TFET-based Operational Transconductance Amplifier Design for CNN Systems

Abstract

A Cellular Neural Network (CNN) is a powerful processor that can significantly improve the performance of spatiotemporal applications such as pattern recognition, image processing, motion detection, when compared to the more traditional von Neumann architecture. In this paper, we show how tunneling field effect transistors (TFETs) can be utilized to enhance the performance of CNNs. Specifically, power consumption of TFET-based CNNs can be significantly lower when compared to MOSFET-based CNNs due to improved voltage controlled current sources (VCCSs) -an important component in CNN systems. We demonstrate that CNNs can benefit from low power conventional linear VCCSs implemented via TFETs. We also show that TFETs can be useful to realize non-linear VCCSs, which are either not possible or exhibit degraded performance when implemented via CMOS. Such non-linear VCCSs help to improve the performance of certain CNN operations (e.g., global maximum/minimum). We provide two case studiesimage contrast enhancement and maximum row selectionthat illustrate the benefits of non-linear VCCSs (e.g., reduced computation time, energy dissipation, etc.) when compared to CMOS-based approaches.

A Mixed Signal Architecture for Convolutional Neural Networks

“…[22,37]. For the work to be discussed in this paper, we assume the OTA design from [40]. OTAs provide a large linear range for voltage to current conversion, and can implement a wide range of transconductances that could be used for di erent 1:4 Q. Lou et al…”

“…4. at said, (i) while non-linear templates are well established in the theory of CeNNs, (ii) the application of a non-linear function has obvious computational utility, and (iii) non-linear templates can be easily simulated, in practice, non-linear operations are much more di cult to realize. While existing hardware considers non-linear template implementations [40], it may still not exactly mimic the behavior of non-linear templates. (We will discuss this in more detail in Sec.…”

“…2, a widely used implementation is based on OTAs [9]. Here, an OTA is built with two-stage operational ampli ers [40]. We use N OTAs with quantized m values (i.e., m0 , 2 m0 , …, 2 N −1 m0 ) to realize N-bit templates (i.e., weights).…”

Efficient Analog Circuits for Boolean Satisfiability

“…While quantum computing is a promising venue, it is far from being brought to practical reality, with many challenges still to be faced, both in physics and engineering. Neuromorphic computing systems, e.g., Cellular Neural Networks (CNNs) [3]- [5] and IBM's TrueNorth [6], have been shown to be promising alternatives for solving a range of problems in, say, sensory processing (vision, pattern recognition) and robotics. Analog mixed-signal information processing systems such as CNNs can offer extremely power/energy efficient solutions to some problems that are costly to solve by digital computers [7].…”

“…Other delay functions are being investigated. It is important to note that the decrease of satisfied clause's associated a m due to this time-delayed form relatively reduces the clause's weight in (5), thus increases other clauses' weights in the focused search space, enhancing the driving capability of unsatisfied clauses.…”