Abstract: Apparatus and methods are described for space-efficient, high-speed data communications for integrated circuits. Bandwidth is multiplied by using multiple individual wireline communications channels coupled to form a communications lane. The data receiver for a channel implements symbol-rate equalization and crosstalk filtering that is space efficient, allowing high-speed data communications to be added as an ancillary function to an IC.

Abstract: In one aspect, a memory circuit is provided. The memory circuit includes a first memory device; a second memory device coupled to the first memory device; a freeze circuit coupled to a first output terminal and a second output terminal, where the first output terminal is an output terminal of the first memory device and the second output terminal is an output terminal of the second memory device; and a test switch coupled to the first output terminal and the second output terminal.

Abstract: A programmable device system includes one or more network-on-chip (NoC) die layers vertically connected to one or more programmable chip dice layers. The NoC die layer includes interconnects, a bus or non-blocking switches, and optionally memory blocks and direct memory access engines. The NoC die layer improves on-chip communications by providing fast and direct interconnection circuitry between various parts of the programmable chip die.

Abstract: An integrated circuit with distributed power using through-silicon-vias (TSVs) is presented. The integrated circuit has conducting pads for providing power and ground located within the peripheral region of the top surface. A number of through-silicon-vias are distributed within the peripheral region and a set of TSVs are formed within the non-peripheral region of the integrated circuit. Conducting lines on the bottom surface are coupled between each peripheral through-silicon-via and a corresponding non-peripheral through-silicon-via. Power is distributed from the conducting pads to the TSVs within the non-peripheral region through the TSVs within the peripheral region, thus supplying power and ground to circuits located within the non-peripheral region of the integrated circuit chip.

Abstract: High-speed serial data transceiver circuitry on a programmable logic device (“PLD”) includes some channels that are able to operate at data rates up to a first, relatively low maximum data rate, and other channels that are able to operate at data rates up to a second, relatively high maximum data rate. The relatively low-speed channels are served by relatively low-speed phase locked loop (“PLL”) circuitry, and have other circuit components that are typically needed for handling data that is transmitted at relatively low data rates. The relatively high-speed channels are served by relatively high-speed PLLs, and have other circuit components that are typically needed for handling data that is transmitted at relatively high data rates.

Abstract: A programmable device system includes one or more network-on-chip (NoC) die layers vertically connected to one or more programmable chip dice layers. The NoC die layer includes interconnects, a bus or non-blocking switches, and optionally memory blocks and direct memory access engines. The NoC die layer improves on-chip communications by providing fast and direct interconnection circuitry between various parts of the programmable chip die.

Abstract: Incoming data at a high-speed serial receiver is digitized and then digital signal processing (DSP) techniques may be used to perform digital equalization. Such digital techniques may be used to correct various data anomalies. In particular, in a multi-channel system, where crosstalk may be of concern, knowledge of the characteristics of the other channels, or even the data on those channels, may allow crosstalk to be subtracted out. Knowledge of data channel geometries, particularly in the context of backplane transmissions, may allow echoes and reflections caused by connectors to be subtracted out. As data rates increase, fractional rate processing can be employed. For example, the analog-to-digital conversion can be performed at half-rate and then two DSPs can be used in parallel to maintain throughput at the higher initial clock rate. At even higher rates, quadrature techniques can allow analog-to-digital conversion at quarter-rate, with four DSPs used in parallel.

Abstract: In one aspect, a memory circuit is provided. The memory circuit includes a first three-terminal (3T) resistive memory device and a second 3T resistive memory device coupled to the first 3T resistive memory device. In another aspect a memory array with memory circuits having 3T devices is provided. In yet another aspect, a method of programming a memory array is provided.

Abstract: A semiconductor system in a package in which at least first and second semiconductor substrates are mounted one above the other on a package substrate. The first substrate is mounted on the package substrate with its active (or front) side facing the package substrate. A plurality of through-silicon-vias (TSVs) extend through one or more peripheral regions of the first substrate; and a redistribution layer is located on the back side of the first substrate and connected to the TSVs. The second substrate is mounted on the first substrate and electrically connected to circuits in the active side of the first substrate through the redistribution layer and the TSVs. Illustratively, one of the substrates is an FPGA and one or more of the other substrates stores the configuration memory and/or other functional memory for the FPGA.

Abstract: The disclosed invention is a technology for producing a recovered clock signal using a multi-mode clock data recovery (CDR) circuit that accommodates a flexible range operating frequencies F and consecutive identical digit requirements CID. In a first mode of operation, a controlled oscillator produces the recovered clock signal, and in a second mode of operation, a phase interpolator produces the recovered clock signal. The multi-mode CDR circuit operates in the first mode if (CID/F) is less than a threshold time value and in the second mode if (CID/F) is greater than the threshold time value.

Abstract: An integrated circuit with distributed power using through-silicon-vias (TSVs) is presented. The integrated circuit has conducting pads for providing power and ground located within the peripheral region of the top surface. A number of through-silicon-vias are distributed within the peripheral region and a set of TSVs are formed within the non-peripheral region of the integrated circuit. Conducting lines on the bottom surface are coupled between each peripheral through-silicon-via and a corresponding non-peripheral through-silicon-via. Power is distributed from the conducting pads to the TSVs within the non-peripheral region through the TSVs within the peripheral region, thus supplying power and ground to circuits located within the non-peripheral region of the integrated circuit chip.

