AN-032: TMC4361A Closed-loop Motor Control for Stepper Motor Drivers

This application note explains functional details of closed-loop motor control of stepper motor (gate) drivers in combination with the motion control IC TMC4361A. It serves as technical background information. Please refer to the specific product documentation for specific parameter information and configuration instructions.

Closed-Loop Motor Control

TMC4361A provides closed-loop motor control capabilities for stepper motor (gate) drivers. Typically, current values or step impulses from TMC4361A for the motor drivers are based only on internal calculations. With the closed-loop unit of the TMC4361A the output currents resp. Step/Dir outputs of the internal step generator will be modified directly in dependence of feedback data. This feedback data can be obtained from different encoder types like incremental ABN encoders or absolute SSI or SPI encoders.

Closed-loop Motor Control is not Field-oriented Control

The classical field-oriented control typically uses a cascade of three control loops. The inner loop controls the motor current. One level beyond velocity is controlled until final position is reached. The current control loop always assigns a current, which has a commutation angle of 90°, to gain maximum torque. Feedback is obtained by encoders or by the measurement of phase currents and voltages; with additional help of mathematical models.

The 2-phase closed-loop control of TMC4361A follows a different approach than PID control cascades, which consider stepper motor driver characteristics. The ramp generator - that assigns target and velocity - is independent of position control (commutation angle control), which is also independent of current control. The closed-loop motor control scheme is depicted in the following picture.

Areas of Special Concern

The encoder should be mounted directly on the motor axis for closed-loop operation to prevent unwanted coupling effects.

Use of PID Regulation Feature

Also, if the encoder is not mounted directly on the motor, the PID regulation feature must be used. For example, PID control will avoid slipping of belt drives, because they cannot be handled correctly with the closed-loop motor control.

Closed-loop Motor Control Prevents Step Loss

Phase currents are assigned directly to the motor drivers, typical for stepper motor drivers. This results in current vector which should be followed by the rotor. The rotor position will be directly sampled by encoder feedback. The closed-loop motor control now monitors the resulting load angle.

The direction of the current vector tracks the rotor position in case the load angle exceeds a certain limit. The result is a load angle, which never exceeds the given limit. As a result no step loss will occur. Thus, the current vector follows an overpowered load until the load is reduced. After overcoming the overload, the current control assigns the required position or velocity.

Note: Typically, the load angle limit will be 90° which is the maximum torque which is one full step, but it is also possible to assign any limit between 0° and 180°.

Closed-loop motor control saves energy

Besides the load angle tracking, it is also possible to assign different current scale values for different load situations. Thus, lower load situation can be used to save energy. This is a result of the free adaptable current scaling feature which assigns the current in dependence on the actual load. During low load phases, lower current levels will be assigned to the motor driver, whereas the maximum current will be applied if the load limit is reached.

How to Setup Closed-loop Motor Control

In the following application note the complete set of closed-loop features of TMC4361A will be explained. The general flow can be divided in three different stages whereas the latter two are optional and can be used independently on each others. Firstly, the closed-loop behavior have to be prepared and established. This general setup process is mandatory and will be clarified in chapter 3. The two optional refinements of the closed-loop motor control will elucidated in chapter 4 (current scaling capabilities) and chapter 5 (back emf considerations). In the last chapter, further settings for closed-loop operations will be explained. But before starting with the closed-loop setup, the whole setup must be connected correctly. In the next chapter 2 the possible connections between µC, TMC4361A, the motor, the encoder and the motor (gate) drivers will be introduced.

Information about correct pinning and layout considerations of the different devices can be found in the data sheets of the particular components. Information on correct SPI and encoder communication schemes for TMC4361A are available in corresponding manuals, providing detailed information.

Note:There is one exception: If circular motion (see 0) is enabled and current velocity is too high for exact positioning during one revolution, it is possible to change from velocity to positioning mode.

Connecting TMC4361A with µC, motor (gate) driver and encoder

Signal Description of the Required Pins

Connecting the Particular Pins for a Minimalist Setup

In the following the required pin signals of TMC4361A will be identified to establish a complete closed-loop system. This is a minimalist setup. If you want to use the other pin signals with its features, please refer to the TMC4361A manual.

For a minimal closed-loop motor control setup please do the following:

• Connect the ground pins and the power supply pins properly. Put at every supply line a capacitor of 100nF to ground.
• Connect the external clock signal to CLK_EXT with a frequency between 4.2 to 32 MHz.
• Disable TEST_MODE by connecting it with ground.
• Connect the INTR output pin to the µC. It is not necessary to evaluate this signal, but it will be very helpful if so. The setups which will be explained in the next chapters will use this pin.
• Establish an SPI connection from the µC to TMC4361A. Thus, connect NSCSIN, SCKIN, SDIIN and SDOIN with the µC (please refer to the TMC4361A manual to establish a correct working SPI communication scheme).
• The feedback signals should be connected to the encoder interface pins (A_SCLK, ANEG_NSCLK, B_SDI, BNEG_NSDI, N, NNEG). Here, three types of encoder feedback signals can be evaluated: incremental ABN encoders, absolute SSI and SPI encoders. Please refer to the TMC4361A manual to connect encoder feedback signals correctly.
• Connect the StepDir output pins (STPOUT and DIROUT) to any motor driver.
• And/Or connect the SPI output signal lines (NSCSDRV_SDO, SCKDRV_NSDO, SDIDRV_NSCLK, SDODRV_SCLK) to the SPI input pins of a particular TMC motor driver.

Note:The encoder should be mounted directly on the motor axis for closed-loop operation to prevent unwanted coupling effects.

General TMC4361A Register Setup Hints

Different setups are possible. Please note that any encoder type can be used with any motor driver. The shown examples in the next section are not determined for the particular encoder type.

The differential signals of incremental ABN and absolute SSI encoder types are optional. If selected, the signals will be differentiated by its digital values. There is no differential amplifier in front of the digital inputs inside the chip. Per default, differential signals are expected. It can be easily switched off by setting the GENERAL_CONF register 0x00 properly (bit12: diff_enc_in_disable).

