Low complexity methodology for complex square-root computation

NXFEE INNOVATION
(SEMICONDUCTOR IP &PRODUCT DEVELOPMENT)
(ISO : 9001:2015Certified Company),
# 45, Vivekanandar Street, Dhevan kandappa Mudaliar nagar, Nainarmandapam,
Pondicherry– 605004, India.
Buy Project on Online :www.nxfee.com | contact : +91 9789443203 |
email : nxfee.innovation@gmail.com
_________________________________________________________________
Low-Complexity Methodology for Complex Square-Root Computation
Abstract:
In this brief, we propose a low-complexity methodology to compute a complex square
root using only a circular coordinate rotation digital computer (CORDIC) as opposed to
the state-of-the-art techniques that need both circular as well as hyperbolic CORDICs.
Subsequently, an architecture has been designed based on the proposed methodology and
implemented on the ASIC platform using the UMC 180-nm Technology node with 1.0 V
at 5 MHz. Field programmable gate array (FPGA) prototyping using Xilinx’ Virtex-6
(XC6v1x240t) has also been carried out. After thorough theoretical analysis and
experimental validations, it can be inferred that the proposed methodology reduces
21.15% slice look up tables (on FPGA platform) and saves 20.25% silicon area overhead
and decreases 19% power consumption (on ASIC platform) when compared with the
state-of-the-art method without compromising the computational speed, throughput, and
accuracy.
Software Implementation:
 Modelsim
 Xilinx 14.2
Existing System:
Complex numbers have been used significantly in scientific community for the real-time
data representation and system modeling, including electronic circuits, electromagnetism,
communication systems, and signal processing algorithms. However, existing real valued
square-root computation methods cannot be used directly to compute complex square
root without requiring additional hardware. On the other hand, the state-of-the-art
architecture for complex square-root computation was designed using the coordinate

NXFEE INNOVATION
_________________________________________________________________
rotation digital computer (CORDIC) involving two circular and one hyperbolic
CORDICs. However, a hyperbolic CORDIC requires more iterations to obtain the same
precision and accuracy when compared with a circular CORDIC, resulting in more
computational complexity in terms of power consumption and silicon area overhead
when implemented on chip. Motivated by the fore mentioned facts, we introduce here in
this brief a low-complexity methodology for computation of the complex square root
using only two circular CORDICs unlike the state-of-the-art method where a hyperbolic
CORDIC is also necessary.
Disadvantages:
 High complexity
 Speed and throughput is low
 Power consumption is high
 Accuracy is low
Proposed System
Proposed methodology and architecture
Consider a complex number z = p + jq, whose magnitude is . From, the
proposed methodology has been divided into three steps. The first step is Cartesian to
polar conversion, which computes the magnitude (R) and arctangent (ψ), as shown in Fig.
1(a). To compute R and ψ, consider the inputs to the CORDIC as x0 = p and y0= q, and
operate the CORDIC in the vectoring mode until the final y-component becomes zero,
which can be expressed as follows:

NXFEE INNOVATION
_________________________________________________________________
Consider |R| ≤ 1. If |R| > 1, it can be easily scaled down to 1 or less by performing a
simple shifting operation. For example, if 2(l-2)
< R ≤ 2 l
where l is an even number, the |R|
is scaled down to less than 1 by right shifting by l bit. After this shifting, when √R
computation is over, the final value is left shifted by (l/2) bit to get the actual value. If
Rlower < R ≤ Rupper , the l value can be computed from R using Table I, where Rlower and
Rupper are the lower and upper boundaries of R, respectively. For example, considering R
= 23, from Table I as 16 < R ≤ 64, then l will be equal to 6. After right shifting R by 6 bit,
R will become 0.359375 and = 0.5994, which is then readjusted to the actual value
= 4.7952 by left shifting 0.5994 by (l/2) = 3 bit
Fig. 1. Geometrical representation of proposed methodology. (a) Cartesian to polar conversion. (b)
Square-root computation. (c) Polar to Cartesian conversion.
Table I
Computation of l From R (b = Word length)

