User interface for tensor builder to construct images for input to machine learning
By designing user interfaces and methods, allowing users to build tensor images and input into deep learning networks, the problem of inefficient construction and use of tensor images in the prior art is solved, and the prediction accuracy and efficiency of machine learning in testing and measurement is improved. .
Patent Information
- Application Number
- CN202410653652.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2024-05-15
- Filing Date
- 2024-05-24
- Publication Date
- 2024-11-26
AI Technical Summary
Existing machine learning systems, when applied to test and measurement, are difficult to effectively construct and use tensor images as input, resulting in limited prediction accuracy and efficiency.
A user interface and method is designed to allow users to select tensor types and related parameters, build tensor images through waveforms or S-parameters, and input them into deep learning networks for prediction in test and measurement applications. This interface provides construction options for waveform vectors, waveform images and S-parameter images, supports a variety of data organization methods and noise processing, and enhances the efficient processing capabilities of data.
Improves the prediction accuracy and efficiency of machine learning systems in testing and measurement, can process more data and improves the vertical resolution of data, and enhances the detection ability of low-level signals.
Smart Images

Figure CN119024997A_ABST
Abstract
Description
[0001] CROSS-REFERENCE TO RELATED APPLICATIONS
[0002] The present disclosure is a non-provisional application of U.S. Provisional Application No. 63 / 468,779, filed on May 24, 2023, and claims the benefit of that provisional application, entitled “USER INTERFACE FOR A TENSOR BUILDER TO CONSTRUCT IMAGES FOR INPUT TO MACHINE LEARNING,” the disclosure of which is incorporated herein by reference in its entirety. Technical Field
[0003] The present disclosure relates to test and measurement systems and methods, and more particularly, to applying machine learning to the measurement of signals from a device under test. Background Art
[0004] U.S. Patent Application No. 17 / 747,954, filed on May 18, 2022, and entitled “WAVEFORMDATABASEBASEDMACHINE LEARNING FORMEASUREMENT” (hereinafter referred to as the “’954 Application,” the contents of which are hereby incorporated by reference into this application), describes systems and methods for applying a trained machine learning system to make predictions about measurements of signals received from a device under test (DUT). When the DUT is, for example, an optical transceiver, an example signal measurement is the transmitter and dispersion eye closure quad (TDECQ). The input to the machine learning system described in the ’954 Application for both training and runtime is in the form of a tensor image, which is constructed using a graphical representation of specific short pattern segments within a long waveform acquired from the DUT.
[0005] U.S. patent application Ser. No. 17 / 701,186, filed on March 24, 2022, and entitled “OPTICAL TRANSMITTER TUNING USING MACHINE LEARNING AND REFERENCE PARAMETERS” (hereinafter referred to as the “’186 application,” the contents of which are hereby incorporated by reference into this application), describes the application of a trained machine learning system to predict optimal tuning parameters of an optical transceiver DUT.
[0006] U.S. Patent Application No. 18 / 199,846, filed on May 19, 2023, and entitled “AUTOMATED CAVITY FILTERTUNING USING MACHINE LEARNING” (hereinafter “the '846 Application,” the contents of which are hereby incorporated by reference into this application), describes systems and methods for applying a trained machine learning system to predict optimal locations of tuning elements in a tunable device under test (e.g., a cavity filter). The input to the machine learning system described in the '846 Application for both training and runtime is in the form of tensor images, which are constructed using graphical representations of scattering parameters (S-parameters) or other parameters that characterize the DUT.
[0007] A tensor represents a way to efficiently provide information to a machine learning system for training or operation. U.S. Patent Application No. 18 / 510,234, entitled "METHOD FOR THREE-DIMENSIONAL TENSORBUILDER FOR INPUT TO MACHINE LEARNING" (hereinafter referred to as the "'234 application", the contents of which are hereby incorporated by reference into this application), describes a system and method for creating an efficient 3D RGB tensor that incorporates reference parameters or S-parameters or other types of data for three channels. BRIEF DESCRIPTION OF THE DRAWINGS
[0008] Figure 1 An embodiment of a test and measurement device is shown.
[0009] Figure 2 An embodiment of a user interface of a tensor builder is shown, which allows a user to select a tensor type and related parameters.
