0% found this document useful (0 votes)
9 views

AWC Unit 4

Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
9 views

AWC Unit 4

Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 11

Unit 4:

Space-Time Trellis Codes

Trellis-coded modulation (TCM) combines modulation and coding to achieve higher coding
gains. It provides significant coding gain and a better performance for a given bandwidth
compared to uncoded modulation schemes.

Spacetime trellis codes (STTCs) combine modulation and trellis coding to transmit information
over multiple transmit antennas and MIMO channels. Therefore, one can think of STTCs as
TCM schemes for MIMO channels.

Partition:

- Let us assume that we use a trellis with I states such that 2l branches leave every state.
- Enumerate the states from top to bottom starting from zero.
- Thus, we have states i = 0, 1, . . . , I − 1.
- To design a TCM for a SISO channel to transmit b bits per channel use, one needs to
utilize a modulation scheme with 2b+1 points.
- Let us denote dmin as the minimum Euclidean distance among all points in a set.
- The first step of TCM design is set partitioning.
- We need to partition the set of constellation points into a hierarchy of subsets such that
the dmin increases at each partition level.
- Figure shows such a set partitioning for an 8-PSK constellation.
- To define the Euclidean distance between two sets of constellation points, we consider
the set of all possible pairs such that the two symbols of the pair are selected from the
two different sets. Then, the Euclidean distance is defined as the minimum Euclidean
distance among all such pairs of symbols.
- The second step in the design is to assign different subsets to different paths in the
trellis.
- At each state, a collection of l input bits picks one of the 2l branches originating from
the state.
- The remaining b − l bits, if any, are utilized to pick one of the 2b−l parallel paths.
- The l bits that select the state transitions are “coded bits” while the b − l remaining bits,
if any, are “uncoded bits.”

Dr.HMV
Figure. Set partitioning for 8-PSK.

Ungerboeck has suggested the following principles, as a rule of thumb, to maximize the coding

gain of the TCM:

- all subsets should be used an equal number of times in the trellis;


- transitions originating from the same state or merging into the same state in the trellis
should be assigned subsets that are separated by the largest Euclidean distance;
- parallel paths, if they occur, should be assigned signal points separated by the largest
Euclidean distance.

Let us denote the state of the encoder at time t by St. The encoder usually starts at state zero,
that is S1 = 0. Then at time t, the transmitted symbol depends on the state at time t, St, and the
b input bits. The b input bits select the next state, St+1, and the constellation point that is
transmitted.

A four-state trellis example of Ungerboeck’s TCM is shown in below figure.

Figure. A four-state trellis; an example of Ungerboeck’s TCM.

Dr.HMV
- Next to every state, there are two sets corresponding to the two branches originating
from that state.
- The first set corresponds to the top branch and the second set corresponds to the bottom
branch.
- If there were more than two branches originating from each state, the sets from left to
right would correspond to branches from top to bottom.
- The b input bits are divided into two parts of l most significant bits and b − l least
significant bits.
- Here b = 2 and l = 1. Therefore, the most significant bit selects the leaving branch,
equivalently the set and next state St+1, and the least significant bit picks a symbol from
the selected set.
- The encoder’s state-machine goes to state St+1 and the same encoding process continues.
- Enough zero-coded bits should be added to the input bitstream in order to return the
state of the encoder to state zero at the end of transmission

To achieve full diversity, a STTC should satisfy the rank criterion.

STTC: Encoding

Let us concentrate on two transmit antennas.

- There are two symbols that are transmitted from these two antennas for every path in
the trellis and there is no parallel path in the STTCs. Therefore, the STTC can be
represented by a trellis and a pair of symbols for each trellis path.
- We need to use the corresponding indices of the symbols to present the transmitted
symbols for each path. For STTCs that send b bits/(s Hz) of information, 2b branches
leave every state. A set of 2b pairs of indices next to every state represents the 2b pairs
of symbols for the 2b outgoing branches from top to bottom.
- The below figure illustrates a rate one space-time trellis code to transmit r = 2 bits/(s
Hz).
- The code uses a QPSK constellation, b = 2, that includes indices 0, 1, 2, 3 to represent
1, j,−1,−j, respectively.
- Let us assume that the encoder is at state St at time t. Then, b = 2 bits arrive at the
encoder to pick one of the 2b = 4 branches leaving state St.

