Convolution final slides
-
Upload
ramyasreessj -
Category
Education
-
view
2.891 -
download
0
Transcript of Convolution final slides
![Page 1: Convolution final slides](https://reader035.fdocuments.in/reader035/viewer/2022062220/555ef73dd8b42a487d8b5029/html5/thumbnails/1.jpg)
Presented By :D.Ramu (09605A0401)M.Sirisha (08601A0471)
P.Ramya Sree (08601A0497)
EFFICIENT FPGA IMPLEMENTATION OF
CONVOLUTION
Under the Esteemed Guidance of
Mr. S.Nagireddy
1
20/06/2011
![Page 2: Convolution final slides](https://reader035.fdocuments.in/reader035/viewer/2022062220/555ef73dd8b42a487d8b5029/html5/thumbnails/2.jpg)
INTRODUCTION
Reduces convolution processing time using hardware
computing
Implements discrete linear convolution of two finite
length sequences (N x N)
2
20/06/2011
![Page 3: Convolution final slides](https://reader035.fdocuments.in/reader035/viewer/2022062220/555ef73dd8b42a487d8b5029/html5/thumbnails/3.jpg)
EXISTING SYSTEM
Convolution is implemented using DSP processor Chips
Consumes more PowerRequires more Chip Area Low Speed
3
20/06/2011
![Page 4: Convolution final slides](https://reader035.fdocuments.in/reader035/viewer/2022062220/555ef73dd8b42a487d8b5029/html5/thumbnails/4.jpg)
PROPOSED SYSTEM
Convolution is implemented using VLSI Architechture
Consumes less PowerRequires less Chip Area High SpeedExtended for Signed and Unsigned Nos.Reduces Processing time
4
20/06/2011
![Page 5: Convolution final slides](https://reader035.fdocuments.in/reader035/viewer/2022062220/555ef73dd8b42a487d8b5029/html5/thumbnails/5.jpg)
PURPOSE
Proves the feasibility of an application specific integrated circuit (ASIC)
Digital images can be modified using Point wise operations
Image processing operations
Provides great significance in discrete signal processing
5
20/06/2011
![Page 6: Convolution final slides](https://reader035.fdocuments.in/reader035/viewer/2022062220/555ef73dd8b42a487d8b5029/html5/thumbnails/6.jpg)
PROJECT OVERVIEW
Identify the Architecture From the literature survey
Model the Architecture into RTL [Register Transfer Level]modeling
Verify the functionality of Modeled architecture in MODELSIM®
Synthesis the verified design in Xilinx ISE
Generation of Bit map file for Dump into Spartan 3E FPGA
Program the Bit map file into FPGA.
Post simulation in ChipScope pro.
6
20/06/2011
![Page 7: Convolution final slides](https://reader035.fdocuments.in/reader035/viewer/2022062220/555ef73dd8b42a487d8b5029/html5/thumbnails/7.jpg)
ADDITIVE PORTION OF LINEAR PROPERTY
7
![Page 8: Convolution final slides](https://reader035.fdocuments.in/reader035/viewer/2022062220/555ef73dd8b42a487d8b5029/html5/thumbnails/8.jpg)
LINEAR CONVOLUTIONImpulse response
Shifted version of the input signal
Scaling aspect of linearity of the system
Additive aspect of linearity of the system.
= y(t) 8
![Page 9: Convolution final slides](https://reader035.fdocuments.in/reader035/viewer/2022062220/555ef73dd8b42a487d8b5029/html5/thumbnails/9.jpg)
BLOCK DIAGRAM
9
20/06/2011
![Page 10: Convolution final slides](https://reader035.fdocuments.in/reader035/viewer/2022062220/555ef73dd8b42a487d8b5029/html5/thumbnails/10.jpg)
Multiplexer Multiplexer referred to as “multiplexor” or “mux”
MUX contains
2n Inputs lines
n Select lines
1 Output line
Working of MUX:
Selects any one of the inputs from 2n inputs
Directs to the output depending on n-select lines.
10
20/06/2011
![Page 11: Convolution final slides](https://reader035.fdocuments.in/reader035/viewer/2022062220/555ef73dd8b42a487d8b5029/html5/thumbnails/11.jpg)
Each input is 4-bit signed form
Each output is also a 4-bit signed form
Convolution design uses two 4*1 Multiplexers
Multiplexer 4*1
11
20/06/2011
![Page 12: Convolution final slides](https://reader035.fdocuments.in/reader035/viewer/2022062220/555ef73dd8b42a487d8b5029/html5/thumbnails/12.jpg)
Multiplexers 4*1
12
20/06/2011
![Page 13: Convolution final slides](https://reader035.fdocuments.in/reader035/viewer/2022062220/555ef73dd8b42a487d8b5029/html5/thumbnails/13.jpg)
Serial in parallel out block(SIPO)
SIPO converts serial input into parallel output Each serial input will be in a 4-bit signed form
Working of SIPO
Takes SIN(0 to 3) as a input Produces four parallel outputs Q0,Q1,Q2,Q3 Each parallel output will be in a 4-bit signed form
13
20/06/2011
![Page 14: Convolution final slides](https://reader035.fdocuments.in/reader035/viewer/2022062220/555ef73dd8b42a487d8b5029/html5/thumbnails/14.jpg)
Serial In Parallel Out Registers
14
20/06/2011
![Page 15: Convolution final slides](https://reader035.fdocuments.in/reader035/viewer/2022062220/555ef73dd8b42a487d8b5029/html5/thumbnails/15.jpg)
Binary multiplier
It is a 4-bit multiplier
Takes two four inputs
Each input is 4-bit signed form and gives an 8-bit output
Special Characteristic of Binary multiplier:
Internal carry will not be forwarded to next stage
So,number of outputs obtained here is seven only
15
20/06/2011
![Page 16: Convolution final slides](https://reader035.fdocuments.in/reader035/viewer/2022062220/555ef73dd8b42a487d8b5029/html5/thumbnails/16.jpg)
Binary Multiplier
16
20/06/2011
![Page 17: Convolution final slides](https://reader035.fdocuments.in/reader035/viewer/2022062220/555ef73dd8b42a487d8b5029/html5/thumbnails/17.jpg)
Multiplexer 8*1
17
20/06/2011
![Page 18: Convolution final slides](https://reader035.fdocuments.in/reader035/viewer/2022062220/555ef73dd8b42a487d8b5029/html5/thumbnails/18.jpg)
Register A Register is a group of flip-flops
It holds information within a digital system
The logic units get access to the Info during the
computing process
It may have combinational gates that perform certain
data-processing tasks.