[0010] Figure 3 Another embodiment of a user interface for a tensor builder is shown, which allows a user to select a tensor type and related parameters.
[0011] Figure 4 Another embodiment of a user interface for a tensor builder is shown, which allows a user to select a tensor type and related parameters. DETAILED DESCRIPTION
[0012] According to embodiments of the present disclosure, many applications can be built using waveforms or S-parameters as input data for deep learning networks in machine learning applications. Embodiments of the present disclosure generally include a user interface and method for building tensor images that are input to deep learning networks for prediction in test and measurement applications.
[0013] Figure 1Embodiments of test and measurement devices providing one or more user interfaces are shown. Embodiments may relate to test and measurement instruments having one or more processors that execute code that causes the processors to perform various tasks. A machine learning system may take the form of a programming model that operates on one or more processors. Figure 1 An embodiment of a test setup in an example of an optical transmitter 30 as a device under test (DUT) is shown. The test setup includes a test and measurement system, which may include a test and measurement instrument such as an oscilloscope 10. The test and measurement instrument 10 receives a signal from the DUT 30 through an instrument probe 32. In the case of an optical transmitter, the probe will typically include a test fiber coupled to an optical-to-electrical converter 34, which provides the signal to the test and measurement instrument through one or more ports 14. Two ports can be used for differential signal transmission, while one port is used for single-ended channel signal transmission. The instrument samples and digitizes the signal to form a waveform. For example, if the test and measurement instrument 10 includes a sampling oscilloscope, a clock recovery unit (CRU) 18 can recover a clock signal from the data signal. Software clock recovery can be used on a real-time oscilloscope.
[0014] The test and measurement instrument has one or more processors represented by processor 12, memory 20 and user interface 16. The memory can store executable instructions in the form of code, which, when executed by the processor, cause the processor to perform tasks. The user interface 16 of the test and measurement instrument allows the user to interact with the instrument 10, such as input settings, configure tests, etc. The test and measurement instrument may also include a reference equalizer and analysis module 24. The user interface 16 may include a display and one or more controllers to allow the user to configure the device. The controller may include one or more switches, knobs or other forms of input, or may include touch buttons on the display.
[0015] The embodiments herein employ machine learning in the form of a machine learning network 26, such as a deep learning network. The machine learning network may include a processor that has been programmed with the machine learning network as part of the test and measurement instrument, or the test and measurement instrument may access the processor through a connection. As test equipment capabilities and processors evolve, one or more processors (e.g., 12) may include both.
[0016] Figure 2One embodiment of a user interface according to some embodiments of the present disclosure is shown. One or more processors will render the user interface on a display. The user interface presents a deep neural network input panel 40 that allows a user to select from different tensor type selections. As shown here, the types include waveform (WFM) vectors, WFM images, and S-parameter images. The user interface menu allows the user to make selections to create waveform vectors for use as input to a deep learning neural network in an application. In Figure 2 In the specific example of , the user has selected WFM vectors, so the other tabs displayed include a define short pattern input panel 42 to allow the user to define a short pattern input. As used herein, the term "short pattern" includes short pattern waveform segments, as described in the '954 application referenced above.
[0017] The tensor comprises a one-dimensional waveform data vector. However, multiple user-defined short pattern waveform sequences are concatenated together to form a one-dimensional waveform data vector. The user interface also provides an average switch 52, in which all similar short patterns found in the waveform can be averaged together before being concatenated together.
[0018] The bar graph input panel 44 allows the user to transform the bar graph into waveform segments and concatenate them into a waveform data tensor vector. Panel 44 includes panels 46, 48 and 50. The user input parameter panel 46 provides options for user-defined input parameters that can be applied to specific test and measurement applications, such as temperature or fiber length or any other desired parameters. The noise bar graph selection panel 48 manages the creation of noise bar graphs created by performing noise measurements on the waveform. The user has the option to control the low pass filtering of each defined short pattern group. The user can also select one of the measurements used to create the bar graph, such as those shown in the measurement panel 50. The measurement options may be different from those shown in the example.
