Clock quartz resonators. How do quartz watches work? Advantages of crystal oscillators
Sometimes it is useful to have a clock in the system that counts time in seconds, and even with high accuracy. Often, special RTC (Real Time Clock) microcircuits like . It’s just that this is an additional case, and sometimes it costs as much as the MK itself, although you can do without it. Moreover, many MKs have a built-in RTC unit. It’s true that AVR doesn’t have it, but it does have an asynchronous timer that serves as a semi-finished product for making a watch.
First of all, we need a clock quartz at 32768 Hertz.
Why is quartz exactly 32768Hz and why is it called a sentry? Yes, everything is very simple - 32768 is a power of two. Two to the fifteenth power. Therefore, a fifteen-bit counter ticking at a frequency of 32768 Hz will overflow once per second. This makes it possible to build a clock using ordinary logical flow without any problems. And in the AVR microcontroller, you can organize a clock with seconds almost without using the brain, using peripheral reflexes.
Asynchronous Timer Mode
Remember how timers work? The clock frequency from the main clock generator (RC external or internal, external quartz or external oscillator) goes to the prescalers, and from the output of the prescalers it already clicks the values of the TCNT register. Or the input signal comes from the counting input Tn and also clicks the TCNT register
To do this, a quartz resonator is hung on the pins TOSC2 and TOSC1. Low frequency, usually an hour quartz at 32768Hz. It is mounted to the right of the controller and connected with jumpers. Moreover, the processor clock frequency must be at least four times higher. We have a clock from the internal oscillator of 8 MHz, so this condition does not bother us at all :)
And you don’t need to calculate the number of cycles of the main quartz, and if it doesn’t exist, then bother with the floating frequency of the built-in RC oscillator. Watch quartz has a much more compact size than regular quartz, and is cheaper.
 |
Also important is the fact that the asynchronous timer can tick on its own, from the clock quartz, because it does not need the processor clock frequency, which means that the clocking of the controller core (the hardest thing it has) can be turned off by hibernating the processor, significantly reducing energy consumption and waking up only when the timer overflows (1-2 times per second) to record new time readings.
Configuration
To turn it on, you just need to set the AS2 bit of the ASSR register - and that’s it, the timer operates in asynchronous mode. But there is one feature here that cost me a lot of headaches at one time. The fact is that when operating from your own quartz, all internal timer registers begin to synchronize using their own quartz. But it is slow and the main program can change an already entered value much faster than it can be processed by the timer.
That is, for example, you preset the value TCNT2, the timer on your 32 kHz thresher hasn’t even had time to chew it yet, but your algorithm has already run through and written something there again - as a result, garbage will probably end up in TCNT2. To prevent this from happening, the recording is buffered. Those. you think that you wrote the data to TCNT2, but in fact it ends up in the temporary register and will only get to the counting register after three clock cycles of the slow generator.
The OCR2 comparison registers and the TCCR2 configuration register are also buffered
How can I find out whether the data has already been entered into the timer or is hanging in intermediate cells? Yes, it’s very simple - using the flags in the ASSR register. These are the TCN2UB, OCR2UB and TCR2UB bits - each is responsible for its own register. When we, for example, write a value to TCNT2, TCNUB becomes 1, and as soon as our number from the intermediate register moves into the real counting register TCNT2 and begins to tick, this flag is automatically reset.
Thus, in asynchronous mode, when writing to the TCNT2, OCR2 and TCCR2 registers, you first need to check the TCN2UB, OCR2UB and TCR2UB flags and write only if they are equal to zero. Otherwise the result may be unpredictable.
Yes, another important point - when switching between synchronous and asynchronous modes, the value in the TCNT counter register may be lost. So, to be safe, we switch like this:
- Disable interrupts from this timer
- Switch to the desired mode (synchronous or asynchronous)
- We set up the timer again as needed. Those. set the TCNT2 preset if necessary, re-configure TCCR2
- If we switch to asynchronous mode, then we wait until all the TCN2UB, OCR2UB and TCR2UB flags are reset. Those. the settings have been applied and are ready to go.
- Resetting the timer/counter interrupt flags. Because with all these perturbations they can accidentally settle
- Enable interrupts from this timer
Failure to follow this sequence leads to unpredictable and difficult to detect glitches.
Sleep modes and asynchronous timer
Because an asynchronous timer is often used in various saving modes, then one feature arises that creates a whole field of rake.
The bottom line is that a timer powered by slow quartz cannot keep up with the main processor, and there are a lot of dependencies on the peripherals - the same interrupts, for example. And when the processor sleeps, these dependencies cannot be implemented, resulting in glitches such as broken interrupts or corrupted values in registers. So the logic for working with an asynchronous timer and sleep mode should be built in such a way that between waking up and putting it into hibernation, the asynchronous timer has time to work out several of its clock cycles and complete all its tasks.
