Istanbul Medipol University

Electrical-Electronics Engineering And Cyber

Systems

Digital Communications Laboratory

Lab 4 Report
Effect of pulse-shaping filters in communication systems.

Written By:
Talha YILMAZ (D3210008)

Date: March 11, 2023

Contents
1 Introduction 3

2 Simulation 3

3 Real Transmission 6

1
 Abstract
Modulated signals in the baseband are represented as discrete impulses. Transmission
of impulses is impossible in practical systems since an infinite amount of bandwidth is
required to generate an impulse. Pulse-shaping is used to shape the power spectrum of
the transmitted signal for efficient transmission over a given bandwidth. Even though the
main objective of pulse-shaping is shaping the power spectrum, there is another impor-
tant aspect of pulse shaping which is preventing inter-symbol interference (ISI). Although
not necessary we would like to have zero ISI caused by transmission(before the channel).
Pulses that satisfy this criterion are called Nyquist pulses. In this lab three mainstream
pulse shapes raised-cosine (RC), root-raised-cosine (RRC), and Gaussian pulses are inves-
tigated.

2
 Preamble Data

127 127 512 10

Guard

Figure 2.1: Transmission frame

1 Introduction
This lab has two parts. The first part is a simulation with RC and rectangular pulses. The
second part of the lab involves Adalm-Plutos to transmit and receive frames with different
pulses.

2 Simulation
The lab starts by building the frame shown in Figure 2.1. The preamble consists of two m-
sequences, data consists of QPSK symbols and the guard is just zeros. The generated frame
is filtered by using an RC filter. Frame generation and filtering are done by the following
MATLAB code:

1 %Filtering Setup
2 flt stp.sps = 8; % samples per symbol, oversampling ratio
3 flt stp.span = 12; %filter span, truncated after this number of sample ...
from both sides of center of the signal
4 flt stp.alpha = 0; %Roll-off factor
5

6 source data = randi([0 mod stp.M-1],mod stp.N,1);

7 symbs tx = qammod(source data,mod stp.M);
8 symbs tx = symbs tx/rms(symbs tx);
9 m seq = mseq(2,7);
10 prembs = [m seq;m seq];
11 frm tx = [prembs; symbs tx; zeros(10,1)];
12
13 upsmpld = upsample(frm tx,flt stp.sps);
14
15 fltr = rcosdesign(flt stp.alpha,flt stp.span,flt stp.sps,'normal');
16 fltr = fltr/max(fltr);
17 tx signal = conv(upsmpld,fltr);

A transmission signal is generated using several roll-off factors as instructed in the lab
manual and some of their properties are compared. The first comparison is done on their
power spectrum as seen in Figure 2.2.
Bandwidth of a RC pulse is given as
Rs
BW = (1 + α) , (2.1)
2
1
where Rs is the symbol rate defined as Rs = Ts
, and α is roll-off factor where 0 ≤ α ≤ 1. By

3
 0

-5
=0
-10 = 0.3
= 0.5
-15

Power (dB)
= 0.7
=1
-20

-25

-30

-35

-40

-1.5 -1 -0.5 0 0.5 1 1.5

Frequency (Hz) 105

(a) %99 occupied bandwidth comparison (b) Null to null bandwidth comparison

Figure 2.2: Power spectrum comparison between different roll-off factor raised-cosine pulses.

inspecting (2.1) we can see that BW increases by the roll-off factor. Results obtained by Figure
2.2 supports this claim. As α increases RC in the time domain becomes more localized and its
spectrum gets closer to a sinc pulse. On the other hand, as α decreases RC in the time domain
effectively approaches a sinc pulse and its spectrum gets closer to a rectangular pulse. Both
properties have their uses and choosing the correct α value becomes an optimization problem.
For example, one effect of low α usage is high side-lobes. Even though these side lobes don’t
affect the orthogonality of the pulse, depending on transmitted symbols these side lobes add
to each other and create large peaks. These would increase the peak-to-average power ratio
(PAPR) value of the transmitted signal and decrease the efficiency of the used power amplifier.
CCDF curves for different roll-off factors are shown in Figure 2.3. By using results from Figure

CCDF Measurement
102
=0
= 0.3
= 0.5
101 = 0.7
=1
Probability (%)

100

10-1

10-2
0 2 4 6 8 10
dB above average power

Figure 2.3: CCDF curves of the transmitted signal with different roll-off factors.

