EEVblog Electronics Community Forum
General => General Technical Chat => Topic started by: Connecteur on October 30, 2019, 03:33:13 pm
-
Personally, I haven't seem much change in the basic digital clock over the last 30 years. I have no doubt some types have evolved, but I'm talking about the basic consumer-grade digital clock, which seems mostly unchanged for decades. Most of these don't keep time all that well in my experience. Some of these clocks can drift tens of minutes over a year.
I have several cheap mechanical clocks, driven by a single AA battery, that are sold in dollar stores, that keep exceptional time, often staying within a minute for months at a time. My uncle had a mechanical clock in the dash of his car that used time adjustment to change the speed of the clock. For example, if the clock is 10 minutes fast, setting it back 10 minutes automatically slows the mechanism by 10 minutes per day. The next day, if it's a minute slow, setting it ahead a minute, speeds it up by a minute a day. That's a level of accuracy we generally don't see in consumer-grade digital clocks.
Is it possible to design circuitry in a digital clock to respond to time correction by changing its speed? Or is digital circuitry generally too variable for greater accuracy to be unachievable? I'm not talking of course, about the more expensive digital clocks that are specifically designed for higher accuracy.
-
The ones we get in the USA are almost exclusively mains synced and keep excellent time long term. The battery powered clocks I have all have WWVB receivers and set themselves to that.
Some of my favorite clocks are mechanical digital clocks like the early Tamura Lumitime series before the name started being used on mundane LED clocks. I like the old flip clocks too.
-
Sure it is. But why would you. I have a clock with epaper display, that connects to my phone every now and then to sync the clock, and to report temperature and humidity. Runs from CR2032. You can build in receiver modules which sync your clock to publicly available long wave radio clocks, like this:
https://en.wikipedia.org/wiki/DCF77
(https://en.wikipedia.org/wiki/DCF77)
-
Personally, I haven't seem much change in the basic digital clock over the last 30 years.
Make that 45 years, e.g. my https://entertaininghacks.wordpress.com/2015/02/23/vetinari-digital-clock/ (https://entertaininghacks.wordpress.com/2015/02/23/vetinari-digital-clock/)
https://www.youtube.com/watch?v=YmscLwBxYJY (https://www.youtube.com/watch?v=YmscLwBxYJY)
-
Thanks, but I wasn't talking about mains-synced, internet-synced or radio-synced. I'm talking about the basic digital clock circuit and whether it could incorporate circuitry that can "learn" if it running fast or slow when someone resets the correct time. Is it feasible, or is that type of circuitry inherently too unstable?
Does anyone know?
Thanks
-
If you use TCXO and calibrate clock well, it can be quite precise. Even good regular calibrated quartz clock can be made pretty stable. Fer example good quartz watches like Swatch drift maybe a few seconds in month. Mechanical clock will suck no matter how expensive it is.
Those cheap electronic clock which drift a lot likely are not calibrated at factory (quartz) or simply use mains frequency as reference.
-
Thanks, but I wasn't talking about mains-synced, internet-synced or radio-synced.
Mains synced is the worst unstable crap you can get. And those clocks which you complained about very likely are of this type.
-
Is it possible to design circuitry in a digital clock to respond to time correction by changing its speed? Or is digital circuitry generally too variable for greater accuracy to be unachievable? I'm not talking of course, about the more expensive digital clocks that are specifically designed for higher accuracy.
You could mess with the load capacitor of the 32.768 kHz crystal that I bet most of the digital clocks are using for timekeeping to make it go faster/slower. There are IC's that do just that automatically to achieve very good accuracy (1 minute per year) such as DS3231: https://datasheets.maximintegrated.com/en/ds/DS3231.pdf (https://datasheets.maximintegrated.com/en/ds/DS3231.pdf)
Or change the crystal itself to a better one. I've bought some random watch crystals from the regular sellers and they were off by several minutes a day, buying from a reputable seller gave me much more accurate results. That's always one option, but really you need some form of compensation to achieve the results of DS3231. Just getting a better crystal or messing with the load capacitors isn't going to cut it.
Mains synced is the worst unstable crap you can get. And those clocks which you complained about very likely are of this type.
It's accurate in the long term though. I remember last year clocks were off by up to 6 minutes: https://www.bbc.com/news/world-europe-43321113 (https://www.bbc.com/news/world-europe-43321113)
Good enough for oven as a clock I guess. :-//
-
Thanks, but I wasn't talking about mains-synced, internet-synced or radio-synced.
Mains synced is the worst unstable crap you can get. And those clocks which you complained about very likely are of this type.
Not in the US.At any given moment, you might see the frequency not precisely 60Hz, but they do control it tightly enough so that over a day it averages to 60Hz. I've been inside one of the locla utility's control rooms. They take it pretty seriously.
My clock it horribly inaccurate. It runs off a single AA battery which, due to the LCD display, lasts years (although modern alkalines will probably leak long before they are dead), and it likely just uses a very poorly calibrated crystal or maybe even something as primitive as an RC oscillator. I really need to get a better one, that syncs from the WWV broadcast. But I've lived with it all these years now, I'm used to it - it runs consistently fast, so worst case is I get up a few minutes earlier than I intended, which isn't a bad thing.
-
It's accurate in the long term though. I remember last year clocks were off by up to 6 minutes: https://www.bbc.com/news/world-europe-43321113 (https://www.bbc.com/news/world-europe-43321113)
Good enough for oven as a clock I guess. :-//
Off by 6 minutes in 1.5 months. Long term it's not good. It's not like low mains frequency is compensated with higher frequency at other time period. It's a pot luck how stable it is.
