Noise Mitigation in Automotive Ethernet Networks

Abstract

The in-vehicle Ethernet physical layer (PHY) transceiver (32) includes an analog front end (FE-46) and a digital processor (44). The FE is configured to receive analog Ethernet signals via a physical Ethernet link (36) while the Ethernet PHY transceiver is operating in the vehicle (24), and is configured to convert the received analog Ethernet signals into digital signals. The digital processor is configured to: maintain one or more noise distributions (84) that characterize noise signals of corresponding predefined noise types that are expected to disrupt the received analog Ethernet signals; classify actual noise signals present in the digital signals into one of the noise types using the noise distributions; and apply noise mitigation operations selected in response to the given noise type in response to determining that the actual noise signal matches a given noise type among the predefined noise types.

Description

[0001] Cross-reference to related applications

[0002] This application claims the benefit of U.S. Provisional Patent Application 62 / 987,204, filed March 9, 2020, the disclosure of which is incorporated herein by reference. Technical Field

[0003] This disclosure relates generally to communication systems, and more particularly to methods and systems for noise reduction in Ethernet transceivers. Background Art

[0004] Various applications, such as in-vehicle communication systems, certain industrial communication systems, and smart home systems, require communication at high data rates over relatively short distances. Several types of protocols and communication media have been proposed for such applications. For example, Ethernet communication over twisted-pair copper wire media is specified in "IEEE 802.3bw-2015-IEEE Standard for Ethernet Amendment 1: Physical Layer Specifications and Management Parameters for 100Mb / s Operation over a Single Balanced Twisted Pair Cable (100BASE-T1)" published in March 2015.

[0005] The above description is presented as a general overview of the relevant technology in the art and should not be construed as an admission that any information contained herein constitutes prior art to this patent application. Summary of the Invention

[0006] The embodiments described herein provide an in-vehicle Ethernet physical layer (PHY) transceiver including an analog front end (FE) and a digital processor. The FE is configured to receive analog Ethernet signals via a physical Ethernet link while the Ethernet PHY transceiver is operating in a vehicle, and is configured to convert the received analog Ethernet signals into digital signals. The digital processor is configured to: maintain one or more noise distributions characterizing noise signals of corresponding predefined noise types intended to disrupt the received analog Ethernet signals; classify actual noise signals present in the digital signals into one of the noise types using the noise distributions; and, in response to determining that the actual noise signal matches a given noise type among the predefined noise types, apply noise mitigation operations selected in response to the given noise type.

[0007] In some embodiments, the actual noise signal includes transient noise signals, and the digital processor is configured to detect an outbreak event of the transient noise signal and to classify the transient noise signal in the digital signal from the outbreak event onwards. In other embodiments, in response to determining that the actual noise signal matches a given noise type, the digital processor is configured to adjust the noise distribution associated with the given noise type based on the actual noise signal. In still other embodiments, the digital processor is configured to measure the quality of the digital signal and is configured to apply noise mitigation operations based on the given noise type and the measured quality by reconfiguring the operation of one or both of the analog FE and the digital processor.

[0008] In one embodiment, the digital processor is configured to apply noise mitigation operations by reconfiguring one or more of the frequency response, gain parameter, and clock rate of the filter in one or both of the analog FE and the digital processor. In another embodiment, the digital processor is configured to generate a sequence of multiple two-dimensional data structures based on a digital signal, each two-dimensional data structure including multiple frequency domain vectors derived from multiple time samples of the digital signal, and is configured to classify the noise signal by applying a machine learning model to the sequence of two-dimensional data structures. In yet another embodiment, the digital processor is configured to specify one or more characteristics selected from a list in the noise distribution, the list including: the source of the noise signal, the frequency pattern of the noise signal, the time progression characteristics of the noise signal, and the signal strength of the noise signal.

[0009] In some embodiments, the digital processor is configured to obtain one or more noise distributions directly from another vehicle, or to obtain one or more noise distributions by accessing cloud storage of noise distributions that share noise types among multiple vehicles. In other embodiments, the in-vehicle Ethernet PHY transceiver is one of multiple interconnected in-vehicle Ethernet PHY transceivers in the vehicle, and the digital processor is configured to maintain noise distributions not used by at least another Ethernet PHY transceiver among the multiple in-vehicle Ethernet PHY transceivers. In still other embodiments, the analog FE is configured to receive, in actual noise signals, electromagnetic interference (EMI) caused by one or more of the vehicle's electronic components, mechanical components, electromechanical components, and electromagnetic radiation sources external to the vehicle.

[0010] In one embodiment, the digital processor is configured to: i) apply noise reduction operations within a predetermined time period, and ii) modify the noise reduction operations within the predetermined time period, based at least on the expected attenuation properties of the actual noise signal determined according to the noise type classification.