2.3 PAPR values can be calculated by

PAPR (dB) = Ppeak (dBm) − Paverage (dBm) (2.2)

resultant PAPR values are shown in Table 1. Continuing on the effect of side-lobes we can
see with polar diagrams there is less transition variation between symbols. This is expected

4
 Table 1: Computed PAPR values.

Roll-off Value (α) Peak-to-Average Power Ratio (PAPR)

0 6.2273
0.3 5.4710
0.5 4.1897
0.7 3.0310
1 1.8728

since by 1 we expect less PAPR with higher α values. A result that is not obvious however is
robustness against timing error. By inspecting eye diagrams of these pulses we can see that a
higher α value pulse has more eye-opening. There still would be interference if you miss the
sampling time since these are overlapped pulses but inter-symbol interference caused by sample
timing error would be way less when higher α values are used. Polar and eye diagrams for these
pulses are shown in Figure 2.4 and in Figure 2.5 respectively.
90 90
120 2 60 120 60
1.5
1.5
150 30 150 1 30
1

0.5 0.5

180 0 0 180 0 0

210 330 210 330

240 300 240 300

270 270

(a) Polar diagram for α = 0 (b) Polar diagram for α = 0.3

90 90
120 1.5 60 120 60

1
1
150 30 150 30

0.5 0.5

180 0 0 180 0 0

210 330 210 330

240 300 240 300

270 270

(c) Polar diagram for α = 0.5 (d) Polar diagram for α = 0.7
90
120 60
1

150 30
0.5

180 0 0

210 330

240 300
270

(e) Polar diagram for α = 1

Figure 2.4: Polar diagrams for different α values.

Next to compare power spectrums between RC pulses and rectangular pulses a frame is
generated using rectangular pulse. The MATLAB does the rectangular filtering is shown below:

1 source data = randi([0 mod stp.M-1],mod stp.N,1);

2 symbs tx = qammod(source data,mod stp.M);
3 symbs tx = symbs tx/rms(symbs tx);
4 m seq = mseq(2,7);
5 prembs = [m seq;m seq];
6 frm tx = [prembs.' symbs tx.' zeros(1,10)];
7

8 temp = repmat(frm tx,flt stp.sps,1);

9 filtered = reshape(temp,1,length(frm tx)*flt stp.sps);

Then the power spectrum of the received frame is plotted. The power spectrum of rectangular
filtering is shown in Figure 2.6. We can see that the spectrum of the rectangular pulse occupies

5
 Eye Diagram for In-Phase Signal Eye Diagram for In-Phase Signal
2
1
1

Amplitude
Amplitude
0 0

-1
-1
-2
-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 -0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5
Time Time
Eye Diagram for Quadrature Signal Eye Diagram for Quadrature Signal
2
1
1

Amplitude

Amplitude
0 0

-1
-1
-2
-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 -0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5
Time Time

(a) Eye diagram for α = 0 (b) Eye diagram for α = 0.3

Eye Diagram for In-Phase Signal Eye Diagram for In-Phase Signal
1
1
0.5

Amplitude
Amplitude

0 0

-0.5
-1
-1
-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 -0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5
Time Time
Eye Diagram for Quadrature Signal Eye Diagram for Quadrature Signal
1
1
0.5

Amplitude
Amplitude

0 0

-0.5
-1
-1
-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 -0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5
Time Time

(c) Eye diagram for α = 0.5 (d) Eye diagram for α = 0.7
Eye Diagram for In-Phase Signal
1

0.5

Amplitude
0

-0.5

-1
-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5
Time
Eye Diagram for Quadrature Signal
1

0.5

Amplitude
0

-0.5

-1
-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5
Time

(e) Eye diagram for α = 1

Figure 2.5: Eye diagrams for different α values.

Figure 2.6: Power spectrum of the rectangular filtered frame.

the whole band. In practical systems, this is unacceptable since there are bands for commercial
uses that are licensed. There are heavy penalties for occupying a licensed band since it would
degrade the performance of other commercial products. The effect of α in a sampling timing
error is mentioned before. The effect of α in the presence of sampling timing error for RC
pulses is demonstrated in Figure 2.7.
By inspecting Figure 2.7 we can see a constellation diagram that hints at some inter-symbol
interference. We know that this ISI happened due to the usage of overlapped symbols and the
presence of symbol timing error. As explained before we see a gradual decrease of interference
power as α increases.