Besides setting proper configuration for the SPI output interface via SPI_{OUT_CONF} register 0x04 to provide correct SPI data for the connected motor drivers, it is also necessary to set the correct encoder type in the GENERAL_CONF register 0x00 (bit 11:10: serial_enc_in_mode). For further information about the GENERAL_CONF and the SPI_{OUT_CONF} register, please refer to the TMC4361A manual.

Further encoder settings will be available with the ENC_IN_CONF register 0x07 which will be explained in detail in the next chapters.

General TMC4361A Register Setup Hints

• All input encoder signals are digital inputs. That means that the differential pairs are evaluated digitally.
• Thus, it may be required to install differential transceivers between TMC4361A and encoder chip.
• Encoder input signals can be digitally filtered using the INPUT_FILT_CONF register 0x03: encoder input sample rate INPUT_FILT_CONF(2:0) and encoder input filter length INPUT_FILT_CONF(6:4). The filter settings can be set freely. Please refer to the TMC4361A manual for further information.

Sample Setups

In the following examples, dashed lines defines optional wires.

Example 1: Motor driver TMC2660/260/261 and incremental ABN encoder

The TMC26x can be configured and driven completely via the SPI output interface of the TMC4361A. The configuration values for the five registers of the TMC26x have to be addressed to the cover register 0x6C of TMC4361A. It will sent automatically after writing to this register. Thus, no connection from µC to the TMC motor driver is required because the communication between both devices can be tunneled through TMC4361A.

If SPI output will be only used for configuration the Step/Dir output pins of TMC4361A can be used to provides steps for the motor driver. The index signal N is also optional.

If no step direction signals are used, the SPI_{OUT_CONF} register 0x04 have to be set accordingly that SPI current datagrams will be sent during motion.

Example 2: Motor driver TMC262 and an Absolute SSI Encoder

This setup depicts the direct configuration of the stepper driver by the µC. Anyhow, the TMC262 can be also configured and driven completely via the SPI output interface of the TMC4361A as mentioned in the example before. Using this setup, the step direction output is necessary to provide step inputs for the motor driver. Also, the negated SSI encoder signal lines are optional.

Note:

• TMC4361A supports absolute SSI encoders which provides multiturn as well as singleturn data.
• In the latter case, calculation of multiturn data have to be turned on internally (TMC4361A) for the most closed-loop motor control applications.

Example 3: Motor driver TMC24x/23x and an absolute SPI encoder

This setup covers the connection between TMC4361A and the TMC motor gate drivers TMC248/239/249. Only SPI current datagrams will be sent, no configuration is necessary. If the stepper motor drivers TMC246/236 are connected no MOSFET power stage is required. All other connections can be transferred without changes from this setup.

The SPI encoder is not differential intrinsically. Thus, differential inputs are switched off automatically if SPI encoders are selected in the GENERAL_CONF register 0x00.

Note:

• TMC4361A supports absolute SPI encoders which provides multiturn as well as singleturn data.
• In the latter case, calculation of multiturn data have to be turned on internally for the most closedloop motor control applications.

Example 4: Motor driver TMC2130 and an incremental ABN encoder

The TMC26130 can be configured and driven completely via the SPI output interface of the TMC4361A. The configuration address values for the registers of the TMC2130 have to be addressed to the cover register 0x6D of TMC4361A, whereas the data value have to be addressed to the cover register 0x6C due to the fact that 40 bits overall have to be sent to TMC2130. It will sent automatically after writing to this register 0x6C. Thus, no connection from µC to the TMC motor driver is required because the communication between both devices can be tunneled through TMC4361A. If SPI output will be only used for configuration the StepDir output pins of TMC4361A can be used to provides steps for the motor driver. The index signal N is also optional.

If no step direction signals are used, the SPI_{OUT_CONF} register 0x04 have to be set accordingly that SPI current datagrams will be sent during motion.

Example 5: Any motor driver and an incremental ABN encoder

Here, TMC4361A is connected with any motor driver. The configuration have to be done via the µC if the driver do not support the TMC SPI configuration scheme. However, if the SPI communication is the same as for TMC drivers, the configuration can be tunneled through TMC4361A. In most cases, step direction output is required. Only if DAC are connected directly to the TMC4361A SPI output current datagrams can be used for motor motion. Please refer to the TMC4361A for further information.

Register access from µC to TMC4361A

Ensuing datagrams (from the µC to TMC4361A) will be written as hexadecimal order extensions to the current values of the particular registers

Please note the most significant bit has to be ‘1’ for writing access to the register. Please refer to the TMC4361A manual for further information.

For example, if bit13 of a given GENERAL_CONF register 0x00 hast to be set to ‘1’ and bit18 to ‘0’, the following sequence result in the required datagram:

Now, 0x8012307678 can be sent to TMC4361A and the particular switches will be set.

Setup SPI output configuration for TMC motor (gate) drivers

If current datagrams should be sent from TMC4361A to TMC motor (gate) drivers and/or configuration datagrams from the µC through TMC4361A to TMC motor (gate) drivers, the SPI output have to set appropriately in the SPI_{OUT_CONF} register 0x04. The setup depends on the motor (gate) driver which is connected to TMC4361A. In the following summary the timing is equally:

• SPI_OUT_BLOCK_TIME = 8 clock cycles
• SPI_OUT_LOW_TIME = 4 clock cycles
• SPI_OUT_HIGH_TIME = 4 clock cycles

This will result in an eightfold slower data rate than the base clock frequency. Faster and slower transfers are possible and depends on the board layout. The COVER_DATA_LENGTH will be set to 0 to enforce automatic assign at which the bit length for cover datagrams is adapted in dependence of the selected TMC motor (gate) driver.

Furthermore, the bits12:4 can be used to adapt the communication with the driver further. For further information about driver specific settings and timing information, please refer to the TMC4361A manual.

Setup INTR Output Configuration

To avoid polling for an executed cover datagram the usage of the INTR output pin is advisable. After configuring the interrupt polarity (default: low active interrupt), the COVER_DONE event (bit25 in events register 0x0E) have to be assigned as INTR output in the INTR_CONF register 0x0D. Then, the status event register 0x0E have to cleared. This can be done easily by reading it.

