Nat. Volatiles & Essent.

Oils, 2021; 8(6): 888-904

Digital Predistortion (Dpd) Algorithm For 5g Applications

Dr. S.G. Shivaprasad Yadav1 , Meghana2 , Bharati R3 , Dr. Bhagya R4

1
Associate Professor, Dept. of Electronics & Telecommunication Engineering, M S Ramaiah Institute of Technology,
Bangalore, India
2
PG Student, Dept. of Telecommunication Engineering, M S Ramaiah Institute of Technology, Bangalore, India
3
Associate Professor, Dept. of Computer Science and Engineering, BMS Institute of Technology & Management,
Bangalore, India
4
Associate Professor, Dept. of Electronics & Telecommunication Engineering, R V College of Engineering, Bangalore,
India

Abstract

The emerging technologies and trends in the Wireless Networks applications have been enormously changing and expect to have
the capability of very high data rates and speeds which maybe hundred times faster with high quality in connection and signal
levels. With this advancement and growth of Wireless communication systems, there is a lot of demand for linearity in power
amplifier. Also the Radio Frequency (RF) power amplifiers, which are expected to work with signals having Wideband Code Division
Multiple Access (WCDMA) and are used extensively in Universal Mobile Telecommunication Systems (UMTS) for the generation of
out-of-band emissions named as “Spectral regrowth”. As there is a need for enhancing the linearity and efficiency of Power
Amplifier, the proposed work used the approach of Digital Predistortion. The proposed work of Digital Pre Distortion was
developed using Matlab/Simulink combining memory polynomial algorithm. This developed algorithm was validated with various
test cases and it was seen that the performance is improved drastically with high non-linearity with the memory polynomials order
of 4 and 5 and having linear output. With the incorporation of memory polynomial digital predistortion in the proposed technique,
the overall performance has been showing very good results. The overall improvement in the performance is observed by the
incorporation of power amplifier with high non-linearity and thereby achieving less distortion with linear output.

Keywords:Power Amplifier, Digital Predistortion, Memory Polynomial, 5G

Introduction
Nowadays, wireless technologies have become most important technology than the wired technology.
Linear and high efficiency Power Amplifiers (PAs) are important components in wireless systems. The
power amplifier is the active element located between the antenna and the transmitter signal circuitry [1].
The PA is the element which uses the low-level RF signal and improves its power, without adding any
change to modulation format, or other factors. In modern telecommunication systems, the Radio
frequency PA’s that are working with the signals having wideband like WCDMA employed in Universal
Mobile Telecommunication Systems (UMTS) are likely to possess spectral regrowth which infers out of
band emissions [2]. The most efficient way is to maintain within the linear portion of the operating curve

888
Nat. Volatiles & Essent. Oils, 2021; 8(6): 888-904

and to inverse the amplifier input level. This method is simple and have the disadvantages of larger size and
higher cost. Another method is the Feed forward method for linearizing the power amplifiers in which the
distortion is subtracted at the output [1]. This involves a considerable increase in cost due to high power
analog combining network.
Another method, known as Linear Amplification with Non-linear Components (LINC), which generated the
two phase-altered nonlinear PAs output to complete linear operation but may still need a high-power
analog combining setup. Alternate way and one of the popular approach is to use the Digital Predistortion
before the amplifier to reduce the effects of nonlinearity. The Digital Predistortion incorporation has led to
several benefits decreasing cost and size as compared with other approaches. One of the commonly used
technique to identify the parameters of a pre-distorter is the Indirect Learning Architecture (ILA).
In this paper, the main focus is on memory models that have effective wideband DPD. Firstly, we discuss
the general process using Volterra Series method and then map out some simple special cases of memory
models which consists of non-linearity coefficients and memory polynomial coefficients.

y(n)
u(n) Predistortion (n) PA

Parameter
Estimation

Fig.1. Generic Predistortion Scheme

Related Work

• Morgan et.al [7]: In this paper work, standard Radio Frequency Power Amplifier which are working
with signals having wideband, like WCDMA in the UTMS is inversed correspondingly from their power
level in order to reduce out of band spectrums. Adjusting these values of amplifiers to operate in
wideband therefore it needs higher cost and larger size that would be needed for the similar power
output. Another way that is attaining attention in wireless communication and have widespread
demand, in order to use digital predistortion before the amplifier to decrease the effects of
nonlinearity, thus by having low spectral regrowth it runs closer to maximum output power. Memory
effects have included in the recent improvements to the technique in the model of predistortion that
is important as there is an increase in the bandwidth. These wideband predistortion model leads to the
GMP (Generalized Memory Polynomial) which integrates the cross terms of memory polynomial
between the signal and leading or lagging terms.

