Proceedings of the World Congress on Engineering 2009 Vol I

WCE 2009, July 1 - 3, 2009, London, U.K.

Advantages of Matched Filter Detection at

Quadrature Baseband Than at Radio Frequency
Lusungu Ndovi., Member, IAENG

for detection and cross-correlates it with the received signal

Abstract

The

continued

advancement

of

[1]. Intuition behind match filtering is that by taking the

software-defined radio (SDR) technology has been a key

convolution of the received signal with the original

factor in furthering research about the implementation

transmitted signal ( a chirp) you are basically sliding across

of most signal processing algorithms at baseband.

your time reversed h(t) across your received signal doing a

Traditionally, most algorithms have been carried out at

radio frequency (RF).

point wise multiplication and then integrating over the area of

With the coming of SDR, the

that product[2].

processing can be done at baseband frequencies which

II. MATCHED FILTER DETECTION AT RF

are more compatible with the fast developing software

radio technology. This paper looks at matched filter

The analysis was firstly carried out at RF and this involved

detection and investigates the possibility and benefits of

generating a simple chirp signal and carrying out the matched

carrying out the detection process at quadrature

filter detection process at RF.

baseband (QBB). A simple chirp signal is considered for

The results will then be

compared with there corresponding results achieved at QBB

the analysis. The analysis is carried out using MatLab

so as to come up with a conclusion about investigating the

simulations at RF and QBB and the results do show the

results obtained at QBB. Consider the block diagram below

possibility of carrying out the detection process at QBB

which illustrates the analysis process carried out at RF.

with the expected benefits as compared to carrying out

the process at RF.

c h irp s ig n a l

x 10

10

m at c hed filt er output

3
4

Index Terms Quadrature, baseband, Radio frequency,

Matched filter, detection.

Matched filter
detector

0
-2
-4

v o lt s2

v o lts

-1

-2

-6

-3

-8

-1 0
0

100

200

300

400

500

600

700

800

900

-4

1000

200

400

fre q (H z )

600

800

1000

1200

1400

1600

1800

2000

t im e (s am ple point num ber)

matched filter output

input chirp

I. MATCHED FILTER DETECTION

Figure 1: RF chirp signal matched filter detection
Before going down to the analysis results, it is necessary to
get a brief background of matched filter detection so as to

The input chirp signal is fed into a matched filter whose

understand the resulting outputs from the simulations carried

output is a convolution between the received signal and its

out. Matched filtering is a process known to be the optimum

time reversed version. Let the chirp input = x , Matlab uses

linear filter for detecting a known signal in random noise.

an inbuilt command to execute the time reversal of the chirp

The DSP responsible for the processing of signals has a copy

input x

of the signal it should detect. It then time reverses the signal

h = fliplr(x)

(1)

The matched filter then convolves the chirp input x with h.

Manuscript received April 11, 2009
The author is with the Department of Electrical Engineering,
Copperbelt University, School of Technology, Zambia (e-mail:
lusungu.ndovi@gmail.com).
paper number: ICSIE_18

ISBN: 978-988-17012-5-1

The equation below illustrates this.

y=conv(x, h)

(2)

WCE 2009

Proceedings of the World Congress on Engineering 2009 Vol I

WCE 2009, July 1 - 3, 2009, London, U.K.

The matched filter output has its peak value at t=T. The
theoretical analysis for the QBB simulations is given in the
next section and the simulation results which follow in
subsequent sections of the chapter will compare the results

A. RF simulation results
The figures given below show the simulation results obtained
for the direct path analysis at RF.
chirp signal
10

obtained at RF and QBB.

MATHCED FILTER

III.

DETECTION AT QBB
4
2
volts

The sole purpose of the study is to investigate the possibility

of carrying out certain signal processing techniques at

-2
-4

quadrature baseband and hence, it is required to outline the

-6

theory used to carryout the simulations at QBB. Consider the

-8

figure below.

-10

100

200

300

400

500
600
freq(Hz)

700

800

900

1000

800

900

1000

1600

1800

(a)
Chirp at QBB
magnitude spectrum of chirp

c h irp s ig n a l

x 10

10

m a t c h e d fil t e r o u t p u t

1000

2 .5
6

LPF

-4
-6

Matched filter
detector
at
QBB

900

1 .5

v o lt s

v o lt s

-2

800

0 .5

-8

-0 . 5

700

-1 0
0

10 0

20 0

3 00

400

50 0

60 0

7 00

800

900

1 0 00

-1

fre q (H z )

2 00

40 0

60 0

800

1 00 0

12 0 0

14 0 0

16 00

1 8 00

2 0 00

t im e (s a m p le p o i n t n u m b e r)

e jct

matched filter output

Figure 2: QBB chirp signal matched filter detection

magnitude

input chirp

600
500
400
300
200
100

From the figure above, it is seen that the chirp is mixed down

100

200

and lowpass filtered to quadrature baseband before being fed

300

400
500
600
frequency(Hz)

700

(b)

into the matched filter which has a time-reversed copy of the

4

QBB input chirp. The output does have its peak at t=T as in

the previous RF analysis. The simulation results will give a

matched filter output

x 10

more detailed comparison of the graphical results obtained

volts2

from the RF and QBB simulations.

xqbb =| x.* e

j c t

| LPF

(3)

1
0
-1
-2

The matched filter output should show that a perfect match

occurs when the QBB chirp input and its time-reversed copy
make a complete overlap at t=T resulting in the large spike at
that point.

