Equalize using Viterbi algorithm
The MLSE Equalizer block uses the Viterbi algorithm to equalize a linearly modulated signal through a dispersive channel. The block processes input frames and outputs the maximum likelihood sequence estimate (MLSE) of the signal, using an estimate of the channel modeled as a finite input response (FIR) filter.
This block supports
The channel estimate takes the form of a column vector containing the coefficients of an FIR filter in descending order of powers. The length of this vector is the channel memory, which must be a multiple of the block's Samples per input symbol parameter.
To specify the channel estimate vector, use one of these methods:
Set Specify channel via to
Dialog and enter the vector in the
Channel coefficients field.
Set Specify channel via to
port and the block displays an additional input port,
Ch, which accepts a column vector input
The Signal constellation parameter specifies the constellation for the modulated signal, as determined by the modulator in your model. Signal constellation is a vector of complex numbers, where the kth complex number in the vector is the constellation point to which the modulator maps the integer k-1.
The sequence of constellation points must be consistent between the modulator in your model and the Signal constellation parameter in this block.
For example, to specify the constellation given by the mapping
set Constellation points to
[1+i, -1+i, -1-i,
1-i]. Note that the sequence of numbers in the vector indicates how
the modulator maps integers to the set of constellation points. The labeled
constellation is shown below.
If your data is accompanied by a preamble (prefix) or postamble (suffix), then configure the block accordingly:
If you select Input contains preamble, then the Expected preamble parameter specifies the preamble that you expect to precede the data in the input signal.
If you check the Input contains postamble, then the Expected postamble parameter specifies the postamble that you expect to follow the data in the input signal.
The Expected preamble or Expected postamble parameter must be a vector of integers between 0 and M-1, where M is the number of constellation points. An integer value of k-1 in the vector corresponds to the kth entry in the Constellation points vector and, consequently, to a modulator input of k-1.
The preamble or postamble must already be included at the beginning or end, respectively, of the input signal to this block. If necessary, you can concatenate vectors in Simulink® software using the Matrix Concatenation block.
To learn how the block uses the preamble and postamble, see "Reset Every Frame" Operation Mode below.
One way that the Viterbi algorithm can transition between successive frames is
Reset every frame mode. You can choose this mode
using the Operation mode parameter.
Reset every frame mode, the block decodes each frame
of data independently, resetting the state metric at the end of each frame. The
traceback decoding always starts at the state with the minimum state metric.
The initialization of state metrics depends on whether you specify a preamble and/or postamble:
If you do not specify a preamble, the decoder initializes the metrics of all states to 0 at the beginning of each frame of data.
If you specify a preamble, the block uses it to initialize the state metrics at the beginning of each frame of data. More specifically, the block decodes the preamble and assigns a metric of 0 to the decoded state. If the preamble does not decode to a unique state -- that is, if the length of the preamble is less than the channel memory -- the decoder assigns a metric of 0 to all states that can be represented by the preamble. Whenever you specify a preamble, the traceback path ends at one of the states represented by the preamble.
If you do not specify a postamble, the traceback path starts at the state with the smallest metric.
If you specify a postamble, the traceback path begins at the state represented by the postamble. If the postamble does not decode to a unique state, the decoder identifies the smallest of all possible decoded states that are represented by the postamble and begins traceback decoding at that state.
Reset every frame mode, the input to the
MLSE Equalizer block must contain at least T symbols, not including an
optional preamble, where T is the Traceback depth
An alternative way that the Viterbi algorithm can transition between successive
frames is called
Continuous with reset option mode. You
can choose this mode using the Operation mode parameter.
Continuous with reset option mode, the block
initializes the metrics of all states to 0 at the beginning of the simulation. At
the end of each frame, the block saves the internal state metric for use in
computing the traceback paths in the next frame.
