International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169

Volume: 5 Issue: 5 389 392

_______________________________________________________________________________________________
Power Amplifier Linearization using Singular Value Decomposition Algorithm

Er. Sanjeev Goyal1 Er. Harpreet Kaur2 Er. Harisharan Aggarwal3

Research Scholar Assistant Professor Head of Department
Deptt: Electronics & Communication Deptt: Electronics & Communication Deptt: Electronics & Communication
Engineering, Guru Gobind Singh College Engineering, Guru Gobind Singh College Engineering, Guru Gobind Singh College
of Engg. & Technology, Guru Kashi of Engg. & Technology, Guru Kashi of Engg. & Technology, Guru Kashi
University, Talwandi Sabo University, Talwandi Sabo University, Talwandi Sabo
Bathinda (PB) Bathinda (PB) Bathinda (PB)
[email protected] [email protected] [email protected]

Abstract---The power amplifier presents itself as a significant device towards the improvement of system. But itself it is the main source of
nonlinearities present in the transmission chain. This effect is appropriate as it produces a spectral regrowth consequently leaking energy into
adjacent channels. To improve the power amplifier efficiency without compromising its linearity, power amplifier linearization is essential. A
common solution to avoid this effect is to drive the PA far from its compression zone, a technique known as back-off.

Keywords---Power Amplifier, Singular Value Decomposition, Quadrature Rotation, Adjacent Channel Leakage Ratio.
__________________________________________________*****_________________________________________________

I. INTRODUCTION used for linearization of power amplifiers [6]. Amplifier has

Power amplifier linearization is currently one of the most a constant gain and phase response for an input power
promising techniques for linearity and efficiency region then the amplifier is said to be linear for this region.
improvement in mobile communication systems. There are Efficiency: Efficiency is another key issue in mobile
numerous techniques which have different levels of communications especially for battery operated mobile
complexity, various advantages and limitations. Different terminals. It has two widely used definitions, drain (or
linearization methods may fit to different communication collector) efficiency and PAE (Power Added Efficiency).
systems [1]. For instance more complicated high Drain efficiency is the ratio of output radio frequency (RF)
performance systems may be used for base station Pas power to input DC power [5].
whereas the systems usable in handsets should have low
complexity, low cost and high efficiency. Although in II. SINGULAR VALUE DECOMPOSITION
general the main reason to implement these systems is to ALGORITHM
linearize the PA, they improve also the efficiency because a Define A R mn , a non-negative square root of AT A .
linearized PA can be driven closer to compression characteristic value are referred as the singular values of A,
A .When A is
(operation with low back-off). Power amplifiers features
singular value of the whole referred as
Gain and Output Power: In mobile communications each
system has its specifications which must be fulfilled. mn T
H
a complex matrix C , A A is expressed as A A .
Obtaining output powers high enough for various Theorem 1 (Singular Value Decomposition Theorem): If
applications is a very important task achieved by Power mn
matrix A R , there must be two orthogonal matrixes
Amplifiers. Commonly the information signal is first
[8]
modulated and up converted and then sent to a Power
Amplifier. This input is multiplied with a gain factor and the U u1 , ,um R mn
(1)
V V1 , ,Vn R
desired output power is obtained[2][3]. mn
(2)
Linearity: Linearity is one of the key issues in Power
Amplifiers used in new generation mobile communication And U and V satisfy the following equation:

systems. The linearity of a Power Amplifier is easily visible r
0
in its gain and phase characteristics. There are many U AV
T r

linearization methods used for power amplifiers which give 0 0 mr
(3)
better linearity but at the cost of poor efficiency i.e. the
Where
method which give good linearity with efficiency should be

389
IJRITCC | May 2017, Available @ http://www.ijritcc.org
_______________________________________________________________________________________
International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169
Volume: 5 Issue: 5 389 392
_______________________________________________________________________________________________
r diag 1 , , r ,1 r 0 (4) e n z n z n (12)

i : the singular values of A

Using the least squares theory, the optimal solution for the
ui : i-th left singular vector memory polynomial coefficients of the predistorter
vi : i-th right singular vector training module is:
a Y H Y Y H z
1
mn
As for complex matrix A C , just put U, V into (13)
unitary matrix, Theorem 1 still holds. For dimension
reduction, the singular value decomposition of the matrix The LS algorithm steps are based on SVD decomposition as
Amn can be understood as a linear transformation. When n- follows:
dimensional space is mapped to r-dimensional space, the
singular vectors i u and i v represent linear transformation According to the observation sampling data
direction or projection direction, and the values of i are obtained from the amplifier output terminal,
equal to projection of variance to the original vectors construct matrix Y, and with the ideal predistortion
signal z(n) configure desired vector z.
projection direction ui or vi . The number of singular
values depends on the balance between performance and
Calculate singular value decomposition of matrix
complexity. Using the minimum mean square error criterion,
Y , yielding the rank of matrix Y H Y nonzero
minimize the squared error of the LS algorithm to train
predistortion device parameters. Predistortion of the input- singular values 1 , 2 , , K , and
output relationship, for convenience of description, corresponding right singular vectors q1 ,q2 , ,qk
transcribe into matrix form [9]:
Calculate the coefficients a .
z
z Xa (5)