18
20/06/2011
![Page 19: Convolution final slides](https://reader035.fdocuments.in/reader035/viewer/2022062220/555ef73dd8b42a487d8b5029/html5/thumbnails/19.jpg)
Register
19
20/06/2011
![Page 20: Convolution final slides](https://reader035.fdocuments.in/reader035/viewer/2022062220/555ef73dd8b42a487d8b5029/html5/thumbnails/20.jpg)
FPGA DESIGN FLOW
20
![Page 21: Convolution final slides](https://reader035.fdocuments.in/reader035/viewer/2022062220/555ef73dd8b42a487d8b5029/html5/thumbnails/21.jpg)
4-INPUT LUT BASED IMPLEMENTATION OF LOGIC BLOCK
21
![Page 22: Convolution final slides](https://reader035.fdocuments.in/reader035/viewer/2022062220/555ef73dd8b42a487d8b5029/html5/thumbnails/22.jpg)
ADVANTAGES
The advantages of convolution by proposed architecture
has following advantages:
Reduce area
Reduce Power
More speed
No data loss
22
20/06/2011
![Page 23: Convolution final slides](https://reader035.fdocuments.in/reader035/viewer/2022062220/555ef73dd8b42a487d8b5029/html5/thumbnails/23.jpg)
Simulation ResultsConvolution Top Level
23
20/06/2011
![Page 24: Convolution final slides](https://reader035.fdocuments.in/reader035/viewer/2022062220/555ef73dd8b42a487d8b5029/html5/thumbnails/24.jpg)
Multiplexers
24
20/06/2011
![Page 25: Convolution final slides](https://reader035.fdocuments.in/reader035/viewer/2022062220/555ef73dd8b42a487d8b5029/html5/thumbnails/25.jpg)
Serial In Parallel Out Registers
25
20/06/2011
![Page 26: Convolution final slides](https://reader035.fdocuments.in/reader035/viewer/2022062220/555ef73dd8b42a487d8b5029/html5/thumbnails/26.jpg)
Binary Multiplier
26
20/06/2011
![Page 27: Convolution final slides](https://reader035.fdocuments.in/reader035/viewer/2022062220/555ef73dd8b42a487d8b5029/html5/thumbnails/27.jpg)
Multiplexer 8*1
27
20/06/2011
![Page 28: Convolution final slides](https://reader035.fdocuments.in/reader035/viewer/2022062220/555ef73dd8b42a487d8b5029/html5/thumbnails/28.jpg)
Register
28
20/06/2011
![Page 29: Convolution final slides](https://reader035.fdocuments.in/reader035/viewer/2022062220/555ef73dd8b42a487d8b5029/html5/thumbnails/29.jpg)
Synthesis ResultsRTL Schematic View
29
20/06/2011
![Page 30: Convolution final slides](https://reader035.fdocuments.in/reader035/viewer/2022062220/555ef73dd8b42a487d8b5029/html5/thumbnails/30.jpg)
RTL Internal View
30
20/06/2011
![Page 31: Convolution final slides](https://reader035.fdocuments.in/reader035/viewer/2022062220/555ef73dd8b42a487d8b5029/html5/thumbnails/31.jpg)
APPLICATIONS Digital image processing(Frequency Filtering)
Real-time signal processing like:
Audio signal processing
Video / Image processing
Large-capacity data processing
In Linear Acoustics
In statistics
In Probability theory
In Optics(The “Blur” is described by Optics) 31
20/06/2011
![Page 32: Convolution final slides](https://reader035.fdocuments.in/reader035/viewer/2022062220/555ef73dd8b42a487d8b5029/html5/thumbnails/32.jpg)
CONCLUSION Optimized implementation of Discrete Linear
Convolution. Uses the mean squared error measurement and
objective measures of enhancement to achieve a more effective signal processing model and accuracy
The proposed circuit uses only 5mw and saves almost 35% area and it takes 20ns to complete. This shows improvement of more than 50% less power.
32
20/06/2011
![Page 33: Convolution final slides](https://reader035.fdocuments.in/reader035/viewer/2022062220/555ef73dd8b42a487d8b5029/html5/thumbnails/33.jpg)
FUTURE SCOPE Extracting a periodic signal from noise. Software Applications: GUI Module. Echo Detection in Linear acoustics. Speech Analysis and pitch. In time-resolved Fluorescense Spectroscopy In Radiotherapy treatment planning systems , most
part of all modern codes can use convolution.
33
20/06/2011
![Page 34: Convolution final slides](https://reader035.fdocuments.in/reader035/viewer/2022062220/555ef73dd8b42a487d8b5029/html5/thumbnails/34.jpg)
Thank You
34
20/06/2011
![Page 35: Convolution final slides](https://reader035.fdocuments.in/reader035/viewer/2022062220/555ef73dd8b42a487d8b5029/html5/thumbnails/35.jpg)
QUERIES??
QUERIES??
35