Forums

Frank Boehme

But it seems that there is a problem which causes the TIMER_A1 to stutter.

Apparently. And since the two work independently, the mos tlikely cause is the way you implemented your ISRs or the control of the PWM.

You should post your code, as there is no known 'TimerA1 PWM fails if TimerA0 is used for TouchSense" hardware bug.

However, I have a guess. Maybe you confused the two timers with the two vectors for each timer.

TIMER0_A0_VECTOR and TIMER0_A1_VECTOR (or their aliases TIMER_Ax_VECTOR) belong both to Timer0.
Timer1 has the vectors TIMER1_A0_VECTOR and TIMER1_A1_VECTOR.

You really should post your code.

Hello,

This is the Code. I stripped it down to the essentials. It is a standard CCS 5 project for the MSP430G2553. All of "my" Code is in main.c. I use the standard files

CTS_HAL.c
CTS_HAL.h
CTS_Layer.c
CTS_Layer.h

from slac489.zip and the mode "RO_PINOSC_TA0_WDTp" with the files structure.c and structure.h from "RO_PINOSC_TA0_WDTp_One_Button" example without modifications.

Comment the line: FeedbackLED(one_button, BIT0, 0); to run only the PWM and it works fine

Comment the line: Timer1_A3_init(); to run only the touch and it works fine

Use both lines and the PWM LED stutters

//******************************************************************************
//          Schematic Description: 
// 
//                         MSP430G2452
//                      +---------------+
//                      |
//           OUT--------|P1.0 (Feedback LED)
//         PWM----------|P2.1 (LED)
//               C------|P2.5 (Touch)
// 
//******************************************************************************
#include "CTS_Layer.h"

// Uncomment to have this compiler directive run characterization functions only
// Comment to have this compiler directive run example application
//#define ELEMENT_CHARACTERIZATION_MODE

#ifdef ELEMENT_CHARACTERIZATION_MODE
// Delta Counts returned from the API function for the sensor during characterization
unsigned int dCnt;
#endif

int up = 1;
int disable = 0;

/*
 *  ======== Timer1_A3_init ========
 *  Initialize MSP430 Timer1_A3 timer
 *  Conf. made with GRACE, Comments removed
 */
void Timer1_A3_init(void)
{
    TA1CCTL0 = CM_0 + CCIS_0 + OUTMOD_0; // + CCIE;
    TA1CCTL1 = CM_0 + CCIS_0 + OUTMOD_7;// + CCIE;
    TA1CCR0 = 9999;
    TA1CCR1 = 500;
    TA1CTL = TASSEL_2 + ID_0 + MC_1;
}

//Feedback LED anschalten
void FeedbackLED(const struct Sensor Sens, int Pin, int disable)
{
	if(disable == 0)
	{
		if(TI_CAPT_Button(&Sens))
			P1OUT |= Pin;                            // Turn on  LED
		else
			P1OUT &= ~Pin;
	}
	else
		P1OUT &= ~Pin;
}

// Main Function
void main(void)
{ 

  // Initialize System Clocks
  WDTCTL = WDTPW + WDTHOLD;             // Stop watchdog timer
  BCSCTL1 = CALBC1_1MHZ;                // Set DCO to 1, 8, 12 or 16MHz
  DCOCTL = CALDCO_1MHZ;
  BCSCTL3 |= LFXT1S_2;                  // LFXT1 = VLO
  
  P1OUT = 0x00;							// Clear Port 1 bits
  P1SEL = BIT6;
  P1DIR |= BIT0 + BIT3 + BIT4 + BIT5 + BIT6 + BIT7;						// Set output pins

  P2SEL &= ~(BIT6 + BIT7);				// Configure XIN (P2.6) and XOUT (P2.7) to GPIO
  P2SEL |= BIT1;						// P2.1 to PWM
  P2OUT = 0x00;							// Drive all Port 2 pins low
  P2DIR = 0xFF;							// Configure all Port 2 pins outputs

  Timer1_A3_init();

  // Initialize Baseline measurement
  TI_CAPT_Init_Baseline(&one_button);
  
  // Update baseline measurement (Average 5 measurements)
  TI_CAPT_Update_Baseline(&one_button,5);
  
  // Main loop starts here
  while (1)
  {
		#ifdef ELEMENT_CHARACTERIZATION_MODE
	  // Get the raw delta counts for element characterization
	  TI_CAPT_Custom(&middle_button,&dCnt);
	  __no_operation(); 					// Set breakpoint here
		#endif

	#ifndef ELEMENT_CHARACTERIZATION_MODE

	  FeedbackLED(one_button, BIT0, 0);

	#endif
  }
} // End Main

You may also have a look on the complete project if you wish 6136.oneButtonPWMTest.zip

Frank

BTW: When I change the ".measGateSource" in structure.c to SMCLK, it looks much better, but the sensitivity decreases dramatically and I am not really happy with that workaround.

I think I have it!

In the original CTS_HAL.c is the code

        if(group->measGateSource == GATE_WDT_ACLK)
        {
            __bis_SR_register(LPM3_bits+GIE);   // Wait for WDT interrupt
        }
        else
        {
	    __bis_SR_register(LPM0_bits+GIE);   // Wait for WDT interrupt
        }

which causes the MCU going in Low Power Mode 3 when using the ACLK which disables the SMCLK and therefore the TIMER_A1 in my code. 

I changed LPM3 to LPM0 and it works.

Can anyone confirm this?

Frank

It's well possible that the funciton TI_CAPT_Custom and TI_CAPT_Button or something they do (especially the 'Custom' looks suspicious) causes a compatibility problem.

Your TIemr setup doesn't use itnerrupts and runs in plain hardware. But it is possible that the CAPT code somehow requires the timer by itself.

IIRC, the touch pins are oscillating and the oscillations are counted by Timer0. However, to make any sense, you'll need to know how many oscillations per time period happen. And since TimerA0 is used for counting pulses, only TimerA1 is available for measuring the period. So both tiemrs are implicitely used for CapSense and your use of TimerA1 conflicts with this.

I don't know the actual TI library code, but I bet it is a good guess.

Frank Boehme

The problem was that the MCU was sent to LPM3 at each touch cycle which disabled my PWM.

Well, as long as Timer1 is using SMCLK, LPM3 shouldn't disable it. But this is on 5x family. I'm not sure for 2x family, I never used LPMs there.

Frank Boehme

TI code, I just used them, hoping that they know what they do

The problem with TI code is not that they don't know what they do, but that you don't know what they do. Lack of detailed documentation. So I always suspect 'foreign' code in a project.
This is not a TI specific problem, I also had it with 3rd party libraries on Flex (Flash), Dojo (JavScript) and elsewhere. But indeed, in some cases it turned out that they didn't know what they were doing too, while after debugging I did know. :)