NXFEE INNOVATION
_________________________________________________________________
Proposed Architecture
Fig. 2 shows the architecture designed based on the proposed methodology, as described.
Here, unlike the state of-the-art design, the proposed architecture has been implemented
by reusing only the circular CORDIC, which eliminates the requirement of the hyperbolic
CORDIC and makes the architecture less-complex. The detailed hardware complexity
analysis is given

NXFEE INNOVATION
_________________________________________________________________
Fig. 2. (a) Without pipelining. (b) Doubly pipelined architecture with a detailed signal flow for the
proposed methodology
Table II
Microrotation table for computation (θ1 + θ2/2)
To enhance the speed of the architecture designed based on the proposed methodology,
we use here the concept of doubly pipelining (DP)—a technique where intermediate
microrotations are directly fed from the vectoring to the rotation-mode circular CORDIC
immediately after these are computed, eliminating the need to wait until the overall angle
is computed explicitly after the entire vectoring is done [see Fig. 2(a) and (b)]. Detailed
discussion on DP is although omitted here due to paucity of page (can be referred for the
same).
However, to apply DP, microrotations should be computed or made available on-the-fly.
From, and Fig. 1(b), doubly pipeline can be used in the design to increase the
computational speed, but the microrotations are not available for angle φ. Hence, a
procedure is introduced here to compute the microrotations for φ from 2φ. Considering
two angles θ1 and θ2 with microrotations μ1 and μ2, respectively, and assuming
microrotation 0 and 1 corresponding to clockwise and anticlockwise directions,
respectively, microrotation μ12 = (θ1 + θ2/2) can be computed as shown in Table II.
Therefore, the microrotations for φ from 2φ can be computed by considering θ1 = 2φ and
θ2 = 0. Similarly, from and Fig. 1(c), the microrotations are required for (ψ/2) instead of

NXFEE INNOVATION
_________________________________________________________________
ψ, which can be calculated by considering θ1 = ψ and θ2 = 0. Hence, to meet the above
requirement, as shown in Fig. 1(b) and (c), the rotation-mode CORDIC is designed for
rotating the given vector by angle (θ/2) instead of θ using microrotations shown in Table
II. The architecture is divided into two modules— Vectoring and Rotation—as shown in
Fig. 3. The pseudocode for the proposed architecture is given in Fig. 4. In the vectoring
module,
Fig. 3. Architecture designed based on the proposed methodology

NXFEE INNOVATION
_________________________________________________________________
Fig. 4. Pseudocode for architectural implementation based on the proposed methodology.
the circular vectoring-mode CORDIC (CVCORDIC) has been used and it takes one of
two inputs based on the selection line s0. When s0 = 0, the input vector to the
CVCORDIC [x0, y0]=[p, q]. The CVCORDIC will rotate the input vector until y-
component becomes zero. Then outputs of the CVCORDIC are = R
and the microrotations μ will correspond to the angle ψ = tan−1(q/p). The output of this
module will become ψ μ = μ. To compute the scaling value l, a combinational circuit is
designed using Table I, which is called the scaling determiner, as shown in Fig. 3. The
magnitude R will be shifted by l bit to right to bring the R value less than 1. Now, (2R−1)
can be computed by shifting R 1-bit left and then subtracting 1. The final outputs of the
circular CORDIC and the hyperbolic CORDIC are needed to be multiplied with CORDIC