Now, every interrupt indicates a transferred cover datagram. The next cover datagram can be initiated then. After every interrupt, the status event register 0x0E have to be read to clear the interrupt. The following sequence will be repeat in most order sequences at startup in the next chapters. It comprises the assignment of the COVER_DONE event as interrupt at INTR and the initial clear of all EVENTS bits after startup.

Note: If certain bits of the EVENTS register should be protected by utilizing EVENT_CLEAR_CONF register 0x0C the bit has to be reset by a writing access to the EVENTS register 0x0E. Please refer to the TMC4361A datasheet for further information.

Sending cover datagrams to setup TMC26x/389

TMC24x/23x are not configurable via register request. Other TMC stepper motor (gate) drivers can be configured using the cover datagram feature of TMC4361A. The mechanism will be explained with the help of TMC26x/389. Thereby, five registers have to be set to put the driver into operation.

The read response of the driver can be read out at the COVER_DRV_LOW register 0x6E. Anyhow, this is not necessary as the status bits will be assign automatically to the status flag register 0x0F of TMC4361A.

In the following the sequence of datagrams from µC to TMC4361A will be displayed with an example for a TMC26x/389 setup. Important values to configure are the chopper settings (CHOP_CONF), the current scale (CS) bit which will determine the maximum current limit at all for the motor (gate) driver, the StepDir disable bit (SDOFF) and the sense resistor value (VSENSE). All register for the TMC26x/389 should be adapted to the current board setup. Please refer the TMC26x/389 manual for detailed information.

Example configurations of appropriate current settings for TMC26x

The current settings are dependent on the used motor type. In the following two examples are explained, how to set correct current settings for CS (current scaling) and VSENSE of TMC26x. The current examples are configured as regards peak current due to the fact that these settings of TMC26x will provide the maximum current which can be reached:

Example 1: Conditions: Motor current I_RMS = 0.85 A, Sense resistor: R_SENSE = 0.22 Ω Settings:

V_SENSE = 0 → Sense resistor voltage V_FS = 305mV (FS - Full scale)

→ Maximum possible current IFS = V_FS / R_SENSE = 1.39 A

I_PEAK = I_RMS ∙ √2 = 1.2 A

→ CS = I_PEAK / I_FS ∙ 32 – 1 = 26

Result:

Cover datagram at sequence no. 8: 0xEC000C001A and
Cover datagram at sequence no. 10a: 0xEC000EF080 or
Cover datagram at sequence no. 10b: 0xEC000EF000

Example 2: Conditions: Motor current IRMS = 0.5, Sense resistor: R_SENSE = 0.15 Ω Settings:

V_SENSE = 1 → Sense resistor voltage VFS = 165mV

→ Maximum possible current IFS = V_FS / R_SENSE = 1.1 A

I_PEAK = I_RMS ∙ √2 = 0.71 A

→ CS = I_PEAK / I_FS ∙ 32 – 1 = 19

Result:

Cover datagram at sequence no. 8: 0xEC000C0013 and
Cover datagram at sequence no. 10a: 0xEC000EF0C0 or
Cover datagram at sequence no. 10b: 0xEC000EF040

Example configuration for TMC21x0

In the following the sequence of cover datagrams from µC to TMC4361A will be displayed with an example for a TMC21x0 setup. The start-up sequence is the same as explained for the TMC26x, except that both cover registers (COVER_LOW 0x6C and COVER_HIGH 0x6D) have to be used as 40 bits overall are sent and received to / from TMC21x0. The read response of the driver can be read out at the COVER_DRV_LOW register 0x6E and COVER_DRV_HIGH 0x6F. Important values to configure are the chopper settings (CHOP_CONF) including VSENSE, the current scale (CS) bit which will determine the maximum current limit at all for the motor (gate) driver, the StepDir disable bit (direct_mode) of GCONF register. Any other registers for the TMC21x0 should be adapted to the actual board setup. Please refer the TMC21x0 manual for detailed information.

Example configuration of appropriate current settings for TMC21x0

The current settings are dependent on the used motor type. In the following example is explained, how to set correct current settings for IRUN and/or IHOLD (current scaling) and VSENSE of TMC21x0. The current example is configured as regards peak current due to the fact that these settings of TMC21x0 will provide the maximum current which can be reached:

Example 1: Conditions: Motor current IRMS = 0.85 A, Sense resistor: R_SENSE = 0.22 Ω Settings:

V_SENSE = 0 → Sense resistor voltage VFS = 320mV (FS - Full scale)

→ Maximum possible current I_FS = V_FS / R_SENSE = 1.33 A

I_PEAK = I_RMS ∙ √2 = 1.2 A

→ CS = I_PEAK / I_FS ∙ 32 – 1 = 27

Result:

Cover datagram at sequence no. 4: set vsense=0
Cover datagram at sequence no. 6a: 0xEC00051B1B or
Cover datagram at sequence no. 6b: 0xEC0005xx1B

Basic encoder setup for closed-loop control

In the following the basic encoder setup for TMC4361A closed-loop operation will be explicated.

General encoder selection

As explained, different encoder types (selection by setting bit11:10 of GENERAL_CONF register 0x00) will be supported. There is also an option to disable differential inputs by enabling bit12 of GENERAL_CONF. Per default, differential inputs are enabled, except for SPI encoders! Following possibilities are practicable:

The N event configuration can be adapted in the ENC_IN_CONF register 0x07: bits(11:2) have to be set particularly. Please note that clear_on_n (bit1) must not be set because clearing the external position ENC_POS will definitely disrupt correct closed-loop behavior in most cases.

Absolute encoder configuration

Before absolute encoders are assigned as active (by setting GENERAL_CONF accordingly), please configure the data transfer scheme properly.

Therefore, changing the serial encoder data input register ENC_IN_DATA 0x08 will lead in a different bit length for the data transfer and the corresponding interpretation of the data by TMC4361A. Further on, multi cycle data and gray code enabling can be set by adapting bit17 and bit18 of ENC_IN_CONF register 0x07. Bit19 will change the interpretation of the alignment of the data.