[0011] According to the embodiments described herein, a communication method is also provided, comprising: receiving an analog Ethernet signal via a physical Ethernet link while an Ethernet PHY transceiver is operating in a vehicle. The received analog Ethernet signal is converted into a digital signal. One or more noise distributions are maintained, which characterize noise signals of corresponding predefined noise types intended to disrupt the received analog Ethernet signal. Using the noise distributions, actual noise signals present in the digital signal are classified into one of the noise types. In response to determining that the actual noise signal matches a given noise type among the predefined noise types, noise mitigation operations selected in response to the given noise type are applied.

[0012] According to embodiments described herein, an in-vehicle system is provided, comprising a central processing unit (CPU) and an Ethernet network. The CPU is configured to be installed in a vehicle. The Ethernet network includes a plurality of in-vehicle Ethernet physical layer (PHY) transceivers configured to connect the CPU and peripheral devices in the vehicle via a physical Ethernet link. Each of the Ethernet PHY transceivers includes an analog front-end (FE) and a digital processor. The FE is configured to receive analog Ethernet signals via the physical Ethernet link and is configured to convert the received analog Ethernet signals into digital signals. The digital processor is configured to: maintain one or more noise distributions characterizing noise signals of corresponding predefined noise types intended to disrupt the received analog Ethernet signals; classify actual noise signals present in the digital signals into one of the noise types using the noise distributions; and, in response to determining that the actual noise signal matches a given noise type among the predefined noise types, apply noise mitigation operations selected in response to the given noise type.

[0013] In some embodiments, the digital processor is configured to maintain a noise distribution that is not used by at least one of a plurality of automotive Ethernet PHY transceivers. In other embodiments, the central processing unit is configured to, on behalf of the digital processor, perform at least a portion of the task of classifying the actual noise signal.

[0014] According to the embodiments described herein, a communication method is also provided, comprising: in an in-vehicle system including a central processing unit installed in a vehicle, and an Ethernet network including a given PHY transceiver comprising a plurality of in-vehicle Ethernet physical layer (PHY) transceivers connected to the central processing unit and peripheral devices in the vehicle via a physical Ethernet link, wherein an analog Ethernet signal is received by a given PHY transceiver via a physical Ethernet link. The received analog Ethernet signal is converted into a digital signal. A digital processor of the PHY transceiver maintains one or more noise distributions, which characterize noise signals of corresponding predefined noise types intended to disrupt the received analog Ethernet signal. Using the noise distributions, actual noise signals present in the digital signal are classified into one of the noise types. In response to determining that the actual noise signal matches a given noise type among the predefined noise types, a noise mitigation operation selected in response to the given noise type is applied.

[0015] This disclosure will be more fully understood from the following detailed description of embodiments thereof, taken in conjunction with the accompanying drawings, in which: Attached Figure Description

[0016] Figure 1 This is a schematic diagram illustrating a vehicle communication system according to an embodiment described herein;

[0017] Figure 2 This is a block diagram schematically illustrating the components of a PHY device for processing analog Ethernet signals according to embodiments described herein;

[0018] Figure 3 This is a schematic diagram illustrating multiple images generated from transient noise signals for classification using a machine learning model, according to embodiments described herein; and

[0019] Figure 4 This is a schematic flowchart illustrating a method for converting a signal into an image sequence for noise classification according to embodiments described herein. Detailed Implementation

[0020] The embodiments described herein provide improved methods and systems for noise mitigation in the physical layer (PHY) interface of Ethernet links. The embodiments described herein are described in the context of automotive applications (e.g., systems that collect data from sensors within a vehicle). However, this choice is merely for clarity. The disclosed techniques are equally applicable to other applications, such as industrial and / or smart home networks.

[0021] In some embodiments, the in-vehicle communication system includes a central processing unit (CPU) and multiple sensors, which are interconnected via an Ethernet network. In an example network implementation, the CPU and sensors are connected to corresponding ports of an Ethernet switch via physical Ethernet links. Each sensor in the CPU and the sensors is connected to the Ethernet link via an Ethernet MAC device coupled to an Ethernet PHY device. For simplicity, the Ethernet PHY device is also referred to as an Ethernet PHY transceiver or simply an Ethernet transceiver. Sensors adapted in the vehicle include, for example, cameras, radar sensors, etc. By communicating over the Ethernet link, the PHY device transmits data collected by the sensors to the CPU and transmits control information from the CPU to the sensors.

[0022] In extreme environments, vehicle Ethernet networks are typically required to transmit large amounts of data at high speed and low latency. On the other hand, for safety reasons, in-vehicle communication must be highly reliable. Electronic components within a vehicle can be susceptible to electromagnetic interference (EMI) from various sources, which can disrupt electrical signals (such as those carrying data on an Ethernet link), leading to communication errors and system malfunctions.

[0023] EMI sources can be located inside or outside the vehicle. Examples of external EMI sources include, for example, radio towers, power lines, and airport radar, among many others. Examples of internal EMI sources include, for example, the vehicle engine and other mechanical and electromechanical components, windshield wipers, mobile phones, infotainment systems, etc.

[0024] In principle, noise generated by EMI can be mitigated using electromagnetic compatibility (EMC) methods designed to ensure proper operation of underlying devices in ordinary electromagnetic environments. However, in automotive Ethernet networks, EMC is often insufficient, and the PHY device must still handle residual EMI. Furthermore, it is desirable for EMI mitigation to be adaptable to various types of interference that may be encountered in the field, as well as future types of interference that cannot be pre-characterized.

