Nonvolatile programmable logic devices

Details Nonvolatile programmable logic devices Nonvolatile programmable logic devices

2000-07-28

2003-04-01

Phan, Trong (Department: 2818)

Electronic digital logic circuitry

Multifunctional or programmable

Array

C326S040000

Reexamination Certificate

active

06542000

ABSTRACT:

BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to nonvolatile programmable logic using integrated magnetic devices such as giant-magneto-resistance (GMR) or spin-dependent-tunneling (SDT) devices. Such programmable logic is useful for programmable logic circuits, adaptive computing systems, programmable analog circuits, magnetic random access memories, and control systems.
2. Problems in the Art
Programmable Logic Devices (PLDS) and Programmable Logic Arrays (PLAs) are widely used for inexpensive prototyping and small volume production runs of digital logic functions. Typically, the term ‘PLD’ is used to describe a relatively small device that can be used to realize perhaps a sum of products or customized state machine and where the program information is usually stored in some form of nonvolatile memory. The term, programmable logic array or PLA on the other hand, usually denotes a more complicated structure that may contain many thousands of equivalent gates but where the programming information is usually stored in a volatile program memory that must be loaded during power up. More recently, some versions of these programmable logic devices that have the capability of being rapidly reconfigured have also been shown useful in high performance and specialized computing.
A significant problem with PLAs, which are largely based upon RAM (random access memory)-like cells, is that the memory is lost upon loss of power. This is contrary to ROM (read only memory) where data is typically stored during manufacturing and is nonvolatile. A computer normally employs random access memory for fast storage of in-use programs or data, although in the case of PLAs it is used for storing the customized programming information.
Nonvolatile memory, besides magnetic disks and tapes, is not new in the art. Computer core memory itself was nonvolatile before the introduction of semi-conductor RAM in the early 1970's. It was assembled from magnetic cores, which were fabricated out of magnetic ferrite materials. These transformer coils were tiny toroidal rings, which were threaded with fine copper wires. Current pulses through the wires would magnetize the cores at either right or left handed to store a 0 or a 1; and thus have a bipolar or binary memory element. Each core was a bit. However, this memory was slow and expensive and was low density by today's standards.
In comparison, present semi-conductor RAM is relatively fast, relatively cheap to fabricate in large quantities, and relatively small. However, it is volatile. Similarly, most semi-conductor digital components, such as latches, counters, flip-flops, etc. have the above-mentioned advantages, yet are also volatile. There is a need for nonvolatile components of this type. Further, regardless of non-volatility, there is room in the art with respect to RAM or other digital logic components that are further reduced in size, speedier, more reliable, and can be operated by and consume less power.
Attempts have been made to use magnetic principles (nonvolatile) as a method of storing binary information. An example is experimentation with anisotropic magnetoresistance (AMR) using one or more layers of AMR magnetic film. Localized portions of the material are magnetized in different directions to store the binary information.
The Giant Magnetoresistance Effect (GMR) results in larger changes in resistance in response to small magnetic fields in certain layered materials than are typically observed with the AMR effect. The GMR effect and a general explanation thereof are discussed in Prinz, Gary A., “Magnetoelectronics,” Science 282, 1660 (1998), incorporated by reference herein. Depending on the magnetism relative to spin polarization of current, the GMR material can be more or less conductive for electrons of specific spin polarization.
It has been shown that the magnetic field generated by even small currents could thus “program” a GMR component to several “logic states,” i.e. higher resistance or lower resistance. Thus, it would be possible to “sense” the logic state by sending current through the programmed GMR component and deriving its resistance (i.e. whether it is the higher or lower resistance). This produces a bipolar memory element that has the advantages of low power read or write, non-volatility, and self-contained. Also, importantly, there are no limitations to the number of read/write cycles known and both reading and writing can be done at high speeds; higher than most existing nonvolatile latches or memory elements.
Spin-polarized tunneling or spin dependant tunneling (SDT) is discussed at Bobo et al., “Spin dependent tunneling junctions with hard magnetic layer pinning”, J.Appl.Phys. 83, 6685 (1998), incorporated by reference herein, as another example of a possible magnetic device. In memory applications, SDT devices may be preferable to GMR devices in that they typically have a larger signal and higher intrinsic resistance for small area components.
However, the mere fact that GMR effect or SDT have been pointed out as a possible memory element is not sufficient for their effective implementation and use in an actual programmable logic circuit, where not only must data be stored in the device but also efficiently retrieved with a minimum of sensing circuitry. To perform adequately, the memory element must not only be programmable or writeable to at least two states; the different states must be reliably readable by the system. Further, it is important that the memory element be able to reliably withstand multiple, and preferably unlimited, read and/or write cycles. There is presently no known nonvolatile memory scheme that satisfactorily meets these criteria in such programmable logic circuits.
There have been some attempts to create nonvolatile, solid state latches or memory elements that improve over the state of the art. Some of the problems with these attempts are as follows. Many of such components require significant write energy. They have limitations regarding read/write speed and number of read/write cycles.
One specific example is called NVRAM or non-volatile random access memory (e.g. from Xicor). It is relatively large in that it actually consists of both a volatile conventional RAM memory and a similarly sized nonvolatile memory that is written with the contents of the volatile RAM during a power failure. In order to have energy to perform this writing of the non-volatile memory, it usually requires an external capacitor. In order to restore the state of NVRAM to what existed prior to power-down, the non-volatile memory is read upon the restoration of power and transferred to the conventional RAM. Because of the dual-memory configuration and the external component, this architecture is not conducive to large scale RAM and the like. Similarly, most PLD devices employ some form of nonvolatile programming technology that may not be changed after it is initially programmed or that may require many microseconds or milliseconds of time to reprogram. Most of these latter types of devices can also only be programmed a limited number of times because repeated programming eventually harms the device.
Programmable logic circuits have evolved since the 1970's. These circuits have revolutionized digital prototyping and design by offering cost effective and rapid hardware solutions to complex digital design problems that now routinely exceed 100K equivalent gates per chip. Historically, these circuits have employed digital memory cells that defined user programmable logic blocks (usually via look-up tables (LUTs)) and customized interconnects between the logic blocks. In traditional implementations of these circuits, the program memory would be changed only in the process of debugging the design and would thereafter be fixed in each chip. In the most advanced forms of field programmable logic circuits (FPGAs), static complementary metal-oxide-semiconductor (CMOS) registers or static random-access memory (SRAM) are used to define the various LUTs and associated interconnects. Be

Affiliated with

Also associated with

Rating

Nonvolatile programmable logic devices does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Nonvolatile programmable logic devices, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Nonvolatile programmable logic devices will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-3067306