Vous êtes sur la page 1sur 6

ANALYSIS OF DECISION FEED BACK EQUALIZER FOR WIRELESS CHANNEL Miss A.PRASANNA VATHANI, Miss K.GIRIJAA. Email id : prasannajuly@yahoo.co.

in Email id: girijaaece@gmail.com Cell:9894614565 Cell :9976970340 Final Year ECE Department of Electronics & Communication Engineering Paavai College of Engineering, Pachal, Namakkal.
ABSTRACT Wireless communication revolutionarily changes the way of communication and provides possibility of communication to anyone, from anywhere, at anytime. While wireless communication technology significantly changes the way people live and work, it adds tremendous challenges to communication engineering design. It is obvious that when a radio signal is transmitted through the air, the signal quality will largely depend on many variables in communication environments that are beyond our control. Unlike a wired communication system, a wireless communication system often has problems with flat or frequency selective fading, and time dispersion. In order to maintain a quality wireless communication signal, the radio channel must be estimated and properly compensated, and one effective means of such channel estimation and compensation is called channel equalization. Due to the time-varying nature of a radio channel, channel equalization is often designed to be adaptive or time-varying, in order to track dynamic channel variation. In this paper, an adaptive fractionally spaced decision feedback equalizer (DFE) for wireless channel is analyzed. An analysis on the optimum performance of the DFE receiver shows the advantages of this system over other methods. A MATLAB simulation of this system is presented to study the convergence properties and implementation considerations of the DFE receiver. INTRODUCTION If we transmit data through the channel with a limited bandwidth, each received signal may suffer from the interference from neighboring signals and it results in inter-symbol interference (ISI). Since the ISI causes high error rates at the receiver, an equalizer is exploited to compensate for the ISI.

If we deal with a band-limited channel that causes ISI, it is convenient to develop an equivalent discrete-time model for the analog (continuous-time) system. The system model shown in Fig. 1 is the cascade of a transmitting filter, a channel, and a matched filter. This is also expressed as an equivalent filter with tap coefficients {fk}, as shown in Fig.2
{I k }

Tx filter

Channel

Matched Filter

T
k

{v k }

Fig.1 Band-limited channel model

{I k }

FIR
{ fk }

{vk }

vk = f n I k n + k
n =0

. Fig.2 Equivalent discrete-time model

k is a white gaussian noise Note: sequence 2 Achieving good performances of equalizer in the case of strongly dispersive or rapidly timevarying channels, is very difficult for that we prefers the Decision Feedback Equalizers (DFEs) [2], [3], [4]. 3 4 A decision feedback equalizer (DFE) is a non-linear equalizer that uses previous detector decisions to eliminate ISI on the current received symbol . In principle, if a number of source symbols are correctly detected and the channel impulse response is known, then the ISI from these symbols can be reconstructed and cancelled from the current received symbol The decision-feedback equalizer (DFE) depicted in Fig. 3 consists of two filters: a feed forward filter and a feedback filter both of which can be linear filters. The non-linearity of the DFE comes with the inclusion of the decision device into the loop. The feed-forward filter can be of any type and generally a minimum meansquare error (MMSE) filter is implemented. The input to the feed forward section is the received

{ }

{vk }

Feedforward filter

{I k }
+

Symbol- bysymbol detector

~ {I k }

Feedback filter

Fig.3 Decision feedback equalizer The output of DFE can be expressed as:
K2

Ik =

j = K1

c j vk j + c j I k j ,
j =1

Where

information symbol, represent the tap coefficients of the filter, and

I k is an estimate of the k-th {c j }

~ ~ {I k 1 , ... , I k K2 }

are previously detected

symbols. The equalizer is assumed to have

( K1 + 1) taps in its feedforward section and K 2 taps in its feedback section.

{ k } . The feedback filter has signal sequence the sequence of decisions on previously detected symbols as its input. Functionally, the feedback filter is used to remove the part of the intersymbol interference from the present estimate caused by previously detected symbols. The basic idea of a DFE that the ISI contributed by the previously detected symbols can be cancelled out exactly from the output of the forward filter by subtracting past symbol values with appropriate weighting. The forward and feed-back tap weights can be simultaneously adjusted to fulfill a criterion such as minimizing the mean-square error.
The DFE has a clear advantage over linear equalizers since it has lower noise enhancement while it processes the received signal. The feedback loop utilizes the detector decisions which are subtracted from the current received symbol and therefore the feed-back signal does not contain any noise components.

