Investigating Parameter Effects in QPSK

*Note: Sub-titles are not captured in Xplore and should not be used

Syed Meesam Syed Ghufran

Habib University Habib University
Karachi, Pakistan Karachi, Pakistan
[email protected] Type your mail

I. INTRODUCTION B. Channel
In this research, we explore the modeling of • AWGN Channel with Frequency Offset and
practical parameters that affect signal processing in Variable Time Delay - Applies the frequency offset,
everyday life, such as static frequency offset, timing drift, a timing drift, and additive white Gaussian noise to
and Gaussian noise (as discussed in our lectures). Starting the signal
from a foundational example model that processes at
complex baseband, we present a practical digital receiver C. Receiver
design incorporating correlation-based coarse frequency • Raised Cosine Receive Filter - Uses a roll off
compensation, PLL-based fine frequency compensation, factor of 0.5
PLL-based symbol timing recovery, frame synchronization, • Coarse Frequency Compensation - Estimates an
and phase ambiguity resolution. Utilizing various approximate frequency offset of the received signal
synchronization algorithms from the Communications and corrects it
Toolbox™, this study aims to address these common • Symbol Synchronizer - Resamples the input signal
impairments, enhancing the reliability and performance of according to a recovered timing strobe so that
digital communication systems. symbol decisions are made at the optimum
sampling instants
II. MOTIVATION
• Carrier Synchronizer - Compensates for the
In the field of digital communications, challenges residual frequency offset and the phase offset
such as frequency shifts, timing errors, and Additive White • Frame Synchronizer - Aligns the frame boundaries
Gaussian Noise (AWGN) often reduce system performance, at the known frame header
particularly in wireless environments. Addressing these • Data Decoding - Resolves the phase ambiguity
issues is vital for ensuring reliable data transmission. The caused by the Carrier Synchronizer, demodulates
motivation behind this research evolves from the need to the signal, and decodes the text message
develop robust communication systems that can maintain
high-quality signal reception, despite above mentioned V. BLOCKS
factors. By modeling and simulating a QPSK-based system
with practical compensation techniques, this study aims to A. Carrier Synchronization
enhance the performance of digital receivers. After symbol synchronization, we expect our signal to
have a consistent amplitude due to AGC(Automatic Gain
III. EASE OF USE Control)-block. And frequency corrected due to Coarse
Frequency Compensation. And reduced inter-symbol
A. Abbreviations & Acronyms interference with the help of symbol synchronization.
[Add list of abbreviations to prevent extra words helping
in page limit] The next block we pass our signal through is Carrier
Synchronizer.
• CFO - Coarse Frequency Compensator
This block has the following parameters:
• FFT – Fast Fourier Transform
• BER – Bit Error Rate • Modulation type: This is required to demodulate the
incoming signal by following the modulation
B. Values Simplification
scheme. In our case QPSK(Quadrature Phase Shift
Values like 1.002e-5 simply means 1.002*10^(-5) Keying) is used and so we’ve selected the specific
demodulation scheme to demodulate the signal.
IV. STRUCTURE OF SIMULINK DIAGRAM:
• Modulation offset: This is the phase offset that
A. Transmitter occurs due to type of modulation that is used and
several other natural reasons (for e.g. Doppler’s
• Bit Generation - Generates the bits for each frame
Effect). Removing this offset is the reason why we
• QPSK Modulator - Modulates the bits into QPSK are using a carrier synchronizer. We may also use a
symbols PLL (will be discussing it after a few parameters). To
• Raised Cosine Transmit Filter - Uses a roll off remove this phase offset the system we are currently
factor of 0.5, and up samples the QPSK symbols by working on has set it to ‘auto’. We can use the
two constellation diagram to analyze the phase offset.
Then we calculate the offset by checking the
difference between measured offset and ideal offset

XXX-X-XXXX-XXXX-X/XX/$XX.00 ©20XX IEEE

 caused by modulation (which is typically 0 or pi/4 affect the BER. Conclusion: We may decrease our normalized
for QPSK modulation scheme). However, this also loop bandwidth for more stability of our PLL.
requires us to check how many phase shifts have
occurred in one second, to determine angular B. Frame Synchronization
velocity and then we proceed (a relatively long The frame synchronizer block in Simulink is essential for
process, and that is why we use PLL). accurately identifying the start of each frame within the
• Samples per symbol: Defines the resolution of the received signal. This block typically uses parameters such as
signal processing. In our case we’ve set it to 2, this a preamble, detection threshold, and output length to perform
allows efficient synchronization and signal tracking. its function.
