Cockpit - User guide

Table of Contents

This user guide explains how to use imperix Cockpit to interact with imperix power converter controllers, namely the B-Box RCP, the B-Board PRO, the Programmable Inverter, and the B-Box Micro. The article describes in detail the tools and features provided by Cockpit as well as how to use them.

For new users, it is recommended to read the following article beforehand to get started with the imperix software development kit (SDK): Programming imperix controllers.

What is Cockpit

Cockpit is the monitoring software included with both the ACG SDK and CPP SDK. It is designed to facilitate the experimental testing of power electronics systems by leveraging the hardware capabilities of imperix programmable controllers.

The software provides multiple tools (modules) for both observing and acting on the user code running on imperix controllers.

The top bar offers the following tabs:

The Modules tab contains several modules that can be placed in the project view to monitor and tune the variables of the control code.
The Import/Export tab provides tools to export signals acquired with the Scope and Rolling plot modules in the CSV or MAT file format or directly as a MATLAB figure.
The Help tab contains tools that help interact with the hardware setup, like the target explorer and the device analog channel configurator, along with the global settings menu and various links to the documentation or other information.

The project bar displays all of the projects created in Cockpit.
In a single target scenario, multiple projects can be created to easily switch between one control algorithm and another or quickly test variants of the same algorithm.
In a multi-controller scenario, it allows users to keep an eye on the status of every controller.

The project pane offers a centralized view from which the user can pilot, configure, and maintain the controllers. It notably offers a quick view of the controller state, connection status, and CPU load.

The project view is the work area where modules can be placed and rearranged at will to tune and monitor control algorithms. Each project has its own project view, which allows the user to tailor the area according to the application.

This is another project-specific area, organized in the form of tabs:

The Logs tab, which is always present in every project, displays messages reported by the connected controller
The Scope and Rolling Plot tabs are generated when their respective modules are added to the project view. They display information related to the signals plotted in them.

Similar to the Scope and Rolling Plot tabs in the bottom bar, the right bar hosts various configuration menus that are created when corresponding modules are added to the project view.

How to create a new project

To interact with an imperix controller, creating a new project is required.
After the steps described in How to program the controller are achieved, Cockpit is automatically launched and a new project is created with a pre-filled project name and path to the user code.
Alternatively, it is also possible to manually open Cockpit and create a new project by clicking on the + button at the bottom of the project bar.

To finalize the project creation, the IP address of the target imperix hardware must be entered. The target’s IP address can be either manually entered or automatically identified using the Target explorer tool (binocular icon).

Please refer to the chapter on How to connect the controller to the host PC to ensure the target is properly connected to the host computer.

Once all three fields are filled in, the project can be created by clicking on the CREATE button. The project pane will then appear in the project bar and a blank project view will be created.

How to interact with the project pane

The project pane offers a centralized view from which the user can pilot his imperix controller. It also allows quick access and an overview of all of the control variables defined in the user code.

Upon creation, the project pane will automatically connect to the specified target, load, and start the user code (.elf file). The following image shows the project pane once it is connected to the target.

This label displays the target’s IP address on which the user code is currently running.
Please note that if several projects are created, only one project at a time can be connected to a given target.

This button displays whether the target is connected to the host computer or not.
Clicking it will disconnect the target from the host computer or connect it, depending on the current state of the link.

Note that when the project is deleted, the user code will continue to run, and if the PWM outputs are enabled, they will not be automatically disabled.

Displays the current operating state of the controller.

Enables/disables the target’s gating signals. Further information on this mechanism is available in the chapter Enabling/disabling PWM signals.

The CPU load expresses how much time the CPU spends in the control task. It is mainly affected by the user code complexity and the control task rate.
Safety mechanisms are implemented to detect CPU overload, which can happen if the control algorithm is too complex or its execution frequency is too high.

This section displays all the user variables defined in the user code. These user variables are signals that are either connected to a Probe block or a Tunable parameter block.

Signals connected to Tunable parameter blocks can be modified directly from this section but also in the Project View, where various Cockpit modules make it easier to visualize and track these user variables in run-time.

Clicking on this entry allows the user to change the user code, change the target associated with the project, or delete this project. The project needs to be disconnected from the target before making any of these changes.

This entry allows the user to delete the project.

Note that when the project is deleted, the user code will continue to run, and if the PWM outputs are enabled, they will not be automatically disabled.