3 Real Transmission
This part of the lab involves the usage of Adalm-Plutos and investigating how these effects are
seen in real life. First with the hardware setup instructed in the lab manual a frame with RC

6
 Scatter plot Scatter plot
1
1

0.5
0.5

Quadrature

Quadrature
0 0

-0.5
-0.5

-1
-1
-1 -0.5 0 0.5 1 -1 -0.5 0 0.5 1
In-Phase In-Phase

(a) Constellation diagram for α = 0 (b) Constellation diagram for α = 0.3

Scatter plot Scatter plot
0.8
0.8

0.6 0.6

0.4 0.4

0.2 0.2
Quadrature

Quadrature
0 0

-0.2 -0.2

-0.4 -0.4

-0.6 -0.6

-0.8
-0.8
-0.5 0 0.5 -0.5 0 0.5
In-Phase In-Phase

(c) Constellation diagram for α = 0.5 (d) Constellation diagram for α = 0.7
Scatter plot
0.8

0.6

0.4

0.2
Quadrature

-0.2

-0.4

-0.6

-0.8
-0.5 0 0.5
In-Phase

(e) Constellation diagram for α = 1

Figure 2.7: Constellation diagrams for different α values.

setup is sent and received. By using EVM of the received data samples SNR is calculated by

SNR = −20 log10 (EVM). (3.1)

SNR for tested transmission is found as 11.5 dB. Then the next transmission is implemented
via the usage of the RRC filter and matched filtering at the receiver. SNR is calculated in the
same way and found as 14.5 dB. For this case, we can see a 3 dB gain via the implementation
of matched filtering. To see what matched filtering does we can inspect the spectrum of the
signal before and after matched filtering. Figure 3.1 shows the spectrum of an RRC shaped
frame without matched filtering and with matched filtering.
By Figure 3.1 we can see that matched filtering attenuates the noise and increases overall
signal power. With RC pulse and no matched filtering, we don’t have this signal power increase
at the receiver. However by implementing RRC pulse shaping in a transmission system matched
filtering becomes a necessity. Because RRC is not a Nyquist filter. In Figure 3.2 there is an eye
diagram comparison of matched and not matched RRC pulse. After a very simple investigation
of 3.2 we can see that there is a very small eye-opening if matched filtering is not implemented.
This small eye-opening is due to ISI caused by the pulse shaping.

7
 -20

-30

-40

-50
Power(dB)

-60

-70

-80

-90 Without matched filtering

With matched filtering
-100
-5 0 5
Frequency(Hz) 105

Figure 3.1: Power spectrum of an RRC shaped frame with and without matched filtering.

10 -3 Eye Diagram for In-Phase Signal 10 -3 Eye Diagram for In-Phase Signal
4
1
2
Amplitude

Amplitude

0 0

-2
-1
-4
-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 -0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5
Time Time
-3 Eye Diagram for Quadrature Signal -3 Eye Diagram for Quadrature Signal
10 10
4
1
2
Amplitude

Amplitude

0 0

-2
-1
-4
-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 -0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5
Time Time

(a) RRC eye diagram without matched filtering (b) RRC eye diagram with matched filtering

Figure 3.2: Eye diagram comparison of matched and not matched RRC pulse.

8
 Lastly, the effect of the Gaussian filter with different time-bandwidth products is tested.
The constellation diagram and power spectrum for each transmission are given in Figure 3.3.
Scatter plot
1.5

0.5
Quadrature

-0.5

-1

-1.5
-1.5 -1 -0.5 0 0.5 1 1.5
In-Phase

(a) Constellation diagram for BT = 0.22 (b) Power spectrum for BT = 0.22
Scatter plot

1.5

0.5
Quadrature

-0.5

-1

-1.5

-1.5 -1 -0.5 0 0.5 1 1.5

In-Phase

(c) Constellation diagram for BT = 0.5 (d) Power spectrum for BT = 0.5
Scatter plot

0.5
Quadrature

-0.5

-1

-1 -0.5 0 0.5 1
In-Phase

(e) Constellation diagram for BT = 1 (f ) Power spectrum for BT = 1

Figure 3.3: Measurements for different BT Gaussian pulse transmissions.

By only observing Figure 3.3 we can say that performance of the system increases with
BT at the cost of occupied bandwidth. BT product was given a relationship between symbol
period and bandwidth. BT = 1 means that bandwidth is exactly T1 . Now if we think about
this relationship for a moment we would come up with this interesting question ”If we are
comparing pulses with fixed sample-per-symbol and sampling frequency what do different BT
values mean for that pulse?” There are two important arguments in this question first being
fixed sample-per-symbol and the second being fixed sampling frequency. If these two are the
same it means that we have fixed our transmission rate. BT product is just a number that tells
us how bandwidth is related to that symbol rate. If the symbol rate is fixed but bandwidth
is changing there is only one sensible explanation which is the pulse in the time domain gets
wider and narrower depending on the BT product. If BT is low it means that bandwidth is
0.2 times the symbol rate. The only way to have a bandwidth of a pulse lower than its symbol
rate is to spread it in time. As a result, as BT decreases the Gaussian pulse spreads in time.

9
This spread causes leakage to the other symbols. That is why low BT causes interference and
degradation in a communication system. On the other hand, you can have a larger spectrum
for a very thin pulse to have better performance.

10