how to build a clock with pulse for the hours!

The "stock" answer for any digital logic problem involving more than a dozen gates is "use a PIC" (or other microcontroller). Whether that's the right answer depends upon how many of these you're planning on making.

If it's just one or two, the cost of a programmer (~$35) and the time taken to learn microcontroller programming will outweigh the cost and effort of a discrete solution. Unless you're planning on making similar projects in the future, in which case you may as well start learning microcontrollers now.

If it's between dozens and many thousands, using a microcontroller (whether PIC, AVR, etc) is the logical option. For very large quantities, a custom chip will eventually win out, but I guess that you're not talking about that kind of scale.

The main question is how you're going to adjust the time without a display. That's a far more involved problem than generating 1-12 pulses every hour.

It does not *have* to be involved at all. It depends on the requirements. You could, for example, have one button that must be pushed at exactly 1 o'clock. Extremely technically simple, but not very practical.

On the other end of the scale you could use GPS or DCF77 to fully eliminate any need for adjustments, even after a power outage.

Finding that perfect balance between simple (cheap) enough and practical enough is the challenge. I can't come up with a better idea than a simple 7 segment display and a couple of buttons.

--
RoRo

The comments already made provide some range to think over. I'm going to jump in and take a specific shot.

I'll assume that you'd really like some precision -- something on the order you've come to expect from a common watch. That's going to be "hard" even for a microcontroller supporting a 32kHz crystal, because watch makers have gone to some lengths now in designing low cost means with relatively high precision. Achieving that in a new design with a microcontroller will take effort and equipment. In fact, doing that with any newly designed circuit will, regardless of whether or not it includes a microcontroller. (So that's out -- unless you can consider something on the order of 20 seconds a day drift, or worse. If this were on the order of some minutes a day drift, it would be cheap and easy with a micro.)

It costs about US$10 at my local shopping store to buy a watch with hands. You might consider doing what any decent bomb-maker may do -- modify the watch to provide an electrical contact for you. These watches have delicate hands but I suspect that the gearing for the hour hand is sufficient that a slight loading would be acceptable. A fine wire, for example. To avoid bounce, relying upon a contact time that is unknown but probably long compared to what you want, etc., you may want to add a second point of contact for resetting a circuit -- a flip flop, for example. And then add a one-shot to the output to give the desired pulse duration.

The watch face cover may be glass or plastic. If plastic, you might simply drill through it with a tiny drill bit to make a penetration near the hour position. If glass, remove it entirely or penetrate it with a drill, too. The fine wire might be 40 gauge, wrapped lightly over the hour hand and arranged so that the tip extends outwards towards the hole. Arrange another wire (any convenient diameter) so that when the hour hand's fine wire goes by the hole that it will have to brush across it. The tip of the hour hand wire should be just barely long enough to be forced to brush across it in its travel.

You'd need to tweak things to get the initial timing of the hour pulse at the right point.

Another possibility would be to use a tiny neodymium magnet on the tip (or even midway, perhaps) of the hour hand. Use a "reed relay" in a glass ampule and locate it near enough to the hour hand that the magnet will be able to close the relay. This one seems good in ways, tricky in others. Good in that you may be able to find a small magnet with enough power that you can close the reed relay even when it is outside the face cover. Which leaves everything sealed. Bad in that if that idea can't work at all, you may still face difficulties finding any magnet tiny enough, yet powerful enough to consistently close the reed relay at any marginal distance apart. Traditionally, the magnets used are rather large, so this would need to be a matter of experimentation before proceeding. But there are companies on the web selling such things. Personally, I'd tear apart a cd-rom drive for the magnets near where the tiny coil near the optical lens is at. They are often VERY small and yet VERY powerful. A mere fragment of one of those may do the trick.

Jon

Which is still significantly more complex than generating the pulses.

--
1. How accurate does it have to be?

2. Can you use the AC mains as a time base?

JF

HUH??? With a 32k xtal and no trimming at all it is easy to achieve

I didn't write that one _must_ get drift that bad. In practice, a few minutes a month does happen and temperature has a particularly noticeable impact. There are some nice data sheets on compensating in software for temperature drift that improve this. I think Dallas also provides a part that is a bit pricey but does pretty good, too.

Cheap watches do all this, and do it cheaply.

It was just a thought. Sorry this bothers you.

