ST AN2804 Application note

AN2804

Application note

Virtual current control loop for L9374-L9375

coil driver kit in ABS-ESC control unit

Introduction

In the conventional hydraulic modulator for ABS-ESC control units, the 6 inlet valves (also known as ISO and/or TC-ISO valves), that is, the 4 devoted to the ABS/TCS functions and the other 2 valves devoted to the ESC functions, need a strict control of the coil energizing current. From the point of view of the electronic components involved in the ABS-ESC system partitioning, the above specification means to have a coil driver kit with 6 current regulated channels. The current regulation specification guarantees to have in the coil, driving the opening/closing of the on-off solenoid valve, the same energizing current against supply voltage changes, temperature gradient, "stiction" phenomena inside the valve due, for instance, to the aging and/or waste of the hydraulic and mechanical components (e.g. brake fluid, ball, armature, spring,…) involved into the valve working.

In this work, we propose a coil driver kit, composed by L9374-L9375, for new generation ABS/ESC control unit. The L9374 is a smart quad low side driver with 2 current regulated channels (accuracy is about 6%) and 2 PWM channels. On the other hand, the L9375 is an octal low side driver having 4 PWM channels and 4 conventional on-off switch channels. In order to allow the L9374-L9375 kit to work as there are 6 current regulated channels, we conceived a SW library, implemented by ST10252M microcontroller, that exploits data (these data are available on the MISO bus of the SPI) coming from the 2 current regulated channels of L9374 to calibrate the 4 PWM channels of L9375 so that they can work as having a current regulation loop. This virtual current control loop has been tested in different working conditions of the coil drivers:

● considering different supply voltages during the calibration and during the actuation of

the virtual current regulated channels of L9375;

● simulating different temperature gradients;

● simulating temperature mismatches between the coil driven by a current regulated

channel of L9374 and the coil driven by a virtual current regulated channel of L9375;

● considering several calibration timing and current set-points;

● taking into account different hydraulic modulators (Bosch, TRW).

Basically, the results show a satisfactory behavior of the virtual current control loop. In all the test conditions taken into account, it comes out an accuracy of the virtual current control loop that is similar to the nominal accuracy of the 2 current regulated channels of the L9374.

July 2008 Rev 1 1/34

www.st.com

Contents AN2804

Contents

1 L9374 and L9375 summary description . . . . . . . . . . . . . . . . . . . . . . . . . 5

2 VCCL: plan of accuracy evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

3 ABS/ESC: load analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

4 Current measurement procedure: validation . . . . . . . . . . . . . . . . . . . . 11

5 VCCL accuracy evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

6 VCCL: computational burden evaluation . . . . . . . . . . . . . . . . . . . . . . . 25

7 VCCL: fixed point arithmetic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

8 Acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

9 Revision history . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

2/34

AN2804 List of tables

List of tables

Table 1. Characterization of the load seen by the L9374 vs. the current setpoint imposed

on the Q3 channel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

Table 2. Characterization of the procedure used to measure the current into the VCCL

validation tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

Table 3. VCCL accuracy evaluation with a spot calibration having a current setpoint

on the Q3 channel of L9374 equal to 700 mA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

Table 4. VCCL accuracy evaluation in option 1A spot calibration versus power supply. . . . . . . . . . 14

Table 5. VCCL accuracy evaluation with a spot calibration having a current setpoint

on the Q3 channel of L9374 equal to 250 mA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

Table 6. VCCL accuracy evaluation in option 1B spot calibration versus power supply. . . . . . . . . . 19

Table 7. VCCL accuracy evaluation in option 1A when a temperature mismatch of 20°

is simulated (Q5 channel of L9375) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

Table 8. VCCL accuracy evaluation in option 1A when a temperature mismatch of 20°

is simulated (Q3 channel of L9375) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

Table 9. VCCL accuracy evaluation in option 1B when a temperature mismatch of 20°

is simulated (Q5 channel of L9375) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

Table 10. VCCL accuracy evaluation in option 1B when a temperature mismatch of 20°

is simulated (Q3 channel of L9375) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

Table 11. Analysis of the error propagation in the VCCL_dc_calculation function call

versus different Isat value and current setpoints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

Table 12. Acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

Table 13. Document revision history . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

3/34

List of figures AN2804

List of figures

Figure 1. L9374 application block diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