If you select Enable the reset input port, the block displays
another input port, labeled
Rst. In this case, the block resets
the state metrics whenever the scalar value at the
Rst port is
The MLSE Equalizer block introduces an output delay equal to the
Traceback depth in the
Continuous with reset
option mode, and no delay in the
The method for specifying the channel estimate. If you select
Input port, the block displays a second input
port that receives the channel estimate. If you select
Dialog, you can specify the channel estimate
as a vector of coefficients for an FIR filter in the Channel
Vector containing the coefficients of the FIR filter that the block uses
for the channel estimate. This field is visible only if you set
Specify channel via to
Vector of complex numbers that specifies the constellation for the modulation.
The number of trellis branches (equivalently, the number of symbols) the block uses in the Viterbi algorithm to construct each traceback path.
The operation mode of the Viterbi decoder. Choices are
Continuous with reset option and
Reset every frame.
When checked, you can set the preamble in the Expected
preamble field. This option appears only if you set
Operation mode to
Vector of integers between 0 and M-1 representing the preamble, where M is
the size of the constellation. This field is visible and active only if you
set Operation mode to
frame and then select Input contains
When checked, you can set the postamble in the Expected
postamble field. This option appears only if you set
Operation mode to
Vector of integers between 0 and M-1 representing the postamble, where M
is the size of the constellation. This field is visible and active only if
you set Operation mode to
frame and then select Input contains
The number of input samples for each constellation point.
When you check this box, the block has a second input port labeled
Rst. Providing a nonzero input value to this port
causes the block to set its internal memory to the initial state before
processing the input data. This option appears only if you set
Operation mode to
This example shows how to equalize the effects of a Multipath Rayleigh Fading Channel block. Maximum Likelihood Sequence Estimation (MLSE) estimates the data the model transmits through a time varying dispersive channel with the least possible number of errors. This model inputs the dynamically evolving channel coefficients of a two-path channel to the MLSE Equalizer block. The model shows the MLSE block being used in a typical multipath wireless Rayleigh channel. It applies the same channel estimate to 50 samples in the frame that is processed by the MLSE Equalizer. This is similar to a practical system, where the training sequence is transmitted in regular intervals and a channel estimate is used until the next training symbol is transmitted.
To open the example, type
doc_mlse_dynamic_coeffs at the MATLAB® command line.
The sample time of the Bernoulli Binary Generator block is set to 5e-6, which corresponds to a bit rate of 200 kbps, and a QPSK symbol rate of 100 ksym/sec.
The Multipath Rayleigh Fading Channel block has a Maximum Doppler shift of 30 Hz, which is a realistic physical value. The Delay vector of the MRFC block is [0 1e-5], which corresponds to two consecutive sample times of the input QPSK symbol data. This reflects the simplest delay vector for a two-path channel. The Average path gain vector is set arbitrarily to [0 -10]. The gain vector is normalized to 0 dB, so that the average power input to the AWGN block is 1 W.
The MLSE Equalizer block has the Traceback depth set to 10 and may be varied to study its effect on Bit Error rate (BER).
The QPSK Demodulator accepts an N-by-1 input frame and generates a 2N-by-1 output frame. This, along with the traceback depth of 10 results in a delay of 20 bits. The model performs frame-based processing with 100 samples per frame. Thus, there is a delay of 100 bits inherent in the model. The combined receive delay of 120 is set in the Receive delay parameter of the Error Rate Calculation block, aligning the samples.
The sample time of the Bernoulli Binary Generator block is set to 5e-6, which corresponds to a bit rate of 200 kbps, and a QPSK symbol rate of 100 ksym/sec. Multipath Rayleigh Fading Channel (MRFC) block: The MRFC block has a max Doppler shift of 30 Hz, which is a realistic physical value. The Delay vector of the MRFC block is [0 1e-5], which corresponds to two consecutive sample times of the input QPSK symbol data. This reflects the simplest delay vector for a two-path channel. The Gain vector of the MRFC block is set arbitrarily to [0 -10]. The gain vector is normalized to 0 dB, so that the average power input to the AWGN block is 1 W.
 Proakis, John G., Digital Communications, Fourth edition, New York, McGraw-Hill, 2001.
 Steele, Raymond, Ed., Mobile Radio Communications, Chichester, England, Wiley, 1996.