STAMP: A Real-Time and Low-Power Sampling Error Based Stochastic Number Generator

In this paper, we introduce STAMP — a real-time and low-power sampling error based stochastic number generator — for stochastic computing circuits. STAMP exploits the stochastic nature of sampling error; its name is derived from ‘stochastic’ and &#x201...

Descripción completa

Guardado en:
Detalles Bibliográficos
Autores principales: Mahmut Burak Karadeniz, Ismail Cevik, Mustafa Altun
Formato: article
Lenguaje:EN
Publicado: IEEE 2021
Materias:
Acceso en línea:https://doaj.org/article/125467ef13a6421b91dd581896c0946e
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
Descripción
Sumario:In this paper, we introduce STAMP — a real-time and low-power sampling error based stochastic number generator — for stochastic computing circuits. STAMP exploits the stochastic nature of sampling error; its name is derived from ‘stochastic’ and ‘sampling’. A unique feature of STAMP which distinguishes it from other random generators is the ability to control the output probability of the generated stochastic bit stream in real-time with no area overhead. STAMP is implemented in 180 nm CMOS. Measurements have shown that STAMP passes all tests in the suit of the National Institute of Standards and Technology (NIST) and outperforms the benchmark random number generators in terms of randomness quality. STAMP performs 140 Mb/s throughput with energy consumption of 77 pJ/bit.