[0019] Figure 3 A user interface is shown when a user selects a tensor type of WFM image to create a waveform image for use as an input to a deep learning neural network in an application. The user interface in this case includes an image builder panel 60. The user can select from two-dimensional or three-dimensional waveform images. In a first embodiment, the tensor includes a two-dimensional waveform data image, where the Z axis contains histogram data that counts how many times each point appears in the waveform short pattern. In a second embodiment, the user can also select the image as a 3D image, where the Z axis represents the amplitude of each waveform, and many more waveforms can fit into many rows of the XY plane. This makes the image more efficient and can handle more data. All user-defined short mode waveform segments can fit into the image.
[0020] It is also possible to create a bar graph as explained above and incorporate it into the tensor image.
[0021] Tensor images can be RGB, with different waveforms in each channel. The arrangement of which waveform goes into which color channel is controlled in the background of the application. However, additional menu controls can be added to specify what data goes into what color channel.
[0022] Another feature is that the bar graph can be transformed into waveform segments and concatenated into waveform data tensor vectors using the bar graph user input panel 44, which is also in Figure 2 The bar graph can be created based on user defined input parameters (such as temperature or fiber length or any other desired parameter, as shown in panel 46), which is also shown in Figure 2 48. The user has the option to control the low pass filtering of each set of short patterns defined. Bar graphs can also be created from a list of user selected measurements that comes with the particular application containing the tensor builder menu, as shown in panel 50.
[0023] The interface includes an image builder panel 60 for controlling various ways of creating the organization of data that goes into the image tensor. There may also be additional menu controls, the ones shown are only provided as examples. Through this user interface, the user can control how the test and measurement instrument organizes the data in the waveform image.
[0024] Figure 4 An embodiment of a tensor builder menu is shown when a user has selected to create an S-parameter image as an input to a deep learning neural network. The tensor includes a three-dimensional S-parameter data image, where the XY image includes real or imaginary or magnitude and phase components. Each row in the XY image represents the real part, imaginary part, or magnitude or phase of one S-parameter in the set. A bar graph can also be created as explained above and incorporated into the tensor image.
[0025] The tensor image may include an RGB image with three different reference S-parameter results from the measured DUT entered into one of the RGB color channels. However, the instrument may allow the user to view the RGB image tensor animation. They do see the image tensors and how they are organized. The user interface also displays a bar graph input panel 46 in the bar graph user input panel 44 to allow the user to specify a bar graph for the transformation based on a measurement or user-defined input parameter (e.g., temperature or other parameter). Figure 2 and 3 The measured noise panel 48 and measurement panel 50 shown in FIG. 5 are not used in the S-parameter tensor builder.
[0026] The S-parameter interface includes an image builder menu panel 70 for controlling various ways of creating the organization of data that goes into the S-parameter image tensor.The user interface may employ additional menu controls in addition to those shown below.
[0027] The following discussion provides a more detailed description of the above user interface elements. As discussed above, the Tensor Builder menu tab comprises the highest level menu in the embodiments herein and contains all other menu items required to construct a waveform vector, waveform image, or S-parameter image for use as input to an applied deep learning neural network. One or more of the following menu items may be included.
[0028] The deep neural network input panel 40 has three options for the type of tensor data input to the neural network. The user interface does not present the algorithm for creating these tensors to the user of the application. These tensors include WFM vectors, which enable the creation of a 1D waveform vector containing many short mode waveform segments, where a large number of waveforms have been discarded and only the parts that are important for training the network are retained. The vector can also contain a bar graph containing input parameters converted to waveform segments representing the bar graph. Another option includes a WFM image, which specifies that the instrument will create a 2D or 3D image tensor to contain a bar graph and multiple short mode waveform data segments. The image can be monochrome or RGB. The last option provided here includes an S-parameter image. This option takes the S-parameters obtained from the device under test (DUT) and builds them into an RGB or monochrome image for use as input to the neural network. The image can also use other parameters, such as S-parameters in the time domain, Z parameters (open circuit or impedance parameters), H parameters (mixed parameters), X parameters (generalization of S-parameters), and many other parameters.