Examples:
The controller uses power saving and core shutdown mode, and wakes up by interrupts from an asynchronous timer. Here we must take into account the fact that if we change the values of the TCNT2, OCR2 and TCCR2 registers, then hibernation should be done ONLY after the TCN2UB, OCR2UB and TCR2UB flags fall. Otherwise, the result will be such a mess - the asynchronous timer has not yet had time to pick up data from the intermediate registers (it is slow, hundreds of times slower than the core), and the core has already been cut off. And it would be nice if the new configuration wasn’t applied, that’s nonsense.
What’s worse is that while the TCNT or OCR registers are being modified, the operation of the comparison unit is blocked, which means that if the core falls asleep earlier, the comparison unit will never start - there will be no one to turn it on. And we will lose the interruption in comparison. The risk is that we will miss the event and lose them until the next awakening from hibernation.
What if the controller is woken up by a comparison interrupt? Then he will fall asleep completely. Oops!
So catch this glitch later.
So before going into power saving modes, you must definitely let the asynchronous timer chew on the entered values (if they were entered) and wait for the flags to reset.
Another joke with the asynchronous mode and energy saving is that the interrupt subsystem, when exiting hibernation, starts in 1 clock cycle of the slow generator. So even if we haven’t changed anything, we can’t go back into hibernation - we won’t wake up, because... interrupts will not have time to run.
So coming out of hibernation and going to sleep when interrupted by an asynchronous timer should look like this:
- Woke up
- They did something necessary
- Fell asleep
And the duration of the operation between Woke up and Falling asleep SHOULD NOT BE LESS than one tick of the asynchronous timer. Otherwise, suspended animation will be eternal. You can set the delay, or you can do it as the datasheet advises:
- Woke up
- They did something necessary
- Just for fun, we wrote something in any of the buffered registers. For example, in TCNT there was 1, and we recorded 1 again. Nothing has changed, but a recording occurred, the TCN2UB flag was raised, which is guaranteed to last three cycles of the slow generator.
- Wait until the flag falls
- We fell asleep.
It is also not recommended to immediately read the TCNT values when exiting hibernation - it can be considered a mess. It's better to wait one tick of the asynchronous timer. Or make a joke with writing to the register and waiting for the flag to fall, as was written above.
Well, the last, but important, point - after applying power, or coming out of deep hibernation, with the shutdown of not only the core, but in general the entire periphery, it is strongly recommended to use a slow generator no earlier than after 1 second(not a millisecond, but a whole second!). Otherwise, the generator may still be unstable and there will be more mess and garbage in the registers.
And, at the end of the article, a small example. Running an asynchronous timer on Atmega16 (How the board is used by the polygon)
The project is standard, based on a dispatcher, the only difference is that the dispatcher has been transferred to timer0 to free up timer2.
| 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 | int main(void) (InitAll(); // Initialize the peripheral InitRTOS() ; // Initialize the kernel RunRTOS() ; // Start the kernel. UDR = "R" ; // Start marker, for debugging SetTimerTask(InitASS_Timer, 1000 ) ; // Since the timer in asynchronous mode // starts slowly, we do // Delay to start initializing the timer. while(1) // Main dispatcher loop( wdt_reset() ; // Reset dog timer TaskManager() ; // Call the dispatcher) return 0 ; ) |
int main(void) ( InitAll(); // Initialize the peripheral InitRTOS(); // Initialize the kernel RunRTOS(); // Start the kernel. UDR = "R"; // Start marker, for debugging SetTimerTask(InitASS_Timer,1000) ; // Since the timer starts slowly in asynchronous mode, we // wait to start initializing the timer while(1) // Main loop of the dispatcher ( wdt_reset(); // Reset the dog timer TaskManager(); // Call dispatcher ) return 0;
The procedure for initializing the timer in asynchronous mode is made in the form of a finite state machine. When it first starts, it sets the asynchronous mode bit and makes preparations, then it starts itself again, through the dispatcher, to give something else the opportunity to slip through the queue without blocking the system while waiting.