-
Not in the US.At any given moment, you might see the frequency not precisely 60Hz, but they do control it tightly enough so that over a day it averages to 60Hz. I've been inside one of the locla utility's control rooms. They take it pretty seriously.
They don't control it to average to exactly 60 Hz long term. They control it to be within spec.
-
Is it possible to design circuitry in a digital clock to respond to time correction by changing its speed?
You can build such clock yourselves. It would be a nice project. With DS3231 (not produced with crystal anymore) you may manipulate its "aging" register value, such your clock runs faster or slower. The DS3231 itself has got the temperature compensation inside, one of my 3231s did +61secs in 3 years.
There are micro-controllers with internal RTC oscillator, where you can also "digitally" slow down or speed up the clocks.
The fundamental problem with the idea of "automatic" adjusting the speed of a clock is its speed depends on many factors. The source of the precision is the xtal (or a fork resonator) oscillator and its frequency depends mainly on temperature, voltage, humidity, position in Earths gravitational field, mechanical vibration, and aging of the crystal. Therefore even you adjust its speed from time to time as your uncle does, the clock will always drift.
The mains frequency precision depends on the country you live in, in EU the providers compensate the mains frequency daily such it does 50Hz in an average pretty precisely.
"Mechanical clocks" with batteries inside are usually "digital" ones, with an Xtal oscillator and digital divider, generating short 1Hz pulses into a coil. Thus their precision is the same as the clocks with the digital display.
-
Thanks, but I wasn't talking about mains-synced, internet-synced or radio-synced.
Mains synced is the worst unstable crap you can get. And those clocks which you complained about very likely are of this type.
Not sure I agree. Even before the dynamos were computer controlled, power stations used an odometer-like revolution-counter to keep track of the number of revolutions per hour or per day, and then they adjusted the speed of the dynamo until the number of revolutions were correct. Over the course of months or years, a mains-controlled clock would remain nearly 100% accurate.
Now I can't speak for your local power, or whether it uses DC power inverted to AC and whether it's accurate enough for a clock, but there are still a lot of mains-regulated clocks around and I'd be surprised if they've lost accuracy.
-
Personally, I haven't seem much change in the basic digital clock over the last 30 years. I have no doubt some types have evolved, but I'm talking about the basic consumer-grade digital clock, which seems mostly unchanged for decades. Most of these don't keep time all that well in my experience. Some of these clocks can drift tens of minutes over a year.
I have several cheap mechanical clocks, driven by a single AA battery, that are sold in dollar stores, that keep exceptional time, often staying within a minute for months at a time. My uncle had a mechanical clock in the dash of his car that used time adjustment to change the speed of the clock. For example, if the clock is 10 minutes fast, setting it back 10 minutes automatically slows the mechanism by 10 minutes per day. The next day, if it's a minute slow, setting it ahead a minute, speeds it up by a minute a day. That's a level of accuracy we generally don't see in consumer-grade digital clocks.
Is it possible to design circuitry in a digital clock to respond to time correction by changing its speed? Or is digital circuitry generally too variable for greater accuracy to be unachievable? I'm not talking of course, about the more expensive digital clocks that are specifically designed for higher accuracy.
sure the "cheap mechanical clocks" aren't really digital ?, i.e. a crystal does the time keeping it is just the "display" that is mechanical
-
By mechanical clocks... do you mean analog? At first I thought this was in reference to pendulum clocks until you mentioned batteries.
I have seen digital clocks keep good time and some that keep poor time. The clocks in my cars have all been pretty rubbish. My presumption is that it is this way because consumers don't really care. Even the worst quartz timekeeper will keep better time than mechanical. But I have been annoyed by digital clocks in my home drifting apart.
One fun aspect of pendulum clocks is that they will sync up if all placed on the same wall. We have a wall where I work with clocks for different time-zones and my OCD kicks in when I see the seconds hands all tick at a different moments. If they were pendulum they would all be ticking at the same time. Although the seconds hand doesn't line up with the markers too so I'm not sure which one to be the most bothered by.
-
Personally, I haven't seem much change in the basic digital clock over the last 30 years. I have no doubt some types have evolved, but I'm talking about the basic consumer-grade digital clock, which seems mostly unchanged for decades. Most of these don't keep time all that well in my experience. Some of these clocks can drift tens of minutes over a year.
I have several cheap mechanical clocks, driven by a single AA battery, that are sold in dollar stores, that keep exceptional time, often staying within a minute for months at a time. My uncle had a mechanical clock in the dash of his car that used time adjustment to change the speed of the clock. For example, if the clock is 10 minutes fast, setting it back 10 minutes automatically slows the mechanism by 10 minutes per day. The next day, if it's a minute slow, setting it ahead a minute, speeds it up by a minute a day. That's a level of accuracy we generally don't see in consumer-grade digital clocks.
Is it possible to design circuitry in a digital clock to respond to time correction by changing its speed? Or is digital circuitry generally too variable for greater accuracy to be unachievable? I'm not talking of course, about the more expensive digital clocks that are specifically designed for higher accuracy.
sure the "cheap mechanical clocks" aren't really digital ?, i.e. a crystal does the time keeping it is just the "display" that is mechanical
Yep.
A 100% mechanical clock is pretty expensive to manufacture these days.
-
One fun aspect of pendulum clocks is that they will sync up if all placed on the same wall.
"Mode locking" happens in electronic circuits too, for analogous reasons.
-
Thanks, but I wasn't talking about mains-synced, internet-synced or radio-synced.
Mains synced is the worst unstable crap you can get. And those clocks which you complained about very likely are of this type.
Not sure I agree. Even before the dynamos were computer controlled, power stations used an odometer-like revolution-counter to keep track of the number of revolutions per hour or per day, and then they adjusted the speed of the dynamo until the number of revolutions were correct. Over the course of months or years, a mains-controlled clock would remain nearly 100% accurate.