As described in the case of the linear adaptive equalizer, the coefficients of the feed forward filter and the feedback filter in a decisionfeedback equalizer may be adjusted recursively by using the LMS (Least-Mean-square) algorithm, which minimizes the MSE (MeanSquare-Error). Based on the assumption that previously detected symbols in the feedback filter are correct, the MSE is defined as:

J ( K ) = E ek = E I k I k = E I k
To find the coefficients that minimize , we set the partial derivative of k= c* respect to k to zero for

^ 2

j = K1

c v

j k j

c j I k j
j =1

K2

J (k )

J (k ) with

i.e.

J (k ) =0 * ck .

K1 ,..., K 2 ,

The LMS algorithm for recursively adjusting the tap weight coefficients yields

c k +1 = c k + ekVk* , Vk = (v k + K1 , ... , v k , I k 1 , ... , I k K 2 )


~ ~

ek = I k ( R2 I k )
The equalizer adaptation based on the CMA does not require training sequences and carrier phase recovery. The algorithm tends to converge slowly. However, it offers the advantage of decoupling the ISI equalization and carrier phase recovery problems from each other. SIMPLE SIMULATION OF DFE: Fig.5 shows a simple simulation model of a decision feedback equalizer, where F(z) is the equivalent filter characteristic
AWGN { k } Random Source (QPSK) {Ik} ISI Channel F(z) Equalizer Decision

where

is the step size.

CMA(Constant Modulus Algorithm) : One approach to removing inter symbol interference in a communications channel is to employ adaptive blind equalization to reduce the symbol error rate (SER). The most popular class of algorithms used for blind equalization are those that minimize the Godard (or constant modulus) criteria. CMA is a steepest-descent algorithm that is widely used in practice when a training sequence is not available. We can implement a blind DFE by using the CMA. In an M-ary PSK system, CMA minimizes a non-convex cost function of the form 2 J (k ) = E[( vk R2 ) 2 ] , Where

Fig.5 Simulation model of DFE The basic parameters are as follow. tap size(N) : 16 step size() : 0.001 SNR : 35dB Iteration :5000 symbols The channel can be implemented by using a filter in the mat lab program. CHOOSING AN ADAPTIVE ALGORITHM:

R2 is a positive real constant defined


E[ I k ] E[ I k ]
2 4

R2 =
as

The CMA algorithm is designed to give a penalty to the deviations of the blind equalizer output from a constant modulus. The constant

R2

is chosen in such a way that the gradient of

the cost function is zero when a perfect equalization is attained. The tap weight coefficients of the equalizer are adapted in accordance with the stochastic gradient algorithm. ck +1 = ck + ekVk* ,

J (k )

~ ~ Vk = (vk + K1 , ... , vk , I k 1, ... , I k K 2 )

Although the best choice of adaptive algorithm might depend on your individual situation, here are some generalizations that might influence your choice: The LMS algorithm executes quickly but converges slowly, and its complexity grows linearly with the number of weights. The RLS algorithm converges quickly, but its complexity grows with the square of the number of weights, roughly speaking. This algorithm can also be unstable when the number of weights is large. The various types of signed LMS algorithms simplify hardware implementation. The normalized LMS and variable-step-size LMS algorithms are more robust to variability of the input signal's statistics (such as power). The Constant modulus algorithm is useful when no training signal is available, and works best for constant-modulus modulations such as PSK.

However, if CMA has no additional side information, it can introduce phase ambiguity. For example, CMA might find weights that produce a perfect QPSK constellation but might introduce a phase rotation of 90, 180, or 270 degrees. Alternatively, differential modulation can be used to avoid phase ambiguity. ERROR PROPAGATION IN DFE 1 Error propagation is a major cause of concern in DFE. The design of a DFE is based on the assumption that the detector decisions are correct. However, due to the presence of noise the decisions made cannot always be correct. The presence of a feed-back loop makes the condition serious as an error in the current decision can cause a burst of errors in future symbol decisions. However, if there are M consecutive correct decisions, where M is the order of the feedback filter, the feed-back part will give appropriate contributions again, and from here on the DFE will operate correctly until the next error. Assuming a binary PAM signal, it can be shown that this happens after K symbols, on average, where

sequence phase helps to synchronize the detector decisions avoiding unreliable adaptation. RESULT AND DISCUSSION: The decision feed back equalizer is simulated in MATLAB for analysis. In simulation we are taken 4 level QAM modulated signal of length 5000 symbol as input signal. The scatter plot of input signal is given in fig.1.this input signal Constellations plot is given in fig.2
Scatter plot 1 0.8 0.6 0.4 Quadrature 0.2 0 -0.2 -0.4 -0.6 -0.8 -1 -1 -0.5 0 In-Phase 0.5 1