[0029] Figure 2 and 3The definition of the waveform image shown in the short mode input panel 42 allows the user to define the code sequence of each short mode waveform segment to be included in the image tensor. There are three groups of mode types. Up to 8 short modes can be used in each group. This supports signals of NRZ to PAM3, PAM4 and up to PAM8 types. The number representing the number of PAM levels defines the signal level of the signal for each unit interval. The levels of NRZ will be 0 and 1. The levels of PAM3 will be 0, 1 and 2, etc. The level mode panel 64 includes a continuous number of unit intervals, where the signal level remains at the same level for each unit interval in the sequence. However, if desired, the user can define any kind of pattern sequence in this menu tab. The definition of the pulse mode panel 66 is intended to be a sequence of all one level, in which one unit interval is set to a different level to create a pulse mode. The user can define up to 8 different pulse modes, and whether to enable each pulse mode to enter the tensor image. The other mode menu panel 68 allows the definition of up to 8 additional short mode waveforms to be included in the tensor input data.
[0030] In these menu panels, a numeric code specifies how many signal unit intervals are included in the length of the short pattern. This menu item may be optional, as the number of UIs in each pattern can be determined individually by the user entering a value for how many unit intervals to use in the edit box that defines the short pattern. As described above, the user interface includes an average switch, such as Figure 2 52 of 52. When off, the application will find all separate instances of all included short patterns and put them into the tensor, depending on the set of internal rules used to build what was selected. Note that although not shown here, other versions of the tensor builder may have a menu for selecting additional rules for tensor construction. When set to "on", all similar patterns found in the waveform are averaged together before allowing the tensor builder to place them into the tensor according to the tensor builder's internal rules. Some of these rules are determined by the tensor builder menu system. Some rules are internal to a specific end application, and some rules may be defined in the future by adding new menus to the tensor builder menu system.
[0031] This is only available if the user selects WFM Image or S-Parameter Image in the DNN Input menu. Figure 3The Image Builder menu panel becomes visible only when the image is in the image builder menu panel. Depending on the user's selection, the image builder menu has different control items. When the user has selected a WFM image, one selection includes selecting the dimensions of the image type. The 2D selection causes the instrument to build a 2D tensor image where the XY plane is the amplitude versus time for each waveform pattern, and the Z axis is the number of sample counts for all instances of different patterns found in the waveform. The 3D Tensor Image selection increases the number of waveforms and data that fit into the image, and can increase the record length of the patterns that fit, and can increase the vertical resolution of the incoming data. The waveform time and amplitude for each pattern fits in the XZ plane. Each waveform fits in one row of the XY plane of the image. The images will also be RGB, with internal controls for what waveform data goes into each color channel. This image builder menu can incorporate additional menu items to control what waveform goes into what color channel.
[0032] Figure 3 The Enhanced Resolution panel in panel 60 in determines which enhanced vertical waveform resolution algorithm is applied when creating the waveform image. This includes single point compression, which controls the remapping of the minimum and maximum values of each individual point in the waveform relative to all instances of the particular short pattern in which the point lies. This increases the gain of low-level signals and reduces the gain of high-level signals to provide better vertical resolution of low-level signals. In some embodiments, additional menu items not shown in this application will appear to control the details of the algorithm. Linear Segments creates a piecewise linear segment gain vector that is multiplied by the short pattern waveform. This algorithm is primarily intended to be applied to pulse waveforms, with the intention of increasing the gain of the low-level decay portion of the pulse as a function of time. This results in enhancing the low-level amplitude portion of the pulse so that it has higher resolution in the tensor image. The selection "None" results in no gain compression or expansion algorithm being used before the short pattern is placed into the tensor.
[0033] The "Row Spacing" selection determines how the various waveform segments are spaced across the rows of the XY image plane. The "Adjacent" selection causes each waveform pattern or winding pattern segment to be placed in adjacent rows across the XY plane of the image. The "Interval Each Group" selection causes blank rows in the XY plane to be placed between each waveform pattern row, or between adjacent spatial waveform groups. For example, all instances of one pattern type will be spaced adjacent to each other. But there will be space for multiple blank rows between one short pattern group and the next. The number of rows will be equal to the dimension of the convolutional filters in the deep learning input layer.
[0034] The "Space Each Pattern" selection causes each pattern in the waveform to be spaced a distance equal to the dimension of the convolution filter, or another menu control can be added to determine the spacing (not shown).
[0035] When the user selects the S-parameter image, the image builder 70 menu has the following. Dimensions include S1P: one-port S-parameter set, S2P: two-port S-parameter set, S3P: three-port S-parameter set, S4P: four-port S-parameter set. RI: use real and imaginary magnitude for each S-parameter waveform in the frequency domain, MP: use logarithmic magnitude and phase for each S-parameter waveform in the frequency domain, Frequency: represent S-parameters in the frequency domain as real and imaginary or magnitude and phase, and Time: represent S-parameters in the time domain as a single waveform of magnitude versus time.