Dr.HMV
- The corresponding indices of the selected branch i1 i2 are used to choose two symbols
ct,1 ct,2 from the constellation. These symbols are respectively sent from the two transmit
antennas simultaneously.
- The encoder moves to state St+1 which is at the right-hand side of the selected branch.
- At the end, extra branches are picked to make sure that the encoder stops at state 0.

Figure: A four-state STTC; r = 2 bits/(s Hz) using QPSK.


- Although the code has been designed manually, there is a logic behind it that guarantees
full diversity. All branches diverging from a state contain the same symbol for the first
antenna while all branches merging to a state contain the same symbol for the second
antenna.
- Using a similar method, one can manually design full-rate full-diversity STTCs

STTC: Decoding

- The maximum-likelihood decoding finds the most likely valid path that starts from state
zero and merges to state zero after T + Q time slots.
- Let us assume that we receive r1,m , r2,m , . . . , rT+Q,m at time slots t = 1, 2, . . . , T + Q
and receive antenna m.
- The Viterbi algorithm can be used for the ML decoding of STTCs.
- If a branch of the trellis transmits symbols s1 and s2 from antennas one and two,
respectively, the corresponding branch metric is given by

- The ML decoder finds the set of constellation symbols that construct a valid path and
solves the following minimization problem:

Dr.HMV
Design rules and properties of STTCs

Two design rules have been suggested to achieve full diversity for two transmit antennas.

- transitions diverging from the same state should differ in the second symbol;
- transitions merging to the same state should differ in the first symbol.

The following lemma provides a lower bound on the number of states that is required to achieve
a particular diversity order.

Lemma: For a STTC with rate b bits/(s Hz) and a diversity r , at least 2b(r−1) states are needed.

Proof: For a rate of b bits/(s Hz), we need 2b branches out of each state.

- At time r − 1, there are 2b(r−1) paths originating from state 0.


- To achieve a diversity of r, all these paths should be separated, that is none of them can
merge at the same state in any time between 0 and r − 1.
- To prove this fact, we assume that there exists two paths originating at state 0 and
merging at time r' where r' < r.
- Let us pick a pair of codewords (C1, C2) that contain these two paths up to time r' and
the same set of branches after time r'.
- Then, the corresponding difference matrix D(C1, C2) contains only r' nonzero rows and
its rank cannot exceed r' < r.
- This is a contradiction that shows all 2b(r−1) paths at time r − 1 should be separate and
therefore we should have at least 2b(r−1) states.
- As a result of the above lemma, we need at least 2b states to design a full diversity STTC
for two transmit antennas.
- In other words, it is impossible to design STTCs with less than 2b states and achieve
full diversity.

Trellis diagram examples:

Example 1: For two states and r = 1 bit/(s Hz) using BPSK is depicted in Figure (i).

Figure: (i) A two-state STTC; r = 1 bit/(s Hz) using BPSK.

The minimum coding gain distance (CGD) for this code is 16.

Dr.HMV
Example 2: A four-state STTC for transmitting b = 2 bits/(s Hz) using QPSK is designed
from the following generating matrix:

- A four-state has 16 levels, that can be represented as in the matrix A:


0000 0100 1000 1100
0001 0101 1001 1101
𝐴=[ ]
0010 0110 1010 1110
0011 0111 1011 1111
- Multiply each row matrix of A with generator matric G, to get trellis code.
T=AxG
00 02 22 24
31 33 53 55
T= [ ]
10 12 32 34
41 43 63 65
- Since, it has only four states, the exceeding values need to be altered to fit within four
states only, therefore:
00 02 22 20
31 33 13 11
T= [ ]
10 12 32 30
01 03 23 21
- The corresponding trellis representation is given in figure below. The CGD of this code
is 8.

Figure: A four-state STTC; r = 2 bits/(s Hz) using QPSK.

