Posted on December 21, 2010 at 22:07Hi Lüfter,
thanks.
Another interesting fact I noticed in the example provided by ST.
If you set the TIM2 INPUT CAPTURE PRESCALER to divide by 4, for example,
TIM2_ICInit(TIM2_CHANNEL_1, TIM2_ICPOLARITY_FALLING, TIM2_ICSELECTION_DIRECTTI, TIM2_ICPSC_DIV4, 0x00);
or less than 8 anyway, it means that you should capture every 4 input events. In this way you should measure the clock frequency of the LSI nearly a HALF than when you set the prescaler IC division by 8 (i.e. 128kHz / 2 = 64 KHz).
Well, it doesn't ! The result you get is always the same 128KHz or so.
The same result when you set the prescaler division by less than 4 down to 1.
WHY?
To get into this apparent anomaly I modified the example code a little bit by toggling and hardware output port (i.e. PE.3) when the capture occured.
while (1)
{
/* wait a capture on CC1 */
while((TIM2->SR1 & TIM2_FLAG_CC1) != TIM2_FLAG_CC1);
/* Get CCR1 value*/
ICValue1 = TIM2_GetCapture1();
TIM2_ClearFlag(TIM2_FLAG_CC1);
/* debug */
GPIO_WriteHigh(GPIOE, GPIO_PIN_3);
/* wait a capture on cc1 */
while((TIM2->SR1 & TIM2_FLAG_CC1) != TIM2_FLAG_CC1);
/* Get CCR1 value*/
ICValue2 = TIM2_GetCapture1();
TIM2_ClearFlag(TIM2_FLAG_CC1);
/* debug */
GPIO_WriteLow(GPIOE, GPIO_PIN_3);
/* Compute LSI clock frequency */
// LSIClockFreq = (8*TIM2ClockFreq) / (ICValue2 - ICValue1);
/* Insert a break point here */
// while (1);
}
So, the capture events were measured with the oscilloscope, an old fashion ''brute force'' method but very straightforward and effective :)
Well, actually the capture frequency of the events didn't changed when the IC prescaler was less then 8 thus confirming that the code captured correctly the LSI only when the IC PRESCALER was 8 and not less than that.
The only way to make things working was to rise up the CPU clock speed by adding the instruction
void main(void)
{
CLK_HSIPrescalerConfig(CLK_PRESCALER_HSIDIV1);
...
at the beginning of the code.
The explanation was, therefore, that capturing the counter by using twice
ICValue1 = TIM2_GetCapture1();
and so on all the entire code made in this way is really time demanding for the cpu, leading to it is not capable to capture correctly the counter when the CPU clock is too slow. That's why ST developer recommended to ''capture only every 8 events !!! ''.
Stefano
