Need Highly Accurate RTC

[Noggin]

I have a project that will be running from DC power and needs a very accurate real time clock. The device is basically a transmitter that will transmit the time to a receiver. The receiver will compare the received time to the actual time as indicated by SNTP servers. The time on the transmitter must be accurate to approximately 1 minute per year with a temperature specification of -40°C to +85°C.

The best time keeper I think of would be GPS based, but I'm not sure I can guarantee that a GPS signal be available so GPS might not work. The second best time keeper I can think of are the NIST time radio stations. I don't think those are standardized world wide, so a single solution might not be doable. The best RTC I can find is the DS3231 / DS3234 with a 3.5ppm, internal, temperature compensated crystal on the -40 to +85°C range. This comes out to about 2 minutes per year, or half of the desired accuracy we're looking to achieve.

I've seen key fobs that display a number on an LCD screen that changes every few minutes. How do those keep accurate time? Or does the server just allow a significant amount of "slop" in the time?

 

[kubeek]

How about https://en.wikipedia.org/wiki/GPS_disciplined_oscillator but the solution really depends on your budget. The key fobs I think are issued for just a few years, and the key could work with a few keys before and after as well.

 

[Noggin]

Budget isn't nailed down, but probably $20 or less at 1k quantity for the time keeping circuit. The GPSDO appears to still rely on receiving a GPS signal, so I'm not sure that would work for us.

 

[kubeek]

The rubidium oscillator should keep the clock pretty stable, and then whenever it finds the GPS signal it corrects the clock to show the right time. Even without the GPS, the rubidium should be a lot more stable than 1min/year.
But that will be way over 20 dollars a piece and requires a lot of energy. You could try looking into temperature compensated and oven crystal oscillators.

 

[AnalogKid]

A disciplined oven or double-oven will cost less than a disciplined rubidium standard (we use those for some MIL stuff), but still more than $20. The idea behind a DO is that it doesn't need continuous GPS. It learns about its own error characteristics when sync'd to GPS, and uses that learning to flywheel very accurately until it gets GPS lock again. If it can see just one GPS satellite for a few hours each day, or even each week, it is an excellent clock.

ak

 

[kubeek]

1 minute in a year is close to 2ppm. Just a quick google found for example the M200F series from here http://www.conwin.com/product_locator.html?ModelFam=Oscillators&ProdType=TCXO&Recommended=Y&Pkg=SM 5x3.2mm which claims +/- 1 ppm stability over one year and 3 ppm in 20 years, however I cannot find any hint on the pricing.

But that is -40/+85, and if you can do your oven for the oscillator, then you should be able to find cheaper models to fit your bill.