The Design of the NLMS Adaptive Filters Using the Fast-Division Approximation with CSD Encoded Divisors

The computational bottleneck of an NLMS algorithm is the division. This brief presents a division approximation method by encoding the divisor in a canonical signed digit (CSD) form. With the CSD-encoded divisor, we can easily approximate the division by multiplication. This approximation can be mad...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:IEEE transactions on circuits and systems. II, Express briefs Express briefs, 2024-04, Vol.71 (4), p.1-1
Hauptverfasser: Yao, Chia-Yu, Huang, Yu-Zhe
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:The computational bottleneck of an NLMS algorithm is the division. This brief presents a division approximation method by encoding the divisor in a canonical signed digit (CSD) form. With the CSD-encoded divisor, we can easily approximate the division by multiplication. This approximation can be made fast. The corresponding percent error will be at most 11.11%. Such percent error has a tiny impact on the performance of the NLMS adaptive filters. The experiment of designing a 16-tap FIR NLMS adaptive filter with the proposed division approximation on an FPGA platform shows that the throughput rate is increased by 136% without any significant performance penalty compared with the NLMS adaptive filter using the conventional SRT division.
ISSN:1549-7747
1558-3791
DOI:10.1109/TCSII.2023.3332692