Vous êtes sur la page 1sur 2

FPGA Implementation of Least Mean Square FIR

Adaptive Filter Using Verilog HDL

Fahad -Al-Ghazali (F-05-CE-011), Muhammad Azam (F-05-CE-012)


Continuing Engineering, Semester FALL 05
CASE, Islamabad

Abstract- FPGA (Field Programmable Gate Array) method of obtaining estimates of filter
implementation of a simple 8-tap Least Mean
Square FIR adaptive filter is described. weights Wk in real time in an easy way.
Architecture is mapped with Verilog and tested The Widrow-Hopf LMS algorithm for
with modelsim. Such type of filter has different updating the weights from sample to
applications in Image processing , speech/audio,
Biomedical, instrumentation and control etc. sample is given by
Wk+1 = Wk + 2µekUk
Keywords-FPGA, Verilog HDL, Modelsim, FIR , The flow chart of this filter is shown in
LMS,etc.
I. INTRODUCTION Initialze Wk(i) and Uk-i
An adaptive filter has an adaptation
algorithm that is meant to monitor the
environment and vary the filter transfer
function accordingly.
The basic operation now involves two Read Uk and d k
processes:
1. a filtering process, which produces an
output signal in response to a given input
signal.
Filter Uk
2. an adaptation process, which aims to yk= Wk(i)Uk-i
adjust the filter parameters (filter
transfer function) to the (possibly time-
varying) environment. The adaptation is
steered by an error signal that indicates
Compute Error
how well the filter output matches some ek= dk - yk
desired response. A brief diagram of
adaptive filter is given in Pic 1.

Final_out Compute Factor


un
Transverse Filter Wk 2 u ek

Adaptive Weight
-
en
Control
Mechanism Update Coefficient
+ Wk+1= Wk + 2 u ek Uk-i

PIC 1 dn

Pic 2. PIC 2
One of the successful adaptive algorithm
is the LMS algorithm. It is a practical
Depending on the application, filter Adaptive Filter FPGA
parameters, the filter output or the error Implementation
signal may be of interest. The Adaptive Filter is implemented
using Verilog HDL the FIR Algorithm
II. MOTIVATIONS Block and the Tree Adder Block are
Real world signal processing problems used as tasks.
often require adaptive solutions. Coefficient Update Block : Updates
Simulation times, using popular tools weights using a variant of the LMS
such as MATLAB, are sometimes algorithm and scalings.
extraordinary (minutes, hours, days, etc). FIR Algorithm Block : Carries out
So a customized solution helps in weight vector multiplications and
maximizing the speed and minimizing scalings
the area aspect of the system.
III. REFERENCES
III. SYSTEM DESCRIPTION
There are various ways to implement the 1) Emmanuel C. Ifeachor,Barrie W.
LMS FIR filter. For example, the LMS Jervis,” Digital signal
unit and the FIR unit may be separated Processing”,Pearson Education, 2nd e,
and run in parallel. In this case, however, 2003.
it should be noted that there is one time
unit of difference. In other words, the 2) Ali H. Sayed,” Fundamentals Of
FIR filtering module uses the current set Adaptive Filtering”, Wiley
of coefficients, while the LMS unit Interscience,2003.
updates the current coefficients for the 3) Chang Choo, “A Memory Reduction
next processing period. Scheme for Multi-Channel Echo
The hardware design of this project Canceller Implementation,” Proc.
mainly consists of two blocks. The ICASSP, Salt Lake City, Utah, May 7-
10, 2001.
transversal filter, which has been
designed using time shared architecture
The coefficients of this filter are
obtained in the real time from the other
module which performs the role of
adaptive weight control mechanism. And
the transversal filter is just like an 8 tap
time shared FIR filter. The other part
takes inputs in the form of the channel
signal and the erroneous signal which is
the difference between the desired
output and output of the transversal
filter. In the adaptive control part
erroneous signal e(n) is multiplied by a
constant “meu” which has
experimentally found to be equal to
0.004. The result multiplied with the
corresponding delayed channel inputs
updates the 8 coefficients that are to be
used for transversal filter.

Vous aimerez peut-être aussi