Figure 2. L9375 application block diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

Figure 3. Block diagram of the VCCL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

Figure 4. Details on the implementation of the VCCL accuracy evaluation plan . . . . . . . . . . . . . . . . . 8

Figure 5. Details on the spot calibration of the VCCL accuracy evaluation plan . . . . . . . . . . . . . . . . . 9

Figure 6. ABS/ESC 8.0 Bosch Hydraulic modulator and INLET valve section . . . . . . . . . . . . . . . . . 10

Figure 7. VCCL accuracy evaluation with a spot calibration having a current setpoint

on the Q3 channel of L9374 equal to 700 mA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

Figure 8. VCCL accuracy evaluation in option 1A spot calibration versus power supply . . . . . . . . . 15

Figure 9. Figure 10. Figure 11. VCCL accuracy evaluation with a spot calibration having a current setpoint

Figure 12. VCCL accuracy evaluation in option 1B spot calibration versus power supply . . . . . . . . . 20

Figure 13. Figure 14.

Figure 15. Sensitivity versus coil temperature in the option 1A VCCL. . . . . . . . . . . . . . . . . . . . . . . . . 23

Figure 16. Sensitivity versus coil temperature in the option 1B VCCL. . . . . . . . . . . . . . . . . . . . . . . . . 24

Figure 17. Details on the implementation of the VCCL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

Figure 18. Details on the implementation of the VCCL_dc_calculation function . . . . . . . . . . . . . . . . . 27

Detail of the VCCL accuracy evaluation in option 1A spot calibration versus power supply 15

Detail of the VCCL accuracy evaluation in option 1A spot calibration versus power supply

on the Q3 channel of L9374 equal to 250 mA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

Detail of the VCCL accuracy evaluation in option 1B spot calibration versus power supply

20 21

4/34

AN2804 L9374 and L9375 summary description

1 L9374 and L9375 summary description

The L9374 is a smart quad low side driver with integrated free-wheeling diodes (see block

diagram in figure 1). The switching of the channels is programmable via SPI (Serial

Peripheral Interface). The main time base is given by an external clock via CLKin. The Clock

Unit monitors this external clock and provides the system clock for all timings. A

Synchronization Unit is used to monitor the SPI communication and provides a sync signal

for the channel activation. The Output Duty Cycle for each channel can be programmed

individually and will be activated by the Set Point Unit. Is possible to program two output

Duty Cycles per channel with a block of 16 SPI commands as well as an individual Duration

Time for each channel actuation. The PWM Controller translates the programmed digital

duty cycle value in a PWM signal which controls the output. For the current regulated

channels the target current value is programmed. It is also possible to program two different

target currents. The target current is compared with the real load current. The output duty

cycle is then calculated with an ALU. As base for the calculation a load model is used to take

into account the variation of the real load versus the load assumption done into the first SPI

transfers setting properly the command register of address 15 (see configuration register 3).

Moreover, the L9374 ALU exploits respectively:

● measurement of the supply voltage (see VD status register of address 5);

● measurement of the voltage drop on the free-wheeling diode (see V_FWD status

● measurement of the sense resistor (see Rs status register of address 11) used to

monitor the current on the Q3 and Q4 channels for the current control loop of the L9374.

All channels are equipped with a load Diagnostic. This allows to detect an open load in off

condition as well as an under current in on condition. The power stage is protected against

over current and over temperature. A weak connection in power ground or in the

recirculation path is monitored. All monitored functions can be read out in a serial diagnostic

protocol dedicated for each channel via SPI.

The L9375 is an octal low side driver with integrated recirculation diodes for PWM controlled

channels, that is, Q5, Q6, Q7 and Q8. On the other hand, the channels Q1, Q2, Q3 and Q4

are configured as switching channels (see figure 2). To achieve a fast switch off a high

voltage output clamp is implemented for a rapid free-wheeling if the inductive load. The

switch on time can be programmed via SPI. The L9375 have the same features of the L9374

except for the current control of the channels of L9374.