This entry opens the target configuration window. From this window, it is possible to configure the target, visualize the various computation and communication delays of the control, and generate an analog front-end configuration.

This entry opens the Timings window. This window provides a graphical representation of the various computation and communication delays involved in the imperix controllers during run-time.

If a fault is triggered on the controller, the project pane interface will be changed to the one displayed in the following image. Further information on how faults work and how to troubleshoot them are detailed in The most common types of faults chapter.

If Cockpit unsuccessfully tries to connect to the target (image below), the project pane will indicate that the connection with the target cannot be established or is lost. In this case, check that the Ethernet cable is correctly connected and the target is powered on. Subsequently, verify that the specified IP address is correct and validate that the target can be pinged.

Using the modules to interact with the controller

Once a project is created, the user has access to various control and monitoring modules that can be dragged and dropped from the top bar to the project view. The user can then freely rearrange and resize these modules inside the project view.

Scope module

This module allows the user to display control signals on an oscilloscope-like interface by capturing and plotting every sample of the scoped user variables. The acquisition is done at the control task rate (i.e. the main interrupt frequency of the controller), ensuring that each and every sample is scoped.

The scope module can monitor up to 100 MBytes of data. The window length and the control task rate will affect how many user variables can be scoped.
If the control task is set to a standard frequency of 20kHz, it represents scoping one variable for 21 minutes or 32 variables (maximum number of scoped variables) for 40 seconds.

To start the acquisition, simply drag and drop a variable from the project pane directly to a plot of the scope module.

Activating the scope will affect the CPU load. This might be of concern if the CPU load is already close to 100% prior to the activation of the scope module.

Scope module interface

The plot preview shows an overview of all the scoped signals. The vertical axis of the preview is automatically scaled so all of the signals fit the plot. The plot preview can be collapsed to optimize space when using Cockpit on a small monitor.

This is the area where the scoped signals are displayed. User variables can be dragged and dropped from the project pane into a plot, which will add them to the acquisition and display the corresponding samples in the area.
The x-axis range (when no zoom is applied) is defined by the Window [ms] field of the Trigger pane (located in the right bar). If multiple plots are created, their x-axis will always remain synchronized, including when zooming.
The user can remove a variable from a plot from the Scope tab of the bottom bar or through the variable plot context menu accessed by right-clicking on the plotted signal.

The plot containing the trigger signal will have two extra cursors added to it. The user can move these cursors to adjust the trigger level and position to set the trigger point (more information in the trigger configuration chapter).

Starts or stops the acquisition. Note that at least one variable must be added to a plot for the acquisition to start.

Force the acquisition of one window, bypassing the trigger mechanism. This is useful when the trigger is configured in Normal or Single mode.

Perform an autoscale on the horizontal and vertical axis, ensuring that the acquired signals fit in the plots.

Perform an autoscale on the horizontal axis of every plot while keeping the vertical axis unchanged.

Perform an autoscale on the vertical axis of every plot while keeping the horizontal axis unchanged.

When switched on, this feature performs a vertical autoscale on every plot after each new acquisition window.

When switched on, a pair of slideable vertical cursors are shown. The cursors allow the user to mark an interval inside which several metrics are calculated on the scoped signals and displayed in the bottom bar.

Clicking this button opens a window that allows the user to create a new plotted variable based on a mathematical formula. The formula can combine any of the variables currently displayed inside the Scope module.

Clicking this button opens a window where the user can examine the signals acquired by the scope in the frequency domain. The displayed spectra correspond one-to-one with the user and math variables displayed in the scope.

Displays the current acquisition state. The possible acquisition states are:

Stopped: the acquisition is stopped. The displayed data is from the last acquisition.
Waiting: the acquisition has started and is now waiting for a trigger event to occur.
Acquiring: the trigger event has occurred and the signals are being captured.
Offline: the target is disconnected from the host computer or the user code is not running on the target.

The loading bar displays the acquisition and transmission status of the acquired window.
Note that it is normal to see the loading bar partially filled and blocked when the acquisition is in Waiting state.

Creates an empty plot area at the bottom of the scope module.

The scoped variables are displayed in the Scope tab of the bottom bar. The bottom bar Scope tab menus allow the user to:

Modify each variable’s display style, scale multiplier, offset value, and measurement unit
Remove a variable from scope
Toggle a variable’s visibility without removing it
Move variables between scope plot areas

In addition to this, the position and the distance between vertical cursors are displayed here, along with the metrics calculated inside the interval defined by the cursors if they are enabled. If the cursors are not enabled, the metrics are calculated over the entire acquisition window.

