Reading Timer A value without stopping it

Tony Payn said:

I'm running MSP430G2553 TimerA from a 32768 Hz watch crystal.  When an event happens on one of the micro pins, I want to read the value of TimerA.

You would want to use timer capture mode for this application.

Hello Ilmars

Thanks for your suggestion.  I had thought about this but my pin which triggers the timer read is not able to be routed to the capture input.

Would you advocate triggering a capture in software in response to the pin activity?  I have an interrupt routine responding to changes on the pin.  Should I get this interrupt routine to implement a software capture? I'd want to do this without going into a capture interrupt routine.  Could I just, in my pin-change interrupt routine, just set a software capture and then read the value of the capture register immediately?  If I could be sure the capture register contained the value immediately, then that would be fine.

Thanks

Tony

Tony Payn said:

but my pin which triggers the timer read is not able to be routed to the capture input.

Can't be so. Out of wires or solder or what? :)

Tony Payn said:

Would you advocate triggering a capture in software in response to the pin activity?

Sure you can initiate capture by software but if you care about precision then hardware capture is way you shall really consider

Tony Payn said:

I was thinking to read it twice (reads 1 and 2) and if the values are the same, then all well and good - use that value.  But if the values are different, then do another read (no 3) and compare this with no 2.

Since you have a good difference in clock rates between CPU and timer, just make it easy and always read it three times consecutively and do the majority vote. Less logic to decide when to read twice vs three times.

I don't think you can ever see the case where 1 & 3 agree but 2 doesn't. So it is 1, 2, & 3 match, 1 & 2 match, or 2 & 3 match.

psuedo-code:

x1 = TAR;
x2 = TAR;
x3 = TAR;

if (x1 == x2)
  value = x1;
else if (x2 == x3)
  value = x3;

In a way, I suppose I am.  The board is already laid out and the pin signal is P2.2.  This can only be used as a capture input on TA1.1, but I am using that capture/compare unit already for another unrelated purpose.  Only TA1.2 is free, but P2.2 isn't able to be assigned to TA1.2.

That's why I was looking for a software method to read the timer value.

I could probably swap the two capture/compare units since TA1.1 is only used for internal timing purposes.  And using the capture/compare unit to do the capture rather than doing it in software would avoid the delay overhead of switching into the pin interrupt routine before doing my software reads of the timer value, improving accuracy as your post states.  I was curious about the race condition and the correct method to read the timer though, if it had to be done this way.

If TA1.2 is not being used but the signal you want to capture is at P2.2, you could use software to make TA1.2 to capture TA1R in TA1CCR2.

Tony Payn said:

my pin which triggers the timer read is not able to be routed to the capture input.

You can do it in software by switching the capture input between VCC and GND:

TA0CCTL1 = MC_1|CCIS_2|CAP; Capture on rising edge, capture input is GND
[...]
TA0CCTL1 = TA0CCTL1 |= CCIS1; // switch capture input to VCC, causes trigger
TA0CCTL1 = TA0CCTL1 &= ~(CCIS1); // switch back to GND
x = TA0CCR1; // read captured timer value.

The majority vote is a good idea too, but it can be simplified:

while(!(x==TA0R)) x=TA0R;

This will read TA0R until two consecutive identical reads. Works only if MCLK much faster than TA0CLK. But this is known at design time.

Jens-Michael Gross said:

while(!(x==TA0R)) x=TA0R;

That has a subtle bug in it, since the goal is to make sure we don't get bits in the middle of changing due to the clock rate mismatch.

That code will make the comparison x==TA0R first... but what is the initial value of x?
 In other words, what if the uninitialized value (or initialized for that matter) of x is equal to TA0R when it is in a transition of bits (the condition we are trying to avoid)?

Answer is that the while loop exits without reading TA0R into x and you have a possible erroneous reading.

Better answer is

x=TA0R;
while(!(x==TA0R)) x=TA0R;

If I use c, I would skin the cat with::
do x=TA0R; while (x!=TA0R);

old_cow_yellow said:

If I use c, I would skin the cat with::
do x=TA0R; while (x!=TA0R);

Well, that's another way. I guess it depends on whether you've had your morning coffee (and woken all the brain cells up yet) or not.

:-)

Assembler gurus could use the IAR macro I set up.
Pause_Routine                               ; example
            tarcopy &TA1CCR2
            add.w   #TA1second/20,&TA1CCR2  ; add the pause before mainloop is run
            ret

;---------- M A C R O S ---------
tarcopy     macro   address
            local gettarmatch
gettarmatch mov.w   &TA1R,R15           ; trash register
            cmp.w   R15,&TA1R
            jne     gettarmatch
            mov.w   R15,address
            endm

Many thanks to all for your suggestions, I am clear on the best way to do this now.

Being new to using the 'Verify Answer' buttons etc, I've just verified the last one in the thread but I don't mean to imply that other suggestions were not equally useful to me, I've learned something from all of them.

Thanks again

Tony

You can verify multiple posts, and as well un-verify them, if appropriate.