Now I can't speak for your local power, or whether it uses DC power inverted to AC and whether it's accurate enough for a clock, but there are still a lot of mains-regulated clocks around and I'd be surprised if they've lost accuracy.
afaik they compare true time and "mains time" once a day and when they differ by more than 20 seconds they run slightly faster or slower for a day to get them back in alignment
-
sure the "cheap mechanical clocks" aren't really digital ?, i.e. a crystal does the time keeping it is just the "display" that is mechanical
You can see the gears and balance wheel running through the translucent plastic case. They are at least electrically driven, as they use a single AA battery. They've been around for decades and sell for around $5 in the shops. They must cost a lot less than that, because I've seen them in the clocks they sell at dollar stores. I'm sure you could buy dozens of them online for a few bucks.
-
I have several cheap mechanical clocks, driven by a single AA battery, that are sold in dollar stores, that keep exceptional time, often staying within a minute for months at a time.
As another poster mentioned, in English we don't call these mechanical clocks, we call them analog quartz clocks. It doesn't matter that they have a few gears between the motor and the hands. The timekeeping is electronic.
My uncle had a mechanical clock in the dash of his car that used time adjustment to change the speed of the clock. For example, if the clock is 10 minutes fast, setting it back 10 minutes automatically slows the mechanism by 10 minutes per day. The next day, if it's a minute slow, setting it ahead a minute, speeds it up by a minute a day.
That's a very unusual design. It would probably be simpler to make the clock more accurate in the first place.
Also, you obviously need a separate mechanism for DST/time zone changes.
Is it possible to design circuitry in a digital clock to respond to time correction by changing its speed?
Sure, but wouldn't it be better to have a separate speed adjustment?
Or is digital circuitry generally too variable for greater accuracy to be unachievable? I'm not talking of course, about the more expensive digital clocks that are specifically designed for higher accuracy.
Analog and digital clocks use exactly the same types of timebase. There is nothing about the digital display that makes the clock inherently less accurate.
-
Yep.
A 100% mechanical clock is pretty expensive to manufacture these days.
From scratch, I agree, but I've had many cheap mechanical clocks that kept amazingly good time. Some of them cost less than a dollar.
-
All those $1 "mechanical" clocks with batteries and big dial with 2-3 hands are digital clocks.
There is a crystal, chip and a coil, the chip includes an oscillator with divider, and the short 1Hz pulses drive the mechanical wheels through an electromagnet.
Mechanical watch without electronics inside with a precision of the DS3231 would cost you at least 350.000 USD.
-
I have seen digital clocks keep good time and some that keep poor time. The clocks in my cars have all been pretty rubbish.
This is because car clocks see a much greater temperature range than home clocks (which in turn see a greater range than a wristwatch). They really need a TCXO to keep good time.
-
I'd call it more of an "electro-mechanical" clock than "electronic,"
but it's just a judgment call.
https://www.youtube.com/watch?v=LGnbOOF1oeo (https://www.youtube.com/watch?v=LGnbOOF1oeo)
-
I have seen digital clocks keep good time and some that keep poor time. The clocks in my cars have all been pretty rubbish.
This is because car clocks see a much greater temperature range than home clocks (which in turn see a greater range than a wristwatch). They really need a TCXO to keep good time.
The car was rubbish to be sure, a Lada made in Russia, but the clock was a masterpiece of craftsmanship.
-
Not in the US.At any given moment, you might see the frequency not precisely 60Hz, but they do control it tightly enough so that over a day it averages to 60Hz. I've been inside one of the locla utility's control rooms. They take it pretty seriously.
They can take it seriously but mains clocks usually gain time due to transients and glitches unless they are very well filtered which most are not. This issue has been discussed here before.
Quartz clocks can be quite regular. For many years I was into celestial navigation and required precision to within one second.
In my experience even cheap watches can be quite stable. They will have a gain or loss rate which is quite constant. The gain or loss is not important as you just account for it. What is important is the stability of the frequency. They are naturally quite stable, and more so after some years of running, but you can improve if you keep it in a stable temperature enclosure.
A clock that gains 3.12 +/- 0.01 seconds/day is much better than one that gains 0.12 +/- 0.82 seconds/day because you account for the gain rate with more certainty.
For years I checked and kept logs of the rates of several timepieces I had. Interestingly one of the best was a pocket calculator.
I also had a microprocessor system where I checked the rate and corrected for it daily so it was always spot-on.
-
Not in the US.At any given moment, you might see the frequency not precisely 60Hz, but they do control it tightly enough so that over a day it averages to 60Hz. I've been inside one of the locla utility's control rooms. They take it pretty seriously.
They can take it seriously but mains clocks usually gain time due to transients and glitches unless they are very well filtered which most are not. This issue has been discussed here before.
Quartz clocks can be quite regular. For many years I was into celestial navigation and required precision to within one second.
In my experience even cheap watches can be quite stable. They will have a gain or loss rate which is quite constant. The gain or loss is not important as you just account for it. What is important is the stability of the frequency. They are naturally quite stable, and more so after some years of running, but you can improve if you keep it in a stable temperature enclosure.
A clock that gains 3.12 +/- 0.01 seconds/day is much better than one that gains 0.12 +/- 0.82 seconds/day because you account for the gain rate with more certainty.
For years I checked and kept logs of the rates of several timepieces I had. Interestingly one of the best was a pocket calculator.
I also had a microprocessor system where I checked the rate and corrected for it daily so it was always spot-on.