[0036] The Include S-Parameters menu panel of panel 70 allows only the desired S-parameters to be enabled for input into the tensor image. Place a check mark in the box next to each S-parameter to include it in the image. Leave the box empty to exclude the S-parameter from the image. The number of check boxes and S-parameters available for selection will vary depending on the selection chosen in the Dimensions menu. The "Row Spacing" menu determines how the S-parameter waveforms in the time domain or frequency domain are spaced into the XY plane rows of the image. The "Adjacent" selection specifies that the image has no spacing between the individual S-parameter data vectors in the image rows placed into the XY plane of the image. "Space Each Group" specifies that the image will have a group of empty rows between each pair of frequency domain S-parameter waveforms. The spacing rules will be determined by an internal algorithm. Enhance Resolution has the same rules as previously described to enhance the vertical resolution of the S-parameter data.
[0037] The Bar Graph Input Menu Panel allows various types of bar graph definitions to be incorporated into the tensor image. User Input Parameters 46 allow the user to define their own parameter names that are included in the metadata tags for training the neural network. This menu allows the creation of four such user input parameters. More or fewer user input parameters may be implemented in the final design. The Measure Noise Panel 48 allows the measurement of how noise is measured on the short pattern and then transformed into a bar graph for inclusion in the tensor. The instrument removes the noise on the spectrum measured from the short pattern before placing it into the tensor. This improves the accuracy of the neural network seeing the desired features in the pattern.
[0038] Another option involves filtering in the measurement noise panel 48. Applying a low pass grade filter to the grade waveform will remove as much noise as reasonably possible. The filtered waveform is used to create each grade image. The cutoff frequency can be fixed, or the cutoff frequency can be controlled according to a menu or PI (program instruction) command. A typical cutoff frequency for this filter will be 1 GHz. The user can edit this frequency. Applying a low pass pulse filter to the pulse short mode waveform will remove as much noise as reasonably possible and still maintain the main shape of the pulse. This filtered waveform is used to create a pulse image. The cutoff frequency can be fixed, or the cutoff frequency can be controlled by the user according to a menu or PI command when performing training of the neural network. Other low pass filters are applied to other modes, and their cutoff frequencies can be edited by the user.
[0039] The measurement selection panel 50 allows the user to select measurements. The instrument transforms the selected measurements into a bar graph for each selected measurement and then includes it in the tensor. Some examples of measurements are as follows, but are not limited to TDECQ OMA, RMA, Extinction Ratio, BER and other measurements.
[0040] In summary, embodiments of the present disclosure include a novel user interface menu representation for a tensor builder. The tensor builder defines how to construct input tensor data to train a neural network, and how to use the trained neural network in the final runtime environment. The embodiments incorporate several novel features, such as where parameter bar graphs are included in the tensor. These parameters can be user defined, or measured noise or other measurements. In addition, some embodiments may include parameters that are transformed into bar graphs represented by segments of waveform vectors. The UI allows selection of three types of input tensors, which are waveform vectors, waveform images, and S-parameters. There are multiple controls for creating short pattern waveform segments that enter the tensor, and there is an image builder menu that controls how the image tensor is assembled.
[0041] Aspects of the present disclosure may be operated on specially created hardware, firmware, digital signal processors, or specially programmed general-purpose computers including processors that operate according to programming instructions. The term controller or processor used herein is intended to include microprocessors, microcomputers, application-specific integrated circuits (ASICs), and dedicated hardware controllers. One or more aspects of the present disclosure may be embodied in computer-usable data and computer-executable instructions, such as one or more program modules executed by one or more computers (including monitoring modules) or other devices. Typically, program modules include routines, programs, objects, components, data structures, etc., which perform specific tasks or implement specific abstract data types when executed by processors in computers or other devices. Computer-executable instructions may be stored on non-temporary computer-readable media, such as hard disks, optical disks, removable storage media, solid-state memory, random access memory (RAM), etc. As will be appreciated by those skilled in the art, the functions of program modules may be combined or distributed in various aspects as needed. In addition, the functions may be embodied in firmware or hardware equivalents in whole or in part, such as integrated circuits, FPGAs, etc. Specific data structures may be used to more effectively implement one or more aspects of the present disclosure, and such data structures are considered to be within the scope of computer-executable instructions and computer-usable data described herein.
