Timer not generating interrupt at desired time...it is varying

It could be interrupt latency variation. Try 8MHz for MCLK and SMCLK, feed timer with SMCLK/8 and see if it helps.

If you want perfect timing then better consider timer PWM output mode.

Sir,

How will I select 8Mhz since in the header file its not mentioned.

The header file has the following information only:(only for 1Mhz)

#ifndef __DisableCalData

SFR_8BIT(CALDCO_1MHZ); /* DCOCTL Calibration Data for 1MHz */
SFR_8BIT(CALBC1_1MHZ); /* BCSCTL1 Calibration Data for 1MHz */

#endif /* #ifndef __DisableCalData */

Indranil said:

How will I select 8Mhz since in the header file its not mentioned.

Oh, you are not using g2553. Then try to run timer at 1MHz/8 clock and then check variation. Supposedly it shall be much smaller. As I said - better consider PWM. Even embedded software is not that deterministic if we talk about timing - especially when you try to count every CPU cycle.

Ok I will check it using PWM output mode..and let you know the results.

One thing I want to know -can I connect external clock of 16Mhz for Msp430g2231 device? 

Thanks.

You could have used one of the Timer output pins to generate a more precise clock. Unfortunately P1.4 is not one of them. Could you use P1.1, P1.2, P1.5, P1.6, or P2.6 instead?

I have tried using other pins, but it shows same behaviour.

I have done several testing to create time interrupt in the range of 18 microseconds to 5 microseconds by

reducing the CCR0 value. But what I have noticed is it is unable to create timer interrupt for small period. Minimum

is more than 20us. Why is that ? Is it because of interrupt latency ? My application needs to generate pulses of

26microseconds duration. Do I need to use high clock source (more than 1 Mhz). I am trying to use external

crystal oscillator clock(16Mhz) to get more accurate timing. But the problem is I am getting oscialltion fault while initializing the clock. Can anybody provide me some help to guide how to initialize 16Mhz external clock for Msp430g2231 device?.  

Even I have tried to use PWM module to generate a duration 26us duration.(it can be any duty cycle). The output what I have noticed is that pulses are varying. Some pulse will come with 26us duration but some pulse will not . If I try the same for 1milliseconds it is somewhat better. But my application requires 26us. I am out of ideas why my device(MSP430g2231) is not supporting less delays. Please help. 

#include "msp430g2231.h"

void main(void) {
WDTCTL = WDTPW + WDTHOLD; // Stop watchdog timer

if (CALBC1_1MHZ == 0xFF || CALDCO_1MHZ == 0xFF)
{
while(1); // If calibration constants erased
// do not load, trap CPU!!
}

DCOCTL = (CALDCO_1MHZ + 0x20) & 0xE0;
BCSCTL1 = CALBC1_1MHZ;

//make Port pin P1.1 as Timer output
P1SEL = BIT1;
P1DIR = BIT1;

TACCR0 = 14-1;
TACCTL0 = OUTMOD_4;
TACTL = TASSEL_2 | MC_1;

while(1);
}

Thanks for the reply and the sample code..:)

What have you done in the following line ? I didn't understand

DCOCTL = (CALDCO_1MHZ + 0x20) & 0xE0;

38kHz*26 = 988kHz; 38kHz*28 = 1064kHz; I took a blind shoot aimed at making the DCO to oscillate closer to 1.064MHz and without modulation. Was I close?

You can actually adjust the setting of DCOCTL to make the frquency at P1.1 even closer to 38kHz.

Jens-Michael Gross said:

I am trying to generate a clock cycle pulses of period "26micro seconds"(1/38khz). [..] in oscilloscope the cycle period varies in between 32, 34 and 38 micro seconds.

If the MSP runs on 1MHz, this means 1 MCLK cycle is 1µs.

Since you're using CCR0 for the interrupt, you don't need to clear CCIE in CCTL0. It is automatically done when the ISR is entered (only for CCR0!). [/quote]

Do you mean CCIFG?

Even though there are no calibration constants for 4, 8, 16 MHz for the msp430g2231, would there be any reason he couldn't run it at 16 MHz and determine the calibration constants himself?

Thanks Michael for clearing my doubt. :) Moreover you said that we can calibrate the controller for high clock(16Mhz).

If such is the case I will try with that, So is this the only solution/workaround for my problem?

I think I understand what the modulation is doing (during every 32 DCO clocks some are fdco and some are fdco+1). But all you are doing is swapping out 1 cycle of slightly larger period and replacing it with 1 cycle of slightly smaller period, succeeding in changing only the phase of 1 cycle, and still ending up with 32 cycles that take the same time. If modx = 2, for example, it's still 1 cycle fdco exchanged for one cycle fdco+1 twice each 32 cycles. Unless of course if fdco+1 is twice the frequency of fdco, then you would be exchanging two cycles of fdco+1 for each fdco cycle. But that doesn't appear to be the case.

Joseph Raslavsky said:

I think I understand what the modulation is doing (during every 32 DCO clocks some are fdco and some are fdco+1). ...

Correct.

Joseph Raslavsky said:

... But all you are doing is swapping out 1 cycle of slightly larger period and replacing it with 1 cycle of slightly smaller period, succeeding in changing only the phase of 1 cycle, and still ending up with 32 cycles that take the same time. ...

Do you mean OCY did that? OCY did not do that at all. What he did with the setting of DCOCTL is not important at all. The important thing is not using interrupt to toggle the port pin and use a Timer Output pin instead.

old_cow_yellow said:

Do you mean OCY did that? OCY did not do that at all. What he did with the setting of DCOCTL is not important at all. The important thing is not using interrupt to toggle the port pin and use a Timer Output pin instead.

No. Not at all. I'm trying to understand how this modulation works (obviously it does), but I don't see how.

Joseph Raslavsky said:

Do you mean OCY did that? OCY did not do that at all. What he did with the setting of DCOCTL is not important at all. The important thing is not using interrupt to toggle the port pin and use a Timer Output pin instead.