Abstract:
A glitch free controlled ring oscillator comprising a programmable delay chain connected to a gating and inverting means wherein a latching means is provided between said delay chain and said gating and inverting means for registering the clock state at the time of disabling the oscillator and setting the output of the oscillator to said registered clock state.
Abstract:
An apparatus for entering and exiting low power mode comprising a processor having a cache; a power management mechanism connected to said processor for controlling a plurality of power management states, at least one of said power management states being a low latency low power state; a memory subsystem including a self sustaining mechanism connected to said processor for retaining data during said low power state; a prefetching means in said memory subsystem for loading instructions into a cache prior to entering said low power state; a disabling mechanism in said processor for disabling any interrupts that may disturb said prefetched instructions; an enabling means in said memory subsystem for initiating the self sustaining operation of said memory subsystem; a detecting means connected to said processor for sensing a trigger to exit from said low power state; and a restoring means in said power management mechanism for restoring the clock of said apparatus; thereby said processor disabling said self sustaining operation and resuming normal operation at the end of said low power state.
Abstract:
The present invention provides a scalable method for implementing FFT/IFFT computations in multiprocessor architectures that provides improved throughput by eliminating the need for inter-processor communication after the computation of the first "log 2 P" stages for an implementation using "P" processing elements, comprising computing each butterfly of the first "log 2 P" stages on either a single processor or each of the "P" processors simultaneously and distributing the computation of the butterflies in all the subsequent stages among the "P" processors such that each chain of cascaded butterflies consisting of those butterflies that have inputs and outputs connected together, are processed by the same processor. The invention also provides a system for obtaining scalable implementation of FFT/IFFT computations in multiprocessor architectures that provides improved throughput by eliminating the need for inter-processor communication after the computation of the first "log 2 P" stages for an implementation using "P" processing elements.
Abstract:
An voltage tolerant protection circuit for input buffer comprising a transmission gate circuit (11) receiving input from a pad for passing the input signal to the input of the input buffer, a control signal generator (12) connected between said transmission gate circuit (11) and the pad to provide a control signal (PMOSCTRL) for operating said transmission gate circuit (11), and an N-Well generation circuit (13) connected between the pad and said transmission gate circuit (11), and also connected to said control signal generator (12) for generating a bias signal (NWELL) for said transmission gate circuit (11) and said control signal generator (12). Thus, the present invention provides a voltage tolerant protection circuit that prevents electrical stress on transistors comprised within the transmission gate circuit (11) and the control signal generator (12), minimizes power supply consumption and transfers signals without any change in amplitude.
Abstract:
An on-chip and at-speed testerfor testing and characterization of different types of memories in an integrated circuit device, comprising a Centralized Flow Controller for automatically controlling the test operations for selected test programs, and Localised Signal Generators located inside each memory block and controlled by said Centralized Flow Controlle for applying specified test patterns on the associated memory array.
Abstract:
An improved method for finding a maximum volume minimum cutset in a network of interconnected nodes, applicable to any system that can be reduced to such network including telecommunication network, traffic network, computer networks, layouts, hydraulic networks etc. According to the invention an equivalent network is derived by replacing all nodes other then source and sink by two interconnected nodes, a conventional method applying augmenting path algorithm identifies then a cutset. If the feasible cutset is not achieved than a reduced network is constructed by directly connecting the member nodes of identified cutset to the source node and repeating the above process for the reduce network until a feasible cutset is achieved.
Abstract:
The present invention relates to an improved semiconductor memory device providing row/column redundancy comprising a plurality of data latches (11) arranged in a row-column matrix connected to a set of bitlines / global bitlines interfacing to read/write circuitry, at least two redundant rows/columns (R1, R2) connected to a redundant bitline / global bitline, a first means (5) for providing a first faulty row/column address in the matrix, a second means (1, 16) for generating other faulty row/column addresses by incrementing or decrementing predetermined numbers from the address provided by the first means (5), a comparison circuitry (3, 13) receiving as its inputs the accessed row/column address and the faulty row/column addresses and a control block (4, 14) connected to the comparison circuitry (3, 13) and receiving a control signal (REN, CS) for normal operation of the memory device, the control block (4, 14) enabling/disabling the redundant rows/columns (R1, R2) and/or other memory cell rows/columns depending upon signals received from the comparison circuitry (3, 13) and the control signal (CS) for normal operation of the memory device.
Abstract:
An improved method for finding a maximum volume minimum cutset in a network of interconnected nodes, applicable to any system that can be reduced to such network including telecommunication network, traffic network, computer networks, layouts, hydraulic networks etc. According to the invention an equivalent network is derived by replacing all nodes other then source and sink by two interconnected nodes, a conventional method applying augmenting path algorithm identifies then a cutset. If the feasible cutset is not achieved than a reduced network is constructed by directly connecting the member nodes of identified cutset to the source node and repeating the above process for the reduce network until a feasible cutset is achieved.
Abstract:
The present invention provides a digital clock modulator providing a smoothly modulated clock period to reduce emitted Electro-Magnetic Radiation (EMR) comprising a plurality of delay elements (14) connected in series receiving an unmodulated clock signal at the input, connected to a multiplexer (11) receiving inputs from unequally spaced selected taps provided between the delay elements. A control block (12) supplies the selection-inputs to said multiplexer (11), and receives a clock signal from said series of delay elements. Further, a predetermined delay element (13) is connected between the clock terminal of the said control block (12) and the last element U(n) of said series of delay elements for enabling glitch free operation by ensuring that the entire delay chain and related signal paths are in the same stable state before the control to the multiplexer changes.
Abstract:
The invention provides an improved method and apparatus for reloading only those frames in which errors are detected during the FPGA configuration. A configuration data frame for a FPGA is sumultaneously loaded to the Frame register of the FPGA and also to an error detection circuit which detects errors with the loaded frame. An error counter value is maintained by the apparatus and is incremented each time an error with a frame is detected. The incremented value is compared by a Comparator circuit with a pre-determined threshold value 'n'. If a match is found then the configuration process is aborted, else the data frame is reloaded in the configuration memory, transferred again to the frame register and recheked for errors. If no error os detected with the reloaded frame, the error counter value is reset and the next frame is loaded until the FPGA configuration process is over.