5/34

L9374 and L9375 summary description AN2804

Figure 1. L9374 application block diagram

Figure 2. L9375 application block diagram

6/34

AN2804 VCCL: plan of accuracy evaluation

2 VCCL: plan of accuracy evaluation

The idea behind the VCCL is to use the information coming from the current feedback of the

Q3 and Q4 channels of L9374 in order to wrap the PWM channels of L9375 with a pseudo

current control loop implemented by the microcontroller of the ABS/ESC control unit. In the

Figure 3, the block diagram of the VCCL is described:

● set-up with a system microcontroller that works as the Master of the SPI

communication with the Slaves, L9374 and L9375;

● diagnostic cycle is applied to the L9374 current controlled channel (Qx), to collect all

possible information;

● diagnostic cycle is applied also to the L9375 PWM channel (Qy);

● using information from Qx and Qy diagnostic, the microcontroller calculates the duty

cycle to be imposed in order to obtain the current setpoint desired on the PWM channels of L9375.

The target of this work is to evaluate the difference between the desired current and the

actual current (precision of the VCCL). The accuracy of the VCCL along different working

conditions have been measured by simulating different disturbances, like battery voltage

changes, load resistance change, temperature mismatches, etc.

The result of this activity is to demonstrate that it is possible to perform a low-end ABS/ESC

system with only 2 current controlled channels instead of 6 required in the conventional

system partitioning of the ABS/ESC control unit.

Figure 3. Block diagram of the VCCL

Figures 4 and 5 describe some details regarding the implementation of the evaluation plan

conceived to quantify the VCCL accuracy. First of all, after the power-up of the

microcontroller and of the L9374-L9375, a spot calibration on the Q3 channel of L9374 is

7/34

VCCL: plan of accuracy evaluation AN2804

done for about 25 ms. During this time, the Q3 channel is driven in order to energize the

corresponding coil mounted on the INLET valve head of the hydraulic modulator used in our

tests. Two current setpoints have been explored for the spot calibration on Q3 channel: 700

mA (option 1A) and 250 mA (option 1B). While the first current setpoint is enough to close

the INLET valve the last isn't. Therefore, the option 1A shows as drawback the need to close

the INLET valve linked to the Q3 channel of L9374 for the spot calibration time. We chose a

spot calibration time of 25 ms because this represents an optimal trade-off, taking into

account the test operative conditions (the cycle-time used is about 5 ms), the filter time for

the CNR condition of the current regulated channels of L9374 (t

is equal to 8 ms) and

CNR

the possibility to wait for a long enough transition time before to start the reading, by MISO

bus of SPI, the VBAT (VD status register), VD (V_FWD status register) and ISAT (ISAT_Qx

status registers) values (see Figure 5).

Figure 4. Details on the implementation of the VCCL accuracy evaluation plan

An internal common used ALU is devoted to guarantee, through the calculation of the

formula (1), the desired accuracy (6% of error against the current setpoint) for the current

control loop of the Q3 and Q4 channels of L9374. The formula (1) implemented in the ALU

takes into account a dedicated load model. This model includes the recirculation path as

well as the resistive value of the load. This value is programmable via SPI (see R

the parameters of the recirculation path and the supply voltage of the load are measured

(see Rs, V

Equation 1

, V

). All these values are readable via SPI.

BAT

DITARGET

--------------------------------------------------------------------------------------------------------------------------=

DVBATITARGET

Rs I

The value of the programmed load (i.e. R

correction value is available via SPI (see I

can be achieved. Starting at time zero a load error is corrected by modifying the

programmed load resistor value with an integrated correction factor I

load model correction is only available for target currents higher than 110mA. In the formula

(1) the factor R

LSPI (ISAT

+ 1) represents, in a first approximation, the real load seen by the Qx channel of L9374 into the current control loop. Anyway, for a more detailed explanation of the current control loop of the L9374 we remand to the L9374 datasheet.

8/34

TARGETRLSPIISAT

1+()++

Rs Rds on()–()++

) is corrected during the control loop. This

LSPI

). With this algorithm a fast controller dynamic

SAT

. This function of

SAT

LSPI

) while

AN2804 VCCL: plan of accuracy evaluation

Figure 5. Details on the spot calibration of the VCCL accuracy evaluation plan

During the spot calibration, once the reading window is opened (see Figure 5), the values of VD, VBAT, ISAT are acquired. These values after an arithmetic mean are stored in some registers of the microcontroller. Care is taken for the ISAT value because it is represented as a two complement number (see L9374 datasheet). The values

BAT

SAT

(a)

, are used into the formula (2) implemented by the system microcontroller in order to calculate, before to energize the coil, the right duty cycle to have a pseudo current control loop on the PWM channels of L9375. Since the L9375 does not have a real current control loop, in the formula (2)) there is not the Rs resistor that we can find in the formula (1) implemented by the ALU of L9374. However, we thought to consider a minimum resistor of 50mΩ. Furthermore, the values of the voltage drop on the free-wheeling diode (i.e. V

) have been scaled in order to

maintain a ratio of 3/4 between the voltage drop on the free-wheeling diode of the PWM channels of L9375 and the voltage drop on the free-wheeling diode of the current controlled channels of L9374.