Trigger configuration

The trigger mechanism of the scope module behaves the same way as the trigger on a regular oscilloscope. The Tigger pane located in the right bar allows the configuration of the scope trigger.

Defines the acquisition window length and changes the x-axis length. This value affects the total number of acquired points.

Defines whether to trigger on rising, falling, or both edges of the trigger signal.

This field defines on which signal the trigger will be applied. This signal will be marked with (T) in the plot legend and can also be changed by right-clicking on a variable and setting it as a trigger through the context menu.
Note that only variables added to the scope module are present in this drop-down list.

Defines the trigger acquisition mode.
In Normal mode, a window is acquired if the input signal reaches the defined trigger point.
In Auto mode (which is set by default), the trigger acts like in normal mode. However, the window acquisition is forced by a timer set to 100 ms, which ensures acquisition even if no trigger event occurs.
In Single mode, a single window is acquired if the input signal reaches the defined trigger point. After that, the acquisition is stopped.

Change these values to set the trigger point. Alternatively, the trigger position and level can be adjusted directly in the plot area by moving the trigger level and position icons.

Transient generator configuration

The transient generator allows the user to apply a predefined sequence of steps to multiple user variables.

In the following example, three transient events are applied to Ig_d_ref. It can be seen that these events are generated simultaneously with the acquisition of Ig_a, Ig_b, Ig_c, and Ig_d.

Scope module containing the result of a transient sequence (on the left). Transient configuration pane (on the right)

This field defines to which variable the transient steps will be applied. In this example, the variable Ig_d_ref is used.

This field is a vector of positions at which the transient steps will be generated. In this example, the steps are generated at 50ms, 100ms, and 150ms.

This field is a vector of values that will be applied to the transient variable ([8, -5, 10] in this example).
Please note that these steps are delta values and therefore based on the initial value of the transient variable. In this example, the value of Ig_d_ref was set to 2 before the transient was fired, so the observed transient jumps to 10, 5, and 15.

The transient generator allows entering multiple transient sequences on different variables.
In this example, the second transient sequence is unused.

These two buttons allow the removal or the addition of transient events.
Note that there is no limit to the number of transient events. However, the number of transient steps throughout all transient events is limited to 32.

Clicking on this button will initiate the transient sequence on all transient variables. Once the sequence is finished, the acquisition will be in Stopped state.

Please take note of the following limitations:

Only user variables connected to the Tunable parameter block can be selected in the Variable field.
The variable in the Variable field must be added to a plot, or the transient will not work.
The number of transient steps throughout all transient variables is limited to 32.

Cockpit Formula Builder

The Formula Builder allows the user to create math variables that are then added to the Scope module. Math variables are created based on a mathematical formula that can include any of the currently Scoped variables.

The math variables are calculated on the PC side after the acquisition of a scope window. This alleviates the need to define every variable to monitor in the user code, saving on hardware resources while still allowing the user to track them live through Cockpit. This also saves time, as new math variables can be added or their formula edited without having to recompile the user code.

With the exception of constant expressions and scalar-valued functions, all of the mathematical operators and functions provided by the Formula Builder are applied element-wise. This means that the formula is applied to each and every sample of the variables included in the expression, creating a new signal to be plotted in the scope.

To create a Math Variable, it needs a valid name. The name has to be unique and cannot be edited after creation.

While not necessary for Math Variable creation, the unit of the new variable can be set while building its formula. If not set here, the unit can always be edited, like with any other variable, from the bottom bar.

This option allows the user to choose where the created Math Variable should be plotted or even create a new area to plot it in. After creation, the Math Variable can be moved around between plots like any other scoped variable.

This button controls the appearance of the list of formulae of previously created Math Variables.

Here, the formula is defined either by typing it or clicking on the options in the menus below.

The list contains all of the formulae of the previously successfully created Math Variables.

The star next to each of the formulae allows us to favorite it, making it available in all projects. Non-favorited formulae disappear when the scope they were used in is deleted.

This menu contains all of the binary operators usable for building a math formula.

This list contains all of the functions usable while building a math formula. The functions are grouped into three categories:

General math is a collection of elementary (abs, exp, log, sgn, sqrt) and rounding functions (ceil, floor, round, trunc)
Trigonometry includes all of the typical trigonometric and hyperbolic functions, along with conversion functions between degrees, radians, and gradians
Scalar-valued functions are grouped based on the fact that they all evaluate into one number.

