Second Order Approximation of the Fractional Laplacian Operator for Equal-Ripple Response Todd J. Freeborn, Brent Maundy, and Ahmed Elwakil* Department of Electrical and Computer Engineering, University of Calgary, Calgary, Alberta, Canada. *Department of Electrical and Computer Engineering, University of Sharjah, P.O. Box 27272, Emirates (elwakil@ieee.org). Abstract—In this paper we propose a modification to a second order approximation of the fractional-order Laplacian operator, s α , where 0 <α< 1. We show how this proposed modification can be used to change the ripple error of both the magnitude and phase responses of the approximation when compared to the ideal case. Equal-ripple magnitude and phase responses that have both less cumulative error and less maximum ripple deviation are presented using this modification. A 1 st order lowpass filter with fractional step of 0.8, that is of order 1.8, is implemented using the proposed approximation. Experimental results verify the operation of this approximation in the realization of the fractional step filter. I. I NTRODUCTION Traditionally the Laplacian operator, s, is raised to an integer order, i.e s, s 2 , ...s n , when used in the design of analog circuits. However, it is also mathematically valid to raise to a non-integer order, s α , where 0 <α< 1; representing a fractional order system. A fractional derivative may be defined, according to the Riemann-Liouville definition [1], as d α dt α f (t) D α f (t)= d dt 1 Γ(1 α) t ˆ 0 f (τ ) (t τ ) α (1) where Γ(·) is the gamma function. Applying the Laplace transform, with zero initial conditions, to (1) yields L { 0 d α t f (t)} = s α F (s) (2) This fractional Laplacian operator has practical applications in numerous areas of engineering [2]. However, at this time there are no commercial fractance devices available. Therefore, to physically realize circuits that make use of the advantages of s α , integer order approximations have been used. There are many methods to create an approximation of s α that include Continued Fraction Expansions (CFEs) as well as numerous rational methods [3]. These methods present a large array of approximations with varying order and accuracy, with the accuracy and approximated frequency band increasing as the order of the approximation increases. The importing of these concepts into circuit theory is relatively new, and has shown applications in power electronics [4], integrator [5], [6] and differentiator circuits [7], multivibrator circuits [8], and filter theory [9], [10] with potentially many other applications. In this paper, we propose a modification to a second- order approximation for the fractional Laplacian, s α , which can be used to manipulate the error of both the magnitude and phase response. We show how this approximation can 10 -2 10 -1 10 0 10 1 10 2 -40 -20 0 20 40 Frequency (Hz) Magnitude (dB) 10 -2 10 -1 10 0 10 1 10 2 0 20 40 60 80 Frequency (Hz) Phase (degrees) Figure 1. Magnitude and phase of ideal (solid) and 2nd order approximation (dashed) of s α when α =0.8 be manipulated to create an equi-ripple error response when compared to the ideal case. A fractional lowpass filter of order 1.8 is verified experimentally to show the application of this approximation. II. LAPLACIAN OPERATOR APPROXIMATION Previous work on the approximation of the fractional Lapla- cian operator has yielded a second order approximation using the CFE method [11] as, s α (α 2 +3α + 2)s 2 + (8 2α 2 )s +(α 2 3α + 2) (α 2 3α + 2)s 2 + (8 2α 2 )s +(α 2 +3α + 2) (3) The magnitude and phase response of (3) compared with s α when α = 0.8 is shown in Fig. 1. From this figure, the magnitude error of the approximation com- pared to the ideal case does not exceed 1.338 dB for ωǫ [0.023, 42.85] rad/s; while the phase error does not exceed 2.173 for ωǫ [0.127, 7.87] rad/s. We note also both the magnitude and phase responses have inner and outer ripples, as highlighted in Fig. 1. To compare the peaks of the inner and outer ripples for both the magnitude and phase, the error of (3) compared to the ideal s α was calculated as Error = s α (α 2 +3α + 2)s 2 + (8 2α 2 )s +(α 2 3α + 2) (α 2 3α + 2)s 2 + (8 2α 2 )s +(α 2 +3α + 2) (4) Using (4) the peaks of the magnitude and phase ripple errors were calculated numerically as shown in Fig. 2, for 0 <α< 1, illustrating that both the inner and outer ripple errors vary significantly with α. Note the size of the outer ripple is 978-1-4244-7773-9/10/$26.00 ©2010 IEEE 1173