[0042] In some cases, the disclosed aspects may be implemented in hardware, firmware, software, or any combination thereof. The disclosed aspects may also be implemented as instructions carried or stored on one or more or non-transitory computer-readable media, which may be read and executed by one or more processors. Such instructions may be referred to as computer program products. As discussed herein, computer-readable media means any medium accessible by a computing device. As an example and not limitation, computer-readable media may include computer storage media and communication media.
[0043] Computer storage media means any medium that can be used to store computer-readable information. By way of example and not limitation, computer storage media may include RAM, ROM, electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital video disk (DVD) or other optical disk storage devices, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, and any other volatile or non-volatile, removable or non-removable media implemented in any technology. Computer storage media does not include the signal itself and transient forms of signal transmission.
[0044] Communication media refers to any medium that can be used for communication of computer-readable information. By way of example and not limitation, communication media can include coaxial cables, fiber optic cables, air, or any other medium suitable for electrical, optical, radio frequency (RF), infrared, acoustic, or other types of signal communication.
[0045] Example
[0046] The following provides illustrative examples of the disclosed technology. Implementations of these technologies may include one or more of the following examples and any combination thereof.
[0047] Example 1 is a test and measurement instrument comprising: one or more ports for allowing the test and measurement instrument to receive data from a device under test (DUT); a connection to a machine learning network; a display configured to display a user interface; one or more controls for allowing the test and measurement instrument to receive input from a user; and one or more processors configured to execute code that causes the one or more processors to: render a menu showing different types of tensors on the display; receive a user selection identifying a selected type of tensor from the one or more controls; and construct a tensor of the selected type based on data from the DUT and send the tensor of the selected type to the machine learning network.
[0048] Example 2 is the test and measurement instrument of Example 1, wherein the data includes one or more of waveform data and bar graph data.
[0049] Example 3 is the test and measurement instrument of Example 1 or 2, wherein the code that causes the one or more processors to render a menu showing different types of tensors on a display causes the one or more processors to perform the following operations: select between displaying waveform vectors, waveform images, and S-parameter images.
[0050] Example 4 is a test and measurement instrument of any one of Examples 1 to 3, wherein the one or more processors are further configured to execute code to cause the one or more processors to: receive a user selection of a waveform vector; and render a second user interface having selections allowing a user to define short pattern input, bar graph input, measured noise, and measurement.
[0051] Example 5 is the test and measurement instrument of Example 4, wherein the code that causes one or more processors to construct a tensor of a selected type based on data from the DUT causes the one or more processors to perform the following operations: construct a one-dimensional waveform vector of multiple user-defined short pattern waveform sequences concatenated together.
[0052] Example 6 is the test and measurement instrument of Example 5, wherein the code causing the one or more processors to render a user interface causes the one or more processors to perform the following operations: render a user interface having a selection for allowing a user to average a sequence of short pattern waveforms before concatenation.
[0053] Example 7 is a test and measurement instrument of any one of Examples 1 to 6, wherein the one or more processors are further configured to: receive a user selection of a waveform image; and render a second user interface having selections for a user to define a short pattern input, an image builder, a bar graph input, measured noise, and a measurement.
[0054] Example 8 is the test and measurement instrument of Example 7, wherein the code that causes one or more processors to construct the selected tensor includes code that causes the one or more processors to perform the following operations: construct a two-dimensional waveform image from a short pattern waveform sequence, wherein the two-dimensional waveform image includes histogram data that counts how many times each point appears in the short pattern waveform sequence.
[0055] Example 9 is the test and measurement instrument of Example 7, wherein the code that causes one or more processors to construct the selected tensor includes code that causes the one or more processors to perform the following operations: construct a three-dimensional image having a z-axis representing the amplitude of each waveform included in the image.
[0056] Example 10 is the test and measurement instrument of any of Examples 1 to 9, wherein the code that causes one or more processors to construct a tensor of a selected type based on data from the DUT causes the one or more processors to: receive a user selection selecting an S-parameter image; and render a second user interface having selections for dimensions, image builder, and bar graph inputs.