This gives the average error probability as

Fig.6 QAM signal scatter plot for input symbol


Input signal constellation 2 1.5 1

Im{s(n)}

Pe,0 is the error probability with no error propagation. It is seen that the DFE error probability grows exponentially with the DFE filter length M. Hence, the feed-back filter length should be kept small in order to guarantee good performance. The error propagation can be kept in control by designing the DFE model with sufficient performance margin to account for the degradation due to error propagation In order to resolve the error propagation phenomenon, in practice a typical operation of adaptive DFE detection involves mode switching between a decision-directed blind transmission phase and a training phase. The training

0.5 0 -0.5 -1 -1.5 -2 -2

-1

0 Re{s(n)}

Fig.7 Input signal Constellations A microwave channel is considered for simulation. it is modeled as a FIR filter with 16

Im{x[n]}

length of impulse response. The impulse response of the channel is given in fig.8.the frequency response of the channel is analyzed using filter analysis tool of MATLAB .the magnitude and phase response of channel is given in.fig.9

Received signal x[n] 40 30 20 10 0 -10 -20 -30 -40 -40

-20

0 Re{x[n]}

20

40

Fig.10 Received signal via channel with complex noise is added for 35 db SNR
1.5 1

0.5

Fig.8 channel impulse response of 16 sample


Magnitude (dB) and Phase Responses 10 5 Magnitude (dB) 0 -5 -10 -15 -20 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 Normalized Frequency ( rad/sample) 0.8 11.0274 7.5107 Phase (radians) 3.9941 0.4774 -3.0393 -6.556

-0.5

-1

-1.5 -2

-1.5

-1

-0.5

0.5

1.5

Fig.11 the complex noise added with signal


output signal constellation 2

-10.0727

1.5 1 0.5

Fig.9 Channel frequency response

A complex noise is generated and added to the input signal after passing through the channel filter such that the SNR to the input of equalizer is 35 db. the fig.10 shows the received signal with the added noise .it shows all the symbol are sitting in the same place the decision device cant detect the symbol without the equalizer.fig,11 shows the complex noise added with the source signal

Im{s(n)}

0 -0.5 -1 -1.5 -2 -2

-1

0 Re{s(n)}

Fig.12 output signal constellation after equalization with presence of strong complex Noise

Fig.13 Semi log plot of MSE values

Fig.12 shows the signal constellation after equalization in presence of strong complex noise. this shows the symbols can be decoded able by decision device after equalization.fig.13 shows the MSE values of the equalizer for the 5000 input symbols CONCLUSION: In this work a MATLAB simulation of decision feed back equalizer for a wireless microwave channel is presented. Here we added a complex noise for analysis and we modeled the channels a FIR filter of order 16 .in the future work the convergence of this equalizer is going to be studied .the analysis of equalizer for various channel data is going to be analyzed. The real time implementation of this on hardware platform is going to be done. REFERENCE:
[1] S. Cherif, S. Marcos, M. Jaidane,Analysis of Blind Decision Feedback Equalizer Convergence: Interest of a Soft Decision, International Journal of Signal Processing 4;3 Summer 2008 [2] O. Macchi and A. Hachicha, Self-adaptive equalization based on a prediction principle, in IEEE Global Telecommunication Conference, Dec. 1986, pp. 16411645. [3] E. Bai and Z. Ding, Blind decision feedback equalization of timevarying channels with DPSK inputs, IEEE Transactions on Signal Processing, vol. 49, no. 7, pp. 15331542, July 2001. [4] A. Rontogiannis and K. Berberidis, Efficient decision feedback equalization for sparse wireless channels, Efficient Decision Feedback Equalization for Sparse Wireless Channels, vol. 2, no. 3, pp. 570581, May 2003. [5] R. Kennedy, B. Anderson, and R. Bitmead, Blind adaptation of decision feedback equalizers : Gross convergence properties, International Journal of Adaptive Control and Signal Processing, vol. 7, pp. 497523, 1993.

Vous aimerez peut-être aussi