This list contains all the variables currently displayed in the scope, including acquired user variables and other math variables. The mathematical formula can use all the variables in this list as operands.

As the math formula is being typed, the errors provided by the formula parser are displayed here. None of these errors are critical to Cockpit operation; they serve as guidance to creating a math expression that can be properly interpreted and used to evaluate the new Math Variable.

Clicking on Save variable will attempt to save the Math Variable with the currently entered parameters. If successful, the window will close, and the saved variable will be displayed in the selected Scope plot area.

Pressing the Enter key while the focus is on the formula field is a shortcut to clicking Save variable.

Clicking on Cancel will close the Formula Builder window without saving anything.

Pressing the Esc key while the focus is anywhere in Cockpit is a shortcut to clicking Cancel.

Cockpit Spectral Analyzer

The Spectral Analyzer allows the user to examine Scope variables in the frequency domain. This includes all of the user and math variables that are currently present in the module. To add a variable to the Spectral Analyzer, add it to the Scope module and it will show up automatically. Conversely, removing a variable from the Scope will make it unavailable in the Spectral Analyzer as well.

The signal spectra are calculated using a variation of the Fast Fourier Transform (FFT) algorithm. Similarly to math variables, the calculations are performed on the PC side, after the acquisition of a scope window. The displayed frequency domain signal is the magnitude of the one-sided spectrum of the scoped variable, with the frequency range \(f \in [0, \frac{f_s}{2}]\). The Scope sampling rate, \(f_s\), corresponds to the control task frequency defined in the user code. The frequency resolution of the resulting spectra depends on the window in time over which the Fourier transform is performed and can be adjusted in the window’s accompanying menus.

This area displays a miniature version of the signals displayed in the Scope module. The main difference compared to the Scope plot preview is that only variables from the currently selected plot are shown here, as opposed to all of the signals added to the Scope. This preview can be used to visually adjust the size and position of the window in time over which the Fourier transform is performed.

Zooming with the mouse scroll wheel sets the size of the Fourier Transform window, while dragging the window with the mouse pointer can set its position within the acquisition window.

The plot shows the scoped signals in the frequency domain. When accessing the Spectral Analyzer from the Scope, typically only one spectrum will be shown. The hide/show options in the bottom bar control the visibility of all of the spectra displayed for a given Scope plot area.

When switched on, an annotation box for the spectral value closest to the mouse cursor is shown. The annotation contains the x and y axis position corresponding to the closest sample and the name of the variable to whom the sample belongs to.

Starts or stops the acquisition of the Scope. The button in the Spectral Analyzer and the start/stop button in the Scope module always have the same state. The same holds for the Acquisition state and the Acquisition loading bar.

The scoped variables for a given Scope plot area are displayed here. To switch between Scope plots areas, click on the tabs in the header of the bottom bar. All of the settings in the Spectral Analyzer are done separately for each of the Scope plot areas.

Creates an empty plot area at the bottom of the scope module. The area is labeled as Subplot so as not to be confused with the Scope plot areas that are used for sorting the spectral signals into tabs.

Besides the usual fields that allow the precise definition of the displayed y-axis range, the Spectral Analyzer offers the option to switch between different scaling settings for the value axis:

‘Absolute, linear’, which displays the magnitudes of the calculated Fourier transform, without any visual or mathematical rescaling
‘Absolute, log10’, which arranges the magnitude values in a log-scale
‘Relative, dB’, which transforms the magnitude values according to the Amplitude dB formula: \(20\text{log}_{10}\big(\frac{|X|}{X_{ref}}\big)\), where \(|X|\) is the magnitude calculated by the Fourier transform for any given \(f \in [0, \frac{f_s}{2}]\)
‘Relative, log10’, which divides all of the magnitude values with \(X_{ref}\) and arranges the resulting values in a log-scale

For the scaling options that require it, \(X_{ref}\) can be set manually in the corresponding field below. If the continuous normalization option is toggled on, \(X_{ref}\) will be set automatically to the biggest magnitude value seen since the option was turned on.

Besides the usual fields that allow the precise definition of the displayed x-axis range, the Spectral Analyzer offers the option to switch between linear and log-scaling for the frequency axis.