• Yasunori Suzuki et.al [8]: In this paper work, configuration of Massive MIMO power amplifiers for
transmitter is proposed. The configuration that is proposed in this work comprises of the path having
linearity and amplifier corresponding to the loop of feed-forward design where the signal cancellation

889
Nat. Volatiles & Essent. Oils, 2021; 8(6): 888-904

takes place. The amplifier path contains the power amplifier and vector regulator while the linear path
contains the delay line. In this paper 5G wireless communication system for Massive MIMO has been
suggested. In Massive MIMO transmitter a decrease propagation loss and provides gigabit data
transmission in high frequency band occurs due to the creation of narrow beam. Other than this the
Massive MIMO is needed to be small enough as WI-FI access because it is developed in small cells that
are presented in hotspot areas. The Massive MIMO transmitter is favorable in using the active and
adaptive antenna technologies to fulfill the specifications that are mentioned in the paper work.

• Jianxin Jing et.al [9]: In this proposed paper, a technique named Digital Predistortion is proposed
which is employed to linearize the target beam restricted with other beams in the multi-beam scenario
of 5G Massive MIMO transmitters. The proposed paper models the interference signal first and then
creates the exact opposite signal and both the signals gets cancelled. The results says that the digital
predistortion technique operates really well to realize the target beam linearization, thus having lot of
advantages for multi user applications in 5G Massive MIMO systems. Massive MIMO multibeam
systems combined with passive beamforming network have been studied due to its simple structure,
low cost and low power consumption.

• MouradDjamai et.al [10]: In this paper work, the RF power amplifiers that have nonlinear model is an
important concept in the development and designing of linear amplifier. A predistortion technique
based on DSP circuit is introduced and the continuous time model defining the power amplifier that is
related to the quadrature modulation and quadrature demodulation method is identified. The model
is designed with two models, the first one is modelled by using polynomial functions which uses
nonlinear phase and amplitude distortion and the second one is description of the power amplifier
dynamics referred to the MIMO continuous filter. The parameters are identified by using a method
based on pseudo random binary sequence and output error as excitation of the signal. Power
amplifiers are one of the significant components in the wireless communication systems. In order to
transmit the amplified and modulated signals power amplifiers are needed. In order to decrease the
power consumption and increase the efficiency of the power amplifier they are needed to be operated
in the saturation region. In such scenarios there may be a possibility of nonlinearity in the power
amplifier. The nonlinear signal components like spectral regrowth caused by the power amplifier are
created which interferes the adjacent channels and bit error rate is also increased due to distortions.

• Dian Xie et.al [11]: In this work, LMS (Least Mean Square) and RLS (Recursive Least Square) algorithm
combined with an improved adaptive algorithm is represented. The results give the comparative
simulation to the mentioned algorithm. By using this method there is a decrease in the calculation
amount which gives convergence stability and also achieves a good linearity performance.

• Compared to the various methods of linearization, one of the most cost effective method is the digital
predistortion because of its better stability, adaptability and its implementation, stability. To calculate
the coefficients of the power amplifier model adaptive algorithms are employed like memory
polynomial model and Volterra model. Both RLS and LMS are adaptive algorithms. LMS algorithm have
less computation and stable convergence whereas RLS algorithm have higher convergence and less
precision. Power amplifier models can be classified into two types: memory and memoryless. The
memory model is for baseband signal while the memoryless model is for narrow band signal and is
defined by Volterra series which is more complex to compute. Moreover, it makes unattractive for the
practical application because of its large coefficient number in Volterra series and thus there appears
some improved model. The memory polynomial model, which is an enhanced version of Volterraseries

890
Nat. Volatiles & Essent. Oils, 2021; 8(6): 888-904

is employed in digital predistortion algorithm to remove the nonlinearity effects in the power
amplifier.

