FOC Architecture

What is FOC?

Field-Oriented Control (FOC), also known as vector control, is a technique used to control Permanent Magnet Synchronous Motor (PMSM) and Brushless-DC Motor (BLDC). FOC provides good control capability over the full torque and speed ranges. The FOC implementation requires the transformation of stator currents from the stationary reference frame to the rotor flux reference frame (also known as the d-q reference frame).

Speed control and torque control are the most commonly used control modes of FOC. The position control mode is less common. Most traction applications use the torque control mode in which the motor control system follows a reference torque value. In the speed control mode, the motor controller follows a reference speed value and generates a torque reference for the torque control that forms an inner subsystem. In the position control mode, the speed controller forms the inner subsystem.

FOC algorithm implementation requires real-time feedback of the currents and rotor position angle. Measure the current and position by using sensors. The sensorless techniques can be used to estimate feedback values (using motor hall sensors and phase angles) instead of the actual sensor-based measurements. This technique is called Sliding-mode based rotor estimation.

More detailed information about FOC can be found at the https://www.mathworks.com/solutions/electrification/field-oriented-control.html or https://www.pmdcorp.com/resources/type/articles/get/field-oriented-control-foc-a-deep-dive-article.

FOC Block Diagram

In Figure 1, the block diagram of the FOC is seen in detail. The Clarke transform converts the time domain components of a three-phase system (in abc frame) to two components in an orthogonal stationary frame (αβ). The Park transform converts the two components in the αβ frame to an orthogonal rotating reference frame (dq). Implementing these two transforms in a consecutive manner simplifies computations by converting AC current and voltage waveform into DC signals. Thanks to the DC magnitudes obtained by these transformations, we obtain the PI current control (to Id and Iq vectors) facility and this allows us to perform torque control.

Thanks to the feedback provided by the motor, the speed feedback(Encoder, Resolver, Hall Sensor, … ) is sent to the speed PI controller for the speed control of the motor. Likewise, after motor PI controls, these two components in the orthogonal rotating reference frame (dq) are converted to the two components in the orthogonal stationary frame (αβ) using the Inverse Park Transform. It is converted to time domain components (in abc frame) thanks to the Space Vector PWM Generator. You may have noticed the absence of the Inverse Clarke Transform. The Inverse Clarke Transform takes place indirectly during the Space Vector Modulation computation. Park Transform and Inverse Park Transform need the instantaneous rotor electrical angle[Θ]'s sin and cos values. Respectively, the instantaneous rotor mechanical angle of the rotor is converted to its electrical angle with relevant calculations, and the sin and cos values of this angle are calculated.

Field-weakening or flux-weakening is a technique for increasing the speed of an electric motor above its rating at the expense of reduced torque. This concept is especially evident when operating at high motor speeds. More information can be found at the https://www.mathworks.com/solutions/electrification/field-weakening-control.html.

• ω_ref - Reference angular speed [rad/s]

• ω_fb - Angular speed feedback of rotor [rad/s]

• T_ref - Reference torque [N.m]

• i_d^ref, i_q^ref - Orthogonal rotating reference frame current components [A]

• i_d, i_q - Orthogonal rotating reference frame current feedbacks [A]

• v_d^ref_, v_q^ref - Orthogonal rotating reference frame voltage components [V]

• V_α, V_β - Orthogonal stationary frame voltage components [V]

• i_a, i_b, i_{c -} Currents flowing in the stator windings [A]

• v_a, v_b, v_c - Applied voltages to the motor [V]

• V_dc - DC Bus Voltage [V]

• θ_m - Instantaneous rotor mechanical angle [rad]

• θ_e - Instantaneous rotor electrical angle [rad]

• G_ah, G_al, G_bh, G_bl, G_ch, G_cl - A,B,C phase MOSFETs' Gate PWM duty cycles [%]

Software Details

This chapter describes how the software is structured and how to configure it according to the requirements.

Folder structure

While the layers are in separate folders by design, the components of that layer are separated as *.c, *.h pair. The folder structure of the FOC project is shown in Figure 2.