The results of the Fourier Transform depend on the size and and position of the window in time over which its performed. While the position is adjusted through the plot preview, the window size can be set more precisely through this menu, either directly or by setting the desired frequency resolution. The typical windowing functions (Rectangular, Hann, Hamming, Blackman, Blackman-Harris and Flat top) are provided and can be applied to the samples within the window before transform.

Two aspects of viewing the same spectral content are provided:

Spectrum mode, where the default FFT window is the same as the acquisition window. This is to give the best overview of the spectral content of the scoped signals, since the widest window results in the highest frequency resolution.
Harmonics mode, where the default FFT window is such that the frequency resolution is 50 Hz. This allows us to visualize the spectral components in a manner typical for the Power Electronics domain. The Total Harmonic Distortion (THD) and Weighted THD (WTHD) metrics are calculated and updated in the bottom bar menu for the variables that are currently displayed in the plot.

Here are some shortcuts for the scope module:

To add multiple variables to a plot, open the user variable section in the project pane.
Keep the ctrl key pressed and click on the desired variables to select them.
Alternatively, click on the first variable to select it, keep the Shift key pressed, and click on the last variable to select. These selected variables can then be dragged and dropped into a plot all at once.
To zoom in and out along the horizontal axis, place the mouse cursor where to zoom. Then, use the mouse wheel to zoom in or out around the location of the mouse cursor.
To zoom in and out along the vertical axis, place the mouse cursor where to zoom. Then press the ctrl key and use the mouse wheel to zoom in or out around the location of the mouse cursor.
To zoom on a specific area, click and drag to draw a blue rectangle over the zoom area.
To achieve a horizontal autoscale, right-click and drag horizontally. A light grey horizontal strip will appear. Release the mouse button to perform the horizontal autoscale.
To achieve a vertical autoscale, right-click and drag vertically. A light grey vertical strip will appear. Release the mouse button to perform the vertical autoscale.
Many of the Scope functionalities can also be accessed through context menus by right-clicking on a plotted variable or on the empty space in the plots.

Scope Variable context menu and plot area context menu

Rolling plot module

The rolling plot module displays a downsampled version of the variables. It allows the monitoring of the selected variables for up to several days. A typical use case is monitoring the long-term evolution of the converter state in order to keep an eye on critical variables at all times.

Rolling plot module interface

The plot preview shows an overview of all of the monitored signals. The vertical axis of the preview is automatically scaled so all of the signals fit the plot. The plot preview can be collapsed to optimize space when using Cockpit on a small monitor.

This is the area where the monitored signals are displayed. User variables can be dragged and dropped from the project pane into a plot to monitor them. If multiple plots are created, their x-axis will always remain synchronized, including when zooming. The user can remove a variable from a plot from the Rolling Plot tab of the bottom bar or through the variable plot context menu accessed by right-clicking on the plotted signal.

Pauses/resumes the rolling of the plot window. Data acquisition will continue in the background regardless of whether the plot window is rolling or not.

Autoscales both the horizontal and vertical axis, ensuring the acquired signals fit in the plots.

Autoscales the horizontal axis of every plot while keeping the vertical axis unchanged. Additionally, if the monitoring is paused, it will be automatically resumed.

Autoscales the vertical axis of every plot while keeping the horizontal axis unchanged.

When switched on, this feature continuously auto-scales every plot, ensuring that the monitored variables never go out of scope.

Displays if the rolling plot is paused or not.

Creates an empty plot area at the bottom of the rolling plot module.

Every monitored variable is sampled at 20 Hz. However, the sampling rate defines a “best effort” and can, therefore, be slower than 20 Hz. Thus, the rolling plot module should be used to monitor slow-moving variables during a long period of time where strict temporal accuracy is not a concern. To acquire variables with strict temporal accuracy, it is recommended to use the scope module instead.

Rolling plot module shortcuts

To add multiple variables to a plot, open the user variable section of the project pane. Keep the ctrl key pressed and click on the desired variables to select them. Alternatively, click on the first variable to select, keep the Shift key pressed, and click on the last variable to select. These selected variables can then be dragged and dropped into a plot all at once.
To zoom in and out along the horizontal axis, place the mouse cursor where to zoom. Then, use the mouse wheel to zoom in or out around the mouse cursor.
To zoom in and out along the vertical axis, place the mouse cursor where to zoom. Then, press the ctrl key and use the mouse wheel to zoom in or out around the mouse cursor.
To zoom on a specific area, click and drag to draw a blue rectangle over the zoom area.
To achieve a horizontal autoscale, right-click and drag horizontally. A light grey horizontal strip will appear. Release the mouse button to perform the horizontal autoscale.
To achieve a vertical autoscale, right-click and drag vertically. A light grey vertical strip will appear. Release the mouse button to perform the vertical autoscale.
Many of the Rolling Plot functionalities can also be accessed through context menus by right-clicking on a plotted variable or on the empty space in the plots.

