Miles, John Richard

A method of receiving a Bluetooth Low Energy (BLE) audio broadcast and a Bluetooth Low Energy audio broadcast receiver is described. A BLE broadcast audio receiver receives a broadcast isochronous stream (BIS) intended for playback by the receiver. The BLE broadcast audio receiver receives a further BIS intended for playback by a different BLE broadcast audio receiver. The further BIS is re-transmitted by the BLE broadcast receiver.

Embodiments of a radio frequency (RF) controller for a 5G modem, a 5G modem, and a method for operating an RF controller of a 5G modem are disclosed. In an embodiment, an RF controller for a 5G modem includes a timing generator configured to generate a symbol clock signal, a memory unit containing execution boxes (Eboxes) of micro-operations and an index of the Eboxes to be executed in response to detecting a symbol idle period of the symbol clock signal, and an Ebox execution engine configured to execute micro-operations in one of the Eboxes based on the index of the Eboxes to perform an RF maintenance or calibration operation during the symbol idle period.

A priority based prefetch by a memory controller is disclosed. A prefetch of data of a low priority master is suspended when a data request associated with a high priority master indicated by a master identification is to be performed. The suspension includes storing a saved state of the prefetch, performing the prefetch of data of the high priority master, and resuming the prefetch of data of the low priority master when the prefetch of data of the high priority master is completed.

A dual-path phase-locked loop (PLL) is disclosed. The dual-path PLL converts, based on a masking value, a pulse signal to a masked pulse signal such that the masked pulse signal comprises one or more unmasked pulses and one or more masked pulses of the pulse signal. An integral current is generated, based on the masked pulse signal in a manner such that that an average current value of the integral current reaches a desired value over a predefined number of cycles of the pulse signal and an optimal peaking in the dual-path PLL is achieved.

The disclosure relates to a touch screen sensing system (1100), example embodiments of which comprise: a touch screen panel (1103) having a plurality pairs of touch sensing electrodes (row<1-Nrow>, col<1-Ncol>) arranged in rows and columns extending across the panel (1103); and a plurality of transceivers (1101<1-Nrow>, 1102<1-Ncol>), each of the plurality of transceivers (1101<1-Nrow>, 1102<1-Ncol>) connected to send and receive sensing signals (701<1-n>) to and from a respective one of the plurality of touch sensing electrodes (row<1-Nrow>, col<1-Ncol>), wherein each one of the plurality of transceivers (1101<1-Nrow>, 1102<1-Ncol>) is configured to transmit a sensing signal (701<1-n>) including a range of frequencies over a signal bandwidth (702<1-n>), the sensing signals (701<1-n>) being orthogonal to each other.

A time to digital converter including multiple programmable buffers coupled in series for receiving a pulse signal, latches configured to provide binary values indicative of the state of the buffers at the end of a timing pulse asserted on the pulse signal, and a phase converter configured to convert the binary values into a digital output value indicative of a measured delay of the timing pulse. Each of the buffers is configured with an adjustable delay based on a delay select input. The adjustable delay is used to select from among multiple different transition delays of each buffer. The buffers may be configured to be compatible with a standard cell layout. The buffers may be configured as standard cell logic gate with modified connections. The buffers may be calibrated by selecting a fastest delay value that does not cause overflow in response to a calibration pulse.

A method of identifying a machine learning model configured for speech synthesis is described. The machine learning model may be included in a speech generator. Input data is received by the machine learning model which synthesizes speech data dependent on the input data. If the input data includes reference data such as a key word or phrase, the speech generator outputs a watermark comprising a predefined image that is visible on an audio spectrogram. The watermark may be generated by training the machine learning model directly or by a separate watermark generator.

A bearing race for a rolling element bearing (130) is described. The bearing race includes a metal region having an outer surface and an inner surface. The race has a groove in at least one of the outer surface and the inner surface. The groove extends around at least part of a circumference the outer surface or the inner surface. A piezoelectric layer (110) and a resistive layer (112) are located in the groove.

A time to digital converter including multiple delay lines and an output adder. Each delay line includes multiple series-coupled buffers receiving a start signal, multiple latches each having an input coupled to an output of a buffer and an output providing one of multiple binary values in response to a stop signal, and an adder. Each adder adds the binary values from corresponding latches in response to the stop signal and provides a corresponding one of multiple digital output values. The output adder adds the digital output values from the delay lines and provides a digital output value indicative of a measured delay. The start and stop signals may be buffered to provide multiple start and stop signals each provided one of the delay lines. Each start signal may be delayed by one of multiple different delays. Select logic may combine multiple delay lines into a longer delay line group.

The disclosure relates to methods of providing a response to a user query. A query is derived from the user query. An embedded query is obtained by passing the query through a first portion of a trained large language model. A semantically relevant element is obtained from an embedded database. The embedded database was obtained by embedding an initial database using the first portion of the trained large language model. The semantically relevant element is combined with the embedded query to form an augmented query. A response is provided (222) to the user query by passing (220) the augmented query through a second portion of the trained large language model.