Simulation of Adaptive Bit Loading for Wireless Network

Channels
Submitted by:

Nithin Dinker Kamath

[email protected]

ABSTRACT
My project involved the study and analysis of Adaptive Bit Loading Technique used in
Wireless Channels using Frequency Division Multiplexing (FDM). To understand this
technique I simulated the same on MatLab platform. The reason for using MatLab was
the ease in with which matrices can be manipulated. Hence I could store the parameters
for the Sub Channel as a matrice.
My simulation consisted of two parts.
a. Generate a random wireless channel. Divide it into Sub Channels and then
transform it into frequency domain using Fast Fourier Transform and Single
Value Decomposition
b. Do Adaptive Bit Loading. The parameter of the Sub Channel which indicates the
whether the Sub Channel is “Good” or “Bad” was the gain it provided.

The Algorithm which was used do Energy Allocation was Chow Algorithm and to do
optimized Bit Allocation was Campello.

1 INTRODUCTION
Wireless channels are inherently heterogeneous, i.e. that is amount of fading and noise
varies considerably from sub channel to sub channel. The reason for this is maybe
various: atmospheric conditions, frequency spectrum selected, selective fading or any
random phenomenon.
Frequency Division Multiplexing (FDM) has been a widely-used technique for signal
transmission in frequency selective channels. In essence, FDM divides the channel
bandwidth into sub channels and transmits multiple relatively low rate signals by carrying
each signal on a separate carrier frequency.
Hence, some sub channels may have good gain: that is, provide relatively less
attenuation, distortion and noise interference, while other sub channels may be bad
because the sub channel Signal to Noise Ratio (SNR) may be high.

2 EFFECTS OF SNR
While it is not really means that the bits traveling through sub channel that has low SNR
will not be received, the probability that such receptions being bad or corrupt is high.
Also, other factors which effects successful reception are
a. Frequency Selected
b. Guard Band
c. Power Allocated
d. Modulation Scheme
In FDM, the data packets are split into smaller packets, each of which is transmitted on a
particular sub channel. Hence if a particular sub channels fails, on the receiver side, it
might be required to retransmit the packet again. Hence there are lots of re-transmissions.
Another variation of FDM is known as OFDM, in which the sub channel frequencies are
selected such a way that they are orthogonal to each other. That is they have a phase
division of 900 between. Even though this reduces Inter Symbol Interference, the loss
because of the channel is still very much present.

Figure 1: OFDM Channel

3 ADAPTIVE BIT LOADING

One of the possible solutions for this is using Adaptive Bit Loading. What this technique
does that
a. More number of bits are made to go through better part of the channel. (Bit
Allocation)
b. More energy is allocated to sub channels which are corrupt. (Energy Allocation)
The reason of allocating more bits to better sub channel is that, since these channels have
better SNR, large number of bits can be received with high degree of error-freeness. At
the same time, by allocating more energy to poorer sub carriers, you are increasing signal
strength of such sub channels so much that they can overcome noise that will be added by
the channel.
Hence the advantages are two pronged:
a. Send more bits through better sub channels so large number of bits are received
error free.
b. Give more energy to bad sub channels so that it improves the chance of successful
reception.
4 SIMULATION
4.1 Channel Creation
A random channel was created. The parameters that were assumed for the channel were:
The channel provided exponential fading. That is the signal strength decayed
exponentially with distance.

Code:
H_int = 1/sqrt(2)*(randn(length(A), 1) + j*randn(length(A), 1));

Then the channel was transformed from Time Domain to Frequency Domain using Finite
Fourier Transform and Single Value Decomposition

Code:
H_f = zeros(1, (N));
h_f = fft(H_int2(1:1:(N-1)+1,1));
for k = 1:(N)
H_f(1,1+(k-1)) = h_f(k);
end

for i = 1:N
[Utmp Stmp Vtmp] = svd(H_f(:,(i-1)+1:i));
S=[S Stmp];
end
Now the parameter S will store the gain of the sub channel. Since S is a matrix, S(i) is the
gain of the ith sub channel.

4.2 Adaptive Bit Loading

Adaptive Bit Loading was done using two algorithm
1. Chow Algorithm [1]
2. Campello Algorithm [2]

Initially each sub channel was allocated energy using Chow Allocation and at the same
time it was allocated maximum number of bits allowable per sub channel. Then bit
allocation was done using Campello Algorithm.
Before running the algorithm, sub channels SNR was determined.

