Abstract: Techniques and systems are provided for implementing a convolutional neural network. One or more convolution accelerators are provided that each include a feature line buffer memory, a kernel buffer memory, and a plurality of multiply-accumulate (MAC) circuits arranged to multiply and accumulate data. In a first operational mode the convolutional accelerator stores feature data in the feature line buffer memory and stores kernel data in the kernel data buffer memory. In a second mode of operation, the convolutional accelerator stores kernel decompression tables in the feature line buffer memory.

Abstract: A voltage regulator is embedded in a circuit intermediate a first node (coupled to a battery) and a second node (supplying power to an external memory). The voltage regulator is activatable in a first mode of operation for startup during which an voltage is applied to the second node that increases towards a supply threshold. In response to the voltage at the second node reaching the supply threshold, the voltage regulator transitions to a second mode of operation where a programmable regulated voltage (higher than the supply threshold) is applied to the second node. In response to receipt of a low-power operation request, a first high-drive regulator circuitry is deactivated and a second low-power regulator circuitry is activated to provide a third mode of operation at low power.

Abstract: A voltage regulator coupled between a first node and second node includes a first (full-power) regulator circuit and a second (low-power) regulator circuit. In a first mode: the first regulator circuit is activated (with the second regulator circuit inactive) when the voltage at the first node is a battery voltage, and the voltage regulator is kept de-activated when the voltage at the first node is a ground voltage. In a second mode: the first regulator circuitry in is active (with the second regulator circuitry inactive) when the voltage at the first node is a battery voltage, and the voltage regulator is inactive when the voltage at the first node is a ground voltage. In a third mode: the second regulator circuitry is active (with the first regulator circuitry inactive) irrespective of the voltage at the first node being at the battery voltage or the ground voltage.

Abstract: A converter circuit converts an input signal applied across a first and a second input node into a converted output signal across a first and a second output node. The converter circuit includes a switching network coupled to the first input node via an inductor having a current flowing therethrough. In a hysteresis current control mode of the switching network, the current flowing through the inductor has a triangular waveform with rising and falling edges between a first current threshold and a second current threshold alternating with a switching frequency. The switching frequency is controlled by varying the distance between the first current threshold and the second current threshold.

Abstract: A control circuit for controlling switching operation of a switching stage of a converter includes a phase detector circuit that generates a pulse-width modulated (PWM) signal in response to a phase comparison of two clock signals. A first clock signal has a frequency determined as a function of a first feedback signal proportional to converter output voltage. A first transconductance amplifier generates a first current indicative of a difference between a reference voltage and the first feedback signal, and a second transconductance amplifier generates a second current indicative of a difference between the reference voltage and a second feedback signal proportional to a derivative of the converter output voltage. A delay line introduces a delay in the first clock signal that is dependent on the first and second currents as well as a compensation current dependent on a selected operational mode of the converter.

Abstract: An apparatus includes a current mirror coupled to an output of an amplifier through control switches, a plurality of capacitors, each of which is coupled to a common node of a leg of the current mirror and a corresponding control switch, a first dipole coupled to a first input of an amplifier, a second dipole coupled to a second input of the amplifier, a third dipole coupled to an output of the apparatus configured to generate the bandgap reference voltage, and groups of switches coupled between the current mirror and the dipoles.

Abstract: A first electronic component, such as a sensor having opposed first and second surfaces and a first thickness, is arranged on a support member with the second surface facing towards the support member. A second electronic component, such as an integrated circuit mounted on a substrate and having a second thickness less than the first thickness, is arranged on the support member with a substrate surface opposed the second electronic component facing towards the support member. A package molding material is molded onto the support member to encapsulate the second electronic component while leaving exposed the first surface of the first electronic component. The support member is then removed to expose the second surface of the first electronic component and the substrate surface of the substrate.

