Sunteți pe pagina 1din 6

# International Journal of Computational Intelligence and Information Security, January 2014 Vol. 5, No.

1 ISSN: 1837-7823

Efficient

## -bit Dynamic Range Reverse Converter

H. Siewobr and K. A.Gbolagade Department of Computer Science Faculty of Mathematical Sciences University for Development Studies GHANA Abstract

This paper proposes a new -bits dynamic range moduli set , and a modulo operation free reverse converter based on New Chinese Remainder Theorem I. The proposed reversed converter is purely adder based and only requires one level of regular carry propagate adder. Theoretical analyses show that the proposed converter outperform existing known state of the art equivalent converters in terms of both area cost and conversion time. Experiments were conducted by implementing the proposed converter and the state of the art equivalent converters using Xilinx ISE 14.3 software to target a Spartan 3 FPGA. The results obtained after design place and route given in terms of the number of FPGA slices and input-to-output propagation delays (in nano seconds) for various dynamic range requirements (different values of n) suggest that, the proposed converter improves the area and delay of the most efficient state of the art equivalent converter by about . and respectively. Keywords: Chinese Remainder Theorem, Moduli Set, Multiplicative Inverses, Reverse Conversion.

1. Introduction

International Journal of Computational Intelligence and Information Security, January 2014 Vol. 5, No. 1 ISSN: 1837-7823

2. Background
For a moduli set with the dynamic range the residue number can be converted into the decimal number , according to the, as follows: (1) where and is the multiplicative inverse of with respect to [4]. , the binary

New Chinese Remainder Theorem I (New CRT-I) ; Given the residue number number can be computed as;

(2) which can further be simplified to; (3) where, (4) (5) . In the next section we present the proposed CRT-I based reverse converters. (6)

3. Proposed Converter
The following property is important to the conversion algorithm to be introduced in this next section; Property 1: Modulo multiplication of a residue number by , where and are positive integers, is equivalent to -bit circular left shifting [4]. Property 2: Modulo integers [11]. of is equal to modulo of , where and are

Given the RNS number with respect to the moduli set in the form the proposed algorithm computes the binary equivalent of this RNS number based on the CRT-I. First, we wish to show that and are relatively prime. the moduli Theorem 1: The moduli set Proof: It has been shown in [10]. Following Equation (3), the CRT-I for two moduli can thus be represented as; (7) contains pairwise relatively prime moduli.

International Journal of Computational Intelligence and Information Security, January 2014 Vol. 5, No. 1 ISSN: 1837-7823 Theorem 2: Given the with and , the following hold true: (8) Proof: substituting the moduli and (8) into (7) we have; (9) We can further simplify (9) as; (10) where, (applying property 1) (11) Thus (7) holds true.

## Finally, (10) can be written as; (13)

4. Hardware Realization
The diagram of blocks in Fig. 1 represents the proposed reverse converter architecture. Considering and as the delay and area of a -bit Full Adder (FA), respectively, we consider also that the delay of a Carry-Propagate Adder (CPA) with End-Around Carry (EAC) is twice the delay of a regular CPA at a similar and hardware cost. Furthermore, we consider that the area and delay of a half adder (HA) are , respectively. The bitwise operations are ignored for area and delay analysis, as they are expected to be negligible regarding the FAs and HAs. From Fig. 1, the hardware structures of the proposed reverse converter is based on (12) and (13). Equation (12) is computed by a CSA with EAC and regular regular CPA. It is worth noting that though CPA is a -bit adder, it contains -bit HAs since one of its operands has a constant -bit 1s. The final binary equivalent of an RNS with -bit circular left shifting of the output of the CPA i.e. at no number is computed by concatenating hardware cost. In terms of area and delay requirements of the proposed scheme, the CSA is made up of , while the CPA imposes a delay of at cost. XNOR/OR pairs and imposes a delay of

International Journal of Computational Intelligence and Information Security, January 2014 Vol. 5, No. 1 ISSN: 1837-7823

