[time-nuts] Time syncing question

Didier Juges didier at cox.net
Tue Aug 29 01:02:22 UTC 2006


I guess it would help to know if the phone system computer runs over a 
general purpose OS, such as Linux or a variation of Windows or other 
GPOS. If so, you may be able to install a good time keeping utility 
designed for that OS and have it steer and update the RTC from an NTP 
time server. If not (not running a GPOS), you are probably out of luck, 
unless the OS allows some sort of scripting that would let you update 
the time periodically.

You did not say if there was any facility at all to update the RTC other 
than when the system boots. If not, there again you are out of luck.

It may be possible (and simpler, while not a 100% fix) to change the 
RTC, or adjust it's crystal for better time keeping.  That would be, 
horror, a 100% hardware solution, but sometimes a hardware fix is more 
effective... Maxim has a couple of chips specially designed with very 
stable 32,368 Hz oscillators and built-in xtal. Check their web site for 
"Extremely accurate RTC" (DS3232 and DS3234). These chips are intended 
to be used with a uC via a serial bus (in which case they provide full 
calendar and other functionality), but you can simply use the 32 kHz 
output to drive your existing RTC and take advantage of the temperature 
compensated TCXO. Now, if you mate the chip with a uC, you could even 
adjust the oscillator for best initial accuracy.

Didier KO4BB

Joseph Gray wrote:
> Tom, thanks for your very clear and simple reply. I may be over thinking 
> about the problem and trying to make it more complicated than it is.
>
> The target device is a phone system. The time is used for call logging and 
> display on the phones. Considering that the device's RTC doesn't keep very 
> good time now, do you think I should worry about adjusting the time 
> incrementally, or just do it in one jump? I'm mostly concerned about a 
> negative time correction. This would certainly throw off the logging. On the 
> other hand, a large + or - time correction would be a one-time situation and 
> would be easily spotted in the log.
>
>   





More information about the Time-nuts_lists.febo.com mailing list