I'm working on a project where I only need the MSP430 (probably a 'F423) to wake up once a day. It looks like the longest I can sleep though is 30s. (TimerA w/ 32k ACLK / 8, updown mode, count to 61440 then back to 0 and interrupt on CCIFG) Any suggestions on how to get a longer delay? Also how can I simulate this? I've set up IAR/TI C-SPY to generate a TimerA interrupt, but when the ISR reads TAIV to determine the interrupt source, it always comes back 0. Regards, Joe
long delays and how to simulate them
Started by ●April 7, 2005
Reply by ●April 9, 20052005-04-09
You don't read TAIV per se, you use it. TAIV is cleared on the first
read. The best way to use it is:_
TA12_ISR:
ADD &TAIV,PC
RETI
JMP TA1_ISR
JMP TA2_ISR
RETI
RETI
JMP TA_OVF_ISR ;You should be using the overflow
On the subject of long delays, these are typically executed as a lot mof
short delays. The MSP430 is particularly good for this becuase of the
DCO and very rapid wake-up (6usecs quoted, but typically about
500nsecs). In your ISR you maintain a second, or even third long
counter. using simply continuous count and rollover interruipt you get
an interrupt every 16 seconds, so a single counter would itself roll
over at 1048576 seconds, or 291 hours, so for a 24 hour interrupt you
need a
TA_OVF_ISR::
BIC #TAIFG, &TACTL
INC &COUNTER ;16 SECOND INTERVAL
COUNTER
CMP #5400,&COUNTER ;24 HOUR TEST
JLO SNOOZE ;IF NOT 24 HOURS
BACK TO SLEEP
BIC #LPM3,0(SP) ;EXIT SLEEP ON RETI,
WHEN OLD SR IS POPPED
SNOOZE:
RETI
This is the entire thing, except for entering sleep mode. I would have
the lowest possible DCO frequency on entry to sleep.
Al
jgb_gt_84 wrote:
>
>I'm working on a project where I only need the MSP430 (probably a
>'F423) to wake up once a day. It looks like the longest I can sleep
>though is 30s. (TimerA w/ 32k ACLK / 8, updown mode, count to 61440
>then back to 0 and interrupt on CCIFG) Any suggestions on how to get
>a longer delay?
>
>Also how can I simulate this? I've set up IAR/TI C-SPY to generate a
>TimerA interrupt, but when the ISR reads TAIV to determine the
>interrupt source, it always comes back 0.
>Regards,
>Joe
>
>
>
>
>
>
>
>
>
>.
>
>
>Yahoo! Groups Links
>
>
>
>
>
>
>
>
>
>
>
Reply by ●April 9, 20052005-04-09
----- Original Message -----
From: "jgb_gt_84" <jgb_gt_84@jgb_...>
To: <msp430@msp4...>
Sent: Thursday, April 07, 2005 2:20 PM
Subject: [msp430] long delays and how to simulate them
>
>
>
> I'm working on a project where I only need the MSP430 (probably a
> 'F423) to wake up once a day. It looks like the longest I can sleep
> though is 30s. (TimerA w/ 32k ACLK / 8, updown mode, count to 61440
> then back to 0 and interrupt on CCIFG) Any suggestions on how to get
> a longer delay?
Just set it to interrupt, say, every second, and use software counters to
get your 24 hour delay.
Leon
Reply by ●April 9, 20052005-04-09
I think he wants to conserve power and only wake up every 24 hours....
At 12:55 AM 4/9/2005, you wrote:
>Just set it to interrupt, say, every second, and
use software counters to
>get your 24 hour delay.
>
>Leon
// richard (This email is for mailing lists. To reach me directly, please
use richard at imagecraft.com)
Reply by ●April 9, 20052005-04-09
Not possible in any conventional MSP430 configuration I'm aware of. Of
course he could externally divide a 32.768kHz clock, or find a very low
frequency stable oscillator, then using the divide by 8 internally he
would need the input frequency to be 6.068Hz. Or set the input frequency
to 6Hz and the overflow counter to 64800. Then, after dividing by 8 the
clock frequency is 0.75hz, so 1/f = 1.333333333 * 64800 = 86400seconds
or 24 hours.
Dividing the 32768 clock would probably be easier than trying to get a
crystal that would neatly divide down to 6Hz. So divide the 32768 clock
by 8192 externally to get 4Hz, divide this by 8 to get 0.5hz, or a 2
second tick, the 43200 as a compare gives 24 hours. Of course you have
probanly consumed more power in the external divider than would be
consumed aking up for a few clock cycles, then going back to sleep
again. using registers for the COUNTER, and to hold the COMPARE value,
my orignal solution requires.
