CCS/TMS320F28035: PWM SW forced CBC trip from CLA

leopold rohrecker

Part Number: TMS320F28035

Tool/software: Code Composer Studio

Hi,

I'm using TMS320F28035 (Piccolo) with CCS Vers. 6.1.0.00104 (and C code both for CPU and CLA) in order to control a boost converter in discontinuous conduction mode with Peak current mode controlPWM(choke current at the beginning of a switch interval should be zero and I'm testing this condition in CLA). CLA task1 executes the C code for controlling PWM1A in order to get constant output voltage from the converter.
If at the beginning of switch interval the current is greater than a threshold value I try to swith PWM1A off for the current switch interval (in the next interval it should turn on again automatically). Doing this via putting 0 into the DAC and with the help of the analog comparator works fine (but might be slower than necessary). I'd like better avoide the analog path and generate a SW trip to turn off PWM1A. Unfortunately this doesn't work and it seems to me that PWM1 simply stops (forever, which isn't what I intended, not at all). Could there be an adressing problem in the CLA (32 bit pointer in CPU, 16 bit pointer in CLA???) when I read and
write EPwm registers the same way in CPU and CLA?

Many thanks in advance for any helpful suggestions.

Regards,
Leo Rohrecker

// ===== initialization in CPU code: =====

        EALLOW;
        // Setup compare: A = PWM max DC, B = ADC SOC
        EPwm1Regs.CMPA.half.CMPA = 0; // DC = 0
        EPwm1Regs.CMPB           = ADC_SOC_OFFSET;

        // Clock ratios to SYSCLKOUT
        EPwm1Regs.TBCTL.bit.HSPCLKDIV = TB_DIV1;
        EPwm1Regs.TBCTL.bit.CLKDIV    = TB_DIV1;

        // period
        EPwm1Regs.TBPRD = PWM_TIMER_TBPRD;
        EPwm1Regs.TBCTL.bit.CTRMODE = TB_COUNT_UP;

        // Set actions
        EPwm1Regs.AQCTLA.bit.ZRO = AQ_SET;   // Set PWM1A on Zero
        EPwm1Regs.AQCTLA.bit.CAU = AQ_CLEAR; // max duty cycle
        EPwm1Regs.AQCTLB.bit.ZRO = AQ_SET;   // Set PWM1B on Zero
        EPwm1Regs.AQCTLB.bit.CBU = AQ_CLEAR; // ADC SOC 1st group

        // CBC trip, define an event (DCAEVT2)
        // (see SPRUGE9E, Ex. 14)
        EPwm1Regs.DCTRIPSEL.bit.DCAHCOMPSEL = DC_COMP1OUT; // DCAH = comparator 1 output
        EPwm1Regs.TZDCSEL.bit.DCAEVT2 = TZ_DCAH_HI;         // DCAEVT2 = DCAH high (will become active as comp. output goes hi)
        EPwm1Regs.DCACTL.bit.EVT2SRCSEL = DC_EVT2;          // DCAEVT2 = DCAEVT2 (not filtered)
        EPwm1Regs.DCACTL.bit.EVT2FRCSYNCSEL = DC_EVT_ASYNC; // take async path
        EPwm1Regs.TZSEL.bit.DCAEVT2 = 1;                    // enable CBC event
        EPwm1Regs.TZCTL.bit.TZA = TZ_FORCE_LO;              // EPWM1A will go low
        EPwm1Regs.TZCTL.bit.TZB = TZ_NO_CHANGE;             // EPWM1B unaltered

        // define SOCs (ADC trigger)
        EPwm1Regs.ETSEL.bit.SOCBEN = 1;               // enable SOC B
        EPwm1Regs.ETSEL.bit.SOCBSEL = ET_CTRU_CMPB;   // SOC from CPMB on upcount
        EPwm1Regs.ETPS.bit.SOCBPRD = 1;               // Generate pulse on 1st event
        EDIS;

// ===== run time actions in CLA code: =====
        // check for CCM by current measurement at beginning of switch interval
        if ((float32)ADC_U_I1 > U_I_START_THRESH_CCM_TIL) {
            // start current too high .. turn off immediately
            EALLOW;
            //EPwm1Regs.TZFRC.bit.DCAEVT2 = 1; // force DCAEVT2 trip condition ... doesn't work, either !!!!!
            EPwm1Regs.TZFRC.bit.CBC = 1; // force CBC trip ... doesn't work !!!!!
            EDIS;
            //Comp1Regs.DACVAL.bit.DACVAL = 0;      // turn off immediately via analog comp. (alternative, works fine, but slower than digital path)
        }
        // controller code etc....
        // ...

// --------------------------------------- end ---------------------------------------

over 8 years ago

0 Vishal_Coelho over 8 years ago

TI__Mastermind 20850 points

Hi,

On the CLA side did you include F2803x_Cla_typedefs.h before any other peripheral header file? this file sets the correct size for data types on the CLA, so that peripheral bit field accesses are done correctly.

