1 /   
 0 
 The first case has higher capacity even with very high value γ3
for compared to γ2 .

 This is because it occurs only 40% of the transmission time

Wireless Communications
Slide 21

Capacity in Flat-Fading

Wireless Communications
Slide 22

١١
 Frequency Selective : Fading Channels

 For TI channels, capacity achieved by water-filling in frequency

 Capacity of time-varying channel unknown
 Approximate by dividing into sub-bands
Each sub-band has width Bc (like MCM) where the channel
H(f) = hj is constant over each sub-band
Independent fading in each sub-band
Capacity is the sum of sub-band capacities

Wireless Communications
Slide 23

Frequency Selective : Fading Channels

 The capacity is given by

2
3  H j Pj 

C  B log 2 1  N B 
max Pj : P j  P 0
j  
 Which yields to

S( ) 10  1    0

S  0 else
1/|H(f)|2

P
Bc
f

Wireless Communications
Slide 24

١٢
  To find the value of γ0 we use

1 1 
  
 j 
1
j  0

 The capacity then becomes

C Blog 
 j  0
2 j /  0 .

Wireless Communications
Slide 25

Wireless Communications
Slide 26

١٣
  For continuous case:

2

 H  f  P  f  
C  max  log 2 1  df
Pj : Pj  P  N0 
j  
 To find the value of γ0 we use
  1 1
0  0   p  1
 And the capacity becomes

C log2   f  /  0 df .

f :  f  0

Wireless Communications
Slide 27

 Example :Consider a time-invariant frequency-selective block

fading channel consisting of three subchannels of bandwidth B =
1 MHz. The frequency response associated with each channel is
H1 = 1, H2 = 2 and H3 = 3.
 The transmit power constraint is P = 10 mW and the noise PSD is
N0 = 10−9 W/Hz. Find the Shannon capacity of this channel and
the optimal power allocation that achieves this capacity
 Solution
 The snrs are found to be γ1 = 10, γ2 = 40 and γ3 = 90.
 To find the optimum γ0:

3 1
1  1.14 OR 0  2.64
0 j j

C  Blog2  j /  0  10.93Mbps.
 j  0

Wireless Communications
Slide 28

١٤