• Varahram et.al [12]: In this work paper, the application of digital predistortion in baseband signal is a
technique used for power amplifier linearization in order to decrease the ACI in the systems which
have different modulation techniques like MQAM, OFDM etc, is employed. To linearize the power
amplifiers digital predistortion is the best way because of its less cost effective and most architectures
presume that memoryless nonlinearity is present in the power amplifiers. For applications like W-
OFDM and WCDMA that has wideband signals the memory effects cannot be ignored. A new adaptive
technique of DPD that involves memory effects of the power amplifier. This approach is combination
of two methods, slope dependent method and memory polynomial predistortion. This new approach
is examined by employing a 1.9GHz 60W LDMOS power amplifier with two or three carrier CDMA.

• Bondar et.al [13]: In this paper work, a new approach of digital baseband predistortion technique of
wide band with improved power amplifiers and high memory quality is presented and investigated. By
computing the iterative digital baseband technique distortion compensation can be increased. The
memory effects are decreased by frequency- domain equalization in baseband. Different methods of
modulation are employed in communication systems where increase in efficiency of spectrum leads in
increase of PAPR and therefore, needs a power amplifier. Hence, linearizing amplifiers is becoming
interesting research. From all the methods of linearization more concentration is given to DPD as it
gains performance of good linearization without extra radio frequency components. A successive
method in solving this problem is employing predistortion method. The first thing can be made is to
minimize the effects of memory using DSP at baseband. An alternate is the overcoming distortion
compensation by applying suitable iterations until the sequence leads to zero or until linearity is
achieved.

• Mahmoud Abdelaziz et.al [14]: In this paper work, reduced complexity DPD solution is proposed. The
proposed DPD removes the unwanted distortions due to in-phase and quadrature modulator
impairments and power amplifier nonlinearity in un the transmitters are reduced by using the
bandwidth that is reduced by filtered basis functions. However, the estimation of parameters in DPD is
based on reduced bandwidth observation thus, by decreasing the overall complexity. In Massive
MIMO systems the digital predistortion can be used with large number of radio transmitters and
receivers and power amplifiers are employed where decreasing digital predistortion complexity is very
hard. However, estimation of closed loop parameter solution is constructed on decreasing the
bandwidth of the power amplifier output signal that is combined with simple correlation preprocessor.
In the earlier context the filtered basis function was used in ILA where only the nonlinearity of the
power amplifier is considered.

• Xin Liu et.al [15]: In this work paper, BO-DPD (Beam Oriented Digital Predistortion) method for
linearizing the power amplifiers is proposed which uses the Massive MIMO technique of hybrid
beamforming at the transmitters, which gains linearization in the direction of main beam of the
transmitted signal and the issue of hybrid beamforming array is reduced by implementing the digital
predistortion. In Massive MIMO hybrid beamforming, the digital predistortion implementation is not
suitable to linearize the power amplifiers because digital chains numbers are lesser than power
amplifier numbers that are used. Moreover, the beam oriented digital predistortion can solve the
problem by linearizing and constructing the signal of virtual main beam instead of the single power
amplifiers. As per the weights of beam forming for the array elements, the output of the power
amplifier to make an estimate the main beam for digital predistortion is combined with the feedback

891
Nat. Volatiles & Essent. Oils, 2021; 8(6): 888-904

path. An average estimation technique is proposed to enlarge the range of the linearized angle in the
direction of main beam

• Lei Ding et.al [16]: In this paper work, the RF power amplifiers uses a popular method called as digital
predistortion which is the used very widely in the current scenarios. In order to linearize the power
amplifiers, it is a cost-effective approach in various wideband signals due to this digital predistortion
technique is employed which working with less out of band spectrums is called as spectral regrowth.
The parameters are estimated in the digital predistortion in order to reduce the noise that is present at
the output of the power amplifier. The final aspect of the noise is to reduce the digital predistortion
performance resulting in high spectral regrowth. By applying the coefficient bias introduced by noise
measurement model reduces the spectral regrowth. This paper proposes two methods for reducing
these effects. The first method is to use the forward estimation followed by offline inverse noiseless
estimation and the other one is to use a specifically designed frame averaging with periodic training
signal. The spectral regrowth is reduced by 10dB is shown in the simulation results.

