Lee Stuart Tavrow - Sunnyvale CA Mark Ronald Santoro - Sunnyvale CA
Assignee:
Micro Magic, Inc. - Sunnyvale CA
International Classification:
G11C 800
US Classification:
3652257
Abstract:
An integrated circuit memory structure includes a plurality of regular columns of memory cells arranged as a sequence such that each regular column except the last in the sequence has an associated adjacent regular column. Each regular column has associated sense amplifier circuitry and write driver circuitry for, respectively, reading output data from and writing input data to the regular column. At least one redundant column of memory cells is also provided. The structure also includes a programmable element that responds to a programming stimulus by providing a programming signal that identifies one of the regular columns as a defective column. Reconfiguration circuitry responds to the programming signal by reconfiguring the memory structure such that the sense amplifier circuitry and the write driver circuitry of each regular column in the sequence, beginning with the defective column, is reconfigured to be associated with the adjacent regular column. The sense amplifier circuitry associated with the last regular column in the sequence is reconfigured to be associated with the redundant column, which has its own associated write driver. The concepts of column redundancy are also applicable to row redundancy.
Robert J. Drost - Santa Clara CA David M. Murata - San Jose CA Robert J. Bosnyak - Sunnyvale CA Mark R. Santoro - Sunnyvale CA Lee S. Tavrow - Sunnyvale CA
Assignee:
Sun Microsystems, Inc. - Mountain View CA
International Classification:
H03K 190175 H03K 19082 H03K 190948
US Classification:
326 66
Abstract:
An efficient high-speed ECL to CMOS logic converter for BiCMOS integrated circuits. In one embodiment, a differential amplifier compares an ECL input signal to an ECL reference voltage and generates a pair of complementary intermediate signals on a corresponding pair of differential output nodes. The differential amplifier has two load resistors coupled in series with a common load resistor which limits the upper voltage swing at the differential output nodes. A regenerative stage coupled to the differential output nodes switches between a partially on state and a fully on state in response to the complementary intermediate signals. A pair of inverter stages convert the complementary intermediate signals into a pair of CMOS level signals. A pair of complementary output drivers coupled to the respective complementary inverter stages provide current driving capability. In this embodiment, each output driver includes a CMOS inverter pair and a bipolar transistor coupled between the respective output node of the driver and V. sub. DD.
Lee S. Tavrow - Sunnyvale CA Mark R. Santoro - Sunnyvale CA Gary W. Bewick - Palo Alto CA
Assignee:
Sun Microsystems, Inc. - Mountain View CA
International Classification:
G11C 1140
US Classification:
36523006
Abstract:
A row select circuit for semiconductor memories is disclosed. The row select circuit includes a decoder portion and a driver portion. The decoder potion of the row select circuit includes a plurality of decoder circuits, each servicing a multiplicity of rows. Two levels of decoding are used to select a row. First, one of the plurality of decoder circuits is selected. Second, a predecoder is provided for simultaneously selecting one of the multiplicity of rows serviced by the selected decoder circuit. A single current source is used to service the multiplicity of rows associated with a particular decoder. The driver portion of the circuit includes a driver circuit for each row. Each driver includes an inverter stage, a driver stage, a clamp and a voltage reference circuit. For a selected row, the driver circuit provides ultra-fast access time. For the deselected rows, the driver circuit consumes minimal power.
Method And Appartus For Detecting Multiple Address Matches In A Content Addressable Memory
Robert J. Bosnyak - Sunnyvale CA Mark R. Santoro - Sunnyvale CA
Assignee:
Sun Microsystems, Inc. - Mountain View CA
International Classification:
G11C 1500
US Classification:
365 49
Abstract:
A circuit for detecting multiple address matches in an associative array includes a match current generator that responds to active match signals generated by the associative array by generating a match current that is linearly proportional to the number of active match signals generated by the array. A reference current source generates a reference current that is between one and two times greater than the match current when a single active match signal is generated by the associative array. A comparator compares the match current and the reference current and generates an active output signal when the match current is greater than the reference current.
Gary W. Bewick - Palo Alto CA Mark R. Santoro - Sunnyvale CA Lee S. Tavrow - Sunnyvale CA
Assignee:
Sun Microsystems, Inc. - Mountain View CA
International Classification:
G11C 1140
US Classification:
36523006
Abstract:
A driver circuit for use in a semiconductor memory array is disclosed. The memory array includes a plurality of the driver circuits, each used to drive a word line in the memory array. The driver circuit of the present invention includes a pull up portion and an active pull down portion. The pull up portion includes a pair of cascaded transistors arranged to pull up an output node coupled to the word line. The active pull down portion includes a pair of cascaded transistors arranged to pull down the output node coupled to the word line. A control feedback path is coupled between the output node and the active pull down portion of the driver circuit. The feedback path controls the activation of the pull down portion of the driver circuit.
David Minoru Murata - San Jose CA Mark Ronald Santoro - Sunnyvale CA Lee Stuart Tavrow - Sunnyvale CA
Assignee:
Micro Magic, INc. - Sunnyvale CA
International Classification:
H03K 19094 H03K 19177
US Classification:
326 44
Abstract:
A programmable logic array (PLA) AND plane receives data input signals from input registers and generates corresponding minterms. The minterms are OR-ed together to form a sum of products, which are provided to output latches and clocked out before the end of each clock cycle by an internal self-timed signal as PLA output data. The OR plane (or the AND plane, or both) includes NOR gates that include a plurality of NMOS transistors. Each NMOS transistor in a gate has its drain connected to a common NOR gate output node, its source connected to ground and its gate connected to receive a corresponding minterm from the AND plane. The NOR gate further includes a PMOS load transistor having its source connected to a voltage supply, its drain connected to the NOR gate output node and its gate connected to receive a timing signal that turns on the PMOS load transistor as the minterms are generated at the output of the AND plane and turns off the PMOS load transistor when the sum of products are provided at the output latches.
Lee Stuart Tavrow - Sunnyvale CA Mark Ronald Santoro - Sunnyvale CA
Assignee:
Micro Magic, Inc. - Sunnyvale CA
International Classification:
G11C 700
US Classification:
365208
Abstract:
The speed of large SRAMs is improved by embedding sense amplifiers into the SRAM core. In this way, the bit line length that the SRAM cells must drive is very short and, thus, the slew rate is fast. An additional layer of metal is employed to route and accumulate the sense amp results vertically over the entire SRAM core. To reduce the required pitch of the additional metal layers, a sense amp muxing scheme is also provided.
Mark R. Santoro - Sunnyvale CA Lee S. Tavrow - Sunnyvale CA Gary W. Bewick - Palo Alto CA
Assignee:
Sun Microsystems, Inc. - Mountain View CA
International Classification:
G11C 800
US Classification:
36523003
Abstract:
An improved approach for breaking the bit lines of a semiconductor memory device into small pieces, referred to herein as Embedded Access Trees (EATs), is introduced. Embedded Access Trees enjoy the principal advantage of the banked approach by dividing long bit lines into several smaller bit lines to decrease the effective load which a selected cell must drive. However, EATs avoid most of the limitations of the banked approach, e. g. , increased size, power and complexity. In a preferred embodiment of the invention, EATs are embedded into the existing full array and do not require additional peripheral decoders, MUXes or complex and costly global routing. For a given processing technology, the present invention permits a full memory array to be subdivided into more subarrays than the banked approach, with corresponding performance improvements.
Snarl-ups will begin soon after, critics fear. Council member Mark Santoro said: "The project will certainly cause traffic issues, but I'm happy to hear Apple's going to work with us on solving these problems."