TY - GEN
T1 - Design and implementation of adaptive Viterbi decoder for using a dynamic reconfigurable processor
AU - Kishimoto, Yuken
AU - Haruyama, Shinichiro
AU - Amano, Hideharu
PY - 2008/12/1
Y1 - 2008/12/1
N2 - Viterbi decoder implemented with hard-wired logic often requires extra cost and consuming power by using individual logic for various constraint length and decode precisions. Although the redundant hardware which is not used in the given condition can be omitted by replacing the hard-wired logic on FPGA, the time for loading configuration data often takes milliseconds and causes too long system stall. In this paper, we implemented the Viterbi algorithms whose constraint length are from 3 to 5 on coarse grained dynamically reconfigurable processor DAPDNA-II and replaced them according to the requirement. A certain thresh-old of BER is set for a fixed SNR and the Viterbi decoder with multiple constraint lengths is simulated. In the result of evaluation, when at least 4.50 Mbps throughput is ensured even with frequent reconfiguration was performed, the power consumption is reduced by 30% - 80% compared with the case when a constraint length of the best performance is utilized.
AB - Viterbi decoder implemented with hard-wired logic often requires extra cost and consuming power by using individual logic for various constraint length and decode precisions. Although the redundant hardware which is not used in the given condition can be omitted by replacing the hard-wired logic on FPGA, the time for loading configuration data often takes milliseconds and causes too long system stall. In this paper, we implemented the Viterbi algorithms whose constraint length are from 3 to 5 on coarse grained dynamically reconfigurable processor DAPDNA-II and replaced them according to the requirement. A certain thresh-old of BER is set for a fixed SNR and the Viterbi decoder with multiple constraint lengths is simulated. In the result of evaluation, when at least 4.50 Mbps throughput is ensured even with frequent reconfiguration was performed, the power consumption is reduced by 30% - 80% compared with the case when a constraint length of the best performance is utilized.
UR - http://www.scopus.com/inward/record.url?scp=62349131395&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=62349131395&partnerID=8YFLogxK
U2 - 10.1109/ReConFig.2008.39
DO - 10.1109/ReConFig.2008.39
M3 - Conference contribution
AN - SCOPUS:62349131395
SN - 9780769534749
T3 - Proceedings - 2008 International Conference on Reconfigurable Computing and FPGAs, ReConFig 2008
SP - 247
EP - 252
BT - Proceedings - 2008 International Conference on Reconfigurable Computing and FPGAs, ReConFig 2008
T2 - 2008 International Conference on Reconfigurable Computing and FPGAs, ReConFig 2008
Y2 - 3 December 2008 through 5 December 2008
ER -