• Jing Zhang et.al [17]: In this paper, in the transmitter’s part of the communication system power
amplifiers are used which are inherently nonlinear device. To enhance the function of communication
channels digital predistortion is employed to decrease the nonlinearities in power amplifier. But most
of the architecture are based on the memory less power amplifier model. The signal of power
amplifier in the transmission cannot be ignored while using memory effects and the memory less
predistortion have the weak performance. In this work, a predistortion technique is analyzed which is
applied to the wideband wireless transmitter and is done by indirect learning architecture is proposed.
Then the performance is verified by simulation results of overall transmitter system. The measurement
shows that good AM/PM and AM/AM performance is obtained when input is a single carrier
Wideband Code Division Multiple Access signal with 5MHz bandwidth and after adding the proposed
predistortion method approximately 20dB improvements out off band at a frequency offset of 5MHz is
gained.

• Raz et.al [18]: In this paper work, Volterra filters are represented in the form of diagonal coordinate
which is designed a utilized to define the effective Volterra filter implementation for processing input
signals. The sum of filters that are linear that is applied to the input signals that are modified in the
diagonal coordinate which is represented as the output. Hence, different methods of linear filtering
are used to implement the nonlinear filter of the input in the baseband. Down sampling is employed to
decrease the complexity of computation. The similar method is used to emerge efficient
implementations for processing FDM input signals, PAM signals and continuous time carrier-based
signals. There are two dominant methods which reduces the complexity of Volterra filter. In the first
method using a cascade model which is contains memory less linearity’s in series with linear filters in
the Volterra filter is approximated. The cascade model outputs in the filter coefficients are nonlinear
and hence defining the nonlinear problem of estimation in the filter coefficients. In the next method
the product approximation which defines the Volterra kernel filter as a combination of linear products
of simple basis vectors. The estimation of basis vectors is a problem of linear estimation and hence,
conditions for uniqueness and optimality are established easily.

Power Amplifires
3.1 Functioning of Power Amplifiers

892
Nat. Volatiles & Essent. Oils, 2021; 8(6): 888-904

Power amplifiers (PA) are one of the very essential blocks in wireless systems, which are responsible of
overcoming the loss between the transmitter and receiver. In order to achieve the adequate transmission
power, PAs are most important in communication systems [1][5]. As the input power increases, high
efficiency levels can be obtained.
However, PAs are inherently non-linear when they are driven in high power modes, and may lead to a
trade-off among the efficiency and linearity. Moreover, latest techniques which are extensively used in
various wireless industries due to their high speed of communication, that are controlled by is using non-
constant envelope techniques. These techniques have high peak to average power ratio (PAPR) as the
frequency of spectrum efficiency is increased. For linear transmission of the symbols with high peak power,
the PA’s average output power needs to be several dB lower than the compression point which decreases
the overall efficiency [2]. Digital Predistortion (DPD) technique is widely used to overcome the problem of
non-linearity. To overcome this non-linearity - high efficiency problem one of the widely accepted
approaches is digital predistortion (DPD).

Heat out

in out
a)

DC power in
Amplifier Transmission system Accelerator

Fig. 2. Block diagram of Power Amplifier

The efficient power amplifiers exhibit an undesired nonlinear behavior as they are operating close to
saturation, generating a distortion spectrum, which will be wider than the allocated channel [16] and this
widening of the transmitted spectrum can be called as spectral regrowth. Polynomial functions of the
carrier frequency are harmonics and polynomial functions of the digital modulation located near the carrier
frequency are intermodulation.
In most RF transmitters, to limit the transmit bandwidth filter is placed after PA. The passband spans,
typically, several channels.
The out-of-band and in-band distortions are regulated by specifications associated with the modulation
format of the input signal. The distortion of the Out-of-band is measured by the amount of distortion
power into adjacent channels neighboring the allocated channel. In general, the specifications make out-of-
band distortion a key measurement of power amplifier linearity particularly for CDMA-based modulation
formats [6]. The distortion of in-band increases the error vector magnitude (EVM). The EVM is the root
mean square (RMS) difference between the demodulated signal prior to quantization and the desired
digital signal. It is a key measurement within OFDM systems, particularly when the individual subcarriers
are modulated using 64-QAM (quadrature amplitude modulation), and is degraded by impairments
including power amplifier nonlinearities [22].
Power amplifier models are of three types:

