Crystals for better accuracy everyday clock

[nukie]

Now let me state my situation here. We don't have radio controlled clock in Australia so I am looking in to 'upgading' my cheap quartz clock.

I have a few cheap digital and analog they can't keep the time very well. It can be off by 5 minutes in a few months. Which is annoying for someone who is very fond of precision multimeters   Having said that, I do have a Seiko analog clock which keeps time well.

So I am thinking it could be those cheap crystals they use in the clock circuit which is not keeping the time well. I have to be honest here I have absolutely no knowledge of how crystals works. All I know is, you get a crystal frequency you need and slap in the two required load capacitance and hook it up to the X1 X2 microcontroller pins. That's as far as I understand.

So I have been searching around the net for such clock upgrades but no return. So I am wondering if I get a trusted Seiko or Citizen +-20ppm 32.768kHz crystals and replace the cheap no brand cylindrical crystal in my clock, do you think this will increase it's accuracy? The issue is, the datasheet specify a load capacitance of 12.5pF, but there's no capacitors to be seen on the clock circuit board. A blob of epoxy covered chip and some wires is all I can see.

The clock circuit accepts 2 pin crystal, so TCXOs are not possible, unless I spend sometime building the supporting circuitry.

[Tony R]

ok well here is what i know on the matter, (be forgiving we are just learning about this stuff now in one of my classes)

a crystal oscillator which are usually made of quarts are very good in terms of stability, they don't drift very far in regrads to time and temperature most other oscillators do.

Now when do you need to use a crystal?

well an LC oscillator (Colpitts, Hartley) use inductors and capacitors, one can mimic an inductor by the use of OpAmps and capacitors. however some times when doing calculators it appears that you need an inductor of a few 100 Henrys, which is rather difficult to make and will defiantly not fit into a watch if you manage to make one.

this is where crystal oscillators come in handy, they can be used instead of a insanity huge inductance.

Now will this work for your application?

I would say not really, you would one need to match closely to what they already have in place, it may work if you can find one, but i view my self as a very practical person who likes to solve the issue the simplest way possible... new clock?

However it would be a fun learning project none the less, but if your goal is to get a more accurate clock, just get a new clock...

[tnt]

Use a GPS module, those provide very  accurate clock as well.

[saturation]

Have you tried calling the US standard time website, even with the net delay to Oz, it will be at most only off a few seconds, then use this to sync your PC clock daily, which can be set from say, within Windows XP and up versions.

If you have a cellphone, nearly all sync to GPS clocks used by the provider networks; they are accurate to within a few seconds from standard cesium clocks 24/7 all year round.

Any cheap quartz clock is actually accurate to seconds a week, what makes it drift most is ambient temperature changes during the day; particularly if you wear it as a wrist watch, warm when on, cold when off.

http://tf.nist.gov/general/pdf/2276.pdf

For keeping standard time, try keeping a reference clock in a refrigerator.  This will keep the clock at a constant 40F or whatever you set your refrigerator too, so once set it will reduce drift caused by changing ambient conditions.  The drift rate can then be calculated after a month or so,  then simply adjust the watch blindly on that day to correct for ambient drift.  If the watch is more inaccurate, then you need to keep it at a steady room temperature instead of cold, see the NIST article.  Note, it still has an internal oscillator drift from the crystal aging, but this is much slower and will need to be compensated for in seconds/year or every so many months.  

[ziq8tsi]

It can be off by 5 minutes in a few months.

So I am wondering if I get a trusted Seiko or Citizen +-20ppm 32.768kHz crystals

20ppm is 1.73 seconds per day, which is 10.5 minutes a year.  So I am not convinced this is worthwhile upgrade.

I do not think you can get very much better than this without going to temperature compensated oscillators.

[nukie]

Maybe I didn't make myself clear in the first place. The digital clock I have is quite shit, it can be off by 3-5 seconds in a 2 hours. I use www.timeanddate.com as my reference. It's the most accurate time source I have next to the windows clock after synchronization with the time servers.

My point is, I was trying to find out if the crystal in the shit digital clock can be upgraded. I have another Seiko alarm clock which I am happy with it's accuracy. Although some opinion here is not optimistic I still went ahead because it will always bug me without proving it. So I have a few old motherboards most of them have a 32768 crystals. As a sys-admin I know that computer can't keep very accurate time without synchronizing with the time servers every once a while but it's all I've got on hand. I traced the circuit to two ceramics along the crystal pins, I believe they are the load capacitors. Four motherboards confirm my findings. They are about 18-22pf I didn't take time to measure accurately. So I modded the digital clock with the crystal and wired the load capacitors to ground. So far it has been many hours and the seconds seem to be accurate, therefore the original crystal must be 250ppm or something   I will have to wait a few more weeks when winter hits. I will place it in the bathroom where there's most temperature swing, ie hot showers.

Yes I have thought of using a GPS module but those can use quite a lot of power, and indoors means you will have connection issues. I have been doing some research, Maxim, NXP makes some nice Real Time Clock with TCXO chips. I may use them to make my own clock.

I like the idea of having a clock in the fridge, but I guess TCXO is the answer to my problem.

[the_raptor]

I would make my own clock with a GPS unit capable of taking an external antenna. Or a grade below that have it sync NTP over wifi. Fiddling about upgrading crystals seems like a waste of time.

I have been thinking of making my own clock with the flip down style display for a few weeks now. That or nixie tubes.

[Zad]

Maxim do a temperature corrected 32.768kHz oscillator IC that you could probably use instead. The crystal is housed inside the IC package.

http://www.maxim-ic.com/datasheet/index.mvp/id/4627/t/al

It does have control lines, but it should produce an output without having to be fed data (not 100% sure, haven't checked).

[jahonen]

Maxim DS32kHz might be also worth of considering.

Regards,
Janne