130 likes | 373 Views
A Parallel Viterbi Decoder Implementation for High Throughput. Muhammad Shoaib Bin Altaf. Outline. Motivation Actual Flow Optimizations Approach Results Conclusion. Motivation. Convolutional coding with Viterbi decoding a powerful method for FEC in Communication Systems
E N D
A Parallel Viterbi Decoder Implementation for High Throughput Muhammad Shoaib Bin Altaf
Outline • Motivation • Actual Flow • Optimizations • Approach • Results • Conclusion
Motivation • Convolutional coding with Viterbi decoding a powerful method for FEC in Communication Systems • Viterbi Algorithm is based on Maximum Likelihood Estimation which is sequential. Thus slow. • Modern Communications Standards like Wimax support very high throughput • Data speed is increasing so is the need for high speed Viterbi decoding • We are looking for such a scheme which gives vectorized output bits
Actual Algorithmic Flow • We have done this stuff in our Homework as well • On building trellis, at each stage path metric will be computed • Best path metric computation at each stage • Traceback decoding done bit by bit • Each clock cycle, one bit will be decoded
Optimization • VA is sequential but the “Good” thing is, it’s Recursive • Various optimization possibilities can be employed for speed-up. • Since the purpose was to have vectorized output, the only viable option is ‘Look Ahead Transformation’ • Discussed Look Ahead transformation for Hoffman decoding in the class • Block processing of the data
Optimization Contd. • Decoding using 2 Look Ahead step.
Optimization Contd. • Increasing the number of Look Ahead steps
Optimization Contd. • Instead of 2 paths, we have to select the minimum among the 4 possible paths • Lookup table needs to be changed
Approach • Matlab Simulation • N=10^5 bits of data • Two implementations of VA • Constraint Length K=3 • One based on simple decoding • Other based on Look Ahead Transformation • Performance comparison to justify the correctness of the suggested approach
Results • Data processing speed nearly doubles on taking a single Look Ahead step.
Results Contd.. • Performance Comaprsion
Conclusion • Look Ahead Transformation is very attractive for increasing the throughput for Recursive Algorithms • No loss in decoding abilities • Depending on the Application Look Ahead step can be increased to any value • The extra hardware cost is nominal as compared to the achieved performance • In this Project the main focus was on speeding up the decoding rate irrespective of the extra hardware cost incurred