Idea for Wall Clock Synchronization

[Tunelabguy]

Most wall clocks run off a crystal oscillator running at 32768 Hz. But the frequency is not exact, so the clocks drift. Of course there are "atomic clocks" that lock to low-frequency radio station WWVB, and in the old days wall clocks locked to the AC power grid. But wouldn't it be nice to make any off-the-shelf wall clock that runs at 32768 Hz lock to a local frequency reference?

My idea is to generate a precise 32768 Hz signal and distribute it to the wall clocks. At each wall clock use the signal to excite a coil that radiates the 32768 Hz only a few inches. Place this coil as close as possible to the wall clock's oscillator, hopefully without even getting inside the clock or knowing anything specific about that clock (other than the fact than it is based on a 32768 Hz oscillator). If the clock's oscillator is already very close to 32768 Hz it should not take very much of an injected signal to make it actually lock to that signal. I wonder how hard it would be to get a clock oscillator to lock to the signal from the coil?

 

[betwixt]

Several problems there:

1. how can you ensure the signal you are generating is any closer to exactly 32768Hz than the clocks own oscillator?
2. the existing crystal will be in a metal screening can and it's wiring very short, can you be sure enough signal would be induced to manage injection locking and how would you know if it suceeded anyway?
3. To generate continuous relatively high powered oscillations and probably from a temperature controlled crystal takes a lot of power, how do you propose to generate it?
4. can you be sure your signal will not have adverse effects on other parts of the clocks circuitry or increase it's own power consumption.
5. not all clocks use 32768Hz, I've got one in front of me right now with a 4.1943MHz crystal in it. I know of others that use 65536Hz crystals.
6. the orientation of the circuit in the clock may not allow your signal to be adequately picked up.

Brian

 

[Tunelabguy]

​

Several problems there:

1. how can you ensure the signal you are generating is any closer to exactly 32768Hz than the clocks own oscillator?

I can use any number of frequency standard techniques to build a local standard, such as locking to GPS, WWVB, or even Internet NTP. The point is I only have to do this once and amortize that effort over multiple clocks.

2. the existing crystal will be in a metal screening can and it's wiring very short, can you be sure enough signal would be induced to manage injection locking and how would you know if it suceeded anyway?

I don't know if the pickup is sufficient without testing it, but I thought I would gather opinions before I tried it to see if someone can dissuade me and save me the trouble of trying it. As for testing, that will take some time - several days maybe. I could deliberately generate a frequency that was slightly high or slightly low (but still within the lock range) and see if the clock responded appropriately. Maybe I would have to let each test run for a month, but it should be possible to do.

3. To generate continuous relatively high powered oscillations and probably from a temperature controlled crystal takes a lot of power, how do you propose to generate it?

"Relatively" is a relative term. I am hoping that the power level need not be that large, but that's what experimentation is for. As I said before, the local standard could be locked to a better standard, like NIST.

4. can you be sure your signal will not have adverse effects on other parts of the clocks circuitry or increase it's own power consumption.

That would be a deal breaker if it turned out to be true. But what if lower power works?

5. not all clocks use 32768Hz, I've got one in front of me right now with a 4.1943MHz crystal in it. I know of others that use 65536Hz crystals.

What fraction of low-cost consumer grade time of day clocks use something other than 32kHz?

6. the orientation of the circuit in the clock may not allow your signal to be adequately picked up.

Good point.

 

[betwixt]

The problem with injection locking is it stops as soon as you remove the dominant frequency source. So even if your idea worked, it would only make the clock run at the correct speed while your source was nearby, it wouldn't correct the time reading and the clock would revert to it's original timekeeping as soon as you removed your device.

Considering the time sources you suggested, it is possible to use GPS and derive 32768Hz from it but you would need to be visible to at least one satellite so for the most part it would only work on outdoor clocks, that limits your market somewhat! WWVB, NTP and other worldwide time standards will not give you sufficient accuracy unless you sample over a period of at least several days, possibly weeks or months. They do not directly produce a 32768Hz reference, only a time code so you would have to count cycles over a very long period before comparing and adjusting your own reference. Internet would be a particularly poor choice as time references will vary slightly depending on the route the data follows to reach you. A 'ping' is normally used to determine the transit delay and correct the time code but that might only be applicable to the route taken on one occasion and couldn't be trusted over longer periods.

I doubt you will be successful in this venture but I can share my experience from a job a did a few years ago which might help. It wasn't to injection lock an oscillator, just to provide a way of calibrating it so you would need access to the crystal and any adjustment capacitor to permanently cure inaccuracy. What I did sounds strange but I assure you it is true, I used a microphone to listen to the crystal! A small microphone, I used one designed for 40KHz ultrasonics, followed by a narrow band filter and high gain amplifier can hear the crystal vibrating froma few cm away. I squared up the signal and compared it against a temperature controlled reference oscillator to produce a frequency/phase difference signal so the inaccuracy could be read. I used a center-zero analog meter but you could equally use a digital display, the needle would swing one way if the 32768 was low in frequency and the other way if it was high. Aligning the oscillator to center point meant it was accurate to within a second per month or so. In my case the crystal I was adjusting was inside very expensive top-brand wristwatches.