Abstract: A circuit includes a current path and a negative bootstrap circuitry coupled to the current path. The current path is coupled between a floating voltage and a reference ground, and includes a current generator coupled through a resistor to the floating voltage at a first node of the current generator. The current generator is controlled by a pulse signal. The negative bootstrap circuitry includes a pump capacitor coupled to a second node of the current generator and to the reference ground. The pump capacitor is configured to provide a negative voltage at the second node of the current generator based on the pulse signal.

Abstract: A plastic material substrate has a die mounting location for a semiconductor die. Metallic traces are formed on selected areas of the plastic material substrate, wherein the metallic traces provide electrically-conductive paths for coupling to the semiconductor die. The semiconductor die is attached onto the die mounting location. The semiconductor die attached onto the die mounting location is electrically bonded to selected ones of the metallic traces formed on the plastic material substrate. A package material is molded onto the semiconductor die attached onto the die mounting location.

Abstract: A semiconductor chip is arranged on a region of laser direct structuring (LDS) material of a laminar substrate. The semiconductor chip has a front active area facing towards, and a metallized back surface facing away from, the laminar substrate. An encapsulation of LDS material on the laminar substrate encapsulates the semiconductor chip with the metallized back surface of the semiconductor chip exposed at an outer surface of the encapsulation of LDS material. Electrically conductive lines and first vias are structured in the region of LDS material to electrically connect to the front active area of the semiconductor chip. A thermally conductive layer is plated over the outer surface of the encapsulation of LDS material in contact with the metallized back surface of the semiconductor chip. A heat extractor body of thermally conductive material is coupled in heat transfer relationship with the thermally conductive layer.

Abstract: A microelectromechanical structure includes a body of semiconductor material having a fixed frame internally defining a cavity, a mobile mass elastically suspended in the cavity and movable with a first resonant movement about a first rotation axis and with a second resonant movement about a second rotation axis, orthogonal to the first axis. First and second pairs of supporting elements, extending in cantilever fashion in the cavity, are rigidly coupled to the frame, and are piezoelectrically deformable to cause rotation of the mobile mass about the first and second rotation axes. First and second pairs of elastic-coupling elements are elastically coupled between the mobile mass and the first and the second pairs of supporting elements. The first and second movements of rotation of the mobile mass are decoupled from one another and do not interfere with one another due to the elastic-coupling elements of the first and second pairs.

Abstract: A MEMS inertial sensor includes a supporting structure and an inertial structure. The inertial structure includes at least one inertial mass, an elastic structure, and a stopper structure. The elastic structure is mechanically coupled to the inertial mass and to the supporting structure so as to enable a movement of the inertial mass along a first direction, when the supporting structure is subjected to an acceleration parallel to the first direction. The stopper structure is fixed with respect to the supporting structure and includes at least one primary and one secondary stopper elements. If the acceleration exceeds a first threshold value, the inertial mass abuts against the primary stopper element and subsequently rotates about an axis of rotation defined by the primary stopper element. If the acceleration exceeds a second threshold value, rotation of the inertial mass terminates when the inertial mass abuts against the secondary stopper element.

Abstract: Semiconductor dice are arranged on a substrate such as a leadframe. Each semiconductor die is provided with electrically-conductive protrusions (such as electroplated pillars or bumps) protruding from the semiconductor die opposite the substrate. Laser direct structuring material is molded onto the substrate to cover the semiconductor dice arranged thereon, with the molding operation leaving a distal end of the electrically-conductive protrusion to be optically detectable at the surface of the laser direct structuring material. Laser beam processing the laser direct structuring material is then performed with laser beam energy applied at positions of the surface of the laser direct structuring material which are located by using the electrically-conductive protrusions optically detectable at the surface of the laser direct structuring material as a spatial reference.

Abstract: An input node is configured to receive a supply signal which may be of a first polarity or a second polarity opposite the first polarity. A high input current circuit couples the input node to an output node through at least one power transistor having a control electrode. A low input current circuit couples a supply current from the input node to control circuit configured to control the power transistor. A circuit is provided to detect polarity reversal with respect to the supply signal. A protection circuit for the low input current circuit operates to decouple the control circuit from the input node if the supply signal has the second polarity. A protection circuit for the high input current circuit operates to short-circuit the control electrode of the power transistor to the current path provided by the power transistor between the input node and the output node.

