Continue to Site

Welcome to our site!

Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

  • Welcome to our site! Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

Which is better for accurate clocking

Status
Not open for further replies.

RetiredHAL

New Member
Hi, I have spent a day or so (on this site and the internet) trying to determine which is the better approach to get an accurate clock. (seconds over a year)

I only want to consider the arguments relating to using a 32.768 Khz crystal as opposed to a 4.194304 Mhz crystal.
I am not interested in special oven or temperature controlled or software driven correction method ICs. I am only really interested in 4000 Ic series usage.

I have looked at almost everything, but readily admit that my maths skills are not sufficient to understand the integration, differentiation and calculus involved to make a valued judgement. I refer to Uni papers and other technical treatises on this subject.

I note (drooling mouth and lips) the Oliverb's Clock and the accuracy he has achieved.

This is an academic exercise. The 32kHz Xtal is used in all battery operated clocks and watches and keeps exceptionally good time.
The 4.19Mhz Xtal is less widely used but is still a natural 2^n divisible number.

Given that the Xtal manufacturers specify +/- 50-100 ppm error in the cut crystal is there an advantage using either Xtal?

I have done numerous exercises calculating seconds in a year, with slightly different error rates, but am unable to derive at a definite answer.

My other question of course is how do you check the time. I do not not have ready access to GPS and I am not sure that the "time pips" in Australia can be traced to a Standard.

Cheers
RetiredHAL
 
hi Hal,
In the UK and Europe we have a radio transmission system that transmits time data accurate to 1sec in a million years!
A small receiver in the clock keeps track of the transmission and the daylight saving changes.

Is there a like system in OZ??

The other thing you can do, is to go on line and get your PC, time 'pinged' with a very accurate time.

I prefer the 4.19Mhz Xtal for a software clock, you need to pull/push it onto a precise frequency with a small variable cap, about 3 to 35pF
Its possible to get an accuracy of about 1 sec/month.

Look here:
http://www.abstime.com/atomic-clock-sync/

This is free: http://bexonsoft.com/bexclock/download.htm
 
Last edited:
Here is a schematic of a digital clock that I have built a number of times that uses a 2.4576 Mhz crystal and some 4000 series IC's to generate a 60Hz time base. I adjusted the value of the capacitors in the oscillator circuit to adjust the output of the oscillator to exactly 2.4576 Mhz on my frequency counter.
Checking the time over a period of time with WWV(U.S.) the clock was never off more then several seconds.
 

Attachments

  • digital clock.GIF
    digital clock.GIF
    66.7 KB · Views: 1,172
Since you want to use 4000 series CMOS chips than the CD4060 would be a good choice. You can trim any crystal, the 32768Hz ones will use less current.
 
I make crystal oscillators for a living, so I know a little bit about this.

32768 Hz watch crystals are tuning fork crystals, so they are shaped like a tuning fork and they oscillate in the same way.

They have a parabolic temperature curve, so they are at their maximum frequency at about wrist temperature, and they fall at -0.035 ppm/°C^2

So they are quite accurate near wrist temperature, but get quite rapidly worse over wide temperature ranges.

Just about all other crystals are AT cut crystals, that operate in thickness shear mode.

The temperature curve of those is S*T + C*T^3 where T is the temperature difference from room temperature.

S is the slope and can be just about anything and is controlled by the way the crystal disk is cut from the bulk crystal. C is the cubic coefficient and it is about 0.00000000011 for all AT cut crystals

With suitable choice of the slope, an AT cut crystal can be more accurate than a watch crystal, especially for wide temperature ranges. Cheap crystals do not control the slope at all well, so the temperature stability can vary a lot.


**broken link removed**

explains it quite well.

However, the most important thing for all oscillators is the initial adjustement.
 
Dallas Semi (maxim) makes a temperature compensated 32 Khz xtal, I think the part number is DS32Khz. It is quite accurate and doesn't cost much.
 
Diver300 said:
I make crystal oscillators for a living, so I know a little bit about this.

32768 Hz watch crystals are tuning fork crystals, so they are shaped like a tuning fork and they oscillate in the same way.

They have a parabolic temperature curve, so they are at their maximum frequency at about wrist temperature, and they fall at -0.035 ppm/°C^2

So they are quite accurate near wrist temperature, but get quite rapidly worse over wide temperature ranges.

Just about all other crystals are AT cut crystals, that operate in thickness shear mode.

The temperature curve of those is S*T + C*T^3 where T is the temperature difference from room temperature.

S is the slope and can be just about anything and is controlled by the way the crystal disk is cut from the bulk crystal. C is the cubic coefficient and it is about 0.00000000011 for all AT cut crystals

With suitable choice of the slope, an AT cut crystal can be more accurate than a watch crystal, especially for wide temperature ranges. Cheap crystals do not control the slope at all well, so the temperature stability can vary a lot.


**broken link removed**

explains it quite well.

However, the most important thing for all oscillators is the initial adjustement.

Good to know.
 
I prefer the 4.19 MHz Crystal which i have used in a number of projects.

OK, I admit that i have a healthy supply of those, complete with 8 pin IC and circuit board to drive a 1 second bipolar clock motor.
These IC's can divide the frequency down to one Hertz.

The accuracy is very good, ± 2 seconds a month in a non temperature controlled environment.

