Interrupt On Change

[Jan Lichtenbelt]

I have a pulse of about 1-2 msec with 50Hz (period time 20 msec), see picture 1.

P1040898_t.jpg (202.81 KiB) Viewed 4610 times

To measure the period time I use the interrupt-on-change (on port B3 of 16F1939). The interrupt is set only to rising edge. The timer used is Timer2 set on 19200 Hz (52 usec). If I use the pulse as given in picture 2 I measure a period time of about 288, which is 288 x 52 = 15 msec (correct see picture 1).

P1040896_t.jpg (204.62 KiB) Viewed 4610 times

However if I change the pulse to the form as given in picture 3, I measure about 22-40, which is equal to 1-2 msec. And indeed the value change with the pulse length (repeating period stays unchanged). That means pulses like figure 3, not only the rising edge is detected, but also the falling edge. How is that possible????

P1040897_t.jpg (204.2 KiB) Viewed 4610 times

Remarks:
1) The high level of 3 Volt of the pulse is within the TTL definition of "high" 2.2-5V
2) Interrupt on INT pin (B0) gives correct period times with INT interrupt instead of interrupt-on-change

Kind regards

Jan Lichtenbelt

[Enamul]

Hi

I think the problem is bcz of the handler code orientation...

Code: Select all

//Handler code for [PORTB]
    #ifndef MX_INTHANDLER_intcon_IOCIF
    #define MX_INTHANDLER_intcon_IOCIF
    char mxtmp;
    if (ts_bit(intcon, IOCIF) && ts_bit(intcon, IOCIE))
    {
        #ifdef USE_FLOWCODE_ICD
            extern char ICD_Interrupt_Enable = 1;
        #endif    
        FCM_Get_Time();
        mxtmp=portb;
        cr_bit(intcon, IOCIF);
        iocbf=0;
        #ifdef USE_FLOWCODE_ICD
            extern char ICD_Interrupt_Enable = 0;
        #endif    
    }

it should be...

Code: Select all

//Handler code for [PORTB]
    #ifndef MX_INTHANDLER_intcon_IOCIF
    #define MX_INTHANDLER_intcon_IOCIF
    char mxtmp;
    if (ts_bit(intcon, IOCIF) && ts_bit(intcon, IOCIE) && ts_bit(iocbf, IOCBF3))
    {
        #ifdef USE_FLOWCODE_ICD
            extern char ICD_Interrupt_Enable = 1;
        #endif    
        cr_bit(intcon, IOCIF);
        cr_bit(iocbf, IOCBF3);
        FCM_Get_Time();
        mxtmp=portb;
        
        #ifdef USE_FLOWCODE_ICD
            extern char ICD_Interrupt_Enable = 0;
        #endif    
    }

In ISR, it should clear IF flag first then should do rest of the things...

Enamul

[Jan Lichtenbelt]

Dear Enamul

Thanks a lot for helping me. It took me a while to get your interrupt handler in my program. But I have to disappoint you, the results did not change. That means still interrupts on the falling edges, while only on the rising edges is wanted.

Kind regards

Jan Lichtenbelt

[Benj]

Hi Jan,

I wonder if there is noise or something on the signal that is causing the IOC to trigger on a falling edge. i.e. a positive going spike during the falling edge. Your scope trace looks quite clean so I am surprised. Any chance you can add a cap or something to slow the signal voltage transition speed down a bit?

The original interrupt handler is recommended as this shouldn't have anything wrong with it. Enamul's interrupt handler has the possibility of an unwanted interrupt but this possibility is very remote.

[Benj]

The C code to enable the interrupt looks to be correct.

Code: Select all

	st_bit(intcon,GIE);
	iocbp=0x8;
	iocbn=0x0;
	st_bit(intcon, IOCIE);

I wonder if there is another setting I am missing to force the rising / falling edge masks to be used. I'm guessing that no other external I/O is causing the interrupt to trigger so at least one of the masks must be working.

[Jan Lichtenbelt]

Hi Ben,

Only one pin is used as input pin for these measurements. I tried an other source for producing the pulses. But even with pulses like figure one, I get different readings, and sometimes the correct one (Vtop is 5 Volt in this case). It seems to me unstable. But what?

Jan

[Jan Lichtenbelt]

Hi All,

That was a mistake made by myself. In the flowcode reseting the counter after the LCD actions was wrong. After removing this line the program works correct.


Kind regards

Jan Lichtenbelt