Jon

On Jun 5, 6:07=A0pm, snipped-for-privacy@gmail.com wrote: > On Jun 5, 2:27=A0pm, Jon Kirwan wrote: >

Or you can do what I did and use the AC line as the primary timebase for a microcontroller and default to the processor crystal during power failures. I'm currently finishing the software to tie in to a C- Max WWVB receiver. The power line as a reference is 0 seconds drift even without the WWVB but with it will auto set DST and take care of leap seconds. The C-Max receive module with the antenna is at DigiKey for $10.70 - single unit price.

G=B2

Thanks for all the replies. A few things. For me, accuarcy here is not important (it would be nice if it was though). This is more of means to exploit the very nice sound this gong-thing makes. Just a light tap (even with finger) makes it sound. I would like to keep it "mechanical" as possible (someone suggested sampling the sound etc) but would rather not go that route. I think maybe the small magnet on a real clock hand might be the one to have go at in the spirit of this project (as the electronics is probably beyond me). I thought maybe there is a some "clock chip" I could use and then somehow (some how :-)) get the n pulses for each of the hours. Thanks again.

There are clock chips, but they're generally designed to either interface directly to 7-segment displays, or to a CPU.

Decoding either would probably be more effort than keeping the time with a string of binary counters.

I agree with you Jon. here is the Dallas device and indeed it is pricey

precision 32kHz clock:

Most 32Khz crystal clocks are indeed only accurate to about 30ppm, so

(30/1,000,000) x 60 x 60 x 24 = 2.6 seconds per day which is 15 minutes per year.... as long as the temperature does not change. If it does, things just get worse.

You could always measure the drift over 1 week and compensate with a lookup table in your software to frig the time as a rough compensation

Watches are compensated to have good accuracy at body temperature

--
Bill Naylor
www.electronworks.co.uk
Electronic Kits for Education and Fun

Clock building is not an uncommon hobby. You can buy the clock mechanics and then build the rest yourself. Maybe some of those kits have some sort of cuckoo clock functionality. I have found some with chime functionality, but I couldn't find any cuckoo stuff.

Alternatively, you could hack something like this:

--
RoRo

I like your thinking. I don't know if a magnet is practical the clock hands are relatively heavy - I notice when testing batteries with a clock and dummy load, the clock almost always stop with the second in the west most quadrant trying to climb past the "9."

A photo interrupter, that the hand moves through, would probably be a better idea. A single one with some method to set the bell count would probably be easier than trying to find both the hour and minute hands and the 13 interrupters that might take.

Or take a digital clock and use logic or a pic to decode the display, or just use the guts from a quartz analog clock to provide an accurate one second pulse to a counter or pic.

I have no idea how practical it is. I imagine (and only imgaine) that it _may_ be possible to get a small enough 'bit' of a neodymium magnet and place it far enough back towards the axis of the hour hand that the combination might function. I also worry that the remaining size of it may be still too big to remain underneath the watch face cover, so that cover may still be a problem. Besides, it just forces the distance to the reed relay that much farther away, which is part of all the factors that need to be optimized together. However, might be fun to find out.

That's good. I have used interrupters. In this case, I might instead paint a highly reflective spot on the hour hand and use reflection instead of interruption. But I like this.

Yes, that crossed my mind. But then that hauls in the coding, tools, etc. So I avoided mentioning it. Just because I have all that stuff doesn't mean others want to buy into all that trouble.

Jon

And, it seems from experience now, surprisingly good measurement accuracy even off my hand.

Years ago, I had wrist watches that were pretty good when I wore them but if I left them off for a few months would show drift I could quickly spot. Lately, because my autistic daughter likes watches for reasons I don't entirely understand, I've gone out and bought all kinds of cheap $10 watches. They are now surprisingly good, off or on. I have one I wear (she removes it and puts it back on me, one or two times a day) that is tied to WWVB and updates itself every night, so that (plus my cell phone) represents my accuracy standard around the home. Sometimes, I wear them just to "bug" her, keep them in my coat pocket, or they just lay around the house. We tend to keep our house open to the outside (I live on a farm), unheated and uncooled to save money, and temperatures in the house vary widely from about 55F to 90F. (We are very comfortable at 55F, but if it gets much below that we will heat. We never cool.) In short, these watches experience temperature differences over the year vastly exceeding their assumed use. And, of course, there are pocket watches that hang from belt hooks, which may (or may not, I don't know) assume that they are placed _in_ the pocket, normally. (Frankly, I think they have to assume that they will simply hang openly, though.)

I've been very much impressed with the consistency in performance of even these cheaper watches. They have certainly improved the technology. Very, very accurate in marking off time.

