Вы находитесь на странице: 1из 33

Dayananda Sagar College of Engineering

Center for Post Graduation Studies


Department of Electronics and Communications Engineering
Project Seminar MTech Project Phase 3
Project Title : EVALUATION OF POLYPHASE FFT
ARCHITECTURE FOR PULSE DETECTION AND MEASUREMENT
Student Name: JEEVITHA T (1DS12LEC06)
Internal Project Guide: External Project Guide:
Prof. Kiran Gupta Hemanth Vasant Paranjape
ECE dept,DSCE Scientist E, DARE,DRDO
Title of the Project
Problem statement
Time of Arrival(TOA) PDW parameter that assigns a time tag to the leading edge of
a received pulse at the receiver input.
For a digital EW receiver, the spectrum estimator (normally an FFT) is usually
the limiting factor in update rate due to its computational complexity.
update rate 1/ frequency resolution.
Consider a monobit receiver:
256 samples at 2.56GHz TOA resolution 100ns.
with 50% overlap at the same Fs TOA resolution 50ns
increasing the number of FFTs.

Title of the Project
Phase 1 Review
Phase 1 Review





Receiver considered is a Multi-bit Electronic Warfare Receiver

Receiver consists of a decimation filter and FFT block along with an encoder that
outputs the PDW(Pulse Descriptor Words).

The shaded block is realized as a polyphase DFT which uses decimation in frequency
domain.




Title of the Project
objective
the proposed method to increase the TOA resolution is to use decimation in the
frequency domain.

implement the decimation in frequency domain through a channelised polyphase
filtering method.

incorporate the channelised polyphase filtering method in the receiver model.

Title of the Project
Frequency Channelization
Filter bank:

Polyphase Structure for FFT filter bank

USN: 1DS12LEC06

6
Title of the Project
Decimation filter design process:
STAGE 1: FFT filter bank

rectangular window in time domain sinc function in frequency domain

filter shape is not desirable

sidelobes are very high

Weighing function can be used in time domain

the detailed response of the single filter and the filter bank is as shown

USN: 1DS12LEC06

8
Title of the Project
Continued..
STAGE 2 : Decimated FFT filter Bank
ALGORITHM: Decimation in Frequency Domain:

Step 1:




Step 2:

Title of the Project
Continued
Step 3:

Title of the Project
Step 4:





Step 5:




Title of the Project
Step 6:















Results:
USN: 1DS12LEC06

13
Title of the Project
Continued..
Stage 3: Windowed Decimated FFT filter Bank

To widen the individual filters and also suppress the sidelobes a window is applied
to the input data.

The corresponding effect in the frequency domain is the wide bandwidth of each
individual filter shape in the filter bank.

ALGORITHM:

Step 1:



Step 2:


Title of the Project
Continued..
Step 3:






Step 4: 32 point FFT is performed on these y values , 16 individual filters are
generated.







Window function generated using an FDATOOL
Design Parameters:

N=256; M=8; 1= 0.01; 2=0.001; fs=3000MHz;
Fp =fs/(N/M) ; Fs = 2*Fp;
USN: 1DS12LEC06

16
Title of the Project
Continued..
Step 1:





Step 2:



Park Mc-Clellans Criteria:
Title of the Project
Continued..
Results of the stage 3:
USN: 1DS12LEC06

19
Title of the Project
Advantages of Polyphase Filtering
First, by parallelizing the filter through polyphase decomposition, the sampling rate of
each individual filter is reduced by a factor of 1/D, where D is the number of filters.