[0025] One challenge in mitigating EMI in a vehicle environment is detecting and eliminating transient noise signals while minimizing their impact on communication throughput and latency. In this context, transient noise signals consist of one or more pulses, each composed of a short-duration, high-amplitude pulse followed by a decaying low-frequency oscillation.

[0026] In some embodiments, the in-vehicle Ethernet physical layer (PHY) transceiver includes an analog front end (FE) and a digital processor. The analog FE is configured to receive analog Ethernet signals via a physical Ethernet link while the Ethernet PHY transceiver is operating in the vehicle, and is configured to convert the received analog Ethernet signals into digital signals. The digital processor is configured to: maintain one or more noise distributions that characterize noise signals of corresponding predefined noise types that can be detected and are expected to disrupt the received analog Ethernet signals; classify actual noise signals present in the digital signals into a noise type using the noise distributions; and, in response to determining that an actual noise signal matches a given noise type from the predefined noise types, apply noise mitigation operations selected in response to the given noise type.

[0027] As mentioned above, the actual noise signal may include transient noise signals. In one embodiment, in mitigating transient noise signals, the digital processor is configured to detect the onset of a transient noise signal and is configured to classify the transient noise signal in the digital signal from the onset of the onset event.

[0028] The characteristics of the actual noise signal typically deviate slightly from those specified in the matched noise distribution; that is, even in the case of an imprecise match, the noise signal can be classified as the closest noise type. In some embodiments, in response to determining that the actual noise signal matches a given noise type (precisely or nearly), the digital processor adapts the distribution of the matched noise type based on the actual noise signal. This can improve classification performance in subsequent classification operations.

[0029] In some embodiments, the digital processor is configured to measure the quality of a digital signal and is configured to apply noise reduction operations based on a given noise type and the measured quality by reconfiguring the operation of one or both of the analog FE and the digital processor. The noise reduction operations may include, for example, reconfiguring one or more of the frequency response, gain parameters, and clock rates of the filters in the analog FE, the digital processor, or both.

[0030] In some embodiments, the classification of noise signals is based on machine learning methods. In such embodiments, a digital processor is configured to generate a sequence of multiple two-dimensional data structures based on a digital signal, each data structure comprising multiple frequency domain vectors derived from multiple time samples of the digital signal. The digital processor classifies the noise signal by applying a convolutional neural network (CNN) machine learning model to the sequence of two-dimensional data structures. Alternatively, any other suitable machine learning model may be used.

[0031] As mentioned above, each noise type has a corresponding noise distribution that includes one or more characteristics. In one embodiment, characteristics may include, for example, one or more of the following: noise signal source, frequency pattern of the noise signal, time progression characteristics of the noise signal, and signal strength of the noise signal, or other suitable characteristics. Vehicles may, for example, use cloud services to efficiently share one or more noise distributions with other vehicles in their vicinity. Alternatively or additionally, in one embodiment, the vehicle supports direct vehicle-to-vehicle data transmission (e.g., using wireless communication) for sharing noise distributions and other data.

[0032] Vehicles typically include multiple PHY devices. However, different PHY devices may suffer from different types of EMI, depending on factors such as the location of the PHY device within the vehicle. In some embodiments, different PHY devices may maintain different noise distribution sets depending on the desired noise type.

[0033] In some embodiments, the digital processor is configured to: i) apply noise reduction operations for a predetermined time period, and ii) modify the noise reduction operations for a predetermined time period, based at least on the expected attenuation properties of the actual noise signal determined according to the noise type classification.

[0034] In the disclosed technology, the PHY device maintains a noise distribution characterizing the expected type of noise signal (including transient noise signals). In response to detecting the presence of a noise signal in the expected signal, the noise signal is classified into the corresponding noise type using the noise distribution, and appropriate noise mitigation operations are applied. For example, a machine learning model is used to train the noise distribution for the expected noise type and use it for classification. The predefined noise distribution is suitable for the actual noise types encountered in the field. Furthermore, future additions to the noise distribution do not require design or implementation modifications.

[0035] Figure 1 This is a schematic block diagram of an in-vehicle communication system 20 according to an embodiment described herein. The communication system 20 is installed in a vehicle 24 and includes multiple sensors 28, multiple microcontrollers 30, multiple Ethernet physical layer (PHY) devices 32 (also referred to as Ethernet transceivers), and an Ethernet switch 34. The PHY devices 32 and the switch 34 are interconnected via point-to-point physical Ethernet links 36. Among other components, the switch 34 includes MAC devices (not shown) coupled to one or more PHY devices 32. Typically, the MAC devices of the switch 34 are replicated on each port of the switch 34, i.e., replicated on each PHY device 32 coupled to the switch.

[0036] In various embodiments, sensor 28 may include any suitable type of sensor. Several non-limiting examples of sensors include cameras, speed sensors, accelerometers, audio sensors, infrared sensors, radar sensors, lidar sensors, ultrasonic sensors, rangefinders, or other nearby sensors.