Dr.HMV
Example 3: A four-state STTC for transmitting b = 2 bits/(s Hz) using QPSK is designed
from the following generating matrix:

- A four-state has 16 levels, that can be represented as in the matrix A:


0000 0100 1000 1100
0001 0101 1001 1101
𝐴=[ ]
0010 0110 1010 1110
0011 0111 1011 1111
- Multiply each row matrix of A with generator matric G, to get trellis code.
T=AxG
00 02 22 24
21 23 43 45
T= [ ]
12 14 34 36
33 35 55 57
- Since, it has only four states, the exceeding values need to be altered to fit within four
states only, therefore:
00 02 22 20
21 23 03 01
T= [ ]
12 10 30 32
33 31 11 13
- The corresponding trellis representation is given in figure below. The CGD of this code
is 8.

Figure: A four-state STTC; r = 2 bits/(s Hz) using QPSK.

Dr.HMV
Design Criteria for Slow Rayleigh Fading STTCs

We now define the basic design criteria for slow Rayleigh STTCs:

- The rank criterion: To obtain maximal diversity, we need to maximize the minimum
rank r of the matrix B over all pairs of distinct code words. A diversity advantage of
rMR is achieved.
- The determinant criterion: Let rMR be the target diversity advantage. Then the design
goal is to maximize the minimum determinant ∏𝑟𝑖=1 𝜆𝑖 of the matrix A along the pairs
of distinct code words with that minimum rank.

These criteria are referred to as rank and determinant criteria.

Performance Analysis:

The performance of the STTC on slow fading channels is evaluated through simulations.

a. Slow Fading Channel:

With the employment of rank and determinant criterion in formulating codes for one receive
antenna, consider the simulation figure (a) which shows that all the codes achieve the same
diversity order of 2. The code performance is improved by increasing the number of states.

Figure: (a) Performance comparison of 4-PSK codes on slow fading channels with two
transmit and one receive antennas.

Dr.HMV
Figure: (b) Performance comparison of 4-PSK codes based on slow fading channels with two
transmit and two receive antennas and four transmit and two receive antennas.

Figure (b) shows the relative performance of 4-PSK 4-state and 8-state codes with MT = MR =
2 and MT = 4 and MR = 2. Figure reveals that increasing the number of transmit antennas also
increases the margin of the coding gain compared with the coding gains in the top- half graph.

From the pairwise error probability (PEP), the value of rMR defines the amount of coding gain.

- Similarly, if we keep MT constant and increase MR we achieve the same result with
something more, in that part of this margin is also due to the array gain through multiple
receive antennas.
- Furthermore, the diversity order realized with this scheme in the lower half is twice that
in the upper half.
- Proceeding logically, as the number of receiver antennas increases, the diversity order
increases proportionately and the channel tends to AWGN due to the increased
diversity.
- A similar effect can be observed by keeping the number of receive antennas constant
and increasing the number of transmit antennas.
- Finally, in the presence of a large number of receive antennas, increasing the number
of transmit antennas does not produce that much of an increase in performance, as seen
in the case when the number of receive antennas is limited and we increase the number
of transmit antennas.

Dr.HMV
b. Fast fading channel:
- Below figure (c) shows the frame error rate (FER) performance of QPSK STTC with a
bandwidth efficiency of 2 bit/s/Hz in Rayleigh channel.

Figure: (c) Performance of the QPSK STTC with two transmit and one receive
antennas
- The number of receive antennas was one in the simulation.
- We can see that 16-state QPSK codes are better than 4-state codes by 5.9 dB at a FER
of 10-2 for two transmit antennas.
- Once again, as the number of states increases, the coding gain increases and so does the
performance.
- Different values yield different coding gains, which are represented by the horizontal
shifts of the FER curves.
- In figure (d) we examine the case of three transmit antennas.
- We note that 16-state QPSK codes are superior to 4-state codes by 6.8 dB at a FER of
10-2.
- This means that the performance relative to two transmit antennas has improved.
- The conclusion here is that as the number of the transmit antennas gets larger, the
performance gain achieved from increasing the number of states becomes larger.

Dr.HMV
Figure: (d) Performance of QPSK STTC with three transmit and one receive antennas.

Dr.HMV

You might also like