On subsequent inputs, the ready flag bits of the timer registers are checked. If they are all zeros, then just in case we reset the timer interrupt flags to avoid glitches and false positives, and then enable the interrupt we need. And we go out.
| 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 | void InitASS_Timer(void ) ( if (ASSR & (1<< AS2) ) //If this is the second input then( if (ASSR & (1<< TCN2UB | 1 << OCR2UB | TCR2UB) ) // check if there is at least one flag bit( SetTask(InitASS_Timer) ; // If there is, then we send it to a repeat waiting cycle) else // If everything is clear, then you can run interrupts(TIFR |= 1<< OCF2 | 1 << TOV2; // Reset interrupt flags, just in case. TIMSK |= 1<< TOIE2; // Enable overflow interrupt return ; ) ) TIMSK &= ~(1<< OCIE2 | 1 << TOIE2) ; // Disable timer 2 interrupts ASSR = 1<< AS2; // Enable asynchronous mode TCNT2 = 0 ; TCCR2 = 5<< CS20; // A prescaler of 128 by 32768 will give 256 ticks per second // Which will give 1 overflow interrupt per second. SetTask(InitASS_Timer) ; // Run it through the dispatcher to log in again. } |
void InitASS_Timer(void) ( if(ASSR & (1< ISR(TIMER2_OVF_vect) // Timer 2 overflow interrupt ( UDR = i; i++; ) It was possible to make variables containing hours:minutes:seconds and click those variables with all their hour/minute overflow logic, but I was too lazy. And so everything is clear. Watch quartz is the term used to refer to a special battery for quartz watches. From a technical point of view, watch quartz is a quartz oscillator for transmitting energy, which is necessary to turn the hands of a watch. When you press the watch quartz, an electrical impulse appears, and when current is applied to it, compression occurs. It is thanks to the technical characteristics of the watch quartz used that watches based on quartz oscillators are famous for their amazing accuracy of readings. In our online store you can purchase high-quality modern versions of quartz watches wholesale and retail from the best global and domestic manufacturers at competitive prices. Preferential delivery conditions are provided for Moscow residents. When using a high-quality quartz oscillator in the watch mechanism, the watch practically does not require additional charging. It is enough to start them once every few years. The main parameter for choosing a watch quartz is its scope of application, its correspondence to the watch mechanism. New generations of watch quartz are maximally adapted to a variety of modifications of quartz watches. We talk about the basic principles of operation of the quartz movement in wristwatches It seems that everyone more or less understands the basic structure of mechanical watches. Of course, rather less than more, but the basis is clear: the source of energy - the spring - acts on the wheel, the frequency of the latter’s oscillations is determined by the spiral, the vibrations are transmitted to the unit, and it is followed by the transmission, which is also the gears (gears) that drive arrows. In fact, everything is incredibly complicated, but that’s the principle. The main thing is that all this can be seen with your eyes. And even touch it with your hands. Well, at least virtually. But quartz watches - how are they built? After all, not everything can be seen there, much less touched. However, if you look into them, everything is simpler. So, in order. The energy source is a battery. The energy is not mechanical, as in the case of a spring, but electrical. Nevertheless, there is still energy. The battery can be a “tablet”, or maybe, for example, solar. These are just details. An oscillatory system is a generator with a quartz resonator, or quartz for short. The current produced by the battery causes the quartz crystal to vibrate (piezoelectric effect). This crystal - an analogue of a spiral - is tuned to a certain vibration frequency, most often 32,768 hertz
This is approximately ten thousand times more than the number of oscillations of the balance in an ordinary mechanical watch.. At this frequency, a generator produces pulses - an analogue of a balance. 32,768 is 2 to the 15th power, which is important, since the circuit also contains a simple binary counter, also known as a divider, at the output of which the frequency is reduced to 1 hertz - to a second clock cycle. At this frequency of 32,768 hertz - once per second - pulses are sent to a stepper motor, which is analogous to a trigger. In this short essay with photographs, I will show the general public how to treat a fairly common “sore” of digital electronic watches - inaccuracy. The clock can lag or rush, and most often we do not pay attention to small errors, but when the clock is behind by 5 (five) minutes a day, it begins to irritate. I bought this watch in order to feel nostalgic for the old Soviet times, when the sun was greener and the grass was brighter... or vice versa?.. it doesn’t matter! The main thing is that there was no joy - the clock was disgustingly behind. More than 5 minutes a day. I need to heal, I thought. To treat the clock we need: Necessarily Preferably Let's disassemble the watch Unscrew the four screws holding the back cover. Carefully remove the cover and remove the piezoelectric resonator (tweeter). We do not paw the squeaker with our fingers; we hold it by the side edges and the metal base. We take the watch out of the case. We unscrew four screws - three hold the 2016 lithium battery, one holds the spring tab for sending a signal to the tweeter. Using tweezers, carefully remove the board from the plastic holder. To replace the quartz, I decided to use donor quartz from the old motherboard, which died ten years ago and I am slowly taking it apart into small components. To replace, simply