Digital Linearizer Based on 1-Bit Quantizations
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , |
|---|---|
| Format: | Preprint |
| Publié: |
2025
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866915501193560064 |
|---|---|
| author | Linares, Deijany Rodriguez Johansson, Håkan |
| author_facet | Linares, Deijany Rodriguez Johansson, Håkan |
| contents | This paper introduces a novel low-complexity memoryless linearizer for suppression of distortion in analog frontends. It is based on our recently introduced linearizer which is inspired by neural networks, but with orders-of-magnitude lower complexity than conventional neural-networks considered in this context, and it can also outperform the conventional parallel memoryless Hammerstein linearizer. Further, it can be designed through matrix inversion and thereby the costly and time consuming numerical optimization traditionally used when training neural networks is avoided. The linearizer proposed in this paper is different in that it uses 1-bit quantizations as nonlinear activation functions and different bias values. These features enable a look-up table implementation which eliminates all but one of the multiplications and additions required for the linearization. Extensive simulations and comparisons are included in the paper, for distorted multi-tone signals and bandpass filtered white noise, which demonstrate the efficacy of the proposed linearizer. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2503_02729 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Digital Linearizer Based on 1-Bit Quantizations Linares, Deijany Rodriguez Johansson, Håkan Signal Processing This paper introduces a novel low-complexity memoryless linearizer for suppression of distortion in analog frontends. It is based on our recently introduced linearizer which is inspired by neural networks, but with orders-of-magnitude lower complexity than conventional neural-networks considered in this context, and it can also outperform the conventional parallel memoryless Hammerstein linearizer. Further, it can be designed through matrix inversion and thereby the costly and time consuming numerical optimization traditionally used when training neural networks is avoided. The linearizer proposed in this paper is different in that it uses 1-bit quantizations as nonlinear activation functions and different bias values. These features enable a look-up table implementation which eliminates all but one of the multiplications and additions required for the linearization. Extensive simulations and comparisons are included in the paper, for distorted multi-tone signals and bandpass filtered white noise, which demonstrate the efficacy of the proposed linearizer. |
| title | Digital Linearizer Based on 1-Bit Quantizations |
| topic | Signal Processing |
| url | https://arxiv.org/abs/2503.02729 |