893
Nat. Volatiles & Essent. Oils, 2021; 8(6): 888-904

i. Saleh Model
ii. Rapp Model
iii. Ghorbani model

3.1.1 SALEH MODEL

The Saleh model is one of the efficient and quasi-memory less model which uses four parameters for fitting
the model to measurement data. Using the Saleh model, the improved estimation approach for modelling
the power amplifier can be done and Saleh model is mainly used to model the DPD in wireless system and
predicts the nonlinear distortion hence in this paper Saleh model is used. The methods of Amplitude-
Amplitude as well as Amplitude-Phase functions are defined by the following set of equations,
α r(n)
g (r(n)) = [1+βa 2 (1)
a ]r(n)

αψ r(n)
f(r(n)) = [1+βψ ]r(n)2
(2)

where,αa ,βa ,αψ ,βψ are model’s parameter

3.1.2 RAPP MODEL

The Rapp model models the amplitude distortion only and uses three parameters. This technique is quite
simple to use and it adopts the performance of power amplifier to be linear until it reaches the saturation
level. The amplitude-amplitude modulation conversion general expression is as follows,

Where the parameter for setting the smoothness of the transitioning to saturation state from linear state is
s and parameters for setting the output saturation level is Osat, the smaller the s the smoother the
transition.

3.1.3 GHORBANI MODEL

In order to fit the model parameters, the Ghorbani model uses 8 parameters. This is same as the Saleh
model but this model cannot be used to model the digital predistortion model. This model is also a simple
and efficient quasi-memory less system which has its Amplitude-Phase modulation and Amplitude-
Amplitude modulation conversions defined in the following equations,

Where,𝑥1, 𝑥2, 𝑥3, 𝑥4, 𝑦1, 𝑦2, 𝑦3 ,𝑦4 are model’s parameters.

894
Nat. Volatiles & Essent. Oils, 2021; 8(6): 888-904

3.1.4 3.1.4 PA MODEL IN MATLAB/SIMULINK

The power amplifier model is designed in the MATLAB/SIMULINK tool and the models are shown below.

Fig. 3. (a) Simulink model of Power Amplifier

(b) Subsystem model of Bandlimited Source

(c) Subsystem model of PA reference

Fig.3. (a) illustrates the main Simulink Power Amplifier model which have the two subsystem models
Bandlimited Source and PA reference model fig (b) refers to the bandlimited sources model and fig (c)
refers to the Power Amplifer reference model.
The power amplifier model comprising of a Saleh amplifier in series with an asymmetrical complex filter as
shown in fig (c). The excitation is an additive white noise (AWGN) signal that has been low pass filtered as
shown in fig (b). The input and output signals are shown in below figure.

895
Nat. Volatiles & Essent. Oils, 2021; 8(6): 888-904

Fig. 4. (a) Simulation result of Power Amplifier with Saleh Model

(b)Simulation result of Power Amplifier with Rapp Model

896
Nat. Volatiles & Essent. Oils, 2021; 8(6): 888-904

(c) Simulation result of Power Amplifier with Ghorbani Model

From the above simulation results we can observe that the Rapp model output in fig.4. (b) becomes linear
only at the saturation point. Hence the other part of the output will remain non-linear. In next simulation
result of Ghorbani model in fig.4. (c) the output obtained is non-linear but this model cannot be used to
model the DPD model. The Saleh model simulation result shown in fig.4. (a) gives the non-linear output
where the output is not similar to input. In order to remove this non-linearity, we use DPD model and
obtain the linear output. The Saleh model is mostly used in RF Amplifiers so in order to design the DPD
Saleh model is designed.

3.2 Digital Predistortion

Digital Predistortion is an efficient approach for reducing the amplifier distortion. The predistorter
linearizes the power amplifier by generating a non- linear transfer function i.e., inverse to the power
amplifier in such a way that when it comes before the power amplifier the overall amplification of the
system is nearly linear. Amplifier magnifies the input signal and generates a larger output signal [7]. Ideal
amplifier will enhance the signal quality in the identical degree irrespective of the signal levels. This
different amplification level depending on signal level results in the signal distortion which may cause
several undesired outcomes.