[0037] PHY device 32 typically operates at least in part according to one or more standards in the IEEE 802.3 Ethernet standard (e.g., IEEE 802.3bw-2015 cited above). Although the techniques described herein primarily relate to the physical layer, in one embodiment, PHY device 32 may also perform media access control (MAC) functions.

[0038] Depending on the applicable Ethernet standard, link 36 may include any suitable physical medium. In the embodiments described herein, although not mandatory, each link 36 includes a single pair of wires, such as a single twisted-pair link that may be optionally shielded. In alternative embodiments, link 36 may include a single-ended wire link, and does not necessarily have to be Ethernet-compatible.

[0039] In this example, each sensor 28 is connected to a corresponding microcontroller 30, which in turn is connected to a corresponding PHY device 32. Each sensor's PHY device 32 is connected to a peer PHY device 32 via a link 36, which is coupled to a port of a switch 34. On the sensor side of a given link, the microcontroller 30 acts as a Media Access Control (MAC) controller. On the switch side of a given link, the MAC function is performed by the switch 34. Therefore, the switch 34 and the microcontroller 30 are also referred to herein as a MAC device, a host, or a system-on-a-chip (SoC). In some embodiments, the PHY circuitry and the circuitry performing the MAC function (e.g., the microcontroller or the switch) are integrated into the same device. In this example, the central processing unit 40 (on the right-hand side of the figure) is connected via the switch 34 but is not directly connected to any sensor. In one embodiment, alternatively or additionally, the central processing unit 40 may be directly connected to one or more sensors (or the microcontroller 30 connected to the sensors).

[0040] PHY device 32, switch 34, and link 36 form an Ethernet network within vehicle 24. Using this vehicle Ethernet network, central processing unit 40 sends control messages to sensor 28 and receives information captured by the sensors. Figure 1 The Ethernet topology shown is given as an example, and other suitable topologies may also be used, such as topologies that include more than one switch (such as switch 34) and / or more than a single central processing unit 40.

[0041] In one embodiment, Figure 1 The illustration at the bottom shows the internal structure of the PHY device 32. The PHY device 32 includes a digital processor 44 and an analog front-end (FE) 46. In one embodiment, the digital processor 44 includes a MAC interface 48 configured to communicate with a MAC device such as a switch 34, a sensor 28, or a central processing unit 40. The analog FE 46 includes a PHY Media Dependent Interface (MDI) 50 configured to transmit and receive analog Ethernet signals over an MDI channel that includes a physical link 36 (e.g., a twisted-pair link).

[0042] In the following description, the terms "transmit direction" and "receive direction" refer to the PHY32 transmitting analog Ethernet signals to the MDI channel and receiving analog Ethernet signals from the MDI channel via the MDI 50, respectively.

[0043] In the transmitting direction, the digital processor 44 of the PHY device 32 receives data in digital form via the MAC interface 48. The data is processed by the digital processor 44 and further processed by the analog FE 46, which generates an analog Ethernet signal carrying the data. The analog FE transmits the analog Ethernet signal via the MDI 50. In the receiving direction, the analog FE 46 receives the analog Ethernet signal carrying the data via the MDI 50. The analog Ethernet signal is processed by the analog FE 46 and subsequently processed by the digital processor 44, which recovers the data. The digital processor then transmits the recovered data to the peer MAC device via the MAC interface 48.

[0044] In one embodiment, the digital processor 44 includes a physical coding sublayer (PCS) 52 configured to perform various digital data processing operations, such as data encoding and decoding, and data scrambling and descrambling, to name just a few.

[0045] The digital processor 44 also includes a digital signal processor (DSP) 54, which primarily functions as a digital receiver for processing signals received from the analog FE 46. The digital processor 44 includes a noise reduction unit 56 configured to detect the presence of noise in the received signal, identify the noise type, and adjust the operation of the analog FE 46 and the DSP according to the noise type. The digital processor 44 includes a digital clock generator 58 configured to generate a digital clock signal used by various components within the digital processor.

[0046] The analog FE 46 includes a digital-to-analog converter (DAC) 62 configured to receive data processed by the PCS 52 and convert that data into an analog Ethernet signal. The hybrid 64 is configured to separate the Ethernet signals in the transmit and receive directions. In the transmit direction, the hybrid 64 transmits the analog Ethernet signal generated by the DAC 62 via the MDI 50. In the receive direction, the hybrid 64 transmits an analog Ethernet signal carrying the data received via the MDI 50 (for processing via the receive path described herein).

[0047] The receiving path comprises analog and digital sections. The analog section includes a high-pass filter (HPF) 66, an analog matched filter (AMF) 68, and an analog-to-digital converter (ADC) 70. The digital section includes a DSP 54 and a PCS 52. The HPF 66 filters the analog Ethernet signal to eliminate frequency components below a predefined cutoff frequency. The filtered signal passes through the AMF 68 to the ADC 70, which is configured to digitize the analog Ethernet signal into a sequence of digital samples provided to the DSP 54.