From Wikipedia: https://en.wikipedia.org/wiki/Utility_frequency (https://en.wikipedia.org/wiki/Utility_frequency)
"In practice, the exact frequency of the grid varies around the nominal frequency, reducing when the grid is heavily loaded, and speeding up when lightly loaded. However, most utilities will adjust the frequency of the grid over the course of the day to ensure a constant number of cycles occur. This is used by some clocks to accurately maintain their time."
-
Not in the US.At any given moment, you might see the frequency not precisely 60Hz, but they do control it tightly enough so that over a day it averages to 60Hz. I've been inside one of the locla utility's control rooms. They take it pretty seriously.
They can take it seriously but mains clocks usually gain time due to transients and glitches unless they are very well filtered which most are not. This issue has been discussed here before.
Maybe this is an issue with Spain's grid, but in the US, not really.
-
If you use TCXO and calibrate clock well, it can be quite precise. Even good regular calibrated quartz clock can be made pretty stable. Fer example good quartz watches like Swatch drift maybe a few seconds in month. Mechanical clock will suck no matter how expensive it is.
Those cheap electronic clock which drift a lot likely are not calibrated at factory (quartz) or simply use mains frequency as reference.
Quartz watches are effectively ovenized due to being strapped to a thermoregulated primate.
-
Maybe this is an issue with Spain's grid, but in the US, not really.
You did not bother searching, did you?
Saying mains power is always clean is absurd as there are lots of devices which will generate a lot of noise. There is a reason power conditioners are sold by the thousands.
As I said, I recall this being discussed here in another thread started by a poster who had that problem and it was well documented and I even recall posting a PDF with a study conducted by some utility company. I cannot find the thread now but Google gives plenty of hits. I posted a PDF which I believe was titled "Solving the Fast Clock Problem". I found it online but now I cannot find it. I probably have in saved in my computer somewhere. I seem to remember in one case it was a neighbor who had some kind of ozone machine and when it was disconnected the clock ran fine again and did not gain time. At any rate, it is a common and well documented problem.
https://www.electronicspoint.com/forums/threads/why-ac-digital-clocks-gain-time.250464/ (https://www.electronicspoint.com/forums/threads/why-ac-digital-clocks-gain-time.250464/)
https://www.electronicspoint.com/forums/threads/digital-clock-speeding-up-ie-gaining-time.134854/ (https://www.electronicspoint.com/forums/threads/digital-clock-speeding-up-ie-gaining-time.134854/)
https://www.electronics-tutorials.ws/power/transient-suppression.html (https://www.electronics-tutorials.ws/power/transient-suppression.html)
Maybe someone with better searching skills can find the thread here and/or the study by the utility that I posted.
E.T.A.: Found the PDF which I am attaching.
At Sierra Blanca, the West Texas engineers discovered random capacitive arcing across the faulty disconnect, which created the oscilla-tions in the power lines of a few rural customers. The oscillations causing their digital clocks to run fast were eliminated when the engineers properly closed the disconnect.
Based on the signature of the distur-bance confirmed by PEAC, knowledge of their substation equipment, and their experience, the Duke Power engineers were able to isolate the problem by inspecting circuits at the substation. They correctly deduced that the arcing was caused by a failing vacuum bottle in a capacitor bank breaker. When they located the problem vacuum bottle, it was blackened and close to failure. Service personnel replaced the vacuum bottle without disruption to Salisbury’s electrical service, avoiding the costly situation of the vacuum bottle failing and disrupting service to the customers.
In Wisconsin, an electric utility customer reported that his parents were plagued by a digital clock that ran at approximately twice the expected rate. The customer had taken a storage oscilloscope to his parents’ home and recorded voltage notches at the outlet feeding the clock. Because he saw the same voltage notches at the main breaker, he believed that they were coming from outside the house. He faxed the engineers at Wisconsin Electric a copy of the voltage wave-form from the main breaker, similar to the waveform shown in Figure 5. The notches in the waveform reminded the engineers of a waveform seen at the site of an industrial customer, where a six-pulse, AC-to-DC converter was causing voltage notching four times per second and ringing at about 10 kHz. The Wisconsin Electric engineers at first believed that the voltage notches at the customer’s house were probably caused by something in the house, but eliminated that possibility when they found out that the notched waveform had been recorded on the line side of the residence main breaker with the breaker off. Because there was no industrial facility fed from the same service transformer, the engineers eliminated industrial processes as a source of distortion.
The engineers recorded the same notched waveform on both phases of the service entrance. Moreover, similar waveforms were present on a neighbor’s meter socket. The notching was present at the secondary of the service transformer as well. However, another nearby transformer, which was connected to the same single-phase primary line, had an undistorted secondary voltage. The engineers concluded that the primary voltage was not to blame.
Because the voltage notches were at the meter even when the main breaker was open and similar impulses were at neighboring meters, the Wisconsin Electric personnel removed a neighbor’s meter as a test. The voltage notches at the clock vanished. By switching the neighbor’s breakers off one by one, the search for the distur-bance was narrowed to an electronic air ionizer.
Several customers of the Public Service Electric and Gas Company (PSE&G) in Newark, New Jersey, found that they had fast-running digital clocks and asked the utility for assis-tance. Field personnel discovered a disturbance on the supply voltage sine wave in every house with a fast clock. In the case of one customer, the source of the disturbance was traced to a neighbor’s electronic air ionizer. The utility obtained an identical ionizer and conducted some experiments, finding that operating it would cause some digital clocks to run fast. PSE&G asked PEAC to conduct system tests to understand and define the problem to encourage the manufacturer of that particular ionizer to modify it. PEAC researchers took a systems approach, not simply to test an individual ionizer but to look at a system that consists of a disturbing load, a sensitive load, and various utility system conditions. The test results confirmed that the high-voltage discharges of the ionizer were respon-sible for the bursts of impulsive transients in the supply voltage.