Open

Figure 2 - Field-Oriented Control Software project’s folder structure

The FOC, MCL, and HAL components are initialized into the main.c file from lower layers to upper layers (HAL → MCL → CORE). Desired API function is used through Handles, which is obtained after FOC initialization. The MATH library can be accessed by all layers. The public folder specifies the header files that are open to the user. The Public folder contains the user-accessible header files.

Program Structure

The FOC application consists of four layers. The “Hardware Abstraction Layer“ (HAL) provides 6 components to communicate with the hardware. The “Timer” is used to create the main application task timer of the FOC application and is designed with the flexibility to adapt to future requirements. The “Analog-to-Digital Converter“ (ADC) provides information from the board sensors (Input voltage, phase voltages, phase current, SinCos Resolver, Board Potentiometer, etc.). The “Motion Pulse-Width Modulation“ (MPWM) serves as the configured PWM for the motor. The “Gate Driver“ provides the control of the driver chip used the drive the motor. The “Quadrature Encoder Interface“ (QEI) is used to estimate the rotor position. The “Hall Sensors“ captures information from the GPIOs that are connected to the Hall-Effect sensors (built into the motor), therefore indicating the current state of the rotor. In other words, the main purpose of the HAL layer is to provide feedback information to the higher layers and to transmit the control request it receives from there to the motor. And while doing this, it is to provide an abstraction with hardware.

The “Motor Control Layer” (MCL) retrieves the latest speed setpoint and operation mode from the “Core Layer“ (FOC API). Depending on the active operation mode, it retrieves relevant feedback from the HAL, and this retrieved information is processed according to the FOC algorithm. This algorithm includes various blocks such as PI controllers, Transformation blocks, Space Vector Modulators, Low-pass filters and etc. The latest set speed is kept at the desired value by the Speed PI Controller using the motor's speed feedback. In addition, the motor’s torque is tried to be held at the maximum level thanks to the d and q Current Vector PI controller. The feedback given to the current PI controllers is the current vectors I_d and I_q obtained by transformations using the measured motor current values and the electrical angle[Θ]. Then the calculated dq vectors are converter into 3-phase modulated amplitudes and thus motor rotation is provided. The algorithm described above is executed in the MCL layer. The “Motion“ component includes motion-related functions. It is actually where all FOC application is coordinated. Motor Speed and Operation functions from the FOC API layer are received by this component, processed, and transmitted to the relevant components. It also coordinates the motor position determination calibration. The “FOC Control” component provides the main FOC control. That is, it is the place where operations such as Torque control, vector transformations, Space Vector Modulation, etc. are done. The “Calibration“ component provides the calibrating process, which is used to determination of the electrical rotor position angle in Hall sensor sectors. The “Position & Speed Estimation“ (PSE) component makes the data received from HAL QEI and HAL Hall Sensors processed.

The “Mathematical Library“ (MATH) provides the FOC-based mathematical functionality used by all layers.

The “Foc Api(Core) Layer” contains the necessary API functions for FOC initialization and control.

The components of the relevant layers, their functionality, and connections can be seen in the block diagram in Figure 3.

Hardware Abstraction Layer (HAL)

The “Hardware Abstraction Layer“ (HAL) contains 6 components to communicate with the hardware. The HAL layer uses motor control-related peripherals embedded in the netX90 IC. More detailed information about these peripherals can be obtained from netMOTION and netX 90 in Motion. For even more detailed information, see section 5.14 from netX 90 - Technical data reference guide.

Timer

It provides the main motion application task timer event of the FOC. It is designed with the flexibility to adapt to future requirements.

ADC(Analog-to-Digital Converter)

It provides information from the netX90-MC board sensors (Temperature, Reference voltages, …), motor position feedbacks (Hall sensor sectors, Resolver, …), board voltages (DC bus voltage,…), and Gate driver feedbacks (Measured phase currents, measured phase voltages, …). In Table 1, the signals connected to the MADC matrix of the netX90-MC Rev. 3 board are shown. Green cells indicate signals used in this project are described in Table 2.