Fig.5. Digital Predistortion (DPD) block diagram

The block diagram of Digital predistortion shown in fig.5 is a technique to reduce the amplifier. The
predistorter linearizes the power amplifier by generating a non-linear transfer function i.e., inverse to the
power amplifier in such a way that when it comes before the power amplifier the overall amplification of
the system is nearly linear [15] [13]. The generation of linear amplification is obtained as the Predistorter
generates an inverse transfer function before the amplification. There are different techniques used in
Digital Predistortion they are LUT based digital predistortion, memory less polynomial digital predistortion
and memory polynomial digital predistortion. Memory less polynomial digital predistortion contains only
the present input data and output depends only on the current input. The Memory polynomial digital

897
Nat. Volatiles & Essent. Oils, 2021; 8(6): 888-904

predistortion contains both the current and previous input data and output depends both on the present
and past input. In this paper Memory Polynomial digital predistortion algorithm [13].

3.3 DPD MODEL IN MATLAB/SIMULINK

The below figure. 6 illustrates the design of Model Digital Predistortion.

Fig. 6. Simulink model of Digital Predistortion

The Simulink model of DPD is same as the basic DPD as shown in fig.5. The DPD block is placed before the
PA which generates the inverse transfer function for power amplifier in order to reduce the non-linearity.
The detailed model of DPD is given in the next sections where the DPD is modelled using ILA and Memory
Polynomial DPD Algorithm [8].

3.3.1 INDIRECT LEARNING ARCHITECTURE (ILA)

Firstly, it is introduced to instruct the neural network controllers and there after it is accepted for DPD
generation [10]. ILA is most commonly used method to recognize the digital predistorter parameters. ILA is
established on the inverse modelling technique, using output signal of the power amplifier to model the
power amplifier input. Once the postdistorter of the power amplifier (also known as post-inverse) has been
done, the post distorter parameters will be loaded to the same model used as the predistorter [11].

Fig. 7.Predistortion by Indirect Learning Architecture

The above figure 7 explains about the Indirect Learning Architecture of digital predistortion. The
predistortion block is connected to the power amplifier, the input of the predistortion is given as u(n) and

898
Nat. Volatiles & Essent. Oils, 2021; 8(6): 888-904

predistortion output is x(n). The output of the predistortion is inputted to the power amplifier, the power
amplifier output along with the gain v(n) is added and is connected to the post distortion which is the copy
of predistortion. In order to estimate the coefficient parameters, copy of predistortion called as
postdistortion is introduced. In the block of Predistortion the Memory Polynomial algorithm is used. The
below figure 8 shows the Simulink model of ILA which is also called as Adaptive digital predistortion the
name adaptive because the power amplifier output is connected as input to the digital predistortion in
feedback form and adaptive algorithm is used. Hence the name Adaptive Digital Predistortion [14].

Fig. 8. Simulink Model of Adaptive Digital Predistortion

3.4 MEMORY POLYNOMIALS

Memory polynomial comprises of several function related delay taps and static functions of nonlinearity.
It’s a part of general Volterra series comprising of its principal diagram parameters which are the
minimized.
Figure 1 represents the generic digital predistortion scheme, with u(n) being the input to the predistortion
is u(n) whose output x(n) in turn will be fed to the power amplifier for generating the output y(n). The M-
tap memory nonlinearity defined by Volterra series, comprises of sum of multi-dimensional convolutions
which can be represented as
y(n) = ∑𝐾𝑦𝑘(𝑛) (6)

(7)
Where, is the kth dimensional convolution of the input with Volterra kernel ℎ𝑘. It can be a general form of a
power series represented with a finite memory of length M [8].
The generalized form of memory polynomial equation can be written as

899
Nat. Volatiles & Essent. Oils, 2021; 8(6): 888-904

For the narrow band case, if we choose the combinations of the x(n)|𝑥(𝑛 − 𝑚)|𝑘−1, we obtain at the Kims
model[19], which is referred to as memory polynomial[20].

This model of memory polynomial has considered to efficient in the case of predistortion of actual power
amplifiers under some operating conditions.
The Simulink model of the memory polynomial DPD which is a subsystem of DPD is as shown in below
figure 9 (a) (b).

