Aristotle University Of Thessaloniki (AUTH)
Σχεδίαση προσαρμοστικών και δυναμικά αναδιατάξιμων αρχιτεκτονικών αντιστρόφου μετασχηματισμού συνημιτόνου 8X8 2-D IDCT, για χαμηλή κατανάλωση ισχύος
Abstract
dc:descriptionThe present dissertation considers the development of the 8×8 Inverse Discrete Cosine Transform (8×8 2-D IDCT) architectures, where the main design objective of the research was to devoted low power architectures.The dissertation presents 11 new 8×8 2-D IDCT and one new 8×8 2-D DCT architectures.Most of the architectures, the 8 out of 11, are built upon Systolic Array Processors concept. Two IDCT structures, and one 8×8 2-D DCT solutions are based on Asynchronous Processing Elements. The remaining 5 are Array Processor-based IDCT architectures that utilize synchronous Processing Elements. All architectures exploit the symmetry inherent inside the transform core, in order to improve the following implementation attributes: the minimal chip area, the required arithmetic operations accuracy, and high processing speed. Thereby, the exploitation of the symmetry properties reduces the energy required for the processing of a specific amount of data.The 3 of the proposed new 8×8 2-D IDCT architectures are based on the Arai-Agui-Nakajima algorithm where the throughput rate is increased through the pipelining technique. A power reduction scheme has been also employed in one of these architectures, that is based on activation / deactivation of arithmetic units according to the presence of zero input DCT coefficients. The remaining two Arai-Agui-Nakajima – based IDCT architectures avoid the multiplications of the transform core, through the Algebraic Integer Encoding of the multiplier's constants.The last, “eleventh” 8×8 2-D IDCT architecture exploits the high percentage of zero quantized DCT coefficients and the symmetry of the basis matrices. The required computational time is variable; it depends on the average number of nonzero DCT coefficients, pre-sent in the transform matrix. Thus, the lowest average number of multiplications per nonzero coefficient, comparably to other solutions in the worldwide literature.Regarding to the power consumption, a new algorithm, oriented towards the estimation of the switching activity at the internal nodes of CMOS circuits, is presented. Based on this algorithm, the distribution of dynamic power consumption is investigated for two of the IDCT architectures (Lee and Chen) which utilize two arithmetic representation systems (2’s complement and sign-magnitude). The simulation results are presented for various levels of observation-abstraction; starting from the system level and concluding to the Full Adder le-vel. Based on the simulation results, and in order to reduce the power consumption, a “hybrid” architecture is presented where the operands of the multiplications are expressed in sign - magnitude format, whereas, the operands of the additions/subtractions are expressed in the 2’s complement.
Degree
thesis:*- Grantor dc:publisher
- Aristotle University Of Thessaloniki (AUTH)
- Year dc:date
- 2013
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Tziortzios, Thomas
- Τζιόρτζιος, Θωμάς
Subjects
dc:subject × 14- Αντίστροφος διακριτός μετασχιματισμός συνημιτόνου
- Χαμηλή κατανάλωση ισχύος
- Συστολικοί πίνακες επεξεργαστών
- Inverse discrete cosine transform
- Low power consumption
- Systolic array processors
- Φυσικές Επιστήμες
- Επιστήμη Ηλεκτρονικών Υπολογιστών και Πληροφορική
- Επιστήμες Μηχανικού και Τεχνολογία
- Επιστήμη Ηλεκτρολόγου Μηχανικού, Ηλεκτρονικού Μηχανικού, Μηχανικού Η/Υ
- Natural Sciences
- Computer and Information Sciences
- Engineering and Technology
- Electrical Engineering, Electronic Engineering, Information Engineering
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/37947