Gate Driver

It provides a Gate Driver control interface. The Gate Driver used in our netX90-MC Rev. 3 board is Texas Instruments DRV8323RSRGZR, 6 to 60V Three-Phase Smart Gate Driver. This driver also drives Texas Instruments' 60-V Half-Bridge Power Block Mosfets named CSD88599Q5DC (1 for each motor phase). More detailed information can be get from DRV832x 6 to 60-V Three-Phase Smart Gate Driver and CSD88599Q5DC 60-V Half-Bridge NexFET™ Power Block. The Gate Driver communicates with netX90 via the SPI channel. The Gate Driver component contains important functions such as fault detection, enabling/disabling the gate driver, coast stop, and brake mode. Especially thanks to fault detection, the burning of the motor windings can be prevented in cases such as over-current conditions.

MPWM (Motion Pulse-Width Modulation)

The MPWM component is used to modulate the 3-phase PWM signal required for motor control. Thanks to the “Event Counter Zero (ECZ)“, it generates events so that measurements, transformations, and PI controls are made and new SVPWM duty cycle values are adjusted. Event Counter Zero (ECZ) is the reduced frequency of “Beginning of Period (BOP)” events variant. If the field evt_cnt_top of register mpwm_cfg is set to a value N, an ECZ event will be sent every N+1 period. Ayrica MPWM can be synchronized to hardware-assisted cyclic network events. For detailed information, see section 5.14.3 from netX 90 - Technical data reference guide.

QEI (Quadrature Encoder Interface)

The QEI component provides information from the Quadrature Encoder, which is mounted on the motor. Method 4 was used as the speed estimation method (see section 5.14.4.11 from netX 90 - Technical data reference guide). It supports working with ENC0(QEI 5V TTL) and ENC1(RS422) encoder sources. The NXHX-DH adapter module serves to connect digital transmitters as TTL or RS422 (see section 3.4 from NXHX 90-MC manual). Also, this component supports the encoder direction inversion functionality. For detailed information, see section 5.14.4 from netX 90 - Technical data reference guide.

Hall Sensor

The Hall Sensor component provides information from the Hall Sensors, which are integrated into the motor. The Rotor position can be known by the currently active Hall Sensor. Hall sensor inputs are connected to GPIOs and according to these GPIO statuses, it can be easily determined in which hall sector the rotor is.

Motor Control Layer (MCL)

The “Motor Control Layer“ (MCL) contains 4 components.

Motion

The “Motion“ component provides basic motion functionality. Thanks to the HAL Timer component, the motion task event is created and called Timer Callback. The period of this timer callback is 1 ms, that is, a timer event occurs every 1 ms. The processes that are handled when the timer event occurs are described below.

• MOTION - Auto Configure Encoder Direction: The direction of the encoder mounted on the motor may differ according to the manufacturer and the motor. In order for the motor direction to be determined correctly for each motor, the motor direction must be determined automatically and reversed if necessary. As seen in Figure 4, this is done by determining the turning direction by the motor’s hall sensor, and thus correcting the encoder direction if necessary.

• PSE - Estimate Rotor Speed: Before handling the Motion Control State Machine, The Rotor Speed must be determined. Because this speed data is the feedback of the speed controller(contains Speed and Field-weakening PI controllers).

• MOTION - Motion State Machine: It is the state machine where all FOC coordination is done. These are calibration coordination, motor speed control, coast stop mode control, fault handling and etc. Stall detection fault support is currently available. However, Gate Driver and Motor fault handling support will be provided with updates in the future. The state machine in Figure 5 is processed according to the motor setpoint and operation mode commands receiver from the FOCAPI layer. The explanations of the states are given in Table 3. Coast stop mode is to bring the MOSFET gates to the Hi-Z and ensure that the motor spins down at whatever speed the load imparts on it. The explanations of the operation modes used in the state machine are given in Table 9. FOC operation mode is asynchronous because it comes from the upper layer. Therefore, the operation value set to ensure synchronization creates a request to be processed on the state machine. According to the set operation, the operation of the FOC control component is coordinated by this state machine. MCL Motion Control State Machine is implemented in the mcl_MotionStateMachine(…) function in the “Components\Mcl\Sources\Mcl_Motion.c”.