Code:
SNR = abs((subchan_gains.^2)./(noise));

From Chow’s Algorithm, compute the energy for the ith sub channel based on the
number of bits
ei(b(i)) = (2b(i) - 1)/SNR(i)

Implementing Chow’s Algorithm

Code:
for i = 1:num_subc
bits_alloc(i) = M; %Allocating Maximum Number of
Bits allowable Sub Channel to each Sub Channel
 energy_alloc(i) = (2^bits_alloc(i)-1)/SNR(i) ; %Allocating energy
to each Sub Channel
end

Then I compiled a matrix of energy increments. That is extra energy required by a sub
channel to transmitting n bits to transmit (n+1) bits. This was calculated using
∆ei(b) = ei(n) - ei(n - 1) = (2n-1)/SNR(i)
This was stored in a matrix so that it can be referenced while implementing Campello’s
Algorithm

Code:
energytable = abs((1./SNR)'*(2.^([1:M+1]-1)));
energytable(:,M+1) = Inf*ones(size(energytable(:,M+1)));
for i = 3:2:M
energytable(:,i) = (energytable(:,i) +energytable(:,i+1))/2;
energytable(:,i+1) = energytable(:,i);
end

Campello’s Algorithm
B’ = 0
for n (1 to N) where N = number of sub channels
while (B’ != B); where B = maximum bits allowable per sub channel
if (B > B)
n = arg max1<j<N ∆ej(bj )
B=B–1
b(n)=b(n) – 1
else
n = arg max1<j<N ∆ej(bj+1 )
B=B+1
b(n)=b(n) + 1

Implementation of Campello’s Algorithm

Code:
bt = M*num_subc;
while (bt ~= total_bits)
if (bt > total_bits)
max_val = 0;
max_ind = ceil(rand(1)*num_subc);
for i = 1:num_subc
if(bits_alloc(i) ~= 0)
temp = energytable(i,bits_alloc(i));
else
temp = 0;
end
if(temp > max_val)
max_val = temp;
max_ind = i;
end
end
if(bits_alloc(max_ind) > 0)
bits_alloc(max_ind) = bits_alloc(max_ind) -1;
 energy_alloc(max_ind) = energy_alloc(max_ind) - max_val;
bt = bt-1;
end
else
min_val = Inf;
min_ind = ceil(rand(1)*num_subc);
for i = 1:num_subc
if(bits_alloc(i) ~=0 & bits_alloc(i) <9)
temp = energytable(i,bits_alloc(i) + 1);
else
temp = Inf;
end
if(temp < min_val)
min_val = temp;
min_ind = i;
end
end
if(bits_alloc(min_ind) < 8)
bits_alloc(min_ind) = bits_alloc(min_ind) + 1;
if (min_val == inf)
min_val = energytable(min_ind,bits_alloc(min_ind));
end
energy_alloc(min_ind) = energy_alloc(min_ind) +min_val;
bt = bt+1;
end
end
end
5 SIMULATION RESULTS
The following input parameters were used for the simulation

Code:
A = [1 1/exp(1) 1/exp(2) 1/exp(3)]; %Power Delay Profile
Nchannel = 64; %Number of Sub Channels
Noise = 1e-3 %Noise in the Channel
Bsub = 8; %Maximum Number of Bits per Sub
Channel
Bmax = Nchannel*Bsub; %Maximum Channel Capacity
Btotal = Bmax/2; %Total Number of Bits to be
Transmitted

Figure 2: Simulation Result

6 FACTORS AFFECTING ADAPTIVE BIT LOADING
The various factors which effect bit loading is
a. Bits transmitted. If the number of bits transmitted is very close to the maximum
channel capacity, then the efficiency of this technique decreases. The reason
being, we have to allocate maximum number of bits possible to the sub channel.
b. Input Power. The lower the input power given the lesser the energy allocated per
sub channel.