Bit Organizer

CSA

CPA

## Figure 1: Schematic Diagram of the Proposed Scheme

5. Performance Evaluation
To evaluate the performance of the proposed converter, we compare our proposal with the equivalent state of the in [5]. First, it is worth art converters for the equal dynamic range moduli set remembering as stated earlier that the proposed moduli set is more advantageous than the related state of the since it eliminates the high delay and high area imposing equal dynamic range moduli set . Next, the theoretical analysis presented in Table 1 shows that the proposed converter modulo outperforms the state of the art in terms of both area and delay.
Table 1: Area and Delay comparison Converters [5] Proposed Area (

Delay (

To validate these theoretical suggestions, we did a HDL specification of both the proposed and equivalent state of the art converter. Using these HDL specifications, experiments was carried out using Xilinx ISE 14.3 software to target a Spartan 3 FPGA. The results obtained after design place and route are given in terms of the number of FPGA slices in Table II and input-to-output propagation delays (in nano seconds) for various dynamic range requirements (different values of n) in Table III. These results suggest that, the proposed converter improves the and delay by about . area demanded by the equivalent state of the art converter in [5] by about

International Journal of Computational Intelligence and Information Security, January 2014 Vol. 5, No. 1 ISSN: 1837-7823
Table 2: Converters' Area [Number Of Slices] Converter [5] Proposed

## Converter [5] Proposed

6. Conclusion
In this paper, we presented a new and efficient 3n-bits dynamic range moduli set . subsequently, we proposed an efficient reverse converter for the moduli set under investigation. The proposed converter does not require any modulo operations and is simple to design. Both theoretical and experimental results show that the proposed scheme outperforms the best known similar state of the art equivalent converter in terms of both area and delay.

International Journal of Computational Intelligence and Information Security, January 2014 Vol. 5, No. 1 ISSN: 1837-7823

References
[1]

[2] [3]

[4]

[5] [6]

[7]

[8] [9]

[10]

A.S. Molahosseini, K. Navi, and M.K. Rafsanjani, "A New residue to binary converter based on mixedradix conversion", 3rd International Conference On Information and Communication Technologies: From Theory to Applications (ICTTA 2008), pp. 1-6, April, 2008. A.P. Vinod and A.B. Premkumar, "A memoryless residue to binary converter for the 4-superset - ", Journal of Circuits, Syst. and Computers, Vol. 10, pp. 85-99,2000. K.A. Gbolagade and S.D. Cotofana, "MRC Technique for RNS to Decimal Conversion for the Moduli ", 16th Annual Workshop on Circuits, Systems, and Signal Processing, pp. 318Set 321, Veldhoven, The Netherlands, November, 2008. K.A. Gbolagade, R. Chaves, L. Sousa, and S.D. Cotofana, "An Improved RNS Reverse Converter for the - Moduli Set", IEEE International Symposium on Circuits and Systems (ISCAS2010), pp. 2103-2106, Paris, France, June, 2010. Y. Wang, X. Song, M. Aboulhamid and H. Shen. "Adder Based Residue to Binary Number Converters ", IEEE Transactions on Signal Processing, Vol. 50, pp.17721779. 2002. for Leonel Sousa, Samuel Antao, "MRC-Based RNS reverse converters for the Four-Moduli sets 1 ,2 1,2 , 22 +11} and {2 +1 ,2 1,22 , 22 +11}", IEEE Transactions on Circuits and Systems 59-II (4), pp. 244-248, 2012. D. Gallaher, F. Petry, and P. Srinivasan, "The digital parallel method for fast RNS to weighted number ", IEEE Trans. Circuits Syst. II, vol. 44, pp. system conversion for specific moduli 5357, Jan. 1997. S. Piestrak, "A high-speed realization of a residue to binary number system converter", IEEE Trans. Circuits Syst. II, vol. 42, Oct. 1995. M. Bhardwaj, A. B. Premkumar, and T. Srikanthan, "Breaking the 2n-bit carry propagation barrier in module set", IEEE Trans. Circuits Syst. II, vol. residue to binary conversion for the 45, pp.9981002, Sept. 1998. K. A. Gbolagade, "An Efficient MRC based RNS-to-Binary Converter for the moduli set", AIMS SA, 2011.