Fig. 9. (a) Simulink model of Memory Polynomial DPD

Fig. (b) Subsystem model of Delay tapped lines of fig (a)

Here in above figure 9 (a) and (b) the memory polynomial coefficients and the nonlinear coefficients are
modelled according to equation (9).

3.4.1 ESTIMATION ALGORITHM

Various algorithms based on lest square are available to estimate the model coefficients which may be
available as linear weightage of non-linear signals [6][18] which is the easiest way to formulate such a
problem. Initially the coefficients are collected into one Jx1 vector and can named as “w”, with J being total
number of coefficients. Every item of the coefficient vector w is mapped with a signal whose samples of
samples for a period, say n = 1,2,….N, are accumulated into a vector. Combining all those vectors ina a
matxi X of the size NxJ the output of the proposed model can be expressed as
𝑦 = 𝑋𝑤 (10)
where𝑦 is a Nx1 vector that represents an estimate of the vector y.
For the technique of predistortion which uses inverse modeling shown in figure 8 we can rewrite equation
(10) as

900
Nat. Volatiles & Essent. Oils, 2021; 8(6): 888-904

𝑥 = 𝑌𝑤 (11)
where the input is an estimate of the sampled output and Y is constituted same as X with x(n- m). This is
the actual estimate used predistortion based scenarios.
From fig. 8 the error of estimation can be written as
e(n) = X(n) – x(n) (12)
The block of N samples in vector form can be written as
e = X– x (13)
Now the least squares solution is simple to write which minimizes ||𝑒||2 as follows:
w = (𝑌𝐻𝑌)−1𝑌𝐻x (14)
The inverse of JxJ sample covariance matrix can be represented as 𝑌𝐻 Y, and can be computed by obtaining
linear equations solution as
𝑌𝐻Yw = 𝑌𝐻x (15)
From the above equations the Simulink model for estimation algorithm is modelled as follows

Fig. 10. (a) Simulink subsystem model of Dpd_Coef from fig 8

Fig. 10. (b) Simulink subsystem model of dpd_compute from 10(a) (estimation algorithm model)
Results

901
Nat. Volatiles & Essent. Oils, 2021; 8(6): 888-904

Fig. 11. Simulation result of Power Amplifier With DPD

Fig. 12. Simulation result of Real and Imaginary coefficients of DPD

From fig 4 (a) and fig 11 we can clearly observe that the output without DPD have more signal distortion
and output is not similar to input. From fig 11 we observe that output with is almost similar to power
amplifier input and is linear. The output spectrum with delay taps order M = 4 and memory polynomial with
nonlinearity order K = 5 with 20 coefficients in total gives the linear output. Hence the memory polynomial
DPD algorithm is most preferable algorithm for wireless networks in order to linearize the power amplifiers.

Conclusion
Linear power amplifiers and high efficiency are vital factors in wireless communications Scenarios. Though
there seems to be a trade-off among linearity and efficiency, the highly efficient power amplifiers will show
some non-linearity and the power amplifiers with linear behaviour show lesser efficiency. To accommodate
both the efficiency as well as the linearity requirements the digital predistortion is normally incorporated.
The proposed work discussed in this paper has considered various aspects of the linear of Pas. The

902
Nat. Volatiles & Essent. Oils, 2021; 8(6): 888-904

implementation of a predistorter has been done in two phases. In the first phase the pre-distorter model is
designed and the parameters for this model is identified in the second phase. ILA technique have been
chosen for the identification of the parameters and this ILA technique allows a greater impact and
controlling of the output signal from the linearized power amplifier. The implementation has several
challenges and one of them being addressing the memory effects by the power amplifiers for wideband
signals or higher power amplifiers. Whatever combination of DPD function and estimator structure passes
the specification with sufficient margin for the manufacturing process is acceptable.

