Quartz crystal frequencies for timekeeping

[jadney]

I'm new to this group and hoping some of you can offer a solution to a digital watch quandry I have. I'm handy with electronics, but new to watches.

The most common frequency for the quartz crystals in digital watches seems to be 32,768 Hz, which is 2 to the 15th power (2**15.) This makes sense for a timepiece that measures in seconds because dividing by 2 is the simplest/cheapest digital division you can do.

For a watch that keeps time to the nearest second, all I have to do is divide by 2 fifteen times and I get one pulse per second.

For a watch that times to 1/100 th of a second, however, I need a 100 Hz signal, but there's no way to get that from a 32,768 signal. Yet on my desk I have 2 watches with timers to 1/100 second, each with 32,768 Hz crystals. How is this done?

I had assumed that such watches must use 25,600 Hz crystals, which would be 100 x 2**10 Hz. I find such crystals to be readily available, but that does not seem to be what's actually used.

I can imagine 2 ways to approximate the hundredths of a second, but both require "fudging" the time slices.

You could have a free running oscillator that ran at about 100 Hz, and let that do the timing under 1 second, but that would not have the same accuracy as the main clock.

Or:

You could pick off the 1024 Hz (2**10) signal and use that to approximate 1 kHz, possibly dropping one count 24 times within each second, then divide by 10 to get 100 Hz. That would work, but 24 of our hundredths would be slightly longer than the others.

None of this seems like the kind of thing an actual watch company would do, plus they are all more difficult and expensive than just starting with a 25,600 Hz crystal.

So, how is this done, and why?

thanks,

[rewolf]

...You could pick off the 1024 Hz (2**10) signal and use that to approximate 1 kHz, possibly dropping one count 24 times within each second, then divide by 10 to get 100 Hz. That would work, but 24 of our hundredths would be slightly longer than the others....

Welcome to the forum, Jim

You started with a good question - and also with one possible answer. They way you described it above is easy to implement in an IC - just a few transistors more, and you can use a standard 32768Hz crystal. For a manually operated stopwatch noone bothers about (or would even notice) the slightly inexact 1/100s.

Other ways to derive 1/10 or 1/100 seconds out of a 32786Hz oscillator are listed here: http://dwf.nu/viewtopic.php?t=3669

[jadney]

The way you described it above is easy to implement in an IC - just a few transistors more, and you can use a standard 32768Hz crystal. For a manually operated stopwatch no one bothers about (or would even notice) the slightly inexact 1/100s.

Other ways to derive 1/10 or 1/100 seconds out of a 32786Hz oscillator are listed here: http://dwf.nu/viewtopic.php?t=3669

So we're agreed that they actually fudged it. I don't see any other way to do it, but I have to admit that I'm astonished that they would go this route when a 25,600 Hz crystal would have made this both easier and intrinsically accurate.

And maybe a more versatile choice would have been 36,000 Hz, (1000 x 2**5) which could have become universal for seconds, tenths, hundredths, and thousandths.

I guess my work which frequently deals in milliseconds and microseconds, just makes me see this as more of a fault.

The watch I'm concerned with at the moment is a Casio 95QR-33, which is nothing special except that I bought it in Japan many years ago, which gives it sentimental value to me. It came out in 1977, so I'm pretty sure there's no microprocessor in it. Also, it was cheap. I recently dropped it and now the display is completely blank. I've been all thru it and the crystal is good. The last time I put it back together it read 12:00 for about 2 seconds then went blank again. I guess I have to take it apart and look closer. There must be a broken trace or component somewhere on that little board.

My web searches for Casio seem to turn up almost no interest among enthusiasts web sites, so my guess is that these are generally considered unworthy of interest.

Is there anyone here with experience or interest in my 95QR-33?



Thanks for the explanation and the link.

[Old Tom]

Actually, practically all Casio watches from module 27 onwards could be said to be "microprocessor", only the early NEC made Casios are really simple counter watches.