I have read a page or two getting into _some_ detail about how this is achieved. But nothing as good as I'd like to see. I drool over trying to understand _all_ the myriad details that have coalesced into a modern, cheap watch. If anyone knows of something talking about all of the manufacturing details, those that are important to control and why, those that are not important to control for and also why, etc., I'd love to see it.

Jon

The idea using a wire wrapped around the hour hand and making contact with another wire is close to _mechanical_. The problem will be that it makes contact for some period of time. Using it only as a simple switch to the solenoid without something else may be a problem. You might arrange things, also mechanically, so that the solenoid action also disturbs the wire connection in just such a way as to move it to the other side and break contact. But my imagination is short on practical ideas there. Maybe someone else has a handy thought about it.

[The problem I have with "doing up a microcontroller solution" for you (which I'd otherwise consider doing free for a moment, at least), is that no matter what I did you'd need some way of "setting the time." This means an input device and some means of letting you know what the clock thinks. Two-way communication of some kind. It's possible, though perhaps difficult, to consider doing this using a single push button and a single LED. Short and long LED blinks can communicate surprising amounts of information. I've done crazier things before. But it would require some thought about making that work well. So that holds me back. Making the interface easier to use means more expense and more design considerations, so that blocks me, too.]

Jon

A piezo speaker is slightly more expensive than an LED, but still only needs one pin, and allows for different frequency beeps. Or even speech synthesis (but that's pushing it on a 12F508, though).

Even simpler electronically: telephone-style pulse dialling. One input, no outputs.

--- Just for grins, let's say you want to drive this thing from 60Hz mains.

Then you'll have a very accurate 60Hz signal that you can use for a timebase, and you can get your gong ringing signal like this:

+----+ +----------------+--------------|S | | +------------+ | U2 | | | | COUNT UP | | COUNT DN | | | | +--------+ | | +--------+ | | | +-|LOAD TC|-+ +--|LOAD TC|--|R Q|-+ +------------+ | | | | | +----+ | 60Hz>---| DIV 216000 |--1Hr-+---|> | +--|> | U3 | +------------+ | | | | | | 0---|D3 Q3|---|--|D3 | | 0---|D2 Q2|---|--|D2 | | 0---|D1 Q1|---|--|D1 | | 1---|D0 Q0|---|--|D0 | | +--------+ | +--------+ | U1 | | | ASTABLE | | +--------+ | +--|OUT E|---------+ | +--------+ | U4 | | +V | | | +----------+ | | |K | [SOLENOID] [DIODE] | | | [R] +----------+ | | | C +---B Q1 E | GND

Here's how it works:

Initially, a switch is pressed at precisely 1 o'clock which resets the divide-by-216000 counter, loads binary 0001 into U1 and also into U2 when the "1" ripples through U1.

The switch isn't shown because it's a PITA to do the ASCII and the circuitry is trivial.

Anyway...

Once 0001 is loaded into U2, its Terminal Count (TC) output will go false, allowing U3-Q to go true since its SET "S" will be true when its RESET "R" input goes false.

Then, when U3-Q goes true it will enable U4, an astable with an "ON" time long enough to allow Q1 to energize the solenoid and create the desired sound from the gong and a period long enough to allow the ringing of the gong to decay as desired before the next strike.

The astable is also used as the clock source for U2, and once U2 counts down to zero its TC output will go true, resetting U3, which will disable U4.

Assuming that the output of U1 was 0001 because of the pushbutton reset, then, when the next pulse comes out of the divide-by-216000 counter, an hour later, U1 will increment to 0010, 0010 will be loaded into U2, U3 will be set, U4 will be enabled, will issue two pulses to the solenoid driver, and then will shut down.

When the next pulse comes in from the one-hour divider, U1 will increment, that value will be loaded into U2, and that number of gong-strikes will be dealt with by U2 until U2 counts down to zero.

JF

Hehe. No, I'm not wanting to engage that kind of programming as a one-off for no charge. I've got other stuff to do. The single LED would be okay. Maybe two?

I used to be perfectly able to dial a phone number merely using the phone's hang-up mechanism. Yes, that can work. What I _have_ done before is used the difference between a short press and a long one to select different paths and used the LED to signal the user what was happening. Kind of a press short, LED blinks in certain way to let you know that's what it interpreted; or press and hold and the LED would go into a different state to indicate that it had registered the long press as a long press. Different menu systems entered either way. Etc. While cheap and easy enough to implement, it can be tedious to the user. ;)

Jon