A second significant advantage to using the channelized polyphase filtering
method is an increase in time resolution, which improves the TOA and PW calculations
in an EW receiver.
Comparison:
USN: 1DS12LEC06

21
Title of the Project
MATLAB simulation
Generation of the input signals:



FILE NAME DESCRIPTION
Known_pattern.m Accepts the number of pulses transfers the control
test_case_generator.
Outputs the details of the input info.txt,info1.txt

Test_case_pattern.m Provides the user a platform to enter the details of the input
pattern.
Transfers the control to cw_generator.m depending on the
user entries.
Cw_generator.m

Used to generate cw pulses.
Info.txt, info1.txt Contains the details of the input signal.
USN: 1DS12LEC06 23
start
Enter the number of
pulses
1
Test_case_generator(1
st

pattern details)
2
T_c_g(1
st

pattern)
4
T_c_g(2
nd

pattern)
3
T_c_g(3
rd

pattern)
T_c_g(2
nd

pattern)
T_c_g(1
st

pattern)
T_c_g T_c_g T_c_g
T_c_g
Make a correct
entry
yes
no
no
no
no
yes
yes
yes
USN: 1DS12LEC06 24
Test_case_generator
Title of the Project
Testing of the input signal
filename description
Complete_model.m Imports the samples. Groups the samples
into frames. Performs decimation in
frequency domain.
Sidelobegaurd.m Peak detection algorithm implementation.
Track_detection.m Tracks the detected peaks.
Parameter_measurement.m Measurement of the input signal parameters
is done here
Pdw_info Results obtained from tha parameter
measurement are stored here.
Title of the Project
COMPLETE_MODEL.M ALGORITHM
1. The input data/ samples are stored in samples.txt

2. Grouping of the data begins :
P = length(A);
P = P - mod(P,256);
L = (P/N);
where A is a variable which contains the imported data (samples.txt)

3. Design an equiripple low pass FIR filter with the following parameters:
Fs=1350MHz
Fp=21.09375MHz
Fs=42.1875MHz
Wp=1/64
Ws=1/32
Fdatool is used to generate the filter co-efficients.
H(n) co-efficients.

Title of the Project
Samples(14080)
L 2
1
1 2 3256
P= P- mod(P,256)
Mod(x,y)= x-n.*y
n = floor(x./y)
Consider an example:
Title of the Project
Continued.
4. Decimation in the frequency domain is applied on the incoming data:
Let x(n) be the input, h(n) be the filter co-efficients :





5. Calculate the 32 point FFT of each frame.

6. Compute the magnitude of complex data samples. Determine the first peak with the
highest magnitude. Done by using a MAX command.
fft_frame1 = (fft(yn,32))
[magn1 pos1]=max(fft_frame1(1:16))

7. Suppose X(k) is the peak sample, determine X(k-1)(magn5) and X(k+1)(magn6), only if the
pos1!=1.

8. Also calculate the inphase(I1) and quadrature phase(Q1) data of the fft_frame1 using the
REAL and IMAG commands respectively.


Title of the Project
Sidelobegaurd.m(peak detection algorithm)
1. The algorithm determine 4 peaks with highest magnitude from every frame, taking
into consideration side lobe rejection.
2. The parameters that are passed to sidelobegaurd are magnitude of the elements of
fft_frame1 and pos1.


3. Concept of bandpass sampling:
Used to sample a continuous bandpass signal centred about some frequency
other than zero.
The sampling freq in this case is calculated as:








Title of the Project
4. Check whether the pos1 lies within the detected dft bins.
5. If yes calculate the first differentialof the magnitude of the elements of the fft_frame
if(D(i-1)> 0 && D(i) < 0)
d(i-1) = c(i-1);
else
d(i) = 0;
Detect a peak and eliminate its sidelobes simultaneously.
6. In case there are two or more pulses
s8 = 0.0003; slg8 = c(pos1) * s8;
s7 = 0.0005;
s6 = 0.001;
s5 = 0.0501;
s4 = 0.1585;
s3 = 0.4585;
s2 = 0.631;
s1 =1;



Title of the Project
For ex: let fc=1000MHz and B=500MHz
Then 1500MHz fs 1250MHz. Fs=1350MHz is used.







Now converting these translated frequencies into DFT bins we have:










fc-B/2 =750MHz fc=1000MHz fc+B/2=1250MHz
fs- (fc-B/2) =600MHz fs-fc=350MHz fs-(fc+B/2)=100MHz
Title of the Project
Track detection:
1. Set four filter indices for four input pulses.


Title of the Project
Continued
THANK YOU

Вам также может понравиться