Rolling Plot Variable context menu and plot area context menu

Variables module

The Variables module gives quick access to selected variables. It also allows modifying the user variables without rebuilding the control code. Additionally, it displays extra information on user variables, such as their minimum and maximum values.

This module can contain one or more sections to sort the user variables conveniently.

To add a new section, click the + button on the bottom left of the module.
The columns displayed in a section can be managed by right-clicking on the section’s header and checking the desired columns.
To add a variable to a section, drag and drop a variable from the project pane directly into the desired section.
To remove a variable from a section, right-click on the variable and select “remove variable”.
Signals connected to a Probe block are listed under the probe type, and signals connected to a Tunable parameter block are listed as parameter. Variables of type parameter can be modified in run-time by double-clicking on their value (Data column).

DAC module

The DAC module allows the user to apply a given variable to one of the four analog outputs of the B-Box RCP. The values applied to the analog output are specified in volts. The output range is -5V to 5V, and the output is saturated beyond these limits. Further hardware specifications are available in the B-Box RCP datasheet.

If a DAC block is used in the user code, and the DAC module is also used in Cockpit, the DAC module in Cockpit will have the upper hand.

Logs tab

The logs tab displays every message reported by the controllers, including useful information such as misconfiguration details, software and hardware faults, or custom user messages.

Export data as a file or MATLAB figure

Cockpit offers the possibility to export the plotted signals acquired with the Scope and Rolling plot modules as a CSV or MAT file, or directly as a MATLAB figure.

For the scope module, Cockpit will export each and every sample of the scoped signals. As a reminder, the signal acquisition is performed at the control task rate. For the rolling plot module, Cockpit will also export each and every sample of the acquired signals. In this case, the signals are acquired at 20 Hz. However, this is a best effort, and the data may be acquired at lower rates. Therefore, it is common to observe timestamps in the exported CSV files that are not at 20 Hz.

To export data from a rolling plot or a scope module, click on the Export CSV, Export MAT, or MATLAB figure button located in the Import/Export tab of Cockpit’s top bar, as displayed below.

From the menu that appears, select the desired plots to export and choose the save location MATLAB figure or the CSV/MAT file.

A MAT file is also saved when exporting data as a MATLAB figure.

Exporting as MAT file

Exporting as MAT file or MATLAB figure is done through the MAT 7.3 file format. This format is based on the HDF5 standard for hierarchical storing of data, allowing for partial saving and loading and better compression. This makes the export process more efficient in terms of time and memory in most cases, compared to CSV files.

A MAT file can be easily loaded to a MATLAB workspace by finding the file through the “Current folder” menu in MATLAB and double-clicking on it. To learn more about the structure of the MAT files we export and different ways to load and save them, read Working with MAT files exported from Cockpit.

Exporting as MATLAB figure

Exporting as MATLAB figure will first create and save a MAT file, as described in the previous section. Once the data has been exported as a MAT file, MATLAB is automatically launched with a new workspace set to the chosen folder. Finally, the MATLAB figure is automatically displayed.

If several instances of MATLAB are installed on the computer, the user will be prompted to select the desired MATLAB version with the following pop-up message.

The export as MATLAB figure feature only supports MATLAB versions R2019a and newer.

Popup message when multiple MATLAB versions are detected

The export process can take up to several minutes when exporting data from a Rolling plot module, especially if the data spans over a prolonged period of time. Moreover, it is also possible that when exporting this data as a Matlab figure, the sheer amount of data could be too large for the available memory of MATLAB. Thus, it is not excluded that MATLAB could crash. In such cases, partial loading of the saved file might help.

Target timings

The Timings window, available from the three dots menu of a project pane, provides a graphical representation of the various computation and communication delays involved in the controller during run-time. It is particularly useful to observe the delays involved in the control dynamics of the system as explained in Identifying the discrete control delay (PN142). The control parameters, such as the \(K_p\) and \(K_i\) of a PI controller, can then be adjusted accordingly, as shown in Basic PI control implementation (TN105).