I think it was me that raised the 1/100 second stopwatch conundrum earlier and since I have come to the conclusion that they use the 1024Hz count and "sleep" for 24 counts at the top of the second rather than a "hit and miss" divider approach.

Your module 95 watch is displaying all the classic signs of crystal failure- the 12:00 display is generated when the power on circuitry puts enough of a kick into the upconverter to power on the display but since there is no timebase to drive the converter the display slowly fades out. Casio crystals are "odd", same frequency as others (mostly 32K but there are exceptions!)- best to thieve one from a tatty old working Casio.

The site for old Casios is pocketcalculatorshow.com- loads of rabid Casio enthusiasts over there! And Seiko, Alba & Pulsar LCD digitals!

[jadney]

Your module 95 watch is displaying all the classic signs of crystal failure- the 12:00 display is generated when the power on circuitry puts enough of a kick into the upconverter to power on the display but since there is no timebase to drive the converter the display slowly fades out. Casio crystals are "odd", same frequency as others (mostly 32K but there are exceptions!)- best to thieve one from a tatty old working Casio.

This display disappeared in a small fraction of a second. It's still possible that the crystal is at fault, and I have a couple other 32768s that I could try in there to see if that fixes it.

I tested the crystal that's in there by driving it with a signal generator and looking at the resulting amplitude. I could see a dip at the series resonance a few Hz below the nominal, but I could never see a rise. Discussing this with a co-worker, we decided that the impedances in my circuit were too small, so this killed the Q of the crystal, making the parallel resonance invisible.

It's possible that the crystal is damaged in such a way that its Q is very low, but it still would have passed my test. Since the Q is essential for operation in the watch, I should make up an active oscillator circuit to test the crystal.

So the 95 in my 95QR-33 is the module number? Is there any chance that replacement 95 modules are still available anywhere? That would be as a last resort.

Thanks for the Casio info. I'll look around to see what else I can find.

[jadney]

Sorry if this has been asked and answered before, but what would be the load capacitance for the quartz crystal in my Casio 95QR-33?

Digikey offers quite a few 32,768 Hz watch crystals at load capacitances of 6, 8, & 12.5 pF. Any recommendations as to what models or brands are most reliable/accurate/stable/etc?

thanks,

[J Thomas]

[rewolf]

Digikey offers quite a few 32,768 Hz watch crystals at load capacitances of 6, 8, & 12.5 pF. Any recommendations as to what models or brands are most reliable/accurate/stable/etc?

12.5pF is standard - if you don't know any better, I'd start with this one. If the oscillator is actually designed for 8pF, it'll run too fast because load capacitance is too small for the 12.5pF crystal.
Note that the load capacitance is not in the crystal, but it is the EXTERNAL capacitance required in the oscillator circuit.

[jadney]

Seeing as you're the technical type, you might find the information at these links valuable in explaining why performance is affected.

http://www.ecliptek.com/crystals/glossary.html

An even more detailed description by Fox:
http://www.foxonline.com/techdata.htm

Regards, Jeff

Thanks, Jeff, but I was already familiar with the basic crystal oscillator fundamentals. I was hoping to be able to narrow down my choices of appropriate crystals. Nevertheless, the foxonline ref was particularly good.

I went ahead and ordered an assortment of 32,768 Hz crystals from Digikey yesterday. It's interesting that most of them were well under $1, but one was $6. I've got them coming in 6, 9, & 12.5 pF loading.

In the meantime, I set up a different circuit to do a passive test on the Casio crystal in its parallel resonant mode. It actually appears to be good, so I'm not sure where to go from here. I guess the next move would be to make an active oscillator that I can put any crystal into. I'll wait for my assortment to arrive so I don't take a chance on damaging my Casio one by overdriving it.

It's not clear to me how one determines the drive level. I can measure the voltage easily, but it's hard to understand how to caluclate the power in something like this where I don't know the series motional resistance. In my tests so far, I've been keeping the AC voltage levels across my crystals down to about 0.1 Vrms, but that seems awfully low to trigger logic level circuits..