• Damping Factor: This is a factor used in PLLs The preamble is a predefined sequence known to both the
(Phase Lock Loops). In carrier synchronization we transmitter and receiver, used to signal the beginning of a
use PLL to accurately rectify the phase offset. This frame. In this case, the Barker Code of length 13 is used as the
happens by a close loop, where the signal phase is preamble. Barker Codes are special sequences of +1s and -1s
constantly measured with the reference phase to with excellent autocorrelation properties, meaning they
match the desired phase. There is a voltage produce a strong peak when aligned correctly and low
controlled oscillator that introduces voltage sidelobes elsewhere, making them highly effective for
depending on dis-similar is the signal. This damping synchronization. The preamble helps the receiver lock onto
factor ensures that the voltage introduced doesn't the beginning of the frame, ensuring the data is correctly
overshoot/overestimate. interpreted.
• Normalized Loop Bandwidth: Another key
parameter for PLL is Normalized loop bandwidth. The detection threshold is a crucial parameter that
This is a dimensionless parameter which represents determines the sensitivity of the frame detection process. It
bandwidth relative to the symbol rate. It is basically sets a threshold level that distinguishes between actual frames
the rate of how quickly a PLL matches the phase and noise. A signal that crosses this threshold is considered a
offset with the reference. valid frame, while anything below it is treated as noise. Setting
We plan to change these parameters to visualize the effect this threshold too low might result in false detections due to
of each parameter on the performance of the system. One by noise, while setting it too high might miss some frames.
one each parameter will be changed. The performance will be Output length refers to the size of the frame, ensuring that
measured differently for each parameter. A global measure of the receiver processes the correct amount of data for each
performance will be the BER. frame. This parameter helps maintain a smooth transition of
While we change other parameters, we understand that data throughout the receiver, ensuring consistent processing.
modulation type is a fixed parameter as this is the scheme that Therefore, we don’t change this parameter as changing it will
was used by the transmitter to modulate the signal. And that hinder the process of demodulating the signal.
modulation offset will be handled by the PLL, therefore there The parameters Barker code and Detection threshold are
is no need to change it. actually interlinked with each other. The detection threshold
We proceed first by changing the resolution, Sample per is set by taking the gaussian curves of the two possibilities,
symbols. We expect that by increasing the resolution we have and then the average of the mid points of the two curves
a lower BER and we observe less scatter in our constellation becomes the boundary value to decide whether the incoming
diagram. Findings: By increasing the resolution the scatter on frame is noise or an actual frame. In our case we’ve attached
the constellation diagram decreases, however, we see an Barker code to our frames. So it all boils down to the fact that
increase in BER. Conclusion: By increasing the resolution we the upcoming signal is a Barker code or noise. In our system
may visualize the signal better but it is not necessary that it we see that our Barker code is first normalized (by multiplying
improves the BER . it by, frac{\ sqrt{2}{2} ) and then phase shifted (multiplied by
(-1 -1i)). And then transpose is taken of the whole array of
Secondly we analyze the Damping Factor. We understand entries for sequence’s orientation. And the detection threshold
that any process that takes less time is preferred. And so by is 20 set by the system. The hard code value gives us a good
increasing the damping factor we should see a quick response approximation about signal to noise ratio that the signal has
in terms of carrier synchronizations, but we also know that after being received. While the effects of changing the
such responses generate overshoots as well. Let's analyze the parameters are discussed already, we understand that every
damping factor. Findings: BER increased significantly, when system will have their specific requirement of Barkar code and
damping factor was 4, it went from 9.017e-05 to 0.0002692. that upon its specific SNR we deduce the detection threshold.