By setting the bit count of the data transfer, the number of clock cycles which will be generated by TMC4361A for the serial encoder are automatically assigned. The timing of the data transfer/clock can be adapted by changing the registers 0x56 (Low and High level duration), 0x57 and 0x58 which will tune required delay times for the absolute encoder data transfer. Please refer to the TMC4361A manual for further information.

Finally, SPI data transfer of SPI encoders can be changed by using bit20 and bit21 of the ENC_IN_CONF register 0x07.

Bit12 and bit13 of ENC_IN_CONF register 0x07 enable multiturn data if the encoder supports this feature. If only singleturn data is transmitted from the encoder, in closed-loop motor control applications multiturn data has to be calculated by TMC4361A. Thus, set bit16 (calc_multi_turn_behav) of register 0x07 to ‘1’. This way, TMC4361A will add or remove one revolution if an overflow for the singleturn data is recognized.

Areas of Special Concern

The difference between encoder values of consecutive data transfers must not exceed the half of one revolution during multiturn calculation mode.

Restrict maximum data variation

To avoid the acceptance of encoder values that differ significantly, please use bit31 of the ENC_IN_CONF register 0x07 and the SER_ENC_VARIATION register 0x63 - bit(7:0). Enabling this serial_enc_variation_limit switch results in an automatic rejection of encoder data that differ more than 1/8 · SER_ENC_VARIATION · ENC_IN_RES compared to actual ENC_POS.

Encoder resolution

A correct setup of the encoder resolution is mandatory. First of all, the full step per revolution have to be setup properly. Thus, the FS_PER_REV value in the STEP_CONF register 0x0A have to be checked (default: 200 full steps per revolution) and, if necessary, changed. Further on, the microsteps per fullsteps MSTEP_PER_FS must not be adapted because 256 microsteps are mandatory for closed-loop operation.

Note:

The microstep per fullsteps MSTEP_PER_FS MUST be set to 256 in register 0x0A in the case of closed-loop control

Necessarily, the encoder resolution ENC_IN_RES (register 0x54) have to be set. This value is defined as encoder steps per revolution.

The internally calculated microstep per encoder step constant will be able to read out at 0x54(30:0) with 15 digits and 16 decimal places.

This value represents the number of microsteps which will be added to the encoder position ENC_POS with every detected encoder step for incremental encoders. For absolute encoder, this constant will be multiplied with the current angle which will also result in an encoder position ENC_POS representing microsteps. The encoder constant will be calculated as follows:

ENC_CONST = MSTEPS_PER_REV ∙ FS_PER_REV / ENC_IN_RES.

Per default, the decimal places represent binaries. But for some application, it will be better use a decimal representation of the decimal places. If so, the decimal places 0x54(15:0) are calculated as 1/10000.

If calculated automatically, TMC4361A will take the representation which will suit the best. Only, if both representation are not able to map the encoder constant to exactly 16 decimal places, bit0 of the ENC_IN_CONF register will define the binary (bit0=’0’) or decimal (bit0=’1’) representation. Please note that a not exact encoder constant can lead the encoder mismatches in the long run. On the next page, two examples clarifies the differences between both representations.

Example 1:

An incremental encoder with 8192 pole pairs will generate 32768 encoder steps per revolution due to 4 possible AB transition per pole pair. Thus, ENC_IN_RES have to be set to 32768. With 256 microsteps per full step and 200 full steps per revolution, this leads to an encoder constant of ENC_CONST = 1.5625. This value can be exactly represented as binary number which can be read out then at 0x54: ENC_CONST = 0x00019000

Example 2:

An incremental encoder with 500 pole pairs will generate 2000 encoder steps per revolution due to 4 possible AB transition per pole pair. Thus, ENC_IN_RES have to be set to 2000. With 256 microsteps per full step and 200 full steps per revolution, this leads to an encoder constant of ENC_CONST = 25.6. Six tenth can not be exactly represented with 16 binary decimal places. Therefore, the decimal representation will be used as ENC_CONST with 6000/10000: ENC_CONST = 0x00191770

If it is necessary to calculate the encoder constant manually, it can be written directly to 0x54 by setting the MSB of this register to ‘1’. For example, if the encoder constant should be set to 25.6 manually, write 0x80191770 to register 0x54. Because this is a decimal representation, bit0 of 0x07 have to be set to ‘1’. Otherwise, the encoder step position will be calculated with a binary encoder constant.

Finally, the encoder direction should match the motor direction. This can be reached by various settings. Thus, please move the motor with in an open loop setup in any direction (the encoder resolution has to be set before). If the internal position X_ACTUAL and the external position ENC_POS will not differ in its signs, the directions of the motor and the encoder match. No change are necessary. As opposed to a match, following possibilities for a correct setup are available if motor and encoder diverge:

• Invert the encoder direction by setting invert_enc_dir bit29 of ENC_IN_CONF register 0x07 or
• Interchange the motor cables or
• Remount the encoder.

It is recommended to use the first setup change as no hardware change will be required.

Areas of Special Concern

Do not use the invert_enc_dir switch in combination with a serial encoder and a manual defined encoder constant because the value of ENC_IN_RES defines the resolution for one revolution.

This combination will result in failures due to the fact that the manual encoder constant equals not the revolution resolution in most cases.

Encoder compensation

Systematical encoder misalignments can be compensated on chip. Especially magnetic encoders have high variances on the output value even if the encoder is correctly mounted. A deficiently installed encoder can send values which do not result in a circle. Often, the deviation from the real position results in a new function which is similar to a sine function. Adding offset that follows a triangular shape can improve the encoder value evaluation significantly (refer to Figure 3.1).

The left graph illustrates the difference between encoder position and real position (blue dots - position deviation) as a µstep function within one encoder revolution of 256 microsteps. Blending in a proper triangular function whose function values will be added to the position deviation values will result in another graph at the right side whose position deviations are now minimized significantly compared to the uncompensated values. TMC4361A provides this simple compensation algorithm which adds automatically an microstep offset to the internal microstep counter ENC_POS if turned on. Only three parameter have to be assigned properly in register 0x7D:

• ENC_COMP_XOFFSET (XOFF): the 16bit unsigned abscissa value which is normalized to the number of microsteps per revolution
• ENC_COMP_YOFFSET (YOFF): the 8bit signed ordinate value of the minimum of the triangular function in microsteps
• ENC_COMP_AMPL (AMPL):the 8bit unsigned maximum amplitude value of the triangular function which must not exceed 127!