[0048] AMF 68 is configured to compensate for Gaussian noise present in the signal, which is typically caused by the underlying channel. In some embodiments, applying AMF is optional and may be omitted. Gain and timing components 72 control the operation of ADC 70 and automatic gain control (AGC) components 74 for optimal adjustment of the sampling rate and gain in ADC 70. Analog FE 46 includes an analog clock generator 76 configured to generate an analog clock signal used by various components within analog FE 46. The digital portion of the receive path includes DSP 54 and PCS 52 for recovering data carried in the analog Ethernet signal.

[0049] Figure 2 This is a schematic block diagram illustrating the components of a PHY device 32 for processing analog Ethernet signals according to an embodiment described herein.

[0050] As described above, PHY device 32 uses analog FE 46 to process Ethernet signals, followed by digital processor 44. The digital processor's DSP 54 receives a digitized version of the Ethernet signal, which may contain noisy signals, from ADC 70. In some embodiments, the analog FE is configured to receive electromagnetic interference (EMI) in the noisy signal caused by one or more sources, such as, for example, electronic components of the vehicle, mechanical components of the vehicle, electromechanical components of the vehicle, and electromagnetic radiation sources outside the vehicle.

[0051] DSP 54 includes an equalizer 78 and an echo canceller 80. Equalizer 78 includes one or more of any suitable type of adaptive equalizer, such as a feedforward equalizer (FFE), a decision feedback equalizer (DFE), or other suitable equalizers. Equalizer 78 is configured to reduce the influence of the underlying channel on the received signal. Echo canceller 80 is configured to reduce echo signals created, for example, by reflections of the desired signal. The output of echo canceller 80 is provided to PCS 52.

[0052] In some embodiments, the DSP 54 monitors the quality of the received signal (e.g., as described below) and detects that the quality has degraded below a predefined quality threshold. In one embodiment, the DSP 54 monitors signal quality by measuring the signal-to-noise ratio (SNR), for example, based on evaluating the opening of the eye diagram of the received signal. In alternative embodiments, the DSP 54 may use any other suitable method to measure signal quality.

[0053] In some embodiments, the noise reduction unit 56 is configured to maintain (e.g., in memory 86) one or more noise distributions 84, which characterize noise signals of corresponding predefined noise types expected to be detected in the operating environment and potentially disrupt received analog Ethernet signals. Each noise distribution specifies one or more characteristics of the noise signal, such as, for example, the noise signal source, the frequency pattern of the noise signal, and the signal strength of the noise signal. Other noise characteristics may include, for example, transition and time progression characteristics of the noise signal. Different noise types may be associated with corresponding noise distributions having the same or different sets of characteristic parameters.

[0054] The noise reduction unit includes a noise classifier 82 configured to: receive a digitized version of an analog Ethernet signal from the analog FE 46; detect actual noise signals contained within the analog Ethernet signal; and classify the actual noise signals into a corresponding noise type that matches a given noise type among predefined noise types. Note that the classification does not require an exact match. In other words, the noise signal can match a given noise type even when the characteristics of the noise signal deviate from the characteristics of a matching distribution. In one embodiment, in response to determining that the actual noise signal matches a given noise type (which may not be an exact match), the noise classifier is configured to adjust the noise distribution associated with the given noise type based on the actual noise signal. This feature allows the noise distribution to be fine-tuned to the actual noise signal encountered in the field.

[0055] In some embodiments, the actual noise signal includes transient noise signals. In such embodiments, the noise classifier 82 is configured to detect the onset of transient noise signals and classify transient noise signals in the digital signal from the onset of the onset event.

[0056] Noise mitigation planner 88 is configured to receive a noise type determined by noise classifier 82. Based on this noise type, noise mitigation planner 88 applies appropriate noise mitigation operations. In some embodiments, by classifying transient noise into noise types, noise mitigation operations can be proactively applied over the expected duration of the transient noise based on the classified noise type. In some noise mitigation operations, the noise mitigation operations can be modified and optimally applied over the expected lifetime of the noise signal, as determined based on the noise type classification.

[0057] For example, in one embodiment, the noise reduction planner controls analog FE 46, DSP 54, or both to adjust their operation to reduce or eliminate noise signals. In some embodiments, the noise reduction planner controls analog FE 46 to adjust one or more of the following: the impulse response (or frequency response) of HPF 66, the gain of AGC 74, gain and timing 72, and the clock rate of analog clock generator 76. In some embodiments, the noise reduction planner controls DSP 54 to adjust one or more of the following: digital gain, the impulse response (or frequency response) of the filter implementing equalizer 78, the impulse response (or frequency response) of the filter implementing echo canceller 80, and the clock rate of digital clock generator 58.

[0058] In some embodiments, DSP 54 is configured to measure the quality of a digital signal as described above (e.g., at the output of equalizer 78, echo canceller 80, or both). In such embodiments, DSP 54 provides the measured quality to noise classifier 82, which uses the quality to enhance classification performance. The measured quality can, for example, be used for reinforcement training on the noise classifier to determine how well the currently classified noise matches the observed actual noise. As another example, the measured quality provides additional feedback to a mitigation planner to improve the noise mitigation operations taken.