Brian.

 

[Tunelabguy]

Brian, that's a great idea - to use ultrasonics instead of electromagnetics. The piezoelectric principle could be used in reverse from how you used it. Instead of listening to the crystal I could yell at it with a 32768Hz transducer. That would get around the problem of possible shielding and emf pickup by other circuits.

Of course you are right that this would only work as a full-time injection. However I don't envision the local frequency standard being located with the wall clock. There could be a cable connecting them. So GPS is not out of the question. The antenna could be on the top floor. Also Internet time is a practical solution too. A microcontroller running a software PLL could stabilize a 32768 Hz VCO so that over the long term it is locked to Internet time. It does not matter if the local 32768 Hz is short-term unstable as long as it has the right number of ticks every day. Same thing for the GPS solution. A good PLL can derive 32768 from 1 PPS.

But what really interests me now are the possibilities with ultrasonics. I thought those "tuning fork" crystals were so well balanced that almost none of the energy leaked out. But you are telling me that enough leaked out to detect with a sufficiently narrow band amp? Then I wonder how much acoustic power it would take to shake the crystal enough to inject a locking signal?

 

[yamato96]

i have crossed with some articles in mags that exploit clock sinc trough a radio signal, where in europe the sistem is called DCF77 and the signal is emited from germany to the rest of center europe. u just get a dcf77 receiver module and use it on ur circuit. very usefull sistem cause u ensure that inside 4 example a big factory, building and / or multiple offices the time of the clocks is always the same.

 

[betwixt]

From a commercial aspect yamato96 is correct.
Although magnetic or acoustic coupling might work in some cases, I can't see it being viable in a real World situation. Adapting existing cheap clocks with expensive technology, especially when so invasive to the building isn't a good solution. It would probably be much cheaper to replace the clock with one locked directly to the reference signal and get the benefit of it being 'self setting' at the same time. Once every two years or so I replace the battery in my wall clock here and it sets itself from the time signal automatically (GBR on 60KHz here, DCF in most of Europe). I think most 'radio clocks' turn their receivers off after setting themselves to save battery power and turn themselves back on once a hour or so recalibrate if necessary. I can tune my radio down to 60KHz and hear the time signal and watch the clock at the same time (I must be really boring person!) and they match perfectly.

What I have never seen on the market is a universal self-setting radio clock, one that uses DCF77, GBR, WWVB, RAI, CHU, JJY and probably others to set the time so perhaps that's something to concentrate on instead. All the present ones are limited by the range of the time standard transmitter, one that used and detected all would work anywhere in the world.

Brian.

 

[yamato96]

cellphone towers broadcast hour and info to cellphones, u get good cellphone signal almost everywere in any contry, but i fear costs of implementing such a great hack in a clock hould make better and cheaper just hang a cellphone on the wall.

 

[RCinFLA]

Generating the accurate freq is the easy part. Non-intrusively coupling into the clock is likely a large variable depending on clock circuit design and layout.

Ultrasonics is interesting concept but the degree of coupling is a large variable. Acoustical coupling is just like electromagnetic but speed of sound is much less. The transfer from one medium to another is just like impedance mismatch in RF transmission lines. There would be a great deal of loss, particularly through any air gaps.

A strong EM (RF field) is more likely to give a better result. You would have to overwelm the oscillator's drive to the xtal. Typically this is a resistor in series from a CMOS driver with a 20 pF cap to ground.

 

[Tunelabguy]

...You would have to overwelm the oscillator's drive to the xtal..

Not necessarily. There is a related phenomenon in pipe organs where pipes that are sounded together, even if individually they are slightly off from each other in pitch, will lock to each other and produce a beatless in-tune sound because of the small amount of non-linear acoustic interaction in the pipes.

The weaker the injection the signal the smaller the lock range. But the lock range never goes completely to zero unless the injection energy goes to zero. I suspect that the same thing will happen when injecting energy into a quartz crystal oscillator. A lower energy level will result in a narrower lock range. We aren't trying to shift the frequency by 1%. We are just trying to take an oscillator with 1 second per day accuracy and make it 0 seconds per day. That is a change of only 1 part in 86400. Even a small amount of injected energy should be able to pull the crystal this much.

However I have found a more serious bug in my plan. According to this NIST paper on wristwatches, many watches use an oscillator that is deliberately too high. Then they custom program the watch to inhibit the divided-down count (which drives the stepper motor) every so often so as to make the time come out right. So forcing such a clock to exactly 32768 would be very wrong.

 

[betwixt]

Bear in mind also that wristwatches normally have metal bodies and a metal dial, it would almost completely screen your injection signal. Mind you, a wristwatch permanently hooked up to a telephone line through a modem probably wouldn't sell very well. :-D

Brian.