Parallel Programmable Asynchronous Neighborhood Mechanism for Kohonen SOM Implemented in CMOS Technology

We present a new programmable neighborhood mechanism for hardware implemented Kohonen self-organizing maps (SOMs) with three different map topologies realized on a single chip. The proposed circuit comes as a fully parallel and asynchronous architecture. The mechanism is very fast. In a medium sized map with several hundreds neurons implemented in the complementary metal-oxide semiconductor 0.18 mu m technology, all neurons start adapting the weights after no more than 11 ns. The adaptation is then carried out in parallel. This is an evident advantage in comparison with the commonly used software-realized SOMs. The circuit is robust against the process, supply voltage and environment temperature variations. Due to a simple structure, it features low energy consumption of a few pJ per neuron per a single learning pattern. In this paper, we discuss different aspects of hardware realization, such as a suitable selection of the map topology and the initial neighborhood range, as the optimization of these parameters is essential when looking from the circuit complexity point of view. For the optimal values of these parameters, the chip area and the power dissipation can be reduced even by 60% and 80%, respectively, without affecting the quality of learning.

Analog Programmable Distance Calculation Circuit for Winner Takes All Neural Network Realized in the CMOS Technology

Rafal Tomasz Dlugosz

This paper presents a programmable analog current-mode circuit used to calculate the distance between two vectors of currents, following two distance measures. The Euclidean (L2) distance is commonly used. However, in many situations, it can be replaced with the Manhattan (L1) one, which is computationally less intensive, whose realization comes with less power dissipation and lower hardware complexity. The presented circuit can be easily reprogrammed to operate with one of these distances. The circuit is one of the components of an analog winner takes all neural network (NN) implemented in the complementary metal-oxide-semiconductor 0.18-mu m technology. The learning process of the realized NN has been successfully verified by the laboratory tests of the fabricated chip. The proposed distance calculation circuit (DCC) features a simple structure, which makes it suitable for networks with a relatively large number of neurons realized in hardware and operating in parallel. For example, the network with three inputs occupies a relatively small area of 3900 mu m(2). When operating in the L2 mode, the circuit dissipates 85 mu W of power from the 1.5 V voltage supply, at maximum data rate of 10 MHz. In the L1 mode, an average dissipated power is reduced to 55 mu W from 1.2 V voltage supply, while data rate is 12 MHz in this case. The given data rates are provided for the worst case scenario, where input currents differ by 1%-2% only. In this case, the settling time of the comparators used in the DCC is quite long. However, that kind of situation is very rare in the overall learning process.

Institute of Electrical and Electronics Engineers2016

Current Mode Euclidean Distance Calculation Circuit for Kohonen's Neural Network Implemented in CMOS 0.18μm Technology

Rafal Tomasz Dlugosz

In this paper we present analog current mode Euclidean distance calculation (EDC) block, which calculates the distance between two current vectors. The proposed circuit is an important part of the CMOS-implemented Kohonen’s neural network (KNN) designed for medical applications. The input data vector is compared with the weights’ vector in each neuron in proposed KNN. The neuron, whose weights are the closest to the input training vector becomes the winner and in the next step changes its weights. Proposed EDC block performs several operations such as: subtraction, squaring and summing of the current signals. The output current is the exact measure of the Euclidean distance. Proposed circuit dissipates power 15 μW from 1.5 V voltage supply, working with 20 MHz input data frequency. The signal frequency as well as the power dissipation may be scaled down to 1 MHz and 300 nA. Proposed EDC circuit, in CMOS 0.18 μm technology, occupies area about 500 μm2.

2007

SOI mixed-mode design techniques and case studies

Marija Blagojevic

Since the emergence of Silicon On Insulator (SOI) technology the research activities have been concentrated on a detailed analysis of SOI devices and their use in different application fields (low-voltage, low-power circuits, high temperature electronics, digital circuits, etc.). The present thesis deals with low-voltage, low-power mixed-mode design using a SOI technology. The goal is to establish the concepts that allow one to design SOI mixed-mode circuits, and to apply these concepts to the design of fully depleted (FD) and partially depleted (PD) SOI circuits and systems. Two studies have been realized in the scope of this work: a Hall sensor based microsystem design using an experimental FD SOI technology a DRAM design using capacitor-less 1T floating-body PD SOI memory cells. At the beginning of this thesis, two major types of SOI devices are introduced, namely FD and PD devices. The most important properties of these devices are then presented and compared to those of CMOS bulk devices. A design methodology based on design retargeting from bulk to SOI technology is further proposed. This methodology is developed with the aim of being implemented for the design of FD SOI circuits. It consists in using the same device dimensions and bias currents for the bulk and the corresponding FD SOI design. The analysis performed proves that such an approach permits one to obtain better circuit performances using the FD SOI technology. The EKV model is chosen for Spice simulations. For that purpose the extraction of the EKV intrinsic parameters has been performed for the experimental 0.5 μm FD SOI technology. The EKV model card obtained from transistor measurements is implemented straightforwardly for circuit simulations. The first study presented in this thesis is an FD SOI Hall sensor front-end for energy measurement. The mixed-mode microsystem design is completely based on the proposed design methodology. The system level solutions are also included, such as the spinning-current method that serves for reduction of the offset and low frequency noise of the analog front-end, and a high resolution analog-to-digital conversion technique. The integrated microsystem is entirely functional. Furthermore, according to the existing literature, this integrated circuit is the first microsystem completely realized using FD SOI technology. The overall system error that is obtained is less than 1.5 %. During the second part of this work, a novel sensing scheme that exploits an automatic reference generation based on successive approximations has been developed for the PD SOI capacitor-less 1T DRAM. The 1T DRAM reference current is generated by an adjustable current source. The electrical calibration of the reference current is performed using a digital-to-analog converter and successive approximations algorithm. The proposed scheme is evaluated in a 2 kb test chip. The circuit integrated using PD SOI technology contains a 2 kb 1T memory array, as well as automatic reference generators and peripheral circuits. The automatic reference generator comprises current mode sense amplifier, M/3M converter, and successive approximations register. All blocks implemented in the test chip are described in detail, and PD SOI design issues are discussed. A number of experimental results demonstrate the potential of the PD SOI 1T DRAM for future embedded DRAM applications. The measured retention time under holding conditions is higher than 1s. In the continuous read mode, a few hundreds of the read cycles can be performed without a refresh operation. The test chip has a measured access time of 25 ns with a read cycle time of 70 ns.

EPFL2005

SOI mixed-mode design techniques and case studies

Marija Blagojevic

EPFL2005