accuracy required in DS1302 based clock

[elecfriend]

I am using RTC DS1302 in my clock. The clock advances after one or two months to about 1 or 2 minutes. I need accuracy like in professional digital clocks that do not shift even in years.

 

[kjennejohn]

**broken link removed**
**broken link removed**

Did a Google search using "atomic clock IC". Both these guys offer ICs/modules in various complexities that receive the atomic clocks around the world. Does time correct to the second for a million years work for you?

Hope this helps,
kenjj

 

[philba]

1 minute in 2 months is 12 PPM which is pretty good accuracy. Getting to very high accuracy can be expensive and tricky.

why not just get a clock that receives a standard time broadcast. depending on where you live, of course. I have one that gets the NBS broadcast out of Boulder CO and it is always spot on. I think it cost me $30. Europe has a similar service.

You can do things to increase the accuracy of your clock but they get increasingly more complex and expensive. I'd start with using a micro to adjust the time - say lose a second every 12 hrs (given 1 minute gain per 2 months). You might be able to find more accurate crystals than the 20 ppm I'd guess you are using. You probably need to factor in temperature and aging.

 

[RODALCO]

A good crystal is the 4,194,304 Hz, divide by 22 gives exactly 1 Hz.
Drift less than 3 seconds per month.
A symple chrystal oven to keep the temperature constant may even make it more accurate.
I have an electromechanical pendulum clock at home which drift less than 20 seconds a month. so what you are using is not that accurate.

 

[philba]

3 seconds per month is 1.157 PPM. who makes this crystal and what does it cost?

 

[RODALCO]

Haven't got a clue, these were fitted in German timeclocks of the MaxiRex range.
type BQT made by WEG Legrand D 477 Soest, West Germany.
You see them sometimes on Ebay for a few dollars.

These have very accurate timing crystals and don't seem to suffer from drift.
I will make a photo of one and will post it in the next few days.

 

[philba]

I looked through several catalogs and only found 20 ppm crystals (or worse).

I'd bet those clocks do suffer drift, though small. Probably have some sort of crystal oven. Aging is another issue - as crystals get older, they drift.

Note that accuracy and drift are two different things. A 20 ppm crystal is spec'd to oscillate + or - 20ppm when using specific load caps at a specific temperature. The crystal frequency will change based on a number of factors - temp, age, load capacitance,... I believe it is possible to characterize these effects and compensate for them. Proper low noise design of the circuit is important as noise can be seen as a oscillation and thus increasing the frequency.

 

[Nigel Goodwin]

Aging is another issue - as crystals get older, they drift.

Heh! - we all do that!

 

[philba]

I don't drift, I just plain forget. so what were we talking about???

 

[tunedwolf]

you could try using a simple radio receiver and a pll to lock the carrier freq, then just divide down to what you need...most of these broadcasts are usually atomic accurate...


rgds

 

[hjames]

TCXO's with ~3ppm are US$10, and OCXO's with .1ppm are in the US$100 range. Obviously all need more power than a simple tuning fork osc. Or just get a GPS receiver/serial data stream, NTP network clock, or the atomic clock radio receiver (I think Digikey just stared carrying them).