-
Quarts controlled clocks can be quite accurate, though you do need to have a very tight control on the oscillator supply voltage and the temperature, as they are going to cause the accuracy to change. For a quartz clock mechanism you will find around 1V4 is almost the ideal voltage, and is easy enough to achieve using a regulated power supply to drive the mechanism. I used a 12V supply and a red LED for this, driving the LED with a very low current, through around 6k8 of resistance, and 1000uF 16V capacitor to provide a buffer for when the clock ticked, as running at 10mA caused the clock to gain time, with supply voltage of 1V6.
If you use just the little oscillator board out of the clock, driving a signal transistor via a 10k base resistor from one of the outputs, and getting a 0.5Hz clock pulse out of it, you can get a cheap very stable oscillator, just use the red led to power it, and use the open collector pulled up to your logic levels as clock drive. That way you get a somewhat aged and stable crystal oscillator, and a low power driver that is designed to interface it. Of course the clock accuracy is a random factor, just grab a whole load of these cheap clocks at the cheap store, and put the cheap battery in them, set the time, and see which keeps best time. Set the oscillator battery voltage to the best one voltage and it will be the most accurate one.
Then divide by 30, and you have your minute drive logic level for whatever operation you want to use it for.
-
If you use TCXO and calibrate clock well, it can be quite precise. Even good regular calibrated quartz clock can be made pretty stable. Fer example good quartz watches like Swatch drift maybe a few seconds in month. Mechanical clock will suck no matter how expensive it is.
Those cheap electronic clock which drift a lot likely are not calibrated at factory (quartz) or simply use mains frequency as reference.
Quartz watches are effectively ovenized due to being strapped to a thermoregulated primate.
Even when you sleep at night? Not to say temperature is hardly stable even when on hand.
-
One fun aspect of pendulum clocks is that they will sync up if all placed on the same wall.
"Mode locking" happens in electronic circuits too, for analogous reasons.
And for a group of women living in the same house, anecdotally at least.
-
Even when you sleep at night? Not to say temperature is hardly stable even when on hand.
Research shows the variation in wrist temperature is in the neighbourhood of 1,5 degree Celsius over several days and 20 degrees Celsius ambient temperature swings, which isn't far from calibration lab tolerances. We don't need research to tell us this though. The detrimental effects of temperature swings on mechanical clocks have been understood by clock makers for centuries and they both realised and noticed watches being worn helps tremendously.
-
I was involved in a commercial project making an alarm clock. The electronic clock was driven from a watch 32.768KHz crystal. Surprisingly, they are not that accurate. There is something like 5% variation in the freq.
We had to build into the software a calibration to allow for this 5% variation.
We had to make the calibration quick to do, I think the target was under a minute.
What I conclude from your questions and my experience is that the cheap electronic clocks did not go through the calibration, they probably assumed that the 32.768KHz crystal is exacte and never bothered adding a calibration in the manufacturing. So every clock will gain/loose time randomly due to the crystal.
-
I was involved in a commercial project making an alarm clock. The electronic clock was driven from a watch 32.768KHz crystal. Surprisingly, they are not that accurate. There is something like 5% variation in the freq.
We had to build into the software a calibration to allow for this 5% variation.
We had to make the calibration quick to do, I think the target was under a minute.
What I conclude from your questions and my experience is that the cheap electronic clocks did not go through the calibration, they probably assumed that the 32.768KHz crystal is exacte and never bothered adding a calibration in the manufacturing. So every clock will gain/loose time randomly due to the crystal.
Crystals, like other components, come out with a certain tolerance range and the best ones are probably chosen to be sold more expensively while the others go for lower prices.
Also, they take some time to stabilize so if you need good stability and precision you can burn them in.
And, as has been said, stabilizing temperature and operating conditions will improve stability.
-
Decades ago, when I was starting in electronics, crystals were rare and very expensive. I somehow got a 5230 KHz crystal and it was worth building a divider by 5230 to get a 1KHz signal.
It was part of a frequency counter which I still have and use.
-
I was involved in a commercial project making an alarm clock. The electronic clock was driven from a watch 32.768KHz crystal. Surprisingly, they are not that accurate. There is something like 5% variation in the freq.
We had to build into the software a calibration to allow for this 5% variation.
We had to make the calibration quick to do, I think the target was under a minute.
What I conclude from your questions and my experience is that the cheap electronic clocks did not go through the calibration, they probably assumed that the 32.768KHz crystal is exacte and never bothered adding a calibration in the manufacturing. So every clock will gain/loose time randomly due to the crystal.
5% ? http://www.farnell.com/datasheets/2175894.pdf (http://www.farnell.com/datasheets/2175894.pdf)
-
I have a mechanical digital clock built into the oven in my kitchen. It uses a synchronous AC motor, getting its time base from the AC mains frequency. The motor drives a purely mechanical mechanism to display the time digitally in the style of an old automotive odometer. I believe the clock and stove are from the mid-1970s, but they came with the house, and were old when I got them.
The clock kept perfect time, within its resolution of one minute, except for the occasional power outage. Most years, I have only set it twice, when daylight saving time starts and ends.
Alas, this week, it stopped working entirely. I suspect it may be so dirty that some of the gears are jammed.
Asking if digital clocks are better than mechanical is asking if this clock is better than itself!
-
There are good and bad clocks, both digital and mechanical. In a museum once, I saw a clock that had a pendulum, and the weight at the bottom was in the form of a glass tube filled with mercury. As the ambient temperature changed, so did the length of the column of mercury, giving it temperature compensation.
-
They don't control it to average to exactly 60 Hz long term. They control it to be within spec.