Notes:

• The triangular function must be opposed to the position deviation function to compensate properly which means that the maximum of the position deviation function will lead to the minimum of the triangular function and vice versa.
• The abscissa value of the maximum of the triangular function is defined a half revolution distant to the minimum!
• The compensation algorithm have not to be turned on explicitly. It is switched on all the time, but as long as the encoder compensation registers are set to 0, all values will be compensated with 0.

To obtain the correct value for ENC_COMP_XOFFSET the minimum of the triangular function have to be assigned. The function argument of this minimum is the X offset. As this register is normalized to the microstep count per revolution, the value have to be divided by the number of microsteps per revolution and then multiplied by 2^16.

Example 1:

A motor with 200 fullsteps and 256 microsteps per fullstep has 51200 microsteps per revolution. The minimum for the triangular function is located at (10000 microsteps; -12 microsteps) with a maximum deviation of 65 microsteps

• ENC_COMP_XOFFSET = 10000 / 51200 * 2^16 = 12800
• ENC_COMP_YOFFSET = -12
• ENC_COMP_AMPL = 65 - (-12) = 77

The maximum of the triangular function will be then at (35600;65)

Example 2:

A motor with 72 fullsteps and 256 microsteps per fullstep has 18432 microsteps per revolution. The minimum for the triangular function is located at (11000 microsteps; -54 microsteps) with a maximum deviation of 8 microsteps

• ENC_COMP_XOFFSET = 11000 / 18432 * 2^16 = 39111
• ENC_COMP_YOFFSET = -54
• ENC_COMP_AMPL = 8 - (-54) = 62

The maximum of the triangular function will be then at (1784;65)

Automatic Linearization of encoder misalignments

Following sequence suggest how to obtain compensation data for a given setup:

1. Setup the motor driver and the encoder properly (no load!, maximum current)
2. Move motor to ENC_POS = 0.
3. Set X_ACTUAL = 0.
4. Move in one direction for several revolutions and save the ENC_POS_DEV values at every fullstep position (MSCNT = 128/384/640/896)
5. Move backwards and save the same data and generate the average over both direction and the several revolutions
6. Calculate the minimum value and the maximum amplitude of the triangular function by comparing the minimum and maximum of the stored position deviation values
7. Set the compensation value register 0x7D properly
8. Move again in both direction and check if the compensation is working properly. If not, repeat from 2.
9. After later executed power-ups, it has to be verified that the positions X_ACTUAL and ENC_POS do not differ to the settings of the initialization process (2.-7.)

Note:

• Due to the fact that the maximum of the triangular function is located a half revolution away from the minimum, it can be advisable to try different settings for the compensation register around the minimum to get the best compensation result if the real position deviations do not fit exactly to sine function!

Closed-loop setup

After setting the TMC4361A encoder properties properly, all closed-loop properties should be setup accordingly. Please be aware that before calibrate and switch on closed-loop operation mode that is explained in detail at the end of this chapter, every configuration setup step has to be finished, also the optional features described in the following chapters 5 and 6. For a basic closed-loop motor control operation, only a few values have to be set:

1. Dependent on the actual mismatch between encoder position ENC_POS and internal position X_ACTUAL, the external field which will be initiated by TMC4361A will be altered by an angle offset to overcome the mismatch. This maximum commutation angle CL_BETA 0x1C(8:0) can be set in a range between 0 and 511 microsteps whereas 255 microsteps are equal to 90° when 256 microsteps per full step are set. As long as the position mismatch exceed CL_BETA microsteps the maximum commutation angle will be used to resolve the position mismatch. Because an offset of 90° to the actual motor angle are equivalent to the largest force which can be applied to the motor, CL_BETA = 255 (default value) would fit for best performance. Higher angle values will result in a field weakening operation which can lead to higher velocities but lesser force.



Note:

• If the maximum difference CL_BETA is reached, regulation will be continued with a maximum commutation angle of CL_BETA! An event CL_MAX will be generated in this case.
2. The closed-loop proportional term CL_DELTA_P 0x5C defines the response to a position mismatch as long as the maximum commutation angle is not reached. The higher this value the stronger the response to position mismatches, but also the higher the possibility for oscillations. Higher values means also that the maximum commutation angle CL_BETA will be reached with a smaller position mismatch (see Figure 4.1). The representation of CL_DELTA_P consists of 8 digits and 16 decimal places, e.g. CL_DELTA_P = 0x018000 = 1.5 will result in a commutation angle which is 1.5 multiplied to the current position mismatch. In the following, three different setups are illustrated in a table to clarify the differences of various settings (microsteps per fullstep = 256):

Setup No:	1	2	3
CL_BETA =	255 (0x0FF)	200 (0x0C8)	275 (0x113)
CL_DELTA_P =	1.5 (0x018000)	0.9375 (0x00F000)	2.75 (0x02C000)
Position mismatch (microsteps/angle)	Commutation angle (microsteps/angle) (red entry - maximum angle reached!)
36 / 12.7°	54 / 19.0°	34 / 12.0°	99 / 34.8°
96 / 33.8°	144 / 50.6°	90 / 31.6°	264 / 92.8°
264 / 92.8°	222 / 78.0°	139 / 48.9°	275 / 96.7°
210 / 73.8°	255 / 90°	197 / 69.3°	275 / 96.7°
266 / 93.5°	255 / 90°	200 / 70.3°	275 / 96.7°

3. Finally, setting CL_TOLERANCE 0x5F (microsteps) properly will result in a mismatch range between -CL_TOLERANCE and +CL_TOLERANCE where CL_DELTA_P is automatically equal to 1.0. It is recommended to set CL_TOLERANCE at least slightly higher than the ENC_CONST to avoid too fast responses (if CL_DELTA_P > 1.0) on encoder transitions due to encoder noise.