[0057] Example 11 is the test and measurement instrument of Example 10, wherein the code that causes one or more processors to construct a tensor of the selected type based on data from the DUT causes the one or more processors to perform the following operations: construct an S-parameter image having rows in an XY image, each row in the XY image representing one of the following: a real part or an imaginary part of an S-parameter, or a magnitude and a phase of an S-parameter.
[0058] Example 12 is the test and measurement instrument of Example 10, wherein the code that causes one or more processors to construct a tensor of a selected type based on data from the DUT causes the one or more processors to perform the following operations: construct an S-parameter image comprising red-green-blue channels, and each channel has a different S-parameter.
[0059] Example 13 is a method for providing a user interface, comprising: rendering a user interface on a display of a test and measurement instrument, the user interface displaying different types of tensors; receiving a user selection identifying a selected type of tensor from one or more controls on the test and measurement instrument; constructing a tensor of the selected type based on data from a DUT; and sending the selected type of tensor to a machine learning network.
[0060] Example 14 is the method of Example 13, wherein rendering the user interface on the display to have a menu of different types of tensors displayed on the display includes: displaying a selection between a waveform vector, a waveform image, and an S-parameter image.
[0061] Example 15 is the method of Example 13 or 14, further comprising: receiving a user selection of a waveform vector; and rendering a second user interface having options for allowing a user to define a short pattern input, a bar graph input, measured noise, and a measurement.
[0062] Example 16 is the method of Example 15, wherein constructing a tensor of the selected type from data from the DUT comprises: constructing a one-dimensional waveform vector of multiple user-defined short pattern waveform sequences concatenated together.
[0063] Example 17 is the method of any one of Examples 13 to 16, further comprising: receiving a user selection of a waveform image; and rendering a second user interface having options for a user to define a short pattern input, an image builder, a bar graph input, measured noise, and a measurement.
[0064] Example 18 is the method of Example 17, wherein constructing the two-dimensional waveform image includes constructing the two-dimensional waveform image including histogram data from the short pattern waveform sequence.
[0065] Example 19 is the method of any of Examples 13 to 18, further comprising: receiving a user selection selecting an S-parameter image; and rendering a second user interface having selections for dimensions, image builder, and bar graph inputs.
[0066] Example 20 is the method of Example 19, wherein constructing the S-parameter image includes: constructing the S-parameter image having rows in an XY image, each row in the XY image representing one of: a real part or an imaginary part of an S-parameter, or a magnitude and a phase of an S-parameter.
[0067] In addition, the written description refers to specific features. It should be understood that the disclosure in this specification includes all possible combinations of those specific features. Where a specific feature is disclosed in the context of a particular aspect or example, the feature may also be used in the context of other aspects and examples to the extent possible.
[0068] Furthermore, when a method having two or more defined steps or operations is referred to in this application, the defined steps or operations may be performed in any order or simultaneously, unless the context excludes those possibilities.
[0069] All features disclosed in the specification (including claims, abstract and drawings), and all steps in any disclosed method or process, may be combined in any combination, except combinations in which at least some of such features and / or steps are mutually exclusive. Unless expressly stated otherwise, each feature disclosed in the specification (including claims, abstract and drawings) may be replaced by an alternative feature serving the same, equivalent or similar purpose.
[0070] Although particular examples of the present invention have been shown and described for purposes of illustration, it will be appreciated that various modifications may be made without departing from the spirit and scope of the invention. Accordingly, the present invention should not be limited, except as in the appended claims.
Claims
1. A test and measurement instrument comprising: One or more ports to allow the test and measurement instrument to receive data from the device under test (DUT); Connections to machine learning networks; a display configured to display a user interface; one or more controls for allowing the test and measurement instrument to receive input from a user; as well as One or more processors are configured to execute code, wherein the code causes the one or more processors to perform the following operations: Rendering a menu showing different types of tensors on the display; receiving a user selection from one or more controls identifying a tensor of a selected type; and Build a tensor of the selected type based on the data from the DUT and send the tensor of the selected type to the machine learning network.
2. The test and measurement instrument of claim 1, wherein the data comprises one or more of waveform data and bar graph data.