And the time taken to complete the task was quicker but it was
hardly noticeable. Note: Even decreasing, increased the BER. C. Coarse Frequency Compensator
Conclusion: Damping factor has to be decided after careful The CFO is a block inside the block Receiver => Coarse
consideration, it is highly likely that going below threshold Frequency Compensation => CFO. The CFO is a block that
will not estimate the desired phase offset, and going over the corrects frequency offsets in a received signal. These
threshold will overestimate it. frequency shifts happen when the receiver does not match
Lastly, we analyze that normalized loop bandwidth. perfectly with the transmitted signal. This mismatch happens
Increasing this factor will let the PLL quickly adapt to the during the conversion of received signal to a lower frequency
changes that occur throughout the signal. Let’s experiment signal for easier processing [2]. Why does this happen? It is
and find out. Findings: By increasing the factor we get the due to two major factors in role a. frequency mismatch and b.
results slightly, with a trade off bad BET. However, Doppler effect [2].
decreasing the factor will slow the process slightly, but won’t
 • Frequency mismatch: Both transmitter and receiver D. Symbol Synchronizer
have oscillators that generate carrier signal. If these The Symbol Synchronizer is a block inside the receiver
oscillators are not aligned correctly (due to design block. Symbol synchronizer is a critical component that helps
changes or any temperature change) then their in minimizing errors caused by timing mismatch. It is the
frequencies would change, causing a mismatch. process of estimating correctly and aligning the timing of
• Doppler effect: When a transmitter or receiver or received signal data in the communication system. This
both are moving the frequency of the signal changes ensures that the receiver can correctly interpret transmitted
slightly due to the doppler phenomenon, which data by identifying start and duration of transmitted symbol
results in a signal that is either compressed or [3]. Why does this happen? It usually happens due to either of
stretched. the reasons: clock mismatch between receiver and transmitter,
Due to the above mentioned causes we use compensators delays due to reflection and refraction, noise and interference,
to fix these problems that make it difficult for the receiver to or Doppler shifts. To prevent such mismatches we use Symbol
correctly understand the signal. synchronizer to sync symbols between transmitter and
The CFO has following parameters [1]: receiver.
(To be noted that since we are working on QPSK modulation Symbol synchronizer has the following parameters [4]:
scheme we discuss their parameters only for simplicity) (To be noted that since we are working on QPSK modulation
• Estimation algorithm: This defines the method to scheme we discuss their parameters only for simplicity)
estimate frequency offsets. CFO supports FFT-based • Modulation Type: It has two options
for modulation schemes like QQPSK and QAM, PAM/PSK/QAM and OQPSK.
while correlation-based supports modulation
schemes like 8PSK, BPSK, PAM, QPSK, QQPSK, • Timing error detector: It has three types Zero-
and QAM. Further understanding of algorithms can Crossing (decision-directed), Gardner (non-data-
be seen in section Algorithms in [1]. aided), Early-Late (non-data-aided), or Mueller-
Muller (decision-directed).
• Maximum Frequency Offset: this is the largest
frequency shift that the algorithm can detect. And is • Samples per symbol: Define samples taken per
only applicable for correlation-based algorithms. symbol (positive integer greater than 1).
• Frequency Resolution: It tells how finely the steps • Damping Factor: This controls the stability of the
are taken when measuring frequency offsets. loop filter (positive scalar).
• Normalized loop bandwidth: It is the normalized
bandwidth of the loop filter (a positive scalar less
a. Effect of Changing Algorithm than 1).
While performing the FFT-based algorithm the BER is • Detector gain: The gain of the phase detector (a
1.124e-3; however, performing the Correlation based positive scalar).
algorithm the BER was found to be 2.047e-4. From changing
algorithms it can be seen that the Correlation algorithm works a. Effect of Changing Timing Error Detector
best.
The effect of Zero crossing on BER results in 0. For
b. Effect of Changing Max Frequency Offset (only for Gardner BER is 0.001125. For Early Late we get a similar
correlation-based) result to that of Gardner of BER 0.001135. And Mueller-
Muller gives BER of 0. Both Zero Crossing and Mueller-
On decreasing the max frequency offset from 6KHz to Muller rely on the previous data (which can be corrupted) to
2KHz the BER significantly increases to 0.4842. Even on adjust timing which causes the BER to be 0. However,
increasing to 5KHz which is the default value for phase offset Gardner and Early Rate do not use previous data that was
in AWGN we see no significant change in value of BER. transmitted; rather, Gardner uses the shape of the signal to
However, as we increase from 5KHz to 6KHz we see 2.047e- estimate the best timing adjustment of the samples. And Early
4, but for 8KHz we see BER of 2.371e-4. Late samples before and after the expected timing and finds
From above results we came to a conclusion that there is the best adjustment by comparing those two points.