Note:

• If the difference between internal and external position is inside of the range created by the CL_TOLERANCE value, a flag CL_FIT_F will be released. If the ENC_POS_DEV have been greater than the tolerance before, an event CL_FIT will be also generated to indicate a concordance of internal and external position after a mismatch

Limiting the maximum correction velocity

After turning on closed-loop motor control, TMC4361A will always try to counter position mismatches. If no other settings are made, the correction will be assigned as fast as possible. To limit the catch-up velocity, the internal PI regulator of TMC4361A can be used. Try to assign different values for the following parameters to fit for the best performance of the closed-loop velocity limitation:

• The proportional term CL_V_MAX_CALC_P register 0x5A and the integral term CL_V_MAX_CALC_I register 0x5B define the PI regulator for the maximum velocity limitation. The PID regulator will be updated every 128th clock cycle (update frequency = f_CLK/128). The integral term is an product of CL_V_MAX_CALC_I / 256 ∙ PID_I_SUM / 256. The current integrator sum PID_I_SUM can be read out at register 0x5B.
• By setting the value of the PID_DV_CLIP register 0x5E, the maximum velocity value (pps - pulses per second, no decimal places!) which will be added/subtracted (in dependence of the internal direction) to the current internal velocity V_ACTUAL will be defined. For instance, if PID_DV_CLIP is set to 20000 pps (0x00004E20) and V_ACTUAL = 100000.0 pps, the resulting maximum catch-up velocity will not exceed 120000 pps. Hence, setting PID_DV_CLIP to 0 will result in a static behavior where no catch-up is possible.
• It is also possible to set the value of the PID_I_CLIP register 0x5D(14:0) which is the clipping parameter for PID_I_SUM the real clipping value PID_I_SUMCLIPPED = PID_I_CLIP ∙ 216 . The maximum value of PID_I_CLIP should meet the condition PID_I_CLIP ≤ PID_DV_CLIP / PID_I. If the error sum PID_I_SUM is not clipped, it is increased with each time step by PID_I • PID_E. This continues as long as the motor does not follow.

If the values are set properly, velocity limitation can be turned on by activating (=’1’) the cl_vlimit_en switch of the ENC_IN_CONF register 0x07:

ENC_IN_CONF && 0x00FE7FFFFF || 0x8708000000.

In case position deviation at the end of an internal ramp calculation is still left, the SPI and/or Step/Dir output ramp for correction is a linear deceleration ramp, independently from the preset ramp type. This final ramp for error compensation is a function of ENC_POS_DEV and PI control parameters (CL_V_{MAX_CALC_P}, CL_V_{MAX_CALC_I}, PID_I_CLIP, and PID_DV_CLIP).

Closed-loop operation during ramp positioning mode

If closed-loop operation is turned on during internal positioning mode, TARGET_REACHED status will not be activated as long as the absolute value of the position mismatch ENC_POS_DEV will exceed the assigned value in the CL_TR_TOLERANCE register 0x52. This ensures that even when the internal ramp has been finished, no target reached flag/event is released as long as the external position will not match the internal position within the given tolerance range.

Closed-loop velocity mode

It is also possible to use turn on the closed-loop velocity mode by switching on bit28 of the ENC_IN_CONF register 0x07:

ENC_IN_CONF && 0x00FE7FFFFF || 0x8710400000.

During this modus, X_ACTUAL will be manipulated if the position difference ENC_POS_DEV exceeds 768 microsteps. If so, X_ACTUAL will be changed for 256 microsteps towards ENC_POS. Thus, the position mismatch should not exceed 768 microsteps. That way, a velocity mode can be achieved where TMC4361A will always try to match the given velocity V_MAX. Further on, if the maximum velocity V_MAX should not be exceeded, it is recommended to set PID_DV_CLIP = 0 and to enable the velocity limit. This setting will allow a catch-up as the position mismatch is relatively small. Anyhow, if the encoder is perfectly mounted and have a high resolution, it is possible that catch-up will be very slow or will not be achieved at all. If so, set PID_DV_CLIP to a low value (e.g. between 1 and 1000) according to your maximum value V_MAX.

Note:

Basically, this closed-loop operation mode fits to the velocity ramp mode. Anyhow, it is not forbidden to use positioning mode with the velocity closed-loop operation mode. Due to the feasible permanent adaptations of the internal position X_ACTUAL, it is possible that the internal ramp can miss the end point of the internal positioning ramp.

Closed-loop calibration and activation

Now, the basic setup is ready to drive in closed-loop motor control modus. Please be aware that before calibrate and switch on closed-loop operation mode, every configuration setup step has to be finished, also the optional features described in the following chapters 5 and 6. Turning on closed-loop operation by setting the regulation_modus switch = b’01 of ENC_IN_CONF register 0x07 should be sufficient as calibration of closed-loop operation will be done automatically when closed-loop modus is switched on. It is only required that X_ACTUAL register 0x21 equals MSCNT register 0x79.

Anyhow, for best performance of a compensated encoder the following calibration sequence should be followed:

1. Disable any current scaling and calibrate with maximum current.
2. Move to any full step position in open loop mode which means that the absolute current values of the sine and cosine waveform lookup table should be equal: |CURRENTA| = |CURRENTB| (CURRENTA register 0x7A(8:0); CURRENTB register 0x7A(24:16)). Using the TMC4361A default waveform and 256 microsteps per full step, full step position is reached if X_ACTUAL mod 256 = 128, e.g. X_ACTUAL = 128 or 384 or -640 or 14976. Here, the most stable position should be reached as the current through all inductors amount the same absolute value which leads to an equal force from each side affecting the rotor. If the motor (gate) driver uses its own sine wave (e.g. using TMC26x with StepDir input), move to a full step position of the motor driver!
3. Ideally, the encoder position ENC_POS should not switch if possible. If the encoder position jiggles, try the next full step position.
4. Then, turn on closed-loop operation (bit23:22 of ENC_IN_CONF 0x07 = b’01) and calibrate immediately (bit24 of ENC_IN_CONF 0x07 = ‘1’): ENC_IN_CONF && 0x00FE3FFFFF || 0x8701400000
5. After waiting for at least 10us, turn off closed-loop calibration: ENC_IN_CONF && 0x00FE3FFFFF || 0x8700400000

Now, closed-loop operation is active and TMC4361A will always try to match the external ENC_POS with the internal position X_ACTUAL. The current difference between both can be read out at the ENC_POS_DEV register 0x52. During closed-loop operation, the value of ENC_POS_DEV considers the closed-loop offset CL_OFFSET which can be read out at register 0x59. This value is the position difference during the calibration process. It is readable, but also writeable.