This control system was my job/responsibility for years, and you are correct. Power plants in the US do NOT track frequency long term to ensure an average of X per day/month/year etc.
-
I have a mechanical digital clock built into the oven in my kitchen. It uses a synchronous AC motor, getting its time base from the AC mains frequency. The motor drives a purely mechanical mechanism to display the time digitally in the style of an old automotive odometer. I believe the clock and stove are from the mid-1970s, but they came with the house, and were old when I got them.
The clock kept perfect time, within its resolution of one minute, except for the occasional power outage. Most years, I have only set it twice, when daylight saving time starts and ends.
Alas, this week, it stopped working entirely. I suspect it may be so dirty that some of the gears are jammed.
Asking if digital clocks are better than mechanical is asking if this clock is better than itself!
With the old Telechron and similar clock motors it's common for the oil to gum up. You can pop the back off some or drill a small hole and inject some light machine oil to resurrect them.
-
They don't control it to average to exactly 60 Hz long term. They control it to be within spec.
This control system was my job/responsibility for years, and you are correct. Power plants in the US do NOT track frequency long term to ensure an average of X per day/month/year etc.
hmm, look like it is mandatory, maybe it is done centrally and the powerplants just follow
https://en.wikipedia.org/wiki/Utility_frequency#Time_error_correction_(TEC)
-
Mains clocks suck because if your clock doesn't filter well all the modern switching supply noise, dimmer switches and C3 buss signal noise on your mains wiring too well, if can accidentally count those 'spikes' making it run fast over the year.
In my kitchen, I have 2 mains clocks, the one on my microwave runs fine throughout the year, but, my clock radio one runs fast in the winter as the multiple power fluctuations throughout all my apartment's tenants has spikes in them as all they baseboard heaters go on and off throughout the winter season.
After our power utility replaced a major transformer which blew up last winter for my district, upgrading an entire section of our distribution, now both my kitchen clocks keep time matched, well, we will see this winter...
Trusting a mains clock assumes it is well enough designed to cope with uncertain power issues in you supply.
-
Technically, a quartz oscillator is still mechanical. Therefore, a "non mechanical" digital clock would probably be RC (very poor accuracy) or atomic (excellent accuracy).
-
I worked with a few power utilities and they do buy and sell power from each other. What little I know about connecting two AC power circuits is that they need be synchronized quite precisely, or else things can go boom-boom.
I'm quite sure that such things are computer-controlled nowadays, or else the AC power is rectified to DC and then inverted back to AC.
-
I worked with a few power utilities and they do buy and sell power from each other. What little I know about connecting two AC power circuits is that they need be synchronized quite precisely, or else things can go boom-boom.
I'm quite sure that such things are computer-controlled nowadays, or else the AC power is rectified to DC and then inverted back to AC.
Don't even need any advanced digital wizardry to synchronize two AC sources (although that helps), all you really need is a PLL.
-
If you want atomic level accuracy, use a GPS stabilized clock. Once these are locked to the satellites, they do not drift by more than few tens of nano seconds. The GPS modules are way too cheap these days. You could get one for $3 or lower. Mostly I2C controlled and you could directly get the UTC time from the receiver.
-
I have not seen this problem at all. I have probably 20 clocks in my house that are synchronized to the AC line and they keep excellent time long term, the ones I've built that display the seconds stay perfectly synchronized and I've never had to adjust any of them between DST changes. The power seems quite clean where I am.
-
I always wondered why digital clocks don't use a small quartz crystal, or do they actually use one and they're just not that accurate over time? (no pun intended)
-
The battery powered ones do, and the old Soviet clocks I have do. The vast majority of North America market digital clocks use the 60Hz line frequency, at least the LED ones do.
-
I always wondered why digital clocks don't use a small quartz crystal, or do they actually use one and they're just not that accurate over time? (no pun intended)
Price....
Believe it or not, a 32Khz tuning fork crystal costs more than the IC in those cheap digital clocks and it needs to be manually tuned to keep proper timer.
-
I always wondered why digital clocks don't use a small quartz crystal, or do they actually use one and they're just not that accurate over time? (no pun intended)
Price....
Believe it or not, a 32Khz tuning fork crystal costs more than the IC in those cheap digital clocks and it needs to be manually tuned to keep proper timer.
Wow I knew manufactures built stuff down to a price but didn't figure a sub dollar crystal was really something worth skimping on lol. I guess it adds up over many units.
-
I built a LED clock with the DS3231 a few years ago, added a supercap to ride power outages... it’s kept exceptional time since it was built.
I wanted a large digit bedroom clock that I could read without my glasses - perfect.
Also, I didn’t like the build or display quality of the retail units
Three buttons Set/Select, Up and Down. Auto dimming etc.
-
Not in the US.At any given moment, you might see the frequency not precisely 60Hz, but they do control it tightly enough so that over a day it averages to 60Hz. I've been inside one of the locla utility's control rooms. They take it pretty seriously.
They don't control it to average to exactly 60 Hz long term. They control it to be within spec.
Don't know about Latvia, but at least here, the point was that it's controlled to be exact in the long term - for the very specific reason of mains-synced clocks that exist. And indeed, I did note that the mains synced clocks in VCRs never required adjustments. The downside is, power cuts require setting the clock again. For our grid here, it was maybe once every 3-4 years on average. Quite reliable. If there are blackouts more frequenctly, these are pain-in-the-ass.
Last time I checked (maybe some 15 years ago?), the grid company even provided the equivalent "clock shift" value on the 'net.
It's never very precice, but IIRC, it never was more than about 30 seconds off. The error is fairly large and fluctuates between positive and negative, but is controlled so that it won't accumulate over a certain limit, and for tasks where you look at minutes, not seconds, it was/is accurate enough.
