Entity

Time filter

Source Type

Milpitas, CA, United States

Patent
Grandis Inc. | Date: 2012-04-11

A non-volatile memory (NVM) system compatible with double data rate, single data rate, or other high speed serial burst operation. The NVM system includes input and output circuits adapted to synchronously send or receive back-to-back continuous bursts of serial data at twice the frequency of any clock input. Each burst is J bits in length. The NVM system includes read and write circuits that are adapted to read or write J bits of data at a time and in parallel, for each of a multitude of parallel data paths. Data is latched such that write time is similar for each bit and is extended to the time it takes to transmit an entire burst. Consequently, the need for small and fast sensing circuits on every column of a memory array, and fast write time at twice the frequency of the fastest clock input, are relieved.


Patent
Grandis Inc. and Hynix Semiconductor Inc | Date: 2012-05-21

A magnetic memory device and a method for manufacturing the same are disclosed. The magnetic memory device includes a plurality of gates formed on a semiconductor substrate, a source line connected to a source/drain region shared between the gates neighboring with each other, a plurality of magnetic tunnel junctions connected to non-sharing source/drain regions of the gates on a one-to-one basis, and a bit line connected to the magnetic tunnel junctions. The magnetic memory device applies a magnetic memory cell to a memory so as to manufacture a higher-integration magnetic memory, and uses the magnetic memory cell based on a transistor of a DRAM cell, resulting in an increase in the availability of the magnetic memory.


A method and system for providing a magnetic junction usable in a magnetic memory are described. The magnetic junction includes first and second pinned layers, first and second nonmagnetic spacer layers, and a free layer. The pinned layers are nonmagnetic layer-free and self-pinned. In some aspects, the magnetic junction is configured to allow the free and second pinned layers to be switched between stable magnetic states when write currents are passed therethrough. The magnetic junction has greater than two stable states. In other aspects, the magnetic junction includes at least third and fourth spacer layers, a second free layer therebetween, and a third pinned layer having a pinned layer magnetic moment, being nonmagnetic layer-free, and being coupled to the second pinned layer. The magnetic junction is configured to allow the free layers to be switched between stable magnetic states when write currents are passed therethrough.


A method and system for providing a magnetic junction usable in a magnetic device are described. The magnetic junction includes a pinned layer, a plurality of nonmagnetic spacer layers, and a plurality of free layers. The free layers are interleaved with the nonmagnetic spacer layers. A first nonmagnetic spacer layer of the nonmagnetic spacer layers is between the free layers and the pinned layer. Each of the free layers is configured to be switchable between stable magnetic states when a write current is passed through the magnetic junction. Each of the free layers has a critical switching current density. The critical switching current density of one of the free layers changes monotonically from the critical switching current density of an adjacent free layer. The adjacent free layer is between the pinned layer and the one of the plurality of free layers.


Patent
Grandis Inc. | Date: 2011-04-11

A write circuit is adapted to provide a same logical bit to each of a multitude of memory cells for storage. Each of the multitude of memory cells stores either the bit or a complement of the bit in response to the write circuit. A read circuit is adapted to receive the bits stored in the multitude of memory cells and to generate an output value defined by the stored bits in accordance with a predefined rule. The predefined rule may be characterized by a statistical mode of the bits stored in the plurality of memory cells. Storage errors in a minority of the multitude of memory cells may be ignored at the cost of lower memory density. The predefined rule may be characterized by a first weight assigned to bits 1 and a second weight assigned to bits 0.

Discover hidden collaborations