Note:

The calibration phase (step 4) can be skipped in case X_ACTUAL equals MSCNT. Set it up after step 3 in case X_ACTUAL can be adapted before closed-loop operation.

Areas of Special Concern

To avoid erroneous closed-loop behavior, do not write to closed-loop offset CL_OFFSET register 0x59 during closed-loop operation except for calibration purpose:

To avoid recalibration at restart after switching off closed-loop operation, turn on closed-loop and then write back the stored CL_OFFSET value to its register. The application should move in the same manner as before. Using CL_OFFSET can also avoid the recalibration of the application after a shutdown. Especially systems with absolute encoders would benefit of this recalibration option. Additionally, if X_ACTUAL and the number of revolutions would written back to ENC_POS (if absolute encoders are turned on) before starting closed-loop operation, the whole application would find its last stable closed-loop position, even if the motor have been moved during off-time.

But beware, do not use this option for recalibration if an incremental encoder is used when the N signal is not involved to get the correct absolute position in one revolution!

Current scaling during closed-loop motor control

The TMC4361A provides a scaling unit which can also be used during closed-loop operation. Thereby, energy can be saved as long as no mismatch occur. Basically, up to four parameters can be set by assigning the SCALE_VALUES register 0x06 properly. The particular real resulting scale value of one of the parameters x (which will be introduced in the following) is calculated by the following equation:

real_scale_value = (x + 1) / 256.

Areas of Special Concern

If TMC26x / TMC21xx motor (gate) driver are used in StepDir mode and the current values should be transferred via cover datagrams, only the 5 least significant bits of the current values are transmitted.

Thus, following equation holds true for this setup:

real_scale_value = (x + 1) / 32

Note:

The subsequently introduced values are based on the current scaling with a maximum value of 255. If the StepDir mode with scale value datagrams are used, please adapt the introduced values accordingly

In register 0x06, the following values can be set:

• Closed-loop maximum current scale value CL_I_MAX = SCALE_VALUES(15:8): By setting this register, the maximum current will be set. This current scale value is reached as soon as the position mismatch ENC_POS_DEV exceeds the CL_BETA limit. If during TMC26x configuration (see section 2.8) the RMS current is assigned as CS scale value, CL_I_MAX can be set to 255. Otherwise, if the peak current I_PEAK is set, CL_I_MAX should set to 180 due to the fact that √2 ∙ 256 = 181. Thus, the maximum current will not exceed IRMS. This is especially crucial if the motion speed is low. Anyhow, larger values are possible (e.g. to overcome heavy loads), but beware of overcurrent over a long run!
• Closed-loop minimum current scale value CL_I_MIN = SCALE_VALUES(7:0): The minimum scale value can be set freely. It will be assigned as long as the position mismatch do not exceed a certain limit. This leads to current saving behavior as long as the position mismatch is small. As good starting point, hold scale current settings for open loop behavior can be assigned. For instance, with CL_I_MIN = ¼ ∙ CL_I_MAX = 180 / 4 = 45, up to 1/16 of the maximum energy can be saved.
• The position mismatch limit in microsteps at which current scale values will be increased is defined by CL_START_UP = SCALE_VALUES(23:16). The current scale value will be calculated according to a linear function between CL_START_UP and CL_BETA. It is recommended to set CL_START_UP higher than CL_TOLERANCE.
• CL_START_DN defines the position mismatch value at which the ramp down of the current values will start. It is recommended, to set this parameter to 0 where this limit is set to CL_BETA. Hence, the downward ramp is the same as the upward ramp. Other settings lead to a hysteresis.

In the figure on the next page, the introduced parameters can be found. If the downward ramp is different compared to the upward ramp (CL_START_DN > 0 CL_START_DN ≠ CL_START_UP), the linear function will be the same, resulting in a smaller position mismatch where the minimum scale value is reached. If CL_START_UP is set to 0, closed-loop minimum will be reached almost never as the position mismatch is almost never equal to 0.

For evaluating different upscaling and/or downscaling ramps the delay parameters CL_UPSCALE_DELAY (register 0x18) and CL_DNSCALE_DELAY (register 0x19) can be assigned, too. These values define the period of time in clock cycles in which the current scale value will be altered for one step towards CL_IMAX resp. CL_IMIN. Figure 5.2 depicts the current scaling timing behavior as a function of CL_UPSACLE_DELAY and CL_DNSCALE_DELAY.

After setting the closed-loop scale values properly, closed-loop scaling can be enabled:

0x8500000080.

Note: Though even other scaling modes are enabled, any scale mode becomes disabled if closed-loop scaling is enabled.

Example for a combined usage of closed-loop angle tracking and current scaling

In the following a sequence is depicted which shows a rising deviation between the target position (internal position X_ACTUAL = red vector) and the real motor position (external position ENC_POS = blue vector). Due to the settings of CL_BETA = 255 (90°) the green dotted commutation angle will not change until the deviation exceed 90° here if CL_DELTA_P = 1.0 (0x010000). Thus, the green commutation angle leads the motor to the target position inside of the 90° deviation, but the green current vector will grow with the rising of the deviation to encounter the mismatch. It reaches its maximum if the limit of 90° is reached. Then, the current vector will follow the external position with full current because no steps should be lost.

Considering the back-EMF

When higher velocities are reached, a phase shift between current and voltage occurs at the motor coils. The current control will be transferred in a voltage control. This motor and setup dependent effect have to be compensated as currents will be still assigned for the motor control. This can be done by the γ-correction where a velocity dependent angle in motion direction will be added to the current commutation angle.