[0059] In some embodiments, to further enhance classification performance, the noise classifier 82 receives additional information after applying noise reduction operations, such as, for example, digitized Ethernet signals provided to the DSP 54 and / or processed signals output by the DSP 54.

[0060] The noise distribution 84 (or a portion thereof) is typically determined and stored in the memory 86 of the digital processor before the noise reduction unit 56 is enabled. In some embodiments, the noise distribution is determined by training a suitable machine learning model using a noise signal of the expected noise type. An example machine learning model suitable for noise type classification is a convolutional neural network (CNN). Alternatively, other suitable machine learning models may be used.

[0061] In one embodiment, instead of training a machine learning model to determine noise distribution 84, or in addition to training a machine learning model to determine noise distribution 84, digital processor 44 is configured to obtain one or more noise distributions from noise distribution 84 by accessing cloud storage that shares noise distributions among multiple vehicles. In such an embodiment, a vehicle aware of the relevant noise type can upload the corresponding noise distribution to the cloud, allowing the noise distribution to be selectively shared with other vehicles nearby or near the noise source. Alternatively or additionally, in one embodiment, a vehicle receives one or more noise distributions from another vehicle using direct data transmission communication.

[0062] In some embodiments, the PHY devices 32 in vehicle 24 use the same set of noise distributions 84. This configuration simplifies the training and distribution of noise distributions among the PHY devices. However, this is not mandatory, and in alternative embodiments, different PHY devices may use different sets of noise distributions. For example, this configuration may be useful when different PHY devices in the vehicle are subjected to different types of EMI or noise. For instance, a PHY device located near the vehicle's engine may be affected by noise types that a PHY device located far from the engine might not experience. In an example embodiment, PHY device 32 maintains a dedicated noise distribution not used by at least one other PHY device in vehicle 24.

[0063] Classifying noise signals into corresponding noise types can be tedious. In some embodiments, such as where the central processing unit 40 is more powerful than the PHY device 32, the central processing unit may represent the digital processor 44 of one or more PHY devices to assist in performing at least part of the task of classifying actual noise signals. This may be applicable, for example, when it is necessary to adjust the matched noise distribution to better characterize the actual noise signals encountered. In some embodiments, the central processing unit retrieves noise signals observed by individual PHYs and performs noise classification based on combined information to improve the reliability of classifying noise signals a experienced by the entire system or vehicle, as well as the reliability of local classification at each individual PHY.

[0064] like Figure 1 and Figure 2The configuration of the communication system 20 and its components (such as the internal structure of the PHY device 32) shown is an example configuration described for clarity only. In alternative embodiments, any other suitable configuration may be used. For example, the disclosed techniques can be used in any other suitable network or link topology, such as in a point-to-point Ethernet link between two hosts (e.g., microcontrollers) that do not actually cross a switch. As another example, the disclosed techniques can be used in a point-to-point Ethernet link between two switches. For clarity, elements that are not essential for understanding the disclosed techniques are omitted from the figures.

[0065] The different elements and various components of the communication system 20 can be implemented using dedicated hardware or firmware (such as hard-wired or programmable logic, for example, in an application-specific integrated circuit (ASIC) or a field-programmable gate array (FPGA). Additionally or alternatively, some functions (e.g., the functions of the digital processor 44, such as the functions of the DSP 54) can be implemented using software and / or using a combination of hardware and software elements.

[0066] In some embodiments, the digital processor 44 includes a programmable processor (e.g., a DSP54 or other suitable processor, such as a vector processor), which is programmed in software to perform the functions described herein. For example, the software can be downloaded electronically to any processor via a network, or alternatively or additionally, the software can be provided and / or stored on a non-transient tangible medium such as magnetic, optical, or electronic memory.

[0067] Figure 3 This is a schematic illustration of multiple images generated from transient noise signals for classification using a machine learning model, according to embodiments described herein.

[0068] Figure 3 A transient noise signal 100 is described, which may undesirably disrupt the analog Ethernet signal received in the PHY device 32. As shown, the transient signal is typically characterized by a sudden rise and a much slower decay. This type of transient noise signal can be created by various electromechanical or mechanical components within the vehicle 24, or by noise sources outside the vehicle. The transient noise signal 100 is given as an example. Alternatively, transient noise signals with other characteristics such as amplitude and shape are also applicable.

[0069] In some embodiments, the noise classifier 82 in the PHY device 32 uses a CNN model to classify noise signals (e.g., 100), which has been previously trained to create one or more noise distributions corresponding to the noise type. In some embodiments, the CNN model classifies input signals that have been captured and converted into a sequence of two-dimensional data structures (also referred to as "images"). Figure 3 The image depicts four images 104 corresponding to the transient noise signal 100. Figure 3 In the example, each image includes multiple column vectors of pixels. In this example, pixels include real-valued numbers corresponding to gray levels within a predefined range and resolution. Alternatively, other types of pixels (e.g., color pixels or complex-valued pixels) may also be used. The column vectors in image 104 correspond to corresponding time periods of the signal. For example, column vector 108 corresponds to time period 112 of the transient noise signal.