3. The test and measurement instrument of claim 1 , wherein the code that causes the one or more processors to render a menu showing different types of tensors on a display causes the one or more processors to perform the following operations: select between displaying a waveform vector, a waveform image, and an S-parameter image.
4. The test and measurement instrument of claim 1 , wherein the one or more processors are further configured to execute code to cause the one or more processors to: receiving a user selection of a selected waveform vector; and A second user interface is rendered having selections allowing a user to define a short pattern input, a bar graph input, measured noise, and a measurement.
5. The test and measurement instrument of claim 4, wherein the code that causes one or more processors to construct a tensor of a selected type based on data from the DUT causes the one or more processors to perform the following operations: construct a one-dimensional waveform vector of multiple user-defined short pattern waveform sequences concatenated together.
6. The test and measurement instrument of claim 5, wherein the code causing the one or more processors to render a user interface causes the one or more processors to perform the following operations: render a user interface having a selection for allowing a user to average the short pattern waveform sequence before concatenation.
7. The test and measurement instrument of claim 1 , wherein the one or more processors are further configured to: receiving a user selection selecting a waveform image; and A second user interface is rendered having selections for a user to define a short pattern input, an image builder, a bar graph input, measured noise, and a measurement.
8. The test and measurement instrument of claim 7 , wherein the code that causes one or more processors to construct the selected tensor includes code that causes the one or more processors to: construct a two-dimensional waveform image from the short pattern waveform sequence, the two-dimensional waveform image including histogram data that counts how many times each point appears in the short pattern waveform sequence.
9. The test and measurement instrument of claim 7, wherein the code that causes the one or more processors to construct the selected tensor includes code that causes the one or more processors to perform the following operations: construct a three-dimensional image having a z-axis representing the amplitude of each waveform included in the image.
10. The test and measurement instrument of claim 1 , wherein the code causing the one or more processors to construct a tensor of the selected type from data from the DUT causes the one or more processors to: receiving a user selection selecting an S-parameter image; and A second user interface is rendered with selections for dimensions, image builder, and bar chart inputs.
11. The test and measurement instrument of claim 10, wherein the code that causes the one or more processors to construct a tensor of the selected type based on data from the DUT causes the one or more processors to: construct an S-parameter image having rows in an XY image, each row in the XY image representing one of: a real or imaginary part of an S-parameter, or a magnitude and phase of an S-parameter.
12. The test and measurement instrument of claim 10, wherein the code that causes the one or more processors to construct a tensor of the selected type based on data from the DUT causes the one or more processors to perform the following operations: construct an S-parameter image containing red-green-blue channels, and each channel has a different S-parameter.
13. A method for providing a user interface, comprising: rendering a user interface on a display of the test and measurement instrument, the user interface displaying different types of tensors; receiving a user selection identifying a tensor of a selected type from one or more controls on the test and measurement instrument; Build a tensor of the selected type based on data from the DUT; as well as Sends a tensor of the selected type to the machine learning network.
14. The method of claim 13, wherein rendering a user interface on a display to display a menu of different types of tensors on the display comprises: Choose between displaying waveform vectors, waveform graphs, and S-parameter graphs.
15. The method according to claim 13, further comprising: receiving a user selection of a selected waveform vector; as well as A second user interface is rendered having selections allowing a user to define a short pattern input, a bar graph input, measured noise, and a measurement.
16. The method of claim 15, wherein constructing a tensor of the selected type from data from the DUT comprises: Constructs a 1D waveform vector of multiple user-defined sequences of short pattern waveforms concatenated together.
17. The method according to claim 13, further comprising: receiving a user selection of a waveform image; as well as A second user interface is rendered having selections for a user to define a short pattern input, an image builder, a bar graph input, measured noise, and a measurement.
18. The method of claim 17, wherein constructing the two-dimensional waveform image comprises constructing the two-dimensional waveform image including histogram data from the short pattern waveform sequence.
19. The method according to claim 13, further comprising: receiving a user selection selecting an S-parameter image; as well as A second user interface is rendered with selections for dimensions, image builder, and bar chart inputs.
20. The method of claim 19, wherein constructing an S-parameter image comprises: An S-parameter image is constructed having rows in an XY image, each row in the XY image representing one of: the real or imaginary part of an S-parameter, or the magnitude and phase of an S-parameter.