Below are screenshots of the Timings window when running the model Central PV inverter (AN006) in two scenarios.

The timing graph accurately represents the delays involved in the execution of the user control code.

CLOCK_0 timer represents the clock generator counter. CLOCK_0 is used as the time base for the sampling events and the control task routine execution. In most cases, the PWM modulators are also based on CLOCK_0.
Sampling shows the sampling events, i.e. the instants where the ADCs sample the input analog signals. In the example above, the sampling phase is set to 0.5, so the sampling occurs in the middle of the period of the PWM signals that are based on CLOCK_0. Multiple sampling events per control period can be configured using oversampling, as explained in Oversampling (PN154).
ADC acquisition delay shows the delay between a sampling event and the availability of the values in the FPGA. It comprises the ADC acquisition delay and the transfer time of the read value to the FPGA. This value is 2000 ns in the B-Box RCP and 500 ns in the B-Box Micro and B-Board PRO. The appropriate value must be set from the CONFIG block.
The Read delay shows the time needed to perform FPGA-to-CPU transfers. These tasks are executed right after the ADC results are available in the FPGA. The values transferred are typically the ADC measurements, the GPI values, or the angle decoder output.
The Processing delay shows the time the CPU spends executing the interrupt routine. To execute the CPU-to-FPGA transfers as early as possible (and reduce the overall response delay), the interrupt routine is separated into two phases:
- The control task execution is the time necessary to execute the user code and update the modulation parameters and other FPGA values. The CPU-to-FPGA transfers are executed right after, in parallel with the post-processing execution.
- The post-processing execution is the time necessary to perform all the tasks that are not directly involved in the control algorithm. It includes the datalogging execution, the CAN communication, etc.
The Write delay shows the time needed to perform the CPU-to-FPGA transfers. These tasks are performed once the control task execution is over. The values transferred are typically the PWM modulation parameters (duty-cycle, phase), the GPO values, or the DAC values.

At the top of the window, the following information is displayed:

CLOCK_0, CLOCK_1, CLOCK_2, and CLOCK_3 show the configured frequency of the four Clock generators.
The CPU state displays the current operating state of the controller.
The Control task routine execution frequency and period are also displayed. Note that the control task is always mapped on CLOCK_0.
The Sampling shows the frequency and period of the ADCs sampling. It allows the user to visualize if the oversampling is configured or not.
The CPU load represents how much time the CPU spends in the interrupt routine relative to the period of CLOCK_0. Safety mechanisms are implemented to detect CPU overload. An overload can result from a control algorithm being too complex or an execution frequency being too high. The min, max, and avg values are computed on a window of one second.
The Cycle delay represents the delay between the sampling event and when the newly computed data are available in FPGA (ADC acquisition delay + FPGA-to-CPU transfers + control task execution + CPU-to-FPGA transfers). This value can be used to compute the control delay and the total loop delay, as explained in Identifying the discrete control delay (PN142). The cycle delay value is precisely measured directly from within the FPGA. The min, max, and avg values are computed on a window of one second.

Other features

How to launch a user code at controller start-up

In some cases, it can be useful to automatically start a user code when turning on the imperix controller. To do so:

Launch the user code and connect to the target as usual.
Once the code is running on the target, open the Target configuration window from the three dots menu of the project pane.
In the section User code saved on SD card, click on the Save the currently running code button to save the current code on the SD card inside of the controller.
Click on the checkbox Load at startup to automatically launch the code when the controller is powered on.

This will save the control algorithm on the SD card inside the controller and automatically load and start it from the SD card at start-up.

How to load a custom FPGA bitstream

In some cases, it is interesting to offload all or parts of the computations from the CPU to the FPGA. Doing so often results in much faster closed-loop control systems.

To do so:

Launch the user code and connect to the target as usual.
Once the code is running on the target, open the Target configuration window from the three dots menu of the project pane.
In the section FPGA bitstream saved on SD card, click on the Browse button and select the desired bitstream. The bitstream will be uploaded into the SD card of the controller. A check in the Load at startup checkbox indicates that the target will load the imported customized bitstream at the next power cycle of the target instead of the standard one.

Further information on how to develop power converter control algorithms in FPGA can be found on the Getting started with FPGA control development page.

NEWSLETTER

CONTROL PRODUCTS

POWER PRODUCTS

COMPANY