Abstract: An integrated circuit with distributed power using through-silicon-vias (TSVs) is presented. The integrated circuit has conducting pads for providing power and ground located within the peripheral region of the top surface. A number of through-silicon-vias are distributed within the peripheral region and a set of TSVs are formed within the non-peripheral region of the integrated circuit. Conducting lines on the bottom surface are coupled between each peripheral through-silicon-via and a corresponding non-peripheral through-silicon-via. Power is distributed from the conducting pads to the TSVs within the non-peripheral region through the TSVs within the peripheral region, thus supplying power and ground to circuits located within the non-peripheral region of the integrated circuit chip.

Abstract: A semiconductor system in a package in which at least first and second semiconductor substrates are mounted one above the other on a package substrate. The first substrate is mounted on the package substrate with its active (or front) side facing the package substrate. A plurality of through-silicon-vias (TSVs) extend through one or more peripheral regions of the first substrate; and a redistribution layer is located on the back side of the first substrate and connected to the TSVs. The second substrate is mounted on the first substrate and electrically connected to circuits in the active side of the first substrate through the redistribution layer and the TSVs. Illustratively, one of the substrates is an FPGA and one or more of the other substrates stores the configuration memory and/or other functional memory for the FPGA. Advantageously, design costs are reduced by using pre-existing designs and modifying them as needed to provide TSVs along the periphery of the circuit.

Abstract: In one aspect, a multiplexer array is described. The multiplexer array includes (1) a first multiplexer coupled to a first address line, where the first multiplexer includes a first plurality of memory devices and (2) a first plurality of input logic devices coupled to the first multiplexer, a first plurality of data lines, and a plurality of bitlines. Each input logic device of the first plurality of input logic devices is coupled to a respective memory device of the first plurality of memory devices and includes a first input terminal and a second input terminal, where, for each input logic device, the first input terminal is coupled to a respective data line of the first plurality of data lines and the second input terminal is coupled to a respective bitline of the plurality of bitlines. Embodiments of methods of programming a multiplexer array are also described.

Abstract: Configurable phase-locked loop circuitry is provided. The phase-locked loop circuitry may include a buffer having a buffer output and a multiplexer having inputs and an output. The phase-locked loop circuitry may include multiple voltage-controlled oscillators. The phase-locked loop circuitry may be configured to switch a desired one of the voltage-controlled oscillators into use. Each voltage-controlled oscillator may be controlled by control signals applied to a control input for that voltage-controlled oscillator. The control input of each voltage-controlled oscillator may be connected to the buffer output. The output of each voltage-controlled oscillator may be connected to a respective one of the multiplexer inputs. Power-down transistors may be used to disable unused voltage-controlled oscillators to conserve power. The power-down transistors and the multiplexer may be controlled by signals from programmable elements.

Abstract: Methods and apparatus are provided for generating a clock signal with relatively high bandwidth and relatively low phase noise. A circuit of the invention can include a pair of transistors serially coupled between a signal of relatively high voltage and a source of relatively low voltage, where a voltage of the signal of relatively high voltage can vary according to a voltage of a variable control signal. A gate of one of the pair of transistors can be coupled to an input clock signal, and an output node between the pair of transistors can be coupled to an output clock signal. The circuit can also include a third transistor, whose drain and source are coupled to the output clock signal, and whose gate can be coupled to a gear input signal. This circuit can advantageously operate under at least two different gears, each with different bandwidth and phase noise characteristics.

Abstract: Circuitry and methods for supporting serial communications over serial interconnects between circuit modules are provided. A data recovery circuit receives incoming serial data from the serial interconnect path with zero delay. The data recovery circuit includes a data sampler that samples the incoming serial data using a multiphase clock. Data samples are provided to a multiplexer that selects an optimum sampled data signal to use as a recovered data signal. The multiplexer has a control input that receives a phase pointer signal. Control circuitry in the data recovery circuit analyzes the sampled data signals and a current value of the phase pointer to compute a clock phase shift error. If the clock phase shift error exceeds a predetermined value, the phase pointer signal can be updated. The data recovery circuit may be implemented using hardwired circuitry or programmable logic.

Abstract: High-speed serial data transceiver circuitry on a programmable logic device (“PLD”) includes some channels that are able to operate at data rates up to a first, relatively low maximum data rate, and other channels that are able to operate at data rates up to a second, relatively high maximum data rate. The relatively low-speed channels are served by relatively low-speed phase locked loop (“PLL”) circuitry, and have other circuit components that are typically needed for handling data that is transmitted at relatively low data rates. The relatively high-speed channels are served by relatively high-speed PLLs, and have other circuit components that are typically needed for handling data that is transmitted at relatively high data rates.

Abstract: A semiconductor system in a package separates those circuits in a field programmable gate array (FPGA) into two substrates. In particular, the logic elements are formed in a first semiconductor substrate and certain non-logic elements are formed in a second semiconductor substrate that is in mechanical and electrical connection with the first substrate. The two substrates are enclosed in a suitable protective package and electrical connections are provided between one or both substrates and the exterior. The non-logic elements formed in the second substrate are located in circuits that would have a signal propagation delay in a conventional FPGA that is more than approximately twice the interconnect delay between the two substrates.

Abstract: High-speed serial data transceiver circuitry on a programmable logic device (“PLD”) includes some channels that are able to operate at data rates up to a first, relatively low maximum data rate, and other channels that are able to operate at data rates up to a second, relatively high maximum data rate. The relatively low-speed channels are served by relatively low-speed phase locked loop (“PLL”) circuitry, and have other circuit components that are typically needed for handling data that is transmitted at relatively low data rates. The relatively high-speed channels are served by relatively high-speed PLLs, and have other circuit components that are typically needed for handling data that is transmitted at relatively high data rates.