FOC Control

The “FOC Control“ component takes care of the vector part of the FOC application. It allows the motor to be driven with closed-looped(current vector control) and open-looped(voltage vector control). With the closed-looped control, it is ensured that the maximum torque value is reached by making the I_d and I_q vectors orthogonal by the current PI controllers (Figure 1). In open-looped control, on the other hand, the v_d and v_q vectors are directly determined and given to the Inverse Park Transform block to control the motor, and maximum torque cannot be guaranteed. It is used in our application during calibration. Above, we talked about the MPWM ECZ event while explaining the MPWM component. The ECZ counter value is 2 in your application. That is, 1 ECZ event will occur in every 2 BOP events and the MPWM Event Callback will be handled. The functions handled (Figure 6) each time the MPWM event callback is invoked are described below.

• HAL GATE DRV - Gate Driver Operation State Machine: The gate driver component has an important role in FOC control. This state machine processes the operation control command given to the Gate Driver. These are respectively to enable/disable the Gate Driver, write the desired configuration registers, read the status registers, and perform fault detection. The corresponding HAL Gate Driver operation state machine is shown in Figure 7 and the behavior of the states is explained in detail in Table 4. HAL Gate Driver Operation State Machine is implemented in the Hal_GateDrvCyclicRoutine(…) function in the “Components\Hal\Sources\Hal_GateDrv.c”.

• MCL FOC Control - FOC Control Main State Machine: It is a large-scale state machine that provides FOC control. The state transitions of this state machine are indirectly controlled by the Motion component using the Transit state function. That is, its control is performed indirectly by the motion control state machine. Please keep this in mind when examining the state machine in Figure 8. The corresponding FOC Control Main state machine is shown in Figure 8 and the behavior of the states is explained in detail in Table 5. IDLE state stops all FOC operation including stopping MPWM when the motor is not operating. The purpose of the VOLTAGE CONTROL state is to provide open-loop vector control for the calibration process. The aim here is to keep the torque constant for accurate measurement during calibration, rather than reaching the maximum torque. The CURRENT CONTROL state provides the desired closed-loop vector control. This is the behavior we want during motor control. It keeps the motor torque at the maximum level thanks to the dq current PI controllers it contains. It is the state that is active when the motor is operating. The FAIL state is triggered after a critical error in the Motion Control State and MPWM is stopped. Then it is checked whether there is an error in the Gate Driver Unit and if there is an error, Fault Handling is performed. If necessary, the Gate Driver Unit is restarted. MCL Foc Control Main State Machine is implemented in the Mcl_FocCtrlMainSMCtrl(…) function in the “Components\Mcl\Sources\Mcl_FocCtrl.c”. It is transited with the Mcl_FocCtrlTransitMainSMState(…) function on the same path.

The CURRENT OFFSET state is used for current offset bias calculation. Phase currents are AC by nature, that is they are bidirectional currents. However, the integrated ADC we measured is DC, so it is unidirectional. In order to measure the phase currents with the ADC and interpret it as signed in the variable, we need to make the Current Offset Bias calculation and compensate for the measured values. In the CURRENT OFFSET state, the state machine, the Current Offset Bias Calculation state machine is handled. The corresponding FOC Control Main state machine is shown in Figure 9 and the behavior of the states is explained in detail in Table 6. MCL Current Offset Bias Calculation State Machine is implemented in the mcl_FocCtrlCurrentOffsetSMCtrl(…) function in the “Components\Mcl\Sources\Mcl_FocCtrl.c”.

Calibration

The purpose of the MCL Calibration component is to make the correct position estimation. To do this, it associates the position information from the feedback device (Encoder, Resolver, etc.) related to each motor hall sensor sector for both rotation directions(CW and CCW). Then it is validated whether the position values determined in the 6 Hall Sensor sector are within certain limits in both directions. If all is well, the values obtained in both directions in each sector are averaged and these values are recorded to be used in correct position estimation. The corresponding Calibration state machine is shown in Figure 10 and the behavior of the states is explained in detail in Table 7. MCL Calibration State Machine is implemented in the Mcl_CalCyclicRoutine(…) function in the “Components\Mcl\Sources\Mcl_Cal.c”.