R4 is a counter to count Timera rollovers
R5 is the literal value that correponds to 24 hours
R6 is the bit pattern used to enter the currtent sleep mode
TA_OVF_ISR:
BIC #TAIFG, &TACTL ;5clocks
INC R4 ;16 SECOND
INTERVAL COUNTER 1Clk
CMP R5,R4 ;24 HOUR TEST
1clock
JLO SNOOZE ;IF NOT 24 HOURS BACK
TO SLEEP 2 clocks
BIC R6,0(SP) ;EXIT SLEEP ON
RETI, WHEN OLD SR IS POPPED 4 clocks
SNOOZE:
RETI ;5 clocks
18 clock cycles plus wake up time, executed every 8 seconds so in every
hour this sequence is executed 450 times. assuming DCOCLK = 500kHz then
the above routine, including the vector jmp table requires 23 clock
cycles, or 46usecs, plus 6usec start up. procedure requires a total of
52usecs every 8 seconds. LPM3 consumes around 2uA. run mode. Run mode at
500kHz is nominally 210uA, so the mean consumption is (7,999,948* 2 + 52
* 210)/8,000,000 or 2.001352uA. I haven't yet found a divider that draws
less than 1.35nA, or a stable external oscillatro that draws less
current either.
Al
Richard wrote:
>I think he wants to conserve power and only wake up
every 24 hours....
>
>At 12:55 AM 4/9/2005, you wrote:
>
>
>
>>Just set it to interrupt, say, every second, and use software counters
to
>>get your 24 hour delay.
>>
>>Leon
>>
>>
>
>// richard (This email is for mailing lists. To reach me directly, please
>use richard at imagecraft.com)
>
>
>
>.
>
>
>Yahoo! Groups Links
>
>
>
>
>
>
>
>
>
>
>
Reply by ●April 11, 20052005-04-11
Everyone, thanks for the replies. You've confirmed what I had already expected. I thought someone might have had a clever idea I had not thought of. The TAIV issue is apparently a deficiency of the KickStart interrupt simulation. regards, Joe --- In msp430@msp4..., Onestone <onestone@b...> wrote: > Not possible in any conventional MSP430 configuration I'm aware of. Of > course he could externally divide a 32.768kHz clock, or find a very low > frequency stable oscillator, then using the divide by 8 internally he > would need the input frequency to be 6.068Hz. Or set the input frequency > to 6Hz and the overflow counter to 64800. Then, after dividing by 8 the > clock frequency is 0.75hz, so 1/f = 1.333333333 * 64800 = 86400seconds > or 24 hours. > > Dividing the 32768 clock would probably be easier than trying to get a > crystal that would neatly divide down to 6Hz. So divide the 32768 clock > by 8192 externally to get 4Hz, divide this by 8 to get 0.5hz, or a 2 > second tick, the 43200 as a compare gives 24 hours. Of course you have > probanly consumed more power in the external divider than would be > consumed aking up for a few clock cycles, then going back to sleep > again. using registers for the COUNTER, and to hold the COMPARE value, > my orignal solution requires. > > R4 is a counter to count Timera rollovers > R5 is the literal value that correponds to 24 hours > R6 is the bit pattern used to enter the currtent sleep mode > > TA_OVF_ISR: > BIC #TAIFG, &TACTL ;5clocks > INC R4 ;16 SECOND > INTERVAL COUNTER 1Clk > CMP R5,R4 ;24 HOUR TEST > 1clock > JLO SNOOZE ;IF NOT 24 HOURS BACK > TO SLEEP 2 clocks > BIC R6,0(SP) ;EXIT SLEEP ON > RETI, WHEN OLD SR IS POPPED 4 clocks > SNOOZE: > RETI ;5 clocks > > 18 clock cycles plus wake up time, executed every 8 seconds so in every > hour this sequence is executed 450 times. assuming DCOCLK = 500kHz then > the above routine, including the vector jmp table requires 23 clock > cycles, or 46usecs, plus 6usec start up. procedure requires a total of > 52usecs every 8 seconds. LPM3 consumes around 2uA. run mode. Run mode at > 500kHz is nominally 210uA, so the mean consumption is (7,999,948* 2 + 52 > * 210)/8,000,000 or 2.001352uA. I haven't yet found a divider that draws > less than 1.35nA, or a stable external oscillatro that draws less > current either. > > Al > > > > Richard wrote: > > >I think he wants to conserve power and only wake up every 24 hours.... > > > >At 12:55 AM 4/9/2005, you wrote: > > > > > > > >>Just set it to interrupt, say, every second, and use software counters to > >>get your 24 hour delay. > >> > >>Leon > >> > >> > > > >// richard (This email is for mailing lists. To reach me directly, please > >use richard at imagecraft.com) > > > > > > > >. > > > > > >Yahoo! Groups Links > > > > > > > > > > > > > > > > > > > > > >