2 i
q , Y H z
K
qi H
Where: a
i 0 i

z z 0 , ,z N 1
T
(6) [14]
T
a a1,0 ,a3,0 , ,aK ,0 , ,a1,Q , ,aK ,Q (7) III. RESULTS AND DISCUSSION
Define: To validate the performance of the QR and SVD algorithm
k 1 in PA modelling using the memory polynomial model value
xkq n x n q x n q (8) of memory length and the nonlinearity order has been taken
as 2 and 7 respectively. Figures are presented for the AM-
Then AM characteristics and other characteristics of actual PA
and proposed PA model. The results show the accuracy of
the PA modeling in terms of AM-AM and AM-PM
X x1,0 ,x3,0 , ,xk ,0 ,x1,Q , ,xK ,Q (9)
characteristics. The performance of proposed PA model
based on ACLR (dB), EVM (dB), memory length and non
Applied to the predistorter training module A,the input- linear order can be inferred from table 1.
output relationship is given by:
k 1 The ACLRs of actual PA are Lower Channel 2 (-32.9432),
K Q
y n q y n q
z n akq (10)
Lower Channel 1 (0.8227), Upper Channel 1 (0.6994) and
k 1 q 0 G G Upper Channel 2 (-33.0136). The ACLRs of modeled PA
using QR are Lower Channel 2 (-32.9431), Lower Channel
Equation can be written in the form of matrix:
1 (0.8227), Upper Channel 1 (0.6994) and Upper Channel 2
(-33.0136). Similarly, the ACLR values of modeled PA
z Ya (11)
using SVD are Lower Channel 2 (-32.9431), Lower Channel
1 (0.8227), Upper Channel 1 (0.6994) and Upper Channel 2
The estimated error between the output of the training
(-33.0136). Thus both QR and SVD algorithm have
module z n and the output of the predistorter z(n) is: performed same in modelling the PA characteristics.

390
IJRITCC | May 2017, Available @ http://www.ijritcc.org
_______________________________________________________________________________________
International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169
Volume: 5 Issue: 5 389 392
_______________________________________________________________________________________________
TABLE 1 MEASUREMENTS OF ACTUAL PA AND
PA MODELLED USING QR AND SVD ALGORITHM
Adjacent Channel Leakage Ratio (dB) Error
Vector
Type Lower Upper
Lower Upper Magnitu
Channel Channel 1 Channel 1 Channel de (RMS)
2 2
Actual
-32.9432 0.8227 0.6994 -33.0136 -98.2509
PA

QR PA -32.9431 0.8227 0.6994 -33.0136 98.2509

SVD
-32.9431 0.8227 0.6994 -33.0136 98.2509
PA

The EVM of actual PA and PA modelled using QD and

SVD algorithms have been calculated as -98.2509, 98.2509
and 98.2509 respectively.
Fig. 3 DPD Characteristics using SVD algorithm

Fig. 1 AM/AM characteristics of PA modelled using SVD Fig. 4 Power Spectral Density DPD/PA modelled using
algorithm SVD algorithm

Fig. 5 Linearized Characteristics using SVD Algorithm

Fig. 2 Power Spectral Density of actual PA Input/Output IV. CONCLUSION

and PA modelled using SVD algorithm This paper presents the problematic of implementing a
linearization technique to improve the efficiency of power
amplifiers. The technique is implemented in the frequency
391
IJRITCC | May 2017, Available @ http://www.ijritcc.org
_______________________________________________________________________________________
International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169
Volume: 5 Issue: 5 389 392
_______________________________________________________________________________________________
domain and using a feedback path, which is needed for SVD
algorithm, with an observational bandwidth which is smaller
than the distorted signal. Therefore to improve the power
amplifier efficiency without compromising its linearity is
likely to be addressed.

REFRENCES

[1] Ai B, Zhong Z D, Zhu G, et al. A Novel Scheme for Power

Amplifier Predistortion Based on Indirect Leaning
Architecture. Wireless Personal Communications, 46(4):523-
530,2008.
[2] Bergqvist, G. Larsson , E. G. The Higher-Order Singular Value
Decompsition: Theory and an Application[J], IEEE Signal
Processing,, vol. 27(3):151-154,2010.
[3] Guan Lei, Optimmized Low-Complexity Implementation of
Least Squares Based Model Extraction for Digital Predistortion
of RF Power Amplifiers, IEEE trans. On Microwave Theory
and Tech., vol. 60(3):594-603,2012.
[4] Biyi Lin1, 2 Yide Wang 1 Bruno Feuvrie1 Qingyun Dai 2, A
Novel Power Amplifier Linearization Technique based on Pre-
distortion Principle and Wiener model, International Journal of
Advanced Science and Technology Vol. 22, September, 2010.
[5] J. Stevenson Kenney, Senior Member, IEEE, and Jau-Homg
Chen, Power Amplifier Linearization and Efficiency
Improvement Techniques for Commercial and Military
Applications, IEEE-2012.
[6] Felice Francesco Tafuri*, Cataldo Guaragnella#, Marco Fiore
and Torben Larsen, Linearization of RF Power Amplifiers
Using an Enhanced Memory Polynomial Predistorter, 978-1-
4673-2223-2/12/$31.00 IEEE, 2012.
[7] Nikolai Wolf, A Novel Model for Digital Predistortion of
Discrete Level Supply-Modulated RF Power Amplifiers. IEEE
Microwave and Wireless Components Letters, Vol 26, No. 2, Pp
146-148, 2016.
[8] MeenakshiRawat,Fadhel M. Ghannouchi, and Karun Rawat.
Three-Layered Biased Memory Polynomial for Dynamic
Modeling and Predistortion of Transmitters With Memory.
IEEE Transactions on Circuits and Systems-I: Regular Papers,
2013 60(3). pp 768-777,2013.
[9] Harjinder Singh and Amandeep Sappal, Modeling and
compensation of thermal effects in power amplifiers. Proc. of
IEEE Conference in Computing and Communication
(IEMCON), 2015 Vancouver, pp. 1 7.2015.

392
IJRITCC | May 2017, Available @ http://www.ijritcc.org
_______________________________________________________________________________________