PSE (Position & Speed Estimation)

The PSE component transforms the data it receives from the relevant feedback devices(Encoder, Resolver, Hall Sensor from HAL layer) into meaningful and provides speed and position information for FOC control. Currently, only Encoder (with QEI interface) and Hall Sensors provide speed and position information. However, in the future, different encoder interfaces and Resolver(SinCos) will be supported.

MCL Specific Configurations

The MCL_CFG_PRM_T structure contains some MCL-specific parameters. It provides the possibility to change configuration based on MCL and is initialized in “Components\Mcl\Includes\Public\Mcl.h”. Normally these parameters are not supposed to be modified because they are tuned for the current demo application which is not expected to be changed once configured properly. Related parameters are shown in Table 8.

FOC API Layer (CORE)

The “Foc Api Layer“ provides an interface for the user to control the FOC application. The functions mentioned below can be accessed from the “Targets\Common\Include\Foc_Api.h” file.

Initialize FOC layer

Initializes the FOC. After successful initialization, creates resources, starts the FOC, and gives the user the corresponding handle.

Deinitialize FOC layer

Deinitializes the FOC. It stops the FOC and frees up the resources for the corresponding handle.

Set Motor Speed

Sets the motor speed in RPM. The set speed is processed by the speed controller if the motor is operating.

Get Motor Speed

Gets the actual motor speed in RPM. The get speed is estimated through the encoder by default. If desired, the hall sensor can also be used for speed estimation, although it is not recommended. Resolver support will be available in the future as well.

Set Operation Mode

Sets the FOC’s operation mode. It allows FOC to perform different operations. The function ”Set operation mode” manipulates the Motion Control state machines through requests indirectly. Related operation mode descriptions can be found in Table 9.

Get Operation Mode

Gets the motion control state machine's state. Since the function ”Set operation mode” manipulates the motion control state machine, the enum type “MCL_MOTION_CTRL_STATE_E“ shows the current operation mode. Related motion control state machine’s state descriptions can be found in Table 3.

Mathematical Library(MATH)

Contains all the transform and calculation functions of the FOC. It is accessible from all layers. In Figure 3, the functions included in the MATH library are shown. At the same time, hardware-related files are located here. The Targets folder structure is shown in Figure 2. The McBoardHwConfig.h file in the Common folder contains the definitions(voltage divider resistors, shunt resistors, etc.) associated with the netX90-MC board used by FOC. Thus, new values can be configured easily after a new revision.

Targets

It is the place where all layers of the FOC component are initialized, configured, started, etc.

The “Targets\Common\Include\McBoardHwConfig.h” file contains the board definitions associated with the netX90-MC board used by the FOC. Thus, definitions can be configured easily after new board revisions.

The “Targets\NXHX90-MC\Sources\main.c” file is where all the initialization, configuration, and motor demo functionality resides. Only parameters that can be changed by the user, in any case, will be mentioned here.

Board configuration

The MCL_MC_BOARD_CFG_PRM_T structure contains some board-specific parameters. It provides the possibility to change configuration based on hardware changes and is initialized in “Targets\NXHX90-MC\Sources\main.c”. Normally these parameters are not supposed to be modified because they are tuned for the current NXHX90-MC board which is not expected to be changed once configured properly. Related parameters are shown in Table 10.

Motor-specific Configuration

The Motor-specific parameters are typically found in the Manufacturer's specification for the motor. In the PROFIdrive example, this configuration is found in the structure. Related parameters are shown in Table 11.

(in Targets/NXHX90-MC/main.c).

FOC Cycle TIme

FOC cycle time is totally programmable for experienced users. When the users adapt the FOC application to their software, they can tune the CPU load according to their own software by playing with the __FOC_CYCLE_TIME_MS parameter (Default cycle time is 1ms).

FOC Demo

There are two demo modes:

1. Cyclical demo mode: Ramp-up, ramp-down, and coast stop demo.

2. Demo mode with potentiometer: The motor speed is adjusted with the potentiometer on the board.

The desired demo mode is activated from the “Targets\wscript” file with manipulation of the __USE_DEMO_POTENTIOMETER parameter.

Compilation and Flashing

See section FOC QEI + Hall demo

Abbreviations