The maximum gamma angle can be set to almost 90° (= 255 microsteps) at CL_GAMMA 0x1C(23:16). Due to the fact that his angle is added to the commutation angle, the complete angle can be reach 180° if CL_BETA is also set 90° (= 255 microsteps).

Areas of Special Concern

If the γ-correction is turn on, the maximum possible commutation will be (CL_BETA + CL_GAMMA ). This value must not exceed 180° (512 microsteps at 256 microsteps per fullstep). Angles of 180° and more will result in unwanted motion direction changes!

Switching on γ-correction, set cl_emf_en to ‘1’ in the ENC_IN_CONF register 0x07. Furthermore, the velocity limits for the γ-correction have to be assigned:

ENC_IN_CONF && 0x00FEFFFFFF || 0x8702000000

• CL_V_MIN_EMF 0x60: Below this velocity value [pps] the γ-correction angle is set to 0.
• CL_V_ADD_EMF 0x61: This velocity value [pps] will be added to CL_V_MIN_EMF to assign the upper limit for maximum γ-correction angle.
• Beyond (CL_V_MIN_EMF + CL_V_ADD_EMF ) the current γ-correction angle GAMMA will be set to the maximum value of CL_GAMMA
• Between CL_MIN_EMF and ( CL_V_MIN_EMF + CL_V_ADD_EMF ) the current γcorrection angle GAMMA will be calculated by a linear function which is depicted in the following figure

Velocity Setup

Setting CL_V_MIN_EMF and CL_V_ADD_EMF properly is motor dependent. Thus, connect a current probe to one phase of the motor cables and a voltage probe on the same signal line. To set appropriate velocity limits please observe the distortion of the motor current curve. Figure 6.3 depicts a oscilloscope shot where no distortion of the motor current is observed which should be the normal case during low velocities.

A good starting value for CL_V_MIN_EMF is the velocity when the back emf voltage reaches the motor supply voltage. At this point first slight distortions of the motor currents should be observed at the oscilloscope as shown in Figure 6.4.

The velocity at which the motor currents are completely distorted firstly is a good starting value for the velocity when GAMMA will be reach its maximum of CL_GAMMA (mostly 90°). Thus, CL_V_ADD_EMF will be this velocity subtracted by CL_V_MIN_EMF. Another indicator for reaching the maximum velocity for γ-correction is the vanishing of chopper cycles in the voltage curve of the sense resistor as it can be also seen in Figure 6.5 if the chopper control is turned on.

It is also possible to plot the maximum motor current (maximum amplitude of the motor current) vs. the ramp velocity. At the breaking points, both velocity limits can be identified.

Note:

• To get most sufficient result for motor behavior with γ-correction usage, all motor lines should be analyzed for the velocity limits.
• Further on, tweaking the first guess limits slightly can also result in better motor motion behavior.

Velocity calculation

As it can be seen from Figure 6.2, the trimming is based on the real motor velocity. This velocity ENC_VEL [pps] is calculated internally by TMC4361A and will be released at register 0x65. For a correct correction algorithm the encoder velocity have to be filtered (V_ENC_MEAN 0x66 [pps]). The filtering is done by the following equation:

Besides the filter exponent, further parameters can be set a register 0x63 for different encoder types to adapt the closed control with back-EMF consideration:

Note:

• A good starting value for ENC_V_MEAN_WAIT is 128 and for ENC_V_MEAN_FILTER is 7. Both values should be adapted in conjunction if ABN encoders are used. Further on, the lower both values are the faster the V_ENC_MEAN is adapted to the current velocity. But this also results in higher gradients of the mean velocity which can lead to regulation jumps if the γ correction is enabled
• To prevent this, check the V_ENC_MEAN velocity values during motion transferring the velocity limits CL_V_MIN_EMF and (CL_V_MIN_EMF + CL_V_ADD_EMF ). If the mean encoder velocity is adapted smoothly during motion γ-correction will be also executed properly

Further on, ENC_VEL0 0x62 assigns the delay in number of clock cycles which sets the encoder velocity V_ENC and V_ENC_MEAN to 0! If the signals of the AB lines will not switch during this time period or the absolute encoder values will not change, the encoder velocity will be set to 0.

Areas of Special Concern

If the absolute encoder switches its direction the first new encoder velocity will be 0! Thus, if the encoder values are toggling the encoder positions in such a way that direction changes are produced during motion, this can lead to V_ENC = V_ENC_MEAN = 0. This will result in an incorrect γ-correction if CL_VMIN_EMF has been already exceeded.

Examples

In the following complete examples are illustrated by assigning the different registers. It is advised to check all four examples, even if the setup do not correspond to given application, because slight differences for closed-loop setup are illustrated among the four examples.

Example 1: Motor driver TMC26x (SPI mode) is used with an incremental ABN encoder; back-emf is not considered; no calibration process

In the following the closed-loop calibration process is omitted after one movement in open loop to the next fullstep position. By setting X_ACTUAL equal to MSCNT, turning on closed-loop operation is sufficient for calibration purpose if a change of X_ACTUAL before closed-loop operation is appropriate (see TMC4361A manual section 16.3.2 option 1B). Though, the simplified motion to the next fullstep can lead to torque loss during closed-loop operation if the calibration point is not selected well.

Preliminary considerations:

• Non-differential ABN encoder
• No back-emf consideration
• Catch-up velocity limit = ±50 kpps
• 400 full steps per revolution
• I_RMS = 1.1 A, Sense resistor: R_SENSE = 0.15 Ω
• V_SENSE = 0 → I_FS = 2.0 A
• I_PEAK = I_RMS ∙ √2 = 1.56 A → CS = 24
• Encoder direction is inverted to motor direction

Example 2: Motor driver TMC26x (SPI mode) is used with an incremental differential ABN encoder; advanced calibration process

Preliminary considerations:

• Differential ABN encoder
• Back-emf consideration
• Catch-up velocity limit = ±50 kpps
• 200 full steps per revolution
• I_RMS = 1.1 A, Sense resistor: R_SENSE = 0.15 Ω
• V_SENSE = 0 → I_FS = 2.0 A
• I_PEAK = I_RMS ∙ √2 = 1.56 A → CS = 24