Xtal Oscillators 32768 Khz

[RetiredHAL]

Hi, what would be an allowable error margin for a Xtal Oscillator. I took one out of an old Battery wall clock and tried to feed it into a 4060 chip to see if I could get a 2Hz "tick".
To my surprise it was way off, Yet gave good time, when it ran. The mechanism was the Xtal ,some very small electricals under a Glob of pink and a 2cm coil with an "armature" to drive the analogue movement. I could not find any trimming capacitors to tune the movement.
My question is this? Is the armature acting as a "slug" to tune it?
My second question? What accuracy can I expect if I buy a 32768 Xtal?

Cheers RetiredHAL

 

[Nigel Goodwin]

For it to be accurate you have to load it correctly, you should have a variable capacitor to adjust it to be exactly correct. Any correctly designed circuit should allow to adjust it both above and below 32768Hz.

 

[KMoffett]

I too wonder about the accuracy of the clock circuits once they're out of the clock. I took one apart that was entirely mounted on a metal framework ... xtal/battery contacts/black blob/driver coils. I cut away the coils and battery contacts so that all that was left was a small "H" shaped frame with the xtal and blob. I diode OR'ed the output to get a 1 second pulse. I also did a divide by 60 circuit to get a 1 second pulse off of the 60Hz power line. I put electronic counters on each of these, started them together, and let them run for three days. The xtal clock counter was about 45 seconds ahead of the power line counter. I have all kinds of these clocks at work and at home, and only need to reset them for Daylight Savings or a dead battery. So their normal accuracy seems much better than that. I'd like to try another one and leave all the parts intact, or find ones that are on phenolic PCBs, and do the same thing to see how they compare. Calls for a trip to Goodwill/Salvation Army/Savers.

Ken

 

[RetiredHAL]

For it to be accurate you have to load it correctly, you should have a variable capacitor to adjust it to be exactly correct. Any correctly designed circuit should allow to adjust it both above and below 32768Hz.

Thank You Nigel, for your quick reply. I understand about the variable cap to tune the Xtal AND circuit. This is evident on the 4060 Datasheet and other documentation. My point being that there was no evidence of a trimming cap in the clock movement unless it was under the glob of Pink paint which I shall carefully scrape away tomorrow. The blob of pink is about 4 mm round. it also means that this being a cheap mass produced clock a lot of manual intervention would be required to produce this item (clock movement driver). And it means it's frequency changes once it is out of the clock.

Cheers
RH

 

[Nigel Goodwin]

Thank You Nigel, for your quick reply. I understand about the variable cap to tune the Xtal AND circuit. This is evident on the 4060 Datasheet and other documentation. My point being that there was no evidence of a trimming cap in the clock movement unless it was under the glob of Pink paint which I shall carefully scrape away tomorrow. The blob of pink is about 4 mm round. it also means that this being a cheap mass produced clock a lot of manual intervention would be required to produce this item (clock movement driver). And it means it's frequency changes once it is out of the clock.

The original clock was probably designed using a circuit accurately tuned to the exact crystal used? - with either fixed capacitors, or fixed capacitance built-in the PCB. Presumably the 'blob of pink paint' is a COB IC (Chip On Board), basically an unencapsulated IC fixed directly to the board, then encapsulated with the 'blob'

 

[RetiredHAL]

Nigel, I suspect you are right about the COB. The part I tested included the PCB and blob and Coil , but not the armature as it sort of "fell" apart and I was unable to find it after it fell under my workbench in the garage! (sawdust) Hence my question about the armature being part of the tuned cct. (acting as a Slug?, at least that is what we called the screw inserts in coils)
Like KMoffett, I shall visit some Salvation Army places to buy some more defunct kitchen clocks for testing. I am intrigued now as to what causes the change in frequency.

Cheers RH

 

[Diver300]

Some circuits are adjusted by fiddling with the divide ratio, rather than adjusting the crystal to the "correct" frequency.

https://www.electro-tech-online.com/custompdfs/2007/10/39747d.pdf

Page 176 (confusingly labled page 174 in text on the page) shows how one microcontroller does it.

 

[audioguru]

You cut away metal parts of the clock so you reduced the capacitance of the crystal's wires to the frame which was grounded. Then of course the clock ran faster.

The mechanical pendulum must swing slightly slower than the electronics so it can be physically pushed to the correct frequency.

 

[RetiredHAL]

Hi, Audioguru, I could not scrape off the "Blob" to see what was underneath.
I agree with your explanation that the Xtal should run faster, if I strip the mechanism of the extra metal (extra capacitance).
My observations were that the period of oscillation was 30.8 uSecs. This means that the Xtal/COB/coil pack (without Clock) resonated at ~32.467 Khz. Not 32.768 Khz or more as expected (30.517 + uSec). Is my maths correct? (It's been a long time).
I have ordered a 32.768 Khz Xtal to see if the scope has a timebase problem. I could only purchase a 10 Mhz Xtal (I live in a small country town and there is only a handful of ready components) and made up a Collpit oscillator (discrete components) and measured the frequency. The scope showed exactly 0.1 uSec period for the oscillation (so it appears that the timebase is correct for that frequency range). I hope to give you some more info when I recieve the 32.768 Khz Xtal later this week and see whether the timebase on the scope is out.
NB, I can't repeat the experiment as the original Xtal is now defunct. (it will only take so much abuse from my soldering iron, etc).