0 leopold rohrecker over 8 years ago in reply to Vishal_Coelho

Prodigy 230 points

Hi Vishal,

I think the necessary defines should be included to my code. The CLA C code file starts with

#include "DSP28x_Project.h"

and DSP28x_Project.h Contents is:

//###########################################################################

// FILE: DSP28x_Project.h

// TITLE: DSP28x Project Headerfile and Examples Include File

//###########################################################################

// $TI Release: F2803x C/C++ Header Files and Peripheral Examples V126 $

// $Release Date: November 30, 2011 $

//###########################################################################

#ifndef DSP28x_PROJECT_H

#define DSP28x_PROJECT_H

#include "DSP2803x_Cla_typedefs.h"// DSP2803x CLA Type definitions

#include "DSP2803x_Device.h" // DSP2803x Headerfile Include File

#include "DSP2803x_Examples.h" // DSP2803x Examples Include File

#endif // end of DSP28x_PROJECT_H definition

and DSP2803x_CLAtypedefs.h Contents is:

//###########################################################################

// FILE: DSP2803x_Cla_typedefs.h

// TITLE: Variable type definitions

//###########################################################################

// $TI Release: F2803x C/C++ Header Files and Peripheral Examples V126 $

// $Release Date: November 30, 2011 $

//###########################################################################

#ifndef CLASHARED_C_H_

#define CLASHARED_C_H_

//---------------------------------------------------------------------------

// For Portability, User Is Recommended To Use Following Data Type Size

// Definitions For 16-bit and 32-Bit Signed/Unsigned Integers:

#ifndef DSP28_DATA_TYPES

#define DSP28_DATA_TYPES

typedef short int16;

typedef long int32;

typedef unsigned char Uint8;

typedef unsigned short Uint16;

typedef unsigned long Uint32;

typedef float float32;

typedef long double float64;

#endif

#ifndef _TI_STD_TYPES

#define _TI_STD_TYPES

* These types are also defined in DSP/BIOS 5.x's <std.h> and the

* SYS/BIOS 6.x's <xdc/std.h> files. We need to protect their

* definition with the #ifndef/#define guard to avoid the duplicate

* definition warning.

* SYS/BIOS requires that the <xdc/std.h> file be included before

* any other .h files.

#endif

// Suppress warnings casting CLA pointers

#pragma diag_suppress 70,770,232

// Macros to manipulate pre-processor to generate a header file name

// at compile time that is based on the test name and can be used as

// an argument to #include

#define XSTRINGIZE(s) STRINGIZE(s)

#define STRINGIZE(s) #s

#define XCONCAT(x,y) CONCAT(x,y)

#define CONCAT(x,y) x##y

#ifdef __TMS320C28XX_CLA0__

struct MSTF_SHADOW_BITS { // bits description

Uint16 LVF:1; // 0 Latched Overflow Flag

Uint16 LUF:1; // 1 Latched Underflow Flag

Uint16 NF:1; // 2 Negative Float Flag

Uint16 ZF:1; // 3 Zero Float Flag

Uint16 rsvd1:2; // 5:4 Reserved

Uint16 TF:1; // 6 Test Flag

Uint16 rsvd2:2; // 8:7 Reserved

Uint16 RNDF32:1; // 9 Rounding Mode

Uint16 rsvd3:1; // 10 Reserved

Uint16 MEALLOW:1; // 11 MEALLOW Status

Uint16 RPCL:4; // 15:12 Return PC: Low Portion

Uint16 RPCH:8; // 23:16 Return PC: High Portion

Uint16 rsvd4:8; // 31:24 Reserved

};

extern __cregister volatile unsigned int MSTF;

#endif //__TMS320C28XX_CLA0__

#if defined(__TMS320C28XX_CLA0__)

typedef struct { Uint32 low32; Uint32 high32; } Uint64;

#else

typedef unsigned long long Uint64;

#endif

#ifndef __TMS320C28XX__

#define __cregister

#endif //__TMS320C28xx__

#define Uint16 unsigned short

#endif /*CLASHARED_C_H_*/

--------------------------------------------

I started code development for Piccolo in 2011/12, therefore these TI files are from 2011. But my code worked fine, until I tried to use SW trip within CLA C code.

Best regards,

Leo

0 leopold rohrecker over 8 years ago in reply to leopold rohrecker

Prodigy 230 points

Hi Vishal,
at the Moment I don't have Access to my test HW, but I realized that my C Compiler (v6.2.8) converts EALLOW / EDIS simply to EALLOW / EDIS in CLA assembler code file, whereas the commands __meallow() / __medis() are converted to MEALLOW / MEDIS in CLA Assembler, as it should be! Maybe my Problem is simply a Compiler bug. I will Keep you informed.
Regards,
Leo

0 leopold rohrecker over 8 years ago in reply to leopold rohrecker

Prodigy 230 points

Hi Vishal,

Access to the PWM Registers from CLA is working when I use the commands __meallow() / __medis() with protected registers. It would be nice if the C Compiler generated a warning or an error message after finding commands like EALLOW or EDIS in a .cla file.

Thank you for your help!

Regards,
Leo

C2000™︎ microcontrollers

C2000 microcontrollers forum

CCS/TMS320F28035: PWM SW forced CBC trip from CLA