The doctoral dissertations of the former Helsinki University of Technology (TKK) and Aalto University Schools of Technology (CHEM, ELEC, ENG, SCI) published in electronic format are available in the electronic publications archive of Aalto University - Aaltodoc.
|
|
|
Dense Implementations of Binary Cellular Nonlinear Networks: From CMOS to Nanotechnology
Jacek Flak
Dissertation for the degree of Doctor of Science in Technology to be presented with due permission of the Department of Electrical and Communications Engineering for public examination and debate in Auditorium S4 at Helsinki University of Technology (Espoo, Finland) on the 29th of June, 2007, at 12 noon.
Dissertation in PDF format (ISBN 978-951-22-8854-0) [1317 KB]
Errata (in PDF format)
Dissertation is also available in print (ISBN 978-951-22-8853-3)
Abstract
This thesis deals with the design and hardware realization of the cellular neural/nonlinear network (CNN)-type processors operating on data in the form of black and white (B/W) images. The ultimate goal is to achieve a very compact yet versatile cell structure that would allow for building a network with a very large spatial resolution. It is very important to be able to implement an array with a great number of cells on a single die. Not only it improves the computational power of the processor, but it might be the enabling factor for new applications as well. Larger resolution can be achieved in two ways. First, the cell functionality and operating principles can be tailored to improve the layout compactness. The other option is to use more advanced fabrication technology – either a newer, further downscaled CMOS process or one of the emerging nanotechnologies.
It can be beneficial to realize an array processor as two separate parts – one dedicated for gray-scale and the other for B/W image processing, as their designs can be optimized. For instance, an implementation of a CNN dedicated for B/W image processing can be significantly simplified. When working with binary images only, all coefficients in the template matrix can also be reduced to binary values. In this thesis, such a binary programming scheme is presented as a means to reduce the cell size as well as to provide the circuits composed of emerging nanodevices with an efficient programmability. Digital programming can be very fast and robust, and leads to very compact coefficient circuits. A test structure of a binary-programmable CNN has been designed and implemented with standard 0.18 µm CMOS technology. A single cell occupies only 155 µm2, which corresponds to a cell density of 6451 cells per square millimeter. A variety of templates have been tested and the measured chip performance is discussed.
Since the minimum feature size of modern CMOS devices has already entered the nanometer scale, and the limitations of further scaling are projected to be reached within the next decade or so, more and more interest and research activity is attracted by nanotechnology. Investigation of the quantum physics phenomena and development of new devices and circuit concepts, which would allow to overcome the CMOS limitations, is becoming an increasingly important science. A single-electron tunneling (SET) transistor is one of the most attractive nanodevices. While relying on the Coulomb interactions, these devices can be connected directly with a wire or through a coupling capacitance. To develop suitable structures for implementing the binary programming scheme with capacitive couplings, the CNN cell based on the floating gate MOSFET (FG-MOSFET) has been designed. This approach can be considered as a step towards a programmable cell implementation with nanodevices. Capacitively coupled CNN has been simulated and the presented results confirm the proper operation. Therefore, the same circuit strategies have also been applied to the CNN cell designed for SET technology. The cell has been simulated to work well with the binary programming scheme applied. This versatile structure can be implemented either as a pure SET design or as a SET-FET hybrid. In addition to the designs mentioned above, a number of promising nanodevices and emerging circuit architectures are introduced.
Keywords: integrated circuits, cellular neural networks, cellular nonlinear networks, cellular array processors, CMOS, nanotechnology, single-electron tunneling
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
© 2007 Helsinki University of Technology
Last update 2011-05-26