unsolder one quartz and solder in another. Reassembly We assemble the mechanism in the reverse order - we place the board on the holder, there are guide pins there. We put the battery on the board, minus pointing down. We turn the block over and look - the clock should start. If this does not happen, it means either the battery is upside down or the quartz is not soldered in or it is not working or the board has been killed by static :) OK it's all over Now! During the day the clock has not moved forward or backward, it runs smoothly and accurately. I'll watch it some more and then report back on the accuracy. It must be said that the procedure for replacing quartz is the same for all quartz watches - digital, dial. But, we must remember that most Chinese quartz watches are assembled on plastic rivets, which are melted with “mushrooms”, i.e. in fact, once the watch is disassembled, it is very difficult to reassemble it. Outside the framework of this “Murzilka” there was a film that the Chinese did not remove from the LCD when they put it in the holder. I removed this film and the screen contrast increased slightly. The film is almost invisible, but it was on my watch. UPD
. The comments provide a recipe for finer-tuning the accuracy by soldering miniature ceramic capacitors. As an alternative to replacing quartz, it is quite viable and sensible. The main thing is that there is a place where to place these capacitors. Well, the presence of them... And in general, friends, the main thing is not the review, the main thing is the comments))) Features of the selection and operation of watch quartz
Image: multi-master.ru
Diagram of the operation of a simple quartz movement (with a second hand at 6 o'clock). Image: Encyclopedia Britannica Battery
Battery (using the example of the ETA Flatline 210.001 mechanism). Image: eta.ch Quartz
Generator with a quartz resonator (using the example of the ETA Flatline 210.001 mechanism). Image: eta.ch 
Microprocessor / binary counter (using the example of the ETA Flatline 210.001 mechanism). Image: eta.ch Electric motor
Stepper motor (using the example of the ETA Flatline 210.001 mechanism). Image: eta.ch
Ready? Go!
Intro
Looking ahead, I would like to note that I did not open a dispute; one hundred rubles is not the same money. The problem is not with the seller who sent a low-quality product. The problem is in a product that the seller cannot check in any way - surely a Chinese man/woman will not sit and measure the accuracy of the move?
+ soldering iron. preferably not very powerful, 25-40 watts is enough. 60 will already be too much.
+ replacement quartz resonator. Sold either in China or in any radio store. It is inexpensive and is called “watch quartz.”
+ thin Phillips screwdriver or thin flat head screwdriver. cross is preferable.
+ tweezers with sharp jaws - pick up the screws (yes, the body is plastic, the frame is also plastic. There are screws everywhere)
+ good lighting and a stationary magnifying glass or jeweler/watchmaker glasses to clearly see Little Red Riding Hood's watch.
We note that the watch does not have a protective gasket, therefore water and sweat will get inside the watch. We understand that the Chinese save on everything for the sake of cheapness, which means that the glass is most likely sitting on double-sided tape and the buttons do not have rubber seals. This means that the watch will need to be removed in bad weather and during physical work.
We put the case, back cover, back cover screws and tweeter aside.
Let's put this all aside. We are considering the fee. You can't see any more screws, which means that's good.
Inside the clip we see a conductive rubber band that transmits a signal to the LCD and the LCD indicator itself.
We don’t touch the elastic with our fingers, because it doesn’t matter. If a speck or dirt gets in, some segment on the indicator falls off and you have to take it apart again... what the hell...
In the blue heat shrink there is a coil that produces sound. There is no need to touch it either. It’s easy to damage, the wiring there is thinner than a hair.
But the metal cylinder on the legs is our quartz resonator, which needs to be changed.
The quartz here is slightly larger than in watches.
Here, for comparison, is already soldered quartz from the motherboard and a clock board. 
We apply quartz to the board. Fits. We put quartz in the cage, it also fits! Great! Let's change! 
There is no polarity, no features. The procedure is simple and does not require special qualifications. 
Voila! quartz replaced. We align the quartz body so that it is just below the board and does not touch the battery.
Place the contact block on top of the battery. In this watch, it simultaneously holds the battery and is the contact group for the buttons. Fasten with three screws. Then a separate contact for the tweeter. We also screw it on.
Well, if everything works, carefully place the board into the watch case, center it so that the numbers are parallel to the edge, then install the beeper back, screw on the cover... 
We have overcome a big problem)))
Well, the size of the quartz also matters - if the quartz from the motherboard did not fit in size, then you would have to look for another, smaller one.
Over the past four days, since the quartz was replaced, the clock has moved forward by two seconds. 15 seconds per month.
For a cheap watch and free quartz, I think the result is satisfactory. Personally, it completely satisfies me)))
You can, of course, look for quartz watches for pennies at flea markets, pick up a bunch of quartz from there and experiment with precision... but we’ll leave that to perfectionists and die-hard freaks)))
Thank you all for valuable ideas and various discussions)))