to be an optimum value for which the algorithm provides a b. Effect of Changing Samples per Symbol
good estimate. Lowering the value from the default frequency
offset of AWGN results in the failure of the algorithm in When samples per symbol were equal to 2 we had a BER
compensating frequency shifts greater than that value as in the of 0.0002047 but as we increased the number to 4 the BER we
case of 2KHz. Similarly, on increasing the value we see that got was 0.5013. This increase in BER is due to when we
the algorithm now has a bigger range to find the estimate increase the number of samples the system becomes
which increases the chances of the error. vulnerable to error meaning it becomes more sensitive to noise
or distortions in the signal.
c. Effect of Changing Frequency Resolution (only for
FFT-based) c. Effect of Changing Damping Factor
On decreasing the value of frequency resolution and even At a damping factor of 0.01 we get a BER of 0.04531. At
increasing its value we see no significant changes in the values 1 we get a BER of 0.0002047. At 10 we get a BER of
of BER which is 0.001125. This shows the insensitivity of 0.0001634. This is due to the reason that the system becomes
frequency resolution on BER. slower but becomes more stable thus we see a decrease in BER
value. And as we decrease the damping factor the system now
becomes faster for correcting timing errors which may cause
overcorrection and reacting too much, resulting in a higher
BER.
d. Effect of Changing Detector Gain
At Detector gain 50 we get a BER of 0.001013 decreasing
to 5.4 results in BER of 0.0002047, further decreasing it to 0.5
gives BER of 5.89e-06. When we increase detector gain the
system becomes more sensitive to small changes so even
small changes due to noise or distortion causes the system to
aggressively correct mistakes, resulting in a higher value of
BER. When we decrease it the system becomes less sensitive
which results in the system not taking its job seriously and
may neglect to detect timing or phase error causing low BER.
VI. FINDINGS
The Coarse Frequency Compensator plays an important
part in correcting frequency shifts caused by oscillator
mismatches or Doppler effect. It was found that the
Correlation-based algorithm outperforms the FFT-based one
in reducing Bit Error Rate, due to which it was taken as
default. Additionally, the maximum frequency offset
parameter significantly impacts the algorithm’s performance,
with lower offset causing an increase in BER due to failure in
providing larger frequency shifts, while higher offsets
introduce a wide range for estimates, potentially increasing
errors.
VII. CONCLUSION
The above findings highlight the importance of selecting
optimal parameters for effective compensation and
synchronization. Since we focused over QPSK we got a
different set of results compared to if we worked over different
modulation schemes that would require a different set of
parameters, like how FFT-based algorithms can be used by
QQPSK and QAM.
REFERENCES
[1] The MathWorks Inc. (2022). Coarse Frequency Compensator
Documentation, Natick, Massachusetts: The MathWorks
Inc.https://www.mathworks.com/help/comm/ref/coarsefrequencycom
pensator.html
[2] https://en.wikipedia.org/wiki/Carrier_frequency_offset#:~:text=Carrie
r%20frequency%20offset%20often%20occurs,contained%20in%20th
e%20received%20signal.
[3] Simon, M. K. (2006). Symbol Synchronization. Autonomous
Software‐Defined Radio Receivers for Deep Space Applications, 321-
390.
[4] The MathWorks Inc. (2022). Communications Toolbox
Documentation, Natick, Massachusetts: The MathWorks Inc.
https://www.mathworks.com/help/comm/ref/symbolsynchronizer.htm
l#References
[5] Simon, M. K., & Hamkins, J. (2006). Autonomous software-defined
radio receivers for deep space applications. Wiley-Interscience
[6] Franks, L. E. (1980). Carrier and bit synchronization in data
communication: A tutorial review. IEEE Transactions on
Communications, 28(8), 1107-1121.
[7] Xu, T., Gao, F., Wang, X., & Blaabjerg, F. (2019). A carrier
synchronization method for global synchronous pulse width
modulation application using phase-locked-loop. IEEE Transactions
on Power Electronics, 34(11), 10720-10732.