Title: Introduction to Information theory channel capacity and models
1Introduction to Information theorychannel
capacity and models
- A.J. Han Vinck
- University of Essen
- May 2011
2This lecture
- Some models
- Channel capacity
- Shannon channel coding theorem
- converse
3some channel models
Input X P(yx) output Y
transition probabilities
memoryless - output at time i depends only on
input at time i - input and output alphabet finite
4Example binary symmetric channel (BSC)
Error Source
E
X
Output
Input
E is the binary error sequence s.t. P(1)
1-P(0) p X is the binary information sequence Y
is the binary output sequence
5from AWGN to BSC
p
Homework calculate the capacity as a function of
A and ?2
6Other models
1-e e e 1-e
0 1
0 (light on) 1 (light off)
0 1
0 E 1
X Y
p 1-p
P(X0) P0
P(X0) P0
Z-channel (optical)
Erasure channel (MAC)
7Erasure with errors
1-p-e
0 1
0 E 1
e
p
p
e
1-p-e
8burst error model (Gilbert-Elliot)
Random error channel outputs independent
P(0) 1- P(1)
Error Source
Burst error channel outputs dependent
P(0 state bad ) P(1state bad )
1/2 P(0 state good ) 1 - P(1state good
) 0.999
Error Source
State info good or bad
transition probability
Pgb
Pbb
good
bad
Pgg
Pbg
9channel capacity
-
- I(XY) H(X) - H(XY) H(Y) H(YX) (Shannon
1948) - H(X) H(XY)
-
- notes
- capacity depends on input probabilities
- because the transition probabilites are fixed
-
-
X
Y
channel
10Practical communication system design
Code book
Code word in
receive
message
estimate
2k
decoder
channel
Code book
with errors
n
There are 2k code words of length n k is the
number of information bits transmitted in n
channel uses
11Channel capacity
Definition The rate R of a code is the ratio
k/n, where k is the number of information bits
transmitted in n channel uses Shannon showed
that for R ? C encoding methods exist with
decoding error probability 0
12 Encoding and decoding according to Shannon
Code 2k binary codewords where p(0) P(1)
½ Channel errors P(0 ?1) P(1 ? 0) p i.e.
? error sequences ? 2nh(p) Decoder search
around received sequence for codeword with ? np
differences
space of 2n binary sequences
13 decoding error probability
- For t errors t/n-pgt ?
- 0 for n ? ?
- (law of large numbers)
- 2. gt 1 code word in region
- (codewords random)
-
14channel capacity the BSC
I(XY) H(Y) H(YX) the maximum of H(Y)
1 since Y is binary H(YX) h(p)
P(X0)h(p) P(X1)h(p)
X Y
Conclusion the capacity for the BSC CBSC 1-
h(p) Homework draw CBSC , what happens for p gt
½
15channel capacity the BSC
Explain the behaviour!
16channel capacity the Z-channel
Application in optical communications
H(Y) h(P0 p(1- P0 ) ) H(YX) (1 - P0 )
h(p) For capacity, maximize I(XY) over P0
0 1
0 (light on) 1 (light off)
X Y
p 1-p
P(X0) P0
17channel capacity the erasure channel
Application cdma detection
1-e e e 1-e
I(XY) H(X) H(XY) H(X) h(P0 )
H(XY) e h(P0) Thus Cerasure 1 e
(check!, draw and compare with BSC and Z)
0 1
0 E 1
X Y
P(X0) P0
18Capacity and coding for the erasure channel
Code 2k binary codewords where p(0) P(1)
½ Channel errors P(0 ?E) P(1 ? E) e
Decoder search around received sequence for
codeword with ? ne differences
space of 2n binary sequences
19 decoding error probability
- For t erasures t/n-egt ?
- 0 for n ? ?
- (law of large numbers)
- gt 1 candidate codeword agrees in n(1-e)
positions after ne positiona are erased
(codewords random)
20Erasure with errors calculate the capacity!
1-p-e
0 1
0 E 1
e
p
p
e
1-p-e
21example
0 1 2
0 1 2
1/3
1/3
- Consider the following example
- For P(0) P(2) p, P(1) 1-2p
- H(Y) h(1/3 2p/3) (2/3 2p/3) H(YX)
(1-2p)log23 - Q maximize H(Y) H(YX) as a function of p
- Q is this the capacity?
- hint use the following log2x lnx / ln 2
d lnx / dx 1/x
22channel models general diagram
P11
y1
x1
P21
Input alphabet X x1, x2, , xn Output
alphabet Y y1, y2, , ym Pji
PYX(yjxi) In general calculating capacity
needs more theory
P12
y2
x2
P22
xn
Pmn
ym
The statistical behavior of the channel is
completely defined by the channel transition
probabilities Pji PYX(yjxi)
23 clue
I(XY) is convex ? in the input
probabilities i.e. finding a maximum is simple
24Channel capacity converse
For R gt C the decoding error probability gt 0
Pe
k/n
C
25Converse For a discrete memory less channel
channel
Xi Yi
Source generates one out of 2k equiprobable
messages
encoder
channel
decoder
source
m Xn Yn
m
Let Pe probability that m ? m
26converse R k/n
k H(M) I(MYn)H(MYn) Xn is a
function of M Fano ? I(XnYn)
1 k Pe ? nC 1 k Pe
1 C n/k - 1/k ? Pe
Pe ? 1 C/R - 1/nR Hence for large n, and
R gt C, the probability of error Pe gt 0
27We used the data processing theorem Cascading of
Channels
The overall transmission rate I(XZ) for the
cascade can not be larger than I(YZ), that is
28Appendix
Assume binary sequence P(0) 1 P(1) 1-p
t is the of 1s in the sequence Then n ? ? ,
? gt 0 Weak law of large numbers Probability
( t/n p gt ? ) ? 0 i.e. we expect with high
probability pn 1s
29Appendix
Consequence
1. 2. 3.
n(p- ?) lt t lt n(p ?) with high probability
Homework prove the approximation using ln N! N
lnN for N large. Or use the Stirling
approximation
30Binary Entropy h(p) -plog2p (1-p) log2
(1-p)
Note h(p) h(1-p)
31Capacity for Additive White Gaussian Noise
Noise
Input X
Output Y
W is (single sided) bandwidth
Input X is Gaussian with power spectral density
(psd) S/2W Noise is Gaussian with psd
?2noise Output Y is Gaussian with psd ?y2 ?
S/2W ?2noise
For Gaussian Channels ?y2 ??x2 ??noise2
32Noise
X
Y
X
Y
33 Middleton type of burst channel model
0 1
0 1
Transition probability P(0)
channel 1
channel 2
Select channel k with probability Q(k)
channel k has transition probability p(k)
34Fritzman model
- multiple states G and only one state B
- Closer to an actual real-world channel
35Interleaving from bursty to random
bursty
Message interleaver channel
interleaver -1 message encoder
decoder
random error
Note interleaving brings encoding and decoding
delay Homework compare the block and
convolutional interleaving w.r.t. delay
36Interleaving block
Channel models are difficult to derive - burst
definition ? - random and burst errors ? for
practical reasons convert burst into random error
read in row wise
transmit column wise
1 0 0 1 1
0 1 0 0 1
1 0 000
0 0 1 1 0
1 0 0 1 1
37De-Interleaving block
read in column wise this row contains 1 error
1 0 0 1 1
0 1 0 0 1
1 e e e e
e e 1 1 0
1 0 0 1 1
read out row wise
38Interleaving convolutional
input sequence 0 input sequence 1 delay of b
elements ??? input sequence m-1 delay of
(m-1)b elements Example b 5, m 3
in
out