Equation 2

-----------------------------------------------------------------------------------------------------------------------------------=

TARGET

BAT

50m Ω I

TARGET

TARGETRLSPI

50mΩ 200mΩ–()++

SAT

1+()++

a. The hat on the variables indicates that they are the result of an arithmetic mean applied on the different values

read via SPI once the reading window is opened (see figure5).

9/34

ABS/ESC: load analysis AN2804

3 ABS/ESC: load analysis

In order to validate the VCCL approach, we used the INLET valves of the ABS/ESC 8.0 Bosch hydraulic modulator. The Figure 6 illustrates the hydraulic modulator and the section of an INLET valve. The coil characteristics have been measured with and without the valves by means of the Hameg LCR Meter HM8018. The obtained values are:

● R = 4.65 Ohm, L = 1.6 mH (without the valve)

● R = 5.35 Ohm, L = 7.35 mH (with the valve)

Coil characteristics have been measured, also during the calibration phase of the regulated channel (i.e. Q3) of L9374, using the estimation of the load done by the L9374 through the I

factor measurement.

SAT

Figure 6. ABS/ESC 8.0 Bosch Hydraulic modulator and INLET valve section

Table 1. Characterization of the load seen by the L9374 vs. the current setpoint

imposed on the Q3 channel

Regulated channel current:

setpoint (mA)

250 7.17 0.03

350 6.72 0.02

450 6.68 0.02

550 6.55 0.06

650 6.52 0.04

750 6.21 0.03

850 6.23 0.06

950 6.08 0.16

1050 5.96 0.03

1150 5.6 0.16

Load resistor seen by L9374:

The Tab l e 1 describes the results of a characterization of the load seen by the L9374 during the spot calibration versus different current setpoints. The mean and the standard deviation have been calculated on a set of 10 repeated measurements done in the same conditions (supply voltage, temperature, current setpoint, etc…).

10/34

mean (Ohm)

Load resistor seen by L9374:

std (Ohm)

AN2804 Current measurement procedure: validation

4 Current measurement procedure: validation

The test-bench layout used to measure the current is based on the following components:

● oscilloscope LeCroy wave pro 7300A in ERES (Enhanced resolution, about 211 bit of

vertical resolution) mode;

● current probe amplifier Tektronix TCPA300 (DC to 100 MHz of bandwidth, DC-gain

accuracy < 1%);

● AC/DC current probe Tektronix TCP312 (Lowest Measurable Current = 1mA, Maximum

Amp-Second = 50A*µs (for 1A/V range)).

Table 2. Characterization of the procedure used to measure the current into the

VCCL validation tests

Regulated channel

current: setpoint (mA)

250 251 1 0.4

350 352 1 0.6

450 454 1 0.9

550 557 1 1.3

650 658 1 1.2

750 755 2 0.7

850 857 2 0.8

950 958 2 0.8

1050 1057 2 0.7

1150 1157 2 0.6

1250 1252 2 0.2

1350 1344 2 -0.4

1450 1462 4 0.8

1550 1566 5 1.0

Current measured on

the Q3 of L9374: mean

(mA)

Current measured on

the Q3 of L9374: std

(mA)

Percent deviation

from the ideal current

control accuracy of the Q3 of L9374 (%)

The Tabl e 2 shows the results obtained from a preliminary characterization of the procedure adopted to measure the current into the VCCL validation tests. The results in the table

Tab le 2 are referred to a set of 10 repeated measurements of the current energizing the coil

during the spot calibration of the L9374 and once the reading window is opened (see

Figure 5). It is important to highlight that the measurements have been done in the same

operative conditions (supply voltage, temperature, current setpoint, etc…). From an analysis of the results reported in theTa b l e 2 , it comes out that, in the worst case, the test bench layout used to measure the current is reliable within the limits of the 1% of accuracy.

11/34

+ 23 hidden pages