7 REQUIREMENTS AND OVERHEADS OF ADAPTIVE BIT LOADING

Requirements include
a. Transmitter and Receiver: The transmitter and receiver should be able to handle
multiple modulation schemes. This is because the modulation scheme used to
transmitted 2 bit may be different for one used to transmit 6 or 8 bits. Also they
should be adaptive since they have to switch between different reception
b. Instantaneous Channel Knowledge: We mush have instantaneous channel
knowledge to implement this technique. This is obtained by transmitting a “pilot”
signal at regular intervals.
Of the overheads
a. Overhead due to transmission of pilot signals
b. Overhead due to reconfiguration of transmitters and receivers
c. Overhead due to synchronization

8 ADVANTAGES OF ADAPTIVE BIT LOADING

The advantages of using Adaptive Bit Loading are
a. Reduction of BER: One of the obvious advantages is that since the probability of
successful reception of bits is increased because of bit and energy allocation lower
BERs can be achieved. If along with Adaptive Bit Loading, Error Control is done,
further reduction of BER is possible
b. Reduction in Retransmissions: Because more and more bits are successfully
received, the need to re transmit data is reduced
c. Higher Data Transfer Rates: Higher data rates can be achieved because of the
above two reasons
d. Energy Conservation: Energy is conserved because lesser energy is allocated to
good sub channels. Also at the same time if a sub channel is extremely bad and is
not transmitting any bits because 0 bits were allocated to it during bit eallocation,
no power needs to be supplied to such sub channel. Also because less number of
re transmissions occur, the overhead because of this is reduced

9 APPLICATIONS IN DATA NETWORKS

Adaptive Bit Loading can be implemented in Wireless Data Networks which uses FDM
or OFDM to transmit data. In fact, it can be implemented at any node in the network
which sends data over multiple channels. However to implement this the node should
know exact channel it uses to communicate with another node. Also both transmitting
node and receiving node should know how exactly the bits are allocated over multiple
channels. Higher data rates can be acehived because of this technique. However there
will be some necessary overheads. Like transmitting some data to find the channel status.
One of the other application is in Ad-Hoc Networks. A Mobile Ad-Hoc Network is an
autonomous collection of mobile users that communicate over relatively bandwidth
constrained wireless links. Since the nodes are mobile, the network topology may change
rapidly and unpredictably over time. The network is decentralized, where all network
activity including discovering the topology and delivering messages must be executed by
the nodes themselves, i.e., routing functionality will be incorporated into mobile nodes.
Now in this case, the nodes needs to continuously need to establish and dis connect links
to other nodes. Hence, we can determine the channel status during such establishment
and implement Adaptive Bit Loading.
Advantages of using Adaptive Bit Loading in such Networks are:
a. No additional overhead in determining th channel condition as this can be found
out during link establishment
b. Energy conservation is one of the important concerns of such networks

10 COMMERCIAL APPLICATIONS
Commercially Adaptive Bit Loading is implemented in various wireless standards that
use OFDM or FDM to transmit in a wireless data network. It is been used in various
mobile networks.
CONCLUSION
While providing error free data transfer is the responsibility of Data Link and Network
Layer of the OSI Model, these layers have no control over the channel’s behavior. While
we can always use parity checks and error correcting codes, there is always a chance that
the bits maybe corrupted during transmission.
While this problem is not that apparent when the nodes are connected using physical
medium like cables, when we use wireless channel for connection this becomes very
apparent.
Wireless channels have unpredictable behavior and are very dynamic. They can be easily
influenced by external conditions.
What Adaptive Bit Loading does, is get channel information and use this information. It
tries to provide reliability into the channel.
Adaptive Bit Loading is very useful in wireless because wireless channels are time
varying and very dynamic. By adapting the transmission system to match the current
channel parameter, it tries to negate this dynamism.
There are some additional costs of implementation of this system but compared to
reliability that it can provide, it is an acceptable overhead.
Adaptive Bit Loading provides Low BER and High Data Rates while at the same time
reduces the energy requirement. Hence, it is very useful for wireless and mobile
networks, especially Ad- Hoc ones, where the channel is very dynamic and varying.
REFERENCES
[1]“A Practical Discrete Multitone Transceiver Loading Algorithm for Data
Transmission Over Spectrally Shaped Channels,"
PS Chow, JM Cioffi, JAC Bingham - IEEE Transactions on Communications, 1995
[2]“Discrete Bit Loading for Multicarrier Modulation Systems,"
J Campello - Information Theory IEEE Proceedings, 1998
“Third-Generation Systems and Intellignet Wireless Networking - Smart Antennas and
Adaptive Modulation”
JS Blogh, L Hanzo
“OFDM Wireless LANs: A Theoretical and Practical Guide”
J Heiskala, J Terry