But I guess mains-synced clocks are becoming obsolete as of now.
-
5% ? http://www.farnell.com/datasheets/2175894.pdf (http://www.farnell.com/datasheets/2175894.pdf)
Yes, 20 PPM is pretty good for a cheap clock but it is almost a minute per month which is not acceptable for some applications. But, again, if it gains a minute per month at a very constant rate that is OK because you can just account for it and compensate. The problem is when you do not know how much it gained or lost. In this sense I find quartz clocks to be much more stable than mechanical clocks.
-
.
-
I always wondered why digital clocks don't use a small quartz crystal, or do they actually use one and they're just not that accurate over time? (no pun intended)
Price....
Believe it or not, a 32Khz tuning fork crystal costs more than the IC in those cheap digital clocks and it needs to be manually tuned to keep proper timer.
Hence this guy:
https://www.eevblog.com/forum/microcontrollers/padauk-pms15a-even-cheaper-mcu-(1-2-cents!)/msg2327466/#msg2327466 (https://www.eevblog.com/forum/microcontrollers/padauk-pms15a-even-cheaper-mcu-(1-2-cents!)/msg2327466/#msg2327466)
A 1.2 cents full MCU with flash, a silly digital clock IC with only a few hundred gates cant cost more. A crystal costs a whole lot more. Also, adjusting the trim-cap to tune the 32.786Khz exact costs time on labor and the price of a trimcap as well.
Just wiring a series cap/resistor to the AC mains is the cheapest and more accurate solution.
Wow I knew manufactures built stuff down to a price but didn't figure a sub dollar crystal was really something worth skimping on lol. I guess it adds up over many units.
-
This control system was my job/responsibility for years, and you are correct. Power plants in the US do NOT track frequency long term to ensure an average of X per day/month/year etc.
hmm, look like it is mandatory, maybe it is done centrally and the powerplants just follow
https://en.wikipedia.org/wiki/Utility_frequency#Time_error_correction_(TEC)
I guess further research required, but I'm 100% sure my plant did not count and attempt to adjust to the hour or day etc. I know this area of the control well because after the Northeast blackout of 2003 all of the control system engineers across the company had to present detailed summaries of how their units would respond to such events. As a side note, my plant in Indiana was about 1 sec from tripping off due to low frequency during that event, it was almost way worse than it was!
Even if done centrally (which didn't come up during months of frequency study across the entire company) it would take a huge amount of influence to push the entire country up or down. :-//
-
All those $1 "mechanical" clocks with batteries and big dial with 2-3 hands are digital clocks.
There is a crystal, chip and a coil, the chip includes an oscillator with divider, and the short 1Hz pulses drive the mechanical wheels through an electromagnet.
Mechanical watch without electronics inside with a precision of the DS3231 would cost you at least 350.000 USD.
The best timekeeper (not GPS/DCF77/WWWV... synced) I always had was a mechanical wristwatch. During its first 2 years it always stayed within +/- 12 s of correct time. And, if we do not count on its first 2 weeks, where the mechanism broke-in, it stayed within +/- 5 s of correct time for almost 2 years. This is better that the guaranteed +/- 2 ppm of the ds3231 (although I never used such IC, so I do not know how it really behaves, maybe it works far better that claimed).
A detailed discussion on this special topic would be rather boring for most posters here - but most pure_mechanical_wristwatch_owners claim usually an unprecedented precision, including myself, that is normal :)
The experts would tell you some of those watches are COSC certified, that is all, the actual precision is matter of luck.
A pure mechanical clock/watch which does less than 20-30secs a year "always", without a need to messing with it every 2-3years, or changing your physical activity in order to slow-down or speed-up them twice a week, would cost you at least the number I mentioned above, you bet :)
-
Digital clocks may be better, but that doesn't mean the people who use them are any happier.
-
Just a simple figure - 30s drift over one year is approx. 0.95ppm, which is already an exceptional stability for a very low-power oscillator.
-
I always wondered why digital clocks don't use a small quartz crystal, or do they actually use one and they're just not that accurate over time? (no pun intended)
Price....
Believe it or not, a 32Khz tuning fork crystal costs more than the IC in those cheap digital clocks and it needs to be manually tuned to keep proper timer.
Wow I knew manufactures built stuff down to a price but didn't figure a sub dollar crystal was really something worth skimping on lol. I guess it adds up over many units.
Significant effort is spent shaving pennies and even fractions of pennies from the BOM of mass production items. When you are building millions of something it adds up quickly.
-
I built a LED clock with the DS3231 a few years ago, added a supercap to ride power outages... it’s kept exceptional time since it was built.
I wanted a large digit bedroom clock that I could read without my glasses - perfect.
Also, I didn’t like the build or display quality of the retail units
Three buttons Set/Select, Up and Down. Auto dimming etc.
This, I need to do. The only thing that saves me is the inaccurate thing I use now is so small I can just pick it up and hold it right up to my face to read without my glasses. Have to hit the button to light up the backlight to see the LCD in the dark anyway. Would be nice to just open my eyes and see the time without touching anything. Part of why I have just learned to live with it being so far off is the every time I think about it, I figure I can just build a decent clock instead of spending money to buy one of perhaps questionable quality. And then I never do go about so much as designing one.... something about the cobbler's family always having the worst shoes. Ultimately, I would never change the arrangement in the bedroom, and there's be a nice light I could activate that would light my path to the bathroom so I don't step on a sleeping dog, dog toy, or worse. The furniture arrangement however is beyond my control.
-
I always wondered why digital clocks don't use a small quartz crystal, or do they actually use one and they're just not that accurate over time? (no pun intended)
Price....
