clock stability

[Ally bagpipes]

I've completed my 24hr clock made by chaining ripplecounters and it works fine, the trouble is that the quartz crystal is temperature instable, has anyone ideas to increase the stability of my clock, if it's any help, I've a 74hc4060 in my divide by 2 counter from the 32768 Hz crystal, my load capacitor is variable.
Any help would be great.

 

[ericgibbs]

I've completed my 24hr clock made by chaining ripplecounters and it works fine, the trouble is that the quartz crystal is temperature instable, has anyone ideas to increase the stability of my clock, if it's any help, I've a 74hc4060 in my divide by 2 counter from the 32768 Hz crystal, my load capacitor is variable.
Any help would be great.

hi.

Do you have the specification for the 32KHz crystal.?

The temperature frequency drift for some crystals is quite high, is it the version designed for say, a watch or is it a general purpose type for say a MCU.?

 

[Mikebits]

You might consider a more stable clock source such as this one.
**broken link removed**

Granted, it is not the greatest of time keeping but I bet it will perform nicely for your project.

 

[marcbarker]

If I remember rightly 32 kHz watch crystals are only "15 seconds a month accurate" for 25 deg C +/- 5 degC or something like that.

 

[MikeMl]

32kHz watch crystals are designed to be driven with very low drive power. They are usually driven by an oscillator that is powered on a 1.5V battery, biased so that the oscillator barely starts, and consumes uW of power.


If driven hard, as they would be in a CMOS or TTL biased gate oscillator operated off 5V, the crystal self-heats, and stability goes to hell.

If you really want long-term stability, you will need to start with a crystal in the 4MHz range and follow that up with a suitable counting chain. Even better, put the 4MHz in an oven. Even better, use a "GPS-disiplined" oscillator.

 

[ericgibbs]

32kHz watch crystals are designed to be driven with very low drive power. They are usually driven by an oscillator that is powered on a 1.5V battery, biased so that the oscillator barely starts, and consumes uW of power.


If driven hard, as they would be in a CMOS or TTL biased gate oscillator operated off 5V, the crystal self-heats, and stability goes to hell.

If you really want long-term stability, you will need to start with a crystal in the 4MHz range and follow that up with a suitable counting chain. Even better, put the 4MHz in an oven. Even better, use a "GPS-disiplined" oscillator.

hi Mike,
Providing the correct value series resistors are chosen for a CMOS osc ie: HEF4060 the watch xtals are not over driven.
This is the way I have used them with good results.

 

[MikeMl]

hi Mike,
Providing the correct value series resistors are chosen for a CMOS osc ie: HEF4060 the watch xtals are not over driven.
This is the way I have used them with good results.

Perhaps if the OP posted a schematic, we could critique it

 

[marcbarker]

If driven hard, as they would be in a CMOS or TTL biased gate oscillator operated off 5V, the crystal self-heats, and stability goes to hell.

That's if the crystal doesn't shatter before it gets time to heat up!

 

[Ally bagpipes]

clock stability continued

thank you all for your comments, now for the next one, does anyone have suggestions for a circuit that will produce 1Hz output (to input to my clock circuit) from a +-4MHz crystal?

Thank you once again

 

[blueroomelectronics]

Your 4060 & 32768Hz crystal will be as stable as a 4MHz clock and a bag of dividers. Your PCB layout is probably the culprit or choice of crystal (what's the load pf?)

 

[Ally bagpipes]

Thanks for the gen of pcb layout, the load pf is 22 either side of the crystal.

 

[marcbarker]

does anyone have suggestions for a circuit that will produce 1Hz output

(Please pardon me for quoting you out of context)

Yes, buy the cheapest analogue digital clock (yes I'm serious) you can find and remove the circuit assembly from it, it runs of 1.5 V, takes nearly zero power, and provides a 1 Hz pulse to drive the movement solenoid with. It's accurate to 15 seconds a month.

Another way to consider is use mains line frequency, like alarm clock radios use, that's suprisingly accurate.

 

[blueroomelectronics]

Is the crystal spec for a 22pf load? Some are higher like 68pf, I use 12.5pf crystals in my Dragonfly kit.

Do you have mains current available or is it battery run? What accuracy are you looking for?

Did you put a GND loop around the pins on the PCB? What temperature is the area it's located in (ie room, outside, North pole?)

 

[Ally bagpipes]

i've used the "clock oscillator" as at Electronics 48, using the 4060 as illustrated, the 2hz output is divided by 2 using a d type flip flop and then inputted to the first stage ripple counter. the whole clock is running from 5v dc, (Transformed from mains), all ics a grounded, accuracy wanted will be what looks to something like +-15s per month. the clock is sited at room temperature away from any obvious heat sources.

Thanks for your interest

 

[Hero999]

You can do this with one IC 4.194304MHz crystal will do. You can replace the IC with a 74HC4521 if you like.

I have a cheap Chinese shaving mirror with an integrated clock: it has extremely poor stability and always looses time. Perhaps you have a cheap crystal.