Cheers
RH

 

[audioguru]

Your oscillator measured nearly 1% slow. Your measured time is 0.017% slow. A huge difference.

 

[ronsimpson]

How are you measuring the oscillator frequency?
Do not use an o-scope to measure frequency. (+/-5%)
Do not use a scope probe on an xtal. It adds 20pf.
In the case of the 4060 measure the frequency at one of the higher Q outputs.

 

[RetiredHAL]

Hi, I've been to the "Opportunity Shops" and purchased some battery wall clocks at a ridiculously low price. $3 for two clocks. I dismantled both and apart from the Clock Hands there was no metal. ALL plastic. I probed the running movement (iow, the movement mechanism was running without hands) and found exactly the same time 30.8 uSecs as before (slow).
I shall have to do a further test on a 4060 ic to see if my scope is out or whether I am loading the cct with my probe (tek p6109 12pF). I agree that 12pF is significant, but I would have thought that if I measured at the output of the emitter of the amplifier, I would be isolated enough to allow an accurate time measurement.
NB, I KNOW that the Xtal must run at its resonant frequency (especially 32.768 Khz, as they are used in hundreds of millions of kitchen wall clocks). So this is only an academic problem trying to understand where the error is. ( I am leaning toward my test equipment or my application of the test equipment) .

 

[Nigel Goodwin]

A scope generally isn't accurate enough for measurements like that, +/-5% timebase accuracy is probably about as good as you get.

 

[ronsimpson]

You should never count on you scope for accurate measurements. I think your scope is off by 1%. (only 1% that is really good) Use a frequency counter! Last time I calibrated my frequency counter with WWV the internal temperature controlled oscillator was off a little. 10,000,000.2hz

32,768 is the number! It divides down to 1Hz with a very simple ripple counter. Like the 4060. A divide by 32467 is much harder to make.

 

[mvs sarma]

Instead of doubting on metal parts influencing the frequecny drift, why not take a measurement of the master crystal frequencyon the test pin of the concerned IC.
Most cheap clocks would not have been tuned, neither they have any arrangment for such fine tuning.

PS: while i was typing , Ronsimpson has posted his idea.

He is very right that the best quality frequecny counters need calibration form a known standard source. afterall wall clock mechanisms, do definitely need adjustments.

I have seen synchronus clocks and these derive time from national standard transmissions at atomic standard clock sources.

 

[RetiredHAL]

Hi All, Thank you for your help and replies.
Today I used a 4060 to time the 32768 Xtal. I had to make a crude amplifier of a BC548, some resistors and capacitors so that the Xtal sinewave would not dampen and "die". In the end after a selection process of caps I wound up with a nice clean 10 V pulse to feed the 4060.
I made the trace very faint so I could get the thinnest of trace and extrapolated the reading to the best of my ability. (I have used a scope for 35 years)
I made the following measurements on the Q4 (÷16) pin
at start 490:mu:S = 2040.82 Hz
after 1 hour 487:mu:S = 2053.39 Hz
after 2 hours 488:mu:S = 2049.18 Hz
Theoretical time should be 488.2:mu: S =2048 Hz

On the Q7 (÷128) I measured 3.89 mS
Theoretical time should be 3.906 mS =256 Hz

On the Q12 (÷4096) I measured 128 mS
Theoretical time should be 125 mS =8 Hz

I then measured the same on Channel 2 and found the same results. So my measurements are not "skewed" by which channel I used.

I then measured the same on Channel 2 AND probed the acual Xtal with Channel 1 to see if the probe was "loading" the crystal and "slewing" the clock rate. I found that there was no percieved difference in the measurements.

The amplified clock rate and the actual Xtal clock remained at 30.8 :mu:S. through all procedures. (I have to measure this on the 5:mu:S scale)

My conclusion is that I have an error on my 5:mu:S scale.
The Xtal amplified (÷16) on a 4060 is very close to perfect after 2 hours.
The probes do not seem to load the crystal cct. (note that this is not a naked crystal. I do not know what sort of amplifier/oscillator is under the "Black Blob" ).
I am amazed that the Xtal can be cut on a very large scale (millions of wall clocks) and be so accurate.
I'm happy that my scope is reasonably accurate.

I shall buy a 4040 to check timings on the Q1, Q2, Q3 divisions and will report times next week or so.

I hope my maths is correct. It has been a long time.

Cheers
RH

 

[RetiredHAL]

Hi, I did the times for Q1,Q2 and Q3 on a 4040 ic
Q1 (÷2) 61:mu:S Theoretical 61.035:mu:S
Q2 (÷4) 122:mu:S Theoretical 122.07:mu:S
Q3 (÷8) 489:mu:S over 2 cycles = 244.5 Theoretical 244.14 :mu:S

Also after 3 hours the clock pulse was 30.5 :mu:S on the 5:mu:S/division scale and 30.1 on the 10:mu:S/didvision scale.

In any case I can safely assume now that the UnTuned garden variety quartz crystal kitchen wall clock is very accurate.

When I pulled apart another wall clock I found an 8pin (555size) chip with the name STP2740CoB The input is still 30.5:mu:S sine wave from a Xtal. The mechanism seems quite robust and I am loath to wreck it. I looked on the internet but could not find a Datasheet for that number. I think this is the oscillator as there is also a discrete 18pF cap. Does anyone know about this chip.

Again thank you all for your input

RH