[0070] Figure 4 This is a schematic flowchart illustrating a method for converting a signal into an image sequence for noise classification according to embodiments described herein.

[0071] This method will be described as being executed by the DSP 54 of the digital processor 44. In some embodiments, the digital processor 44 executes continuously. Figure 4 This method helps detect noise signal outbreaks, rather than being triggered by a DSP. In one embodiment, the method may be activated, for example, in response to the detection of a transient noise signal outbreak.

[0072] The method begins at window reception operation 150, where the DSP 54 collects a time window comprising N samples of an analog Ethernet signal sampled at a suitable sampling rate Fs. The window size N and sampling rate Fs are typically predefined design parameters. At time-to-frequency domain transformation operation 154, the DSP 54 calculates the spectral density (PSD) of the signal within the time window. The DSP can calculate the PSD, for example, by applying any suitable transform operation such as the Discrete Fourier Transform (DFT) to the samples within the time window. At image generation operation 158, the DSP 54 supplements M consecutive past PSDs into column vectors to produce an M-by-N image. The pixels of the image are derived from the PSD values ​​in the supplemented M vectors.

[0073] In the next PSD calculation operation 162, DSP 54 calculates the subsequent PSD of the signal in the subsequent time window. Let IM(n) and IM(n+1) represent the most recent and subsequent images, respectively. In the subsequent image generation operation 166, based on the IM(n) image, DSP 54 generates the IM(n+1) image by (i) removing the leftmost vector from IM(n) and (ii) appending the subsequent PSD vector to the rightmost vector of IM(n+1).

[0074] At the termination query operation 170, DSP 54 checks whether the noisy signal region has ended. If not, it loops back to operation 162 to calculate the subsequent PSD. Otherwise, the current region of interest containing noise has ended, and DSP 54 continues to apply noise classification as described above.

[0075] It should be noted that the above embodiments are cited by way of example, and the invention is not limited to what has been specifically shown and described above. More precisely, the scope of the invention includes combinations and sub-combinations of the various features described above, as well as variations and modifications thereof that would occur to those skilled in the art upon reading the foregoing description and not disclosed in the prior art. Documents incorporated herein by reference are considered part of this application, and the definitions in this specification should be considered only to the extent that any terms are defined in these incorporated documents in a manner that conflicts with the express or implied definitions in this specification.

Claims

1. A vehicle-mounted Ethernet physical layer PHY transceiver, characterized in that, include: The analog front-end FE is configured to receive analog Ethernet signals via a physical Ethernet link while the Ethernet PHY transceiver is operating in the vehicle, and is configured to convert the received analog Ethernet signals into received digital signals. as well as The digital processor is configured as follows: One or more noise distributions are maintained, which characterize noise signals of corresponding predefined noise types that are intended to disrupt the received analog Ethernet signals, wherein each noise distribution includes a two-dimensional 2D representation of the time-domain noise signal, the two-dimensional representation including multiple frequency-domain vectors; Based on the received digital signal, a sequence of two-dimensional data structures is generated, each two-dimensional data structure including multiple frequency domain vectors; By attempting to match the characteristics of the actual noise signal present in the received digital signal with the characteristics of the noise distribution, the actual noise signal present in the received digital signal is classified into one of the noise types using the noise distribution; as well as In response to determining that the received digital signal includes an actual noise signal that matches a given noise type among the predefined noise types, a noise reduction operation is selected in response to the given noise type, and the selected noise reduction operation is applied. In response to determining that the actual noise signal matches the given noise type, the digital processor is configured to adjust the noise distribution associated with the given noise type based on the actual noise signal.

2. The vehicle-mounted Ethernet PHY transceiver of claim 1, wherein the actual noise signal in the received digital signal includes a transient noise signal, and wherein the digital processor is configured to detect an event of occurrence of the transient noise signal and is configured to classify the transient noise signal in the received digital signal from the event of occurrence.

3. The vehicle-mounted Ethernet PHY transceiver of claim 1, wherein the digital processor is configured to measure the quality of the received digital signal and is configured to apply the noise mitigation operation by reconfiguring the operation of one or both of the analog FE and the digital processor based on the given noise type and the measured quality.

4. The vehicle-mounted Ethernet PHY transceiver of claim 3, wherein the digital processor is configured to apply the noise mitigation operation by reconfiguring one or more of the frequency response, gain parameter, and clock rate of the filter in one or both of the analog FE and the digital processor.

5. The vehicle-mounted Ethernet PHY transceiver of claim 1, wherein the digital processor is configured to classify the actual noise signal by applying a machine learning model to the sequence of the two-dimensional data structure.

6. The vehicular Ethernet PHY transceiver of claim 1, wherein the digital processor is configured to specify one or more characteristics selected from a list in the noise distribution, the list comprising: The noise signal source, the noise signal frequency pattern, the noise signal time progression characteristics, and the noise signal strength.

