For time counts driven by a crystal the types of errors you are talking about should result in time variations of seconds per day. What I am seeing is seconds per minute. A repeating counter set to 1000 milliseconds should count seconds with some jitter. I don't expect the events to be exactly on the second, I do expect the number of events per hour to be correct. Basically the timers are driven off of the watch crystal which counts the ATmega128 counters. I expect to schedule an event every time the ATmega128 counter passes the appropriate number of ticks. I do not expect the events to get lost, nor do I expect extra events. I expect the events to be near in time to the tick that triggered them but not exact in time, because there may be other tasks that compete for the exact time. In any case counting fast indicates a flaw, competing for resources and failing to get them would result in slow counts.
_____ From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Cory Sharp Sent: Wednesday, January 03, 2007 6:01 PM To: Michael Newman Cc: tinyos-help@millennium.berkeley.edu Subject: Re: [Tinyos-help] Time counting Of course the device capable for some definition of capable. But the keyword isn't capability but calibration: ensuring the mote records accurate time across varying conditions and manufacture. Keep in mind that "varying conditions" even includes things like turning on a high power sensor causing a drop in system voltage. A "causal relation" to wall clock time stands to reason, though perhaps you meant "casual relation". Beyond that, I'm not sure of what precisely both your observations and expectations are. Cory On 1/3/07, Michael Newman <[EMAIL PROTECTED]> wrote: I actually discovered this problem when I started to implement a time synchronization protocol. I am trying to log the time a sample was actually taken on the mote and when that mote comes back in radio range report the list of samples and times they were taken. The problem is that my logs of when the samples were taken bear only causal relation to wall clock time. The hardware is capable of much more than this. I think there is a watch crystal on the Mica2 and Mica2Dot motes and time could be accurately counted and reported. _____ From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Cory Sharp Sent: Wednesday, January 03, 2007 5:27 PM To: Michael Newman Cc: tinyos-help@millennium.berkeley.edu Subject: Re: [Tinyos-help] Time counting The timer system in TinyOS does not guarantee any particular accuracy, only that "accuracy MUST remain reasonable", probably something like +/-10% of the declared units. This is for various reasons including but not limited to temperature, voltage, and manufacturing variations that affect hardware clocks, particular features or lack thereof offered by a microcontroller or its hardware platform, and timer computation choices within TinyOS. Probably the time units are actually 1024 ticks per second not 1000. Even so, as listed above, there are many reasons why it may not be for your particular platform in general or given its current environmental and electrical conditions. If you need more accuracy, it is best that you incorporate a time synchronization protocol into your application. Best, Cory On 1/3/07, Michael Newman <[EMAIL PROTECTED]> wrote: I have a complex application with a half dozen timers. One of the timers is used to count a clock once per second. The timer is set up with a call that looks like this: call TimerClock.start(TIMER_REPEAT,1000); The timer is never stopped or changed. I count a variable in the TimerClock.fired event. I expect this variable to reflect the number of seconds since the node has booted. What I observe is that the counter counts faster than the actual number of elapsed seconds. Given other activities and the possibility of interrupts being off for some periods I do not expect the timer to fire at an exact interval, what I do expect is that the frequency of the timer will be exact with some jitter to the individual firing events. I also expect the timer to never fire early. Has anyone verified that timers actually fire on an exact interval and that time is neither lost nor gained? _____ Michael Newman Dragonnorth Group 319 Dean Rd. Brookline, MA 02445 Cell: 617 821-4608 Home: 617 566-7975 Fax: 617 566-7975 www.dragonnorth.com <http://www.dragonnorth.com/> [EMAIL PROTECTED] _______________________________________________ Tinyos-help mailing list Tinyos-help@Millennium.Berkeley.EDU <mailto:Tinyos-help@Millennium.Berkeley.EDU> https://mail.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help <https://mail.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help>
_______________________________________________ Tinyos-help mailing list Tinyos-help@Millennium.Berkeley.EDU https://mail.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help