Believe it or not, a 32Khz tuning fork crystal costs more than the IC in those cheap digital clocks and it needs to be manually tuned to keep proper timer.
Wow I knew manufactures built stuff down to a price but didn't figure a sub dollar crystal was really something worth skimping on lol. I guess it adds up over many units.
Really poor, noise-sensitive implementations notwithstanding, mains-frequency clocks are more accurate in the long run than consumer-grade quartz clocks. So spending the 30 cents on a quartz crystal would actually be spending money to get an inferior result.
And as it is, many clocks actually use both: line frequency when available, and quartz when running off battery backup. (Remember when alarm clocks would forget the time when they lost power, causing you to not wake up? I remember when I was a kid and battery-backed clocks began to become common, how happy I was to get one...)
-
To be fair, I haven't tried the cheap mechanical watches that sell at discount stores for about $10, and sometimes go on clearance for a dollar, but I wouldn't expect accuracy better than a cheap digital clock, so it may have seemed like I was comparing apples to oranges.
But really, I was wondering why a self-learning feature couldn't have been part of every digital clock, regardless of quality. What I mean is the ability of the circuit to alter the rate of the clock, based on the time adjustments made by the user. For example, if the clock lost 5 minutes in a week, the user would advance the time by 5 minutes after a week, and the clock should be able to speed up the rate by 5 minutes per week. After a few adjustments, it should be keeping excellent time.
This would not be the case however, if the circuitry itself were inherently inconsistent and unable to run at a stable rate, once adjusted. I'm not sure if anyone has answered this yet.
-
.. This would not be the case however, if the circuitry itself were inherently inconsistent and unable to run at a stable rate, once adjusted. I'm not sure if anyone has answered this yet.
I answered that to you (see above). You may do any voodoo with your digital clock settings, the "stable rate" depends on many factors which change all the time. So your AI clock has to learn how to respond to all factors I mentioned above.
Just a simple figure - 30s drift over one year is approx. 0.95ppm, which is already an exceptional stability for a very low-power oscillator.
Long time back I bought a few of DS3231SN (mind "S/SN") cheapo modules off ebay. One of them I observed long time does +20secs/year (+61secs in 3years). The temperature of the junkbox was 20-29C based on the season. There is an "aging" register you may finetune the frequency of the internal 32kHz crystal in 0.1ppm step. The crystal aging 1ppm first year, 5ppm 10years. From what I saw on the web that <1ppm accuracy is typical in ambient temperature range.
Fantastic RTC, the best one low power RTC ever produced :-+
Now back to the "S/SN". The production of DS3231S/SN with the crystal inside has been finished several years back. They sell DS3231M with MEMS oscillator instead, with lower spec.
-
Now back to the "S/SN". The production of DS3231S/SN with the crystal inside has been finished several years back. They sell DS3231M with MEMS oscillator instead, with lower spec.
I've heard several people say this but I don't think it's true. Maxim does list the SN variants as "active" on their website: https://www.maximintegrated.com/en/products/analog/real-time-clocks/DS3231.html/tb_tab3 (https://www.maximintegrated.com/en/products/analog/real-time-clocks/DS3231.html/tb_tab3)
Active
The product is in production and available from Maxim Direct and authorized distributors.
-
But really, I was wondering why a self-learning feature couldn't have been part of every digital clock, regardless of quality. What I mean is the ability of the circuit to alter the rate of the clock, based on the time adjustments made by the user. For example, if the clock lost 5 minutes in a week, the user would advance the time by 5 minutes after a week, and the clock should be able to speed up the rate by 5 minutes per week. After a few adjustments, it should be keeping excellent time.
I guess the real reason is fairly obvious. Although it may add nothing to the cost of a cheap electronic clock circuit to incorporate a learning circuit which self-adjusts the rate, there is simply no financial incentive to do so. The unit cost of these chips has dwindled to a few cents apiece, and tend to be "good enough" for most of the lower-end consumer electronics that use them. It would take an industry-wide commitment to upgrade to better clock chips that don't start your coffee maker 5 minutes later each month requiring you to constantly reset it. Even my car, which cost over $40,000 uses a cheap clock chip that goes out of time by at least 5 minutes every time I check it.
If it ever becomes a real issue, owing to our expectations of perfect time, such as that we get from our smartphones and consequently smart-watches, I doubt the solution will be a self-learning rate-adjustment circuit. It will probably come in the form of a more expensive option, using a higher quality crystal oscillator instead. Either that or a wireless connection to an external time source.
-
I've heard several people say this but I don't think it's true. Maxim does list the SN variants as "active" on their website: https://www.maximintegrated.com/en/products/analog/real-time-clocks/DS3231.html/tb_tab3 (https://www.maximintegrated.com/en/products/analog/real-time-clocks/DS3231.html/tb_tab3)
https://www.maximintegrated.com/en/products/nla/index.mvp/view/EOL/pg/15 (https://www.maximintegrated.com/en/products/nla/index.mvp/view/EOL/pg/15)
-
I've heard several people say this but I don't think it's true. Maxim does list the SN variants as "active" on their website: https://www.maximintegrated.com/en/products/analog/real-time-clocks/DS3231.html/tb_tab3 (https://www.maximintegrated.com/en/products/analog/real-time-clocks/DS3231.html/tb_tab3)
https://www.maximintegrated.com/en/products/nla/index.mvp/view/EOL/pg/15 (https://www.maximintegrated.com/en/products/nla/index.mvp/view/EOL/pg/15)
Ohh I see, DS3231SN is not available anymore. But DS3231SN# is! Why the hashtag? From the datasheet:
#Denotes an RoHS-compliant device that may include lead