Can you give me some idea about what range of voltage levels one would normally see across the crystal in watches like this, or in any quartz watches, for that matter?

thanks,

[jadney]

Can you give me some idea about what range of voltage levels one would normally see across the crystal in watches like this, or in any quartz watches, for that matter?

I think I just answered my own question. I have another cheap watch here that I was able to get a pair of scope probes on the crystal and do a differential measurement while the watch was running. I got about 2.25 Vp-p.

It may have been a bit higher before I added the probes, but at least this gives me the right idea.

[rewolf]

Can you give me some idea about what range of voltage levels one would normally see across the crystal in watches like this, or in any quartz watches, for that matter?

I think I just answered my own question. I have another cheap watch here that I was able to get a pair of scope probes on the crystal and do a differential measurement while the watch was running. I got about 2.25 Vp-p.

It may have been a bit higher before I added the probes, but at least this gives me the right idea.

I haven't ever measured on watches, but RTC (real time clock) chips and microcontrollers usually drive the 32kHz crystals rail-to-rail, that is ~3V. Power consumption of these oscillators is also in the 1µA range.
Most low power oscillators stop if you try to probe them with a scope at the input - a standard 1MOhm||25pF probe will at least detune the oscillator.

[jadney]

.
Most low power oscillators stop if you try to probe them with a scope at the input - a standard 1MOhm||25pF probe will at least detune the oscillator.

The probes I used were 10 MOhm||7 pF. I assume the extra 3.5 pF (2 probes in series) loaded the crystal down a bit more toward the series resonance.

[jadney]

I failed to post any followup to this thread, but here goes....

I replaced the 32,768 crystal with one from Digikey (12.5 pF was the correct load capacitance.) That got the watch working again. Once it worked, I let it "break in" for several weeks, then used a very nice frequency counter I have to measure the frequency it was running at.

At the same time, I set the watch to the correct time as received over the web.

Then I wore the watch normally for 2 weeks.

After 2 weeks, I checked the time difference between my watch and the web time, and calculated the error in parts per million (ppm.)

I then went back and measured the watch frequency again, and adjusted it by the same ppm to compensate for the error.

For the last year the watch has been keeping time to better than 1 second per month. Until last month, that is. At some point, I glanced at it and saw all the LCD segments showing dimly at the same time and while I watched they dimmed more and disappeared.

I figured it needed a new battery, but that didn't help.

So today I finally got around to looking into it more closely and I found that my solder joints to the replacement crystal didn't seem as stiff as I thought they should be, so I resoldered them.

That brought it back to life. I have a display, but it's mostly garbage that changes about once per second. I'm guessing that I need to reset it, but I can't figure out which contact to touch to +. I've actually tried all of them, but I'm reluctant to short them as long as 2 seconds, for fear of damaging something else.

This is a 95QR-33. I don't see any "AC" readily visible, so I'm about to take the battery out and inspect it all over. I'll try to check the PC board vias for continuity at the same time..

I'd appreciate any suggestions.

thanks,

[Old Tom]

Look on the front of the module- a few Casios have the AC terminal here- seem to remember 95 does have an AC- only 52 & 75 come to mind as not having resets.

[jadney]

Look on the front of the module- a few Casios have the AC terminal here- seem to remember 95 does have an AC- only 52 & 75 come to mind as not having resets.

I've looked it over everywhere and don't find "AC" anywhere, front or back, inside or out. There are 4 test points: 1 square & 1 round near the adj capacitor and 1 square & 1 round 180 deg from that cap. Does this help?

However, I've now come to an even worse problem. There are 4 gold plated spring contacts where the outside buttons are expected to touch down. They just slide into grooves in the sides of the plastic module body. There was some corrosion near one of them and that contact has crumbled. Looks like they are gold over steel rather than gold over stainless.

Is there any source for replacement contacts? I have some .002" and .003" stainless which I can try to cut to shape, but the pieces are sooo small that I'm skeptical that I can do this. Were the same contacts used in lots of modules?

What are my chances of finding a replacement 95 module? Anyone got one, working or not?

thanks,