-3
-4

200

400

600
800 1000 1200 1400
time (sample point number)

2000

(c)

Figure 3: (a) Input chirp (b) Chirp signal magnitude

spectrum (c) Matched filter output.

IV. MATLAB SIMULATION RESULTS

From the figure as explained under the theoretical analysis, it
This part of the paper gives the simulation results for the

is seen that the peak of the matched filter output appears at

analysis carried out at RF and QBB. The results at RF and

t=T which in the current case due to the discrete domain

QBB are then compared and discussed.

being used, the point T corresponds to the sample point

number n= 1000. The matched filter output has a peak value

ISBN: 978-988-17012-5-1

WCE 2009

Proceedings of the World Congress on Engineering 2009 Vol I

WCE 2009, July 1 - 3, 2009, London, U.K.

of

5e4 at this point and this value can be verified

mathematically. Let y(n) represent the matched filter output

1
| y (n) |= (1000 100) = 5 104
2

in the discrete domain. It should be noted that the convolution

(9)

is a point-wise multiplication and integration which in the

discrete domain is represented by the following operation (2).

Therefore, the output of the matched filter is correct as has

been verified numerically.

y ( n) = x ( n) h( n)

(4)

B. QBB simulation results

n =1

The analysis then moves down to quadrature baseband where

Now for the current simulation, the input chirp x was
sinusoidal having a peak value of 10 and the sampling
frequency fs=1000. Replacing the expressions for x(n) and
h(n) with the discrete domain versions of x(t) and h(t) and
ignoring other parameters of the chirp expect for the

the matched filter detection is now being implemented at

QBB. The figures shown below are the simulation results
obtained from the QBB simulation. The parameters of the
input chirp are the same as that used for the RF simulation
and hence referring to Figure 4 for the input chirp does
suffice.

amplitude and signal form( i.e. sine or cosine)

we have;

magnitude spectrum of chirp

1000

y (n) =

900

N =1000

x (n) h(n)

(5)

800

n =1

magnitude

700

or

600
500
400
300

y (n) =

N =1000

(10 sin[...])(10 sin[...]) *

(6)

200

n =1

100
0

100

200

300

Note that the * denotes the conjugate of the time-reversed

400
500
600
frequency(Hz)

700

800

900

1000

800

900

1000

(a)

version of the chirp. It is seen that a product of the two inputs

magnitude spectrum of down-mixed chirp

1000

of equation 5 summed from n= 1 to N results in a

900

trigonometric expression given below which upon applying a

800

trigonometric identity reduces to the equation 7 below.

magnitude

y (n) =

700

10 0 0

1 0 0 sin 2 [....]

600
500
400

n =1

300

(7)

200

or

100
0

y (n) =

1000

100( 2 2 cos 2[....])

(8)

n =1

100

200

300

400
500
600
frequency(Hz)

700

(b)

Figure 4: (a) Chirp signal magnitude spectrum before

down-mixing (b) after down-mixing

The absolute value of the matched filter output thus reduces

to;

ISBN: 978-988-17012-5-1

WCE 2009

Proceedings of the World Congress on Engineering 2009 Vol I

WCE 2009, July 1 - 3, 2009, London, U.K.

As mentioned earlier, the main advantage of operating at

magnitude spectrum of QBB chirp

magnitude

1000
900

quadrature baseband is that lower frequency of operation is

800

enhanced which also facilitates for lower sampling

700

frequencies.

600
500

REFERENCES

400
300

[1]

200

[2]

100
0

100

200

300

400
500
600
frequency(Hz)

700

800

900

1000

1600

1800

2000

http://www.cambridge-en.com, Newsletter 5 -March

1999.html.
Mahafza, Bassem R. Radar Systems Analysis and
Design Using Matlab Chapman &Hall/CRC, 2000.

(c)
4

matched filter output

x 10

2.5
2

volts2

1.5
1
0.5
0
-0.5
-1

200

400

600
800 1000 1200 1400
time (sample point number)

(d)

Figure 4: (c) After lowpass filtering down to QBB (d)

Matched filter output.

From the figures above, it is seen that the output of the

matched filter at QBB is half the size of the output obtained at
RF. This is due to the fact that the down mixing frequency
used is equal to the carrier frequency and looking at the
spectrum before and after downmixing, it is seen that the
spectrum remains symmetrical at QBB and hence the factor
of 2 is not required after the lowpass filtering action during
the conversion to quadrature baseband and so this reduced
the output by a factor of 2.

However, it is should be

emphasized that this factor does not overshadow the main

advantage of working at lower baseband frequencies and it
therefore suffices to say that the objective of the direct path
analysis has been achieved.
V. CONCLUSION
The analysis carried out does verify the possibility of
carrying out matched filter detection at quadrature baseband.

ISBN: 978-988-17012-5-1

WCE 2009