3
\$\begingroup\$

A device that I'm working on is controlled by ATmega328P microcontroller. This will be some sort of a watch, therefore I want it to be as precise as possible. For instance if left to run for a whole year I want the time deviation to be not more than absolute value of +/- 10 seconds by the end of the year. (I'm not saying the battery power will be enough for the whole year..)

Is crystal oscillator more precise than crystal with caps driven my micro? If so, are crystal oscillators more commonly used in time sensitive applications?

Or just a regular crystal will be sufficient and more suitable? (Especially when they will not consume power like oscillators) I don't need atomic precision of +/- 0.000001 sec/year.

\$\endgroup\$
  • 1
    \$\begingroup\$ You're not going to get that precise an oscillator on a battery. \$\endgroup\$ – Ignacio Vazquez-Abrams Aug 26 '14 at 3:53
2
\$\begingroup\$

10 seconds / year is 0.3 parts per million error (or as the time-nuts would usually put it, an error of 3 * 10^-7). A decent watch crystal is usually specified as +/- 20 ppm (2 * 10^-5), which is far worse than what you're asking for. The higher-frequency crystals used for clocking CPUs are usually worse than that.

Temperature-compensated oscillators (TCXOs) are better, in the 10^-6 range, but still worse than your spec.

Ovenized oscillators (OCXOs) can be better than 10^-7 but require watts of power, likely more than you can afford running on battery.

\$\endgroup\$
  • 1
    \$\begingroup\$ A watch is like a poor man's OCXO, your wrist keeps the device at approximately little under 37°C. \$\endgroup\$ – jippie Aug 26 '14 at 5:14
  • \$\begingroup\$ The Bulova Precisionist, which uses a three-prong crystal running at 262.144 kHz, is claimed to be accurate to ± 10 seconds/year, compared to ± 15 seconds/month for most quartz crystal watches. Independent testing shows the accuracy is excellent at room temperature but goes down at higher temperatures. \$\endgroup\$ – tcrosley Aug 26 '14 at 9:11
2
\$\begingroup\$

A GPS receiver is an excellent timing source. There are small SMD modules which require between 30 and 100mA from a 3.3Vdc source, depends on the specific module. Beside the power requirement, it also needs an antenna and the GPS signal (Can't work well underground or inside buildings).

\$\endgroup\$
1
\$\begingroup\$

If the device is not portable, you could use the AC power line as a timing reference, as the average frequency of the power distribution system is held to extremely tight tolerance.

\$\endgroup\$
  • \$\begingroup\$ Depends on the application. AC net frequency can have a significant short term drift (I read about >500ppm over a period of several hours). However, it tends to average out over long term. \$\endgroup\$ – Rev1.0 Aug 26 '14 at 9:01
0
\$\begingroup\$

A high precision real time clock like the DS3231 is another option. It can get you up to ±2ppm. That is pretty nice, but would still result in around ±60s/year.

Are you sure you need ±10s/year?

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.