These clock parts i got from superseaded electromechanical MDI metering with newer T.O.U. metering read via cellphone or landline.
These clocks had to be very accurate to meet correct interval timing over peak demand periods.

I can not comment on the 32 kHz Crystals which are more commenly used now these days. I have not used those because of the reasons listed above.

Regards, Raymond
 
Diver300 said:
I make crystal oscillators for a living, so I know a little bit about this.

You are the right guy for my question on crystals. This is about frequency markings on crystals.

I have several ACTV crystal oscillators which has the frequency marked on the can.

It said "7.9365079 MHz" so it seems it is right down to 0.1Hz and that's amazingly precise.

My question is, how accurate would these frequency marking be in real life? So if I give it +5V and at 25 degree C, what frequency output should I expect from it?
 
Frequencies like that are not accurate to the number of digits. I would expect about ±25ppm at room temperature, but it will drift with time an temperature.

The number is just the frequency of a waveform with a period of exactly 126ns, writen to too many digits.

If you can tell me what else is written on the oscillators, if there is anything, I might be able to work out the accuracy.
 
The crystal oscillator has the follow markings:
ACTV v0
AABSJP
7.9365079MHz
S1

I asked the question because for some crystal oscillators we see markings like 8.00MHz but some marked as 8.00000MHz.

So is it a good assumption that the latter would be more precise(i.e. has less frequency error at 25 degreeC) because it has more significant figures in its frequency marking?
 
eblc1388 said:
The crystal oscillator has the follow markings:
ACTV v0
AABSJP
7.9365079MHz
S1

I asked the question because for some crystal oscillators we see markings like 8.00MHz but some marked as 8.00000MHz.

So is it a good assumption that the latter would be more precise(i.e. has less frequency error at 25 degreeC) because it has more significant figures in its frequency marking?

No, it's just designed to be that frequency, the numbers of digits makes no difference.

A common crystal is 4.43361875MHz (PAL subcarrier) - and this is exact - because it's phase locked to the transmitted colour burst! :D however, you still need to adjust a trimmer capacitor to ensure it's in the centre of the PLL capture range.
 
eblc1388 said:
The crystal oscillator has the follow markings:
ACTV v0
AABSJP
7.9365079MHz
S1

**broken link removed**

It was made in week 19 of 2001 (that comes from the S1) but I don't know much more about it. The AABSJP might indicate ±25 ppm (A) and Japan (JP) but those are guesses.
 
I drive a slave clock with a PIC. It keeps time to better than a second per month. Using a 4mhz crystal I simply adjusted the programmed loops until I achieved the desired accuracy.
 
Thank You all for your replies.
Ericgibbs-- I have downloaded the Atomic Clock on a trial basis and compared it to www.timeanddate.com over a period of time ( I altered the bios clock to ensure that it actually pulled the PC clock into correct time, I dont understand how it has access to Bios). The time difference between between the two sites is neglible. BUT there is a perceptible difference some times.(less than 1/4 second). So for long term adjustment in a clock this signal would be quite acceptable.
mc551mc, Yes I looked at that site earlier, but I think that may require a dedicated link to recieve these signals. FYI we have 50Hz in Aus.
Diver300, thank you for your input , I went through ALL your article and skipped the maths (very good reading, and took almost an hour). I do not pretend to understand it all.
Rodalco, I wish I had your access to these 4.19 Mhz xtals and ics (can I buy a set from you?)
Brevor-- DS32kHz is not readily available in Aus.
k7elp60-- mm5314N is not readily available in Aus . BTW what is the IC in the L/H corner prior to RN1?

Further , I cannot make a (salvaged) 32Khz clock Xtal work as per the 4060 datasheets (pierce ocillator setup using pins 10 and 11). The xtal works Ok if I use the COB of the (salvaged) kitchen clock and feed that pulse into Colpitts Oscillator to feed the 4060 IC

NB This Xtal is from a 1.5V AA battery clock.

Cheers
RH
 
RetiredHAL said:
Further , I cannot make a (salvaged) 32Khz clock Xtal work as per the 4060 datasheets (pierce ocillator setup using pins 10 and 11). The xtal works Ok if I use the COB of the (salvaged) kitchen clock and feed that pulse into Colpitts Oscillator to feed the 4060 IC

My guess is that you are most likely overdriven the crystal by using too low a series resistor.

See if the following circuit helps. (From page 5 of the pdf in the link)
 

Attachments

  • 4060_osc.gif
    4060_osc.gif
    16.4 KB · Views: 1,570
Just picked up this thread. If it is of any interest here are my findings.

I designed a crystal controlled timing system for racing. I used a programmed ST62 series micro as the clock and I used a 7.3728MHz crystal because that was a suitable frequecy to divide for the comms I required.

My ultimate goal was an accuracy of 1ms over 15minutes - ie. 1.44secs per day. I used crystals with the lowest temperature coefficient I could get - £12.50 each!. The crystal's parallel capacitors made an enormous difference and I ended up adding a 1" twisted pair across one of them as a fine trimmer. Untwisting just 1/8" made the difference of several seconds a day. In the event, I gave up the attempt as unworkable.

By comparison, my Casio wristwatch gets altered twice a year - when the clocks go back/forward, and it is never more than 30secs out in that time.

So my advice would be, go for the watch crystal.
 
Last edited:
Status
Not open for further replies.

Latest threads

New Articles From Microcontroller Tips

Back
Top