I am using the C0851F500 and want to measure time with the PCA.
The PCA operation is set to 'capture on positive edge' mode.

Does any of you know what will happen to a already captured PCA value if all of the following conditions are true:
- the corresponding interrupt has not been serviced yet and therefore CCFn is still set
- another positive edge occurs on CEXn

Will the value in PCA0CPn be preserved or overwritten with the new value?

excerpt from datasheet:
27.3.1. Edge-triggered Capture Mode
In this mode, a valid transition on the CEXn pin causes the PCA to capture the value of the PCA counter/timer and load it into the corresponding module's 16-bit capture/compare register (PCA0CPLn and PCA0CPHn).
...
When a capture occurs, the Capture/Compare Flag (CCFn) in PCA0CN is set to logic 1.
...

From this description I'd say that the value is always captured and therefore overwrites the value in the PCA register.
This question could be also answered with "there is no hardware buffering implemented".

But, from your question I assume that you're experiencing some data loss with your current software. Or is there another reason for your question? If so, post your code and describe the hardware circuit and the input signals.

Generally, a ISR should be kept as short as possible. In best cases a ISR saves the data to a buffer and informs the main application about the new data. This ensures that each interrupt can be serviced.

Regards,

The reason for my question is that i DON'T wont each of the PCA CCFn interrupts to be serviced.

I try to explain why.

The timestamps i want to measure are used by a subroutine to control a dc motor speed. This subroutine is called from within Timer2_ISR which is serviced all 10ms. I don't need the timestamps to be measured all the time but rather two times between each Timer2_ISR.
I programmed the PCA0_ISR to mask its interrupt flag once it captured two timestamps. At the end of Timer2_ISR this flag is unmasked again.

Now we assume that a capture event occurs at the execution time of Timer2_ISR. Even though the CCFn interrupt is masked a value will be captured in PCA0CPn.
If this value would be overwritten at each capture event the motor speed should be calculated correctly.
But when i debugged the code i found some numerical distortions of the speed value appearing periodically. So i assume that the value will NOT be overwritten unless the CCFn flag is cleared by software.

I solved the problem by clearing the CCFn flag at the end of Timer2_ISR. No more distortions so far.

Nevertheless, i'm not sure a about this. I hoped that someone would know a document which is clear about that point.

Regards,
Jacob

[This message has been edited by jacobk (edited August 05, 2010).]

i don't see your point. Why should i use a counter in the Timer2_ISR? What should i count there except how often this ISR has been serviced?

Besides, my code is working now as i expect it to work. So the answer to my original question is no longer essential to complete my task.

But i don't like it not to be really sure about such fundamental questions.

Just in case, in my timezone we have 8:50pm right now. So don't wonder about my delays of posting.

Regards,
Jacob

I don't see your point. Why should i use a counter in the Timer2_ISR? What should i count there except how often this ISR has been serviced?
because
I don't need the timestamps to be measured all the time but rather two times between each Timer2_ISR.

if (count)
{
count--
}
else
{
timestamp
count = 2
}

from your code snippet i would say that in this way a "timestamp" would be captured every third Timer2 interrupt.

But with the term "timestamp" i was referring to a PCA counter value captured at a time when a positive edge is sensed on a CEXn pin. This signal comes from the dc-motor quadrature encoder and is asynchronous to the Timer2_ISR intervals.

I already solved that problem like i have described in my second post:
I programmed the PCA0_ISR to mask its interrupt flag once it captured two timestamps. At the end of Timer2_ISR this flag is unmasked again.
The number of captured PCA0 values (timestamps) are counted inside the PCA0_ISR. When the PCA0_ISR captured its second value it deactivates itself. The corresponding interrupt enable bit can be used as a semaphore inside the Timer2_ISR to test if two values were captured. If so, the difference of the two values can be calculated. This difference represents the speed of the motor. At the end of Timer2_ISR the PCA0_ISR is activated again.

Here some pseudo code:

main {
while(1) {
SLEEP_AND_WAKE_ON_INTERRUPT;
}
}

Timer2_ISR{
if (PCA0 interrupt is off) {
calc speed;
}
edgecount = 0;
activate PCA0 interrupt;
clear capture flag;
}

PCA0_ISR{
if (edgecount == 0) {
save first captured value;
}
if (edgecount == 1) {
save second captured value;
deactivate PCA0 interrupt;
}
edgecount++;
}

Interrupt execution time plan:

Timer2_ISR|.|-----|...............|-----|
..........|..............................
PCA0_ISR..|........|--|..|--|............
..........|------------------------------> t

[This message has been edited by jacobk (edited August 06, 2010).]

[This message has been edited by jacobk (edited August 06, 2010).]

Erik