REFERENCES

GopinathGampala, C.J. Reddy – “Massive MIMO Behind 4G and Basis of 5G”, International Applied
Computaional Electromagnetics Society Symposium, 2018
R. Neil Braithwaite – “General Principles and Design Overview of Digital Predistortion”, Cambridge
University, 2011
P.B.Kenington, “High – Linearity RF Amplifier Design” Boston MA; Artech House, 2000
A.E.Nordsjo and L.H.Zetterberg, “Identification of Certain Time-Varying Nonlinear Wiener and Hammerstein
Systems” , IEEE Transactions on Signal Processing, Mar 2001.
Aliya Sheik, ManinderJeetKaur – “Comprehensive in Survey of Massive MIMO Communications”, Advances
in Science and Engineering Technology International Conferences, 2019.
Marina Jordao, Nuno Borges Carvlho – “MIMO Antenna Transmitting Characterization”, IEEE MTT – S
International Microwave Workshop Series on 5G Hardware and System Technologies, 2018.
D.R. Morgsn, Zhengxiang MA, Jaehyeong Kim, Michael G. Zierdt, and John Pastalan, “A Generalized
Memory Polynomial Model for Digital Predistortion of RF PowerAmplifiers”, IEEE Transactions on
signal processing, October 2006.
Yasunori Suzuki and ShoichiNarahashi – “Power Amplifier Configuration for Massive – MIMO Transmitter”,
the 22nd Asia – Pacific Conference on Communications, 2016.
Jianxin Jing, Chao Yu – “Digital Predistortion Technique for 5G Massive MIMO Transmitters”, Proceedings of
the 1st European Microwave Conference in Central Europe, May 2019.
MouradDjamai, SmailBachir, Claude Duvanaud, “Behavioral Modeling And Digital Predistortion of RF Power
Amplifiers”, IEEE Transactions on Microwave Theory, May 2006.
Dian Xie, ZhibinZeng – “An Improved Adaptive Algorithm for Digital Predistortion”, International
Conference on Computer Science and Electronics Engineering, 2012.
P. Varahram, Z. Atlasbaf – “Adaptive Digital Predistortion for High Power Amplifiers with Memory Effects”,
Department of Electronic and Electrical Engineering TarbiatModares University, 2012.
D. Bondar and D. Budimir – “Digital Baseband Predistortion of Wideband Power Amplifiers with Improved
Memory Effects” - Wireless Communications Research Group, University of Westminster, 2009.
Mahmoud Abdelaziz, LauriAnttia and MikkoValkama – “Reduced – Complexity Digital Predistortion for
Massive MIMO”, Tampere University of Technology, Tampere, Finland, IEEE Transactions, 2017.
Xin Liu, Qian Zhang – “ Beam – Oriented Digital predistortion for 5G Massive MIMO Hybrid Beam- forming
Transmitters”, IEEE Transactions on Microwave Theory and Techniques, July 2018
D.R.Morgan, Z Ma and L Ding, “Reducing Meausrements Noise Effects in Digital Predistortion of RF Power
Amplifiers”, IEEE Transactions International Conference, 2003.
Jing Zhang, Songbai He, Shirong Yin – “A Memory Polynomial Predistortion for Compensation of Non-
Linearity with Memory Effects in WCDMA Transmitters”, IEEE Transactions on Microwave Theory,
2009.

903
Nat. Volatiles & Essent. Oils, 2021; 8(6): 888-904

Y. W. Fang L. C. Jiao, X.D. Zhang and J.Pan, “Analysis of Conjugate Gradient Algorithms for Adaptive
Filtering”, IEEE Transactions on Signal processing, Feb 2000.
J. Kim and K. Konstantinu, “Digital Predistortion of Wideband Signals Based on Power Amplifier with
Memory Effects”, Electron.Lett, Nov 2001
L.Ding, G.T. Zhou, D.R. Morgan, Z. Ma, J. S. Kenny, J. Kim and C.R. Giardina, “A Robust Digital Baseband
Predistorter Constructed using Memory Polynomials”, IEEE Transactions on Communication, Jan 2004.
Jessica Chani-Cahuana, Christian Fager, Thomas Eriksson, “A New Variant of the Indirect Learning
Architecture for the Linearization of Power Amplifiers”, Proceedings of 10th European Microwave
Integrated Circuits Conference, 2015.
Annika Esther Mayer – “ Comparison of Digital Predistortion Identification Methods for RF Power
Amplifiers”, Department of Signals and Systems Division of Communication Engineering Chalmers
University of Technology, Sweden, 2013.

904