On Mon, 28 Jan 2013 23:11:22 -0800 (PST)
garre...@gmail.com wrote:
> That 10ppm is a lot worse than it seems...it's the rating at a steady
> 25C. When the temperature varies, it's going to go WAY off the rails.
> The DS3231's 2ppm is rated across the chip's entire temperature
> operating range. In practice, and in varying temperature conditions,
> there's just no comparison.
10 ppm is about 5 minutes per year. 2 ppm is about 1 minute per year.
That's pretty decent.
Yes, the DS1339 and similar RTCs have an error curve that depends on
temperature, as seen in [1], so if wide temps are important to you it's
definitely worth understanding that curve.
[1]:
http://www.maximintegrated.com/app-notes/index.mvp/id/58
The DS3231 is a better RTC. But if you compare any RTC with no RTC,
any RTC is better, too :)
At really cold temperatures the batteries (assuming you don't use a
super cap) powering the RTC are going to suck anyway. But again, going
battery vs super cap is another trade off since batteries can hold
quite a lot more energy than a super cap of the same physical size.
My desire for an RTC is to keep close to the correct time if the
product sits on a shelf for 1 or 2 years. Close means we're reasonably
accurate and that every boot is correctable by ntp if it's available.
Having a board think it's the year 2000 is bad, but having the clock
off by 10 minutes isn't a problem for me. If your application is
different, the DS3231 might be a better choice or some other time
keeping mechanism might even be a better choice than that.
It's all trade offs, there's no right answer.
-Andrew