Abstract: A redistribution layer for an integrated circuit is made by forming a conductive interconnection layer; forming a conductive body in electrical contract with the interconnection layer; and covering the conductive body with a first coating layer having a thickness less than 100 nm. The first coating layer is configured to provide a protection against oxidation and/or corrosion of the conductive body. To carry out an electrical test of the integrated circuit, a testing probe locally perforates the first coating layer until the conductive body is electrically contacted by the testing probe.

Abstract: A microelectromechanical device includes a substrate, a first structural layer, and a second structural layer of semiconductor material. A sensing mass extends in the first structural layer and is coupled to the substrate by first elastic connections to enable oscillation of the sensing mass in a sensing direction perpendicular to the substrate by a maximum amount relative to a resting position of the sensing mass. An out-of-plane stopper structure includes an anchorage fixed to the substrate and a mechanical end-of-travel structure, which extends in the second structural layer, faces the sensing mass, and is separated therefrom by a gap having a width smaller than the maximum displacement distance of the sensing mass. The mechanical end-of-travel structure is coupled to the anchorage by second elastic connections that enable movement of the mechanical end-of-travel structure in the sensing direction in response to an impact of the sensing mass.

Abstract: One or more semiconductor dice are arranged on a substrate. The semiconductor die or dice have a first surface adjacent the substrate and a second surface facing away from the substrate. Laser-induced forward transfer (LIFT) processing is applied to the semiconductor die or dice to form fiducial markers on the second surface of the semiconductor die or dice. Laser direct structuring (LDS) material is molded onto the substrate. The fiducial markers on the second surface of the semiconductor die or dice are optically detectable at the surface of the LDS material. Laser beam processing is applied to the molded LDS material at spatial positions located as a function of the optically detected fiducial markers to provide electrically conductive formations for the semiconductor die or dice.

Abstract: A process for manufacturing a combined microelectromechanical device includes forming, in a die of semiconductor material, at least a first and a second microelectromechanical structure, performing a first bonding phase to bond a cap to the die via a bonding region or adhesive to define at least a first and a second cavity at the first and, respectively, second microelectromechanical structures, the cavities being at a controlled pressure, forming an access channel through the cap in fluidic communication with the first cavity to control the pressure value inside the first cavity in a distinct manner with respect to a respective pressure value inside the second cavity, and performing a second bonding phase, after which the bonding region deforms to hermetically close the first cavity with respect to the access channel.

Abstract: A merged-PN-Schottky, MPS, diode includes an N substrate, an N-drift layer, a P-doped region in the drift layer, an ohmic contact on the P-doped region, a plurality of cells within the P-doped region and being portions of the drift layer where the P-doped region is absent, an anode metallization on the ohmic contact and on said cells, to form junction-barrier contacts and Schottky contacts respectively. The P-doped region has a grid-shaped layout separating from one another each cell and defining, together with the cells, an active area of the MPS diode. Each cell has a same geometry among quadrangular, quadrangular with rounded corners and circular; and the ohmic contact extends at the doped region with continuity along the grid-shaped layout.

Abstract: Semiconductor chips are arranged on an elongated substrate and encapsulated by an insulating encapsulation. Electrically conductive formations and electrically conductive plating lines are plated on the insulating encapsulation using, for example, Laser Direct Structuring (LDS) or Direct Copper Interconnect (DCI) material. The electrically conductive plating lines include first transverse plating lines as well as second plating lines branching out from the first plating lines towards the electrically conductive formations. A first partial cutting step is then performed to form grooves which remove the first plating lines. An insulating material is dispensed in the grooves to encapsulate the end portions of the second plating lines. A second cutting step median along the groove and through the elongate substrate is performed to produce singulated semiconductor devices (such as “die pad up” Quad-Flat No-lead (QFN) packages). End portions of the second plating lines are encapsulated by the insulating material.