NXFEE INNOVATION
_________________________________________________________________
scaling factors Kc = 1.646760258121 and Kh = 1.207497067763095, where c and h
denote circular and hyperbolic CORDIC, respectively. When s0 = 1, the input vector to
CVCORDIC [x0, y0] will be [1, 0]. Now the CVCORDIC rotates the input vector until y-
component becomes (2R−1). The output of CVCORDIC microrotation μ corresponds to
angle 2φ. Then the vectoring module output is φμ. The microrotations from the vectoring
module ψμ, φμ will be inputs to the rotation module. Circular rotation mode CORDIC
(CRCORDIC), like the vectoring mode, is used that takes one of the two inputs based on
the selection line s1 like the vectoring module. When s1 = 0, the input vector to
CRCORDIC [x0, y0]=[1, 0] and the input microrotation μ = φμ. The outputs of
CRCORDIC are xn = cos φ = and yn = sin φ = . Now, will be
brought to its original value by shifting (l/2) bit to left. When s1 = 1, the input vector to
CRCORDIC [x0, y0]=[√R, 0] and input microrotations μ = ψμ. Then the outputs of
CRCORDIC are xn = √R cos(ψ/2) = a and yn = √R sin(ψ/2) = b
Advantages:
 Low complexity methodology
 Speed and throughput is high
 Power consumption is low
 Accuracy is high
References:
[1] J. Xiang, L. Guo, Y. Chen, and J. Zhang, ―Study of GPS adaptive antenna technology based on
complex number AACA,‖ in Proc. WiCOM, Oct. 2008, pp. 1–4.
[2] M. Sima, M. Senthilvelan, D. Iancu, J. Glossner, M. Moudgill, and M. Schulte, ―Software solutions
for converting a MIMO-OFDM channel into multiple SISO-OFDM channels,‖ in Proc. WiCOM, Oct.
2007, p. 9.

NXFEE INNOVATION
_________________________________________________________________
[3] K.-I. Ko and F. Yu, ―On the complexity of computing the logarithm and square root functions on a
complex domain,‖ J. Complex., vol. 23, no. 1, pp. 2–24, 2007.
[4] D. Wang and M. D. Ercegovac, ―A design of complex square root for FPGA implementation,‖ Proc.
SPIE, vol. 7444, p. 74440L, Sep. 2009.
[5] D. Wang, N. Zheng, and M. D. Ercegovac, ―Design of high-throughput fixed-point complex
reciprocal/square-root unit,‖ IEEE Trans. Circuits Syst. II, Express Briefs, vol. 57, no. 8, pp. 627–631,
Aug. 2010.
[6] W. Kahan, ―Branch cuts for complex elementary functions or much ado about nothing’s sign bit,‖ in
The State of the Art in Numerical Analysis. Oxford, U.K.: Clarendon Press, 1987, ch. 7.
[7] X. Wang, Y. Zhang, Q. Ye, and S. Yang, ―A new algorithm for designing square root calculators
based on FPGA with pipeline technology,‖ in Proc. 9th Int. Conf. Hybrid Intell. Syst., vol. 1. 2009, pp.
99–102.
[8] M. Ye, T. Liu, Y. Ye, G. Xu, and T. Xu, ―FPGA implementation of CORDIC-based square root
operation for parameter extraction of digital pre-distortion for power amplifiers,‖ in Proc. WiCOM, Sep.
2010, pp. 1–4.
[9] I. Park and T. Kim, ―Multiplier-less and table-less linear approximation for square and square-root,‖
in Proc. IEEE ICCD, Oct. 2009, pp. 378–383.
[10] T. Sutikno, ―An efficient implementation of the non restoring square root algorithm in gate level,‖
Int. J. Comput. Theory Eng., vol. 3, pp. 46–51, Feb. 2011.
[11] T. Sutikno, A. Z. Jidin, A. Jidin, and N. R. N. Idris, ―Simplified VHDL coding of modified non-
restoring square root calculator,‖ Int. J. Reconfigurable Embedded Syst., vol. 1, pp. 37–42, Mar. 2012.
[12] R. V. W. Putra, ―A novel fixed-point square root algorithm and its digital hardware design,‖ in
Proc. Int. Conf. ICT Smart Soc., Jun. 2013, pp. 1–4

Low complexity methodology for complex square-root computation

Related slideshows

More Related Content

Low complexity methodology for complex square-root computation