7. The vehicular Ethernet PHY transceiver of claim 1, wherein the digital processor is configured to: obtain one or more noise distributions from the noise distribution directly from another vehicle, or obtain one or more noise distributions from the noise distribution by accessing cloud storage of noise distributions that share noise types among multiple vehicles.

8. The vehicle Ethernet PHY transceiver of claim 1, wherein the vehicle Ethernet PHY transceiver is part of a plurality of interconnected vehicle Ethernet PHY transceivers in the vehicle, and wherein the digital processor is configured to maintain a noise distribution not used by at least another Ethernet PHY transceiver among the plurality of vehicle Ethernet PHY transceivers.

9. The vehicle-mounted Ethernet PHY transceiver of claim 1, wherein the analog FE is configured to: receive, in the actual noise signal, electromagnetic interference (EMI) caused by one or more of the vehicle's electronic components, the vehicle's mechanical components, the vehicle's electromechanical components, and electromagnetic radiation sources outside the vehicle.

10. The vehicular Ethernet PHY transceiver of claim 1, wherein the digital processor is configured to: at least based on the expected attenuation attribute of the actual noise signal determined according to the noise type classification: i) apply the noise mitigation operation for a predetermined time period, and ii) modify the noise mitigation operation during the predetermined time period.

11. The vehicle-mounted Ethernet PHY transceiver of claim 1, wherein the digital processor is configured to: detect the actual noise signal present in the received digital signal, and classify the detected actual noise signal into one of the noise types.

12. A communication method, characterized in that, include: In the automotive Ethernet physical layer PHY transceiver While the Ethernet PHY transceiver is operating in the vehicle, it receives analog Ethernet signals via a physical Ethernet link and converts the received analog Ethernet signals into received digital signals. One or more noise distributions are maintained, which characterize noise signals of corresponding predefined noise types that are intended to disrupt the received analog Ethernet signals, wherein each noise distribution includes a two-dimensional 2D representation of the time-domain noise signal, the two-dimensional representation including multiple frequency-domain vectors; Based on the received digital signal, a sequence of two-dimensional data structures is generated, each two-dimensional data structure including multiple frequency domain vectors; By attempting to match the characteristics of the actual noise signal present in the received digital signal with the characteristics of the noise distribution, the actual noise signal present in the received digital signal is classified into one of the noise types using the noise distribution; as well as In response to determining that the received digital signal includes an actual noise signal that matches a given noise type among the predefined noise types, a noise reduction operation is selected in response to the given noise type, and the selected noise reduction operation is applied. The communication method further includes: In response to determining that the actual noise signal matches the given noise type, the noise distribution associated with the given noise type is adjusted based on the actual noise signal.

13. The communication method of claim 12, wherein the actual noise signal in the received digital signal includes transient noise signals, and wherein classifying the actual noise signal includes: Detect the occurrence event of the transient noise signal; And classify the transient noise signals in the received digital signals from the start of the outbreak event.

14. The communication method of claim 12, wherein applying the noise reduction operation comprises: Measure the quality of the received digital signal; Based on the given noise type and the measured quality, the noise mitigation operation is applied by reconfiguring the operation of one or both of the analog FE and the digital processor.

15. The communication method of claim 14, wherein applying the noise reduction operation comprises: One or more of the frequency response, gain parameters, and clock rate of the filter are reconfigured in one or both of the analog FE and the digital processor.

16. The communication method according to claim 12, wherein classifying the actual noise signal comprises: The machine learning model is applied to the sequence of the two-dimensional data structure.

17. The communication method of claim 12, wherein maintaining the noise distribution comprises: In the noise distribution, one or more characteristics selected from a list are specified, the list including: the source of the noise signal, the frequency pattern of the noise signal, the time progression characteristics of the noise signal, and the signal strength of the noise signal.

18. The communication method of claim 12, wherein maintaining the noise distribution comprises: One or more noise distributions in the noise distribution can be obtained directly from another vehicle, or by accessing cloud storage that shares noise distributions of the same type among multiple vehicles.

19. The communication method of claim 12, wherein the in-vehicle Ethernet PHY transceiver belongs to a plurality of interconnected in-vehicle Ethernet PHY transceivers in the vehicle, and wherein maintaining the noise distribution comprises: Maintain the noise distribution that is not used by at least one of the plurality of automotive Ethernet PHY transceivers.

20. The communication method according to claim 12, wherein receiving the analog Ethernet signal comprises: In the actual noise signal, electromagnetic interference (EMI) is received caused by one or more of the vehicle's electronic components, mechanical components, electromechanical components, and electromagnetic radiation sources outside the vehicle.

21. The communication method according to claim 12, comprising: Based at least on the expected attenuation properties of the actual noise signal determined according to the noise type classification: i) applying the noise reduction operation for a predetermined time period, and ii) modifying the noise reduction operation during the predetermined time period.

22. The communication method according to claim 12, wherein classifying the actual noise signal comprises: The actual noise signal present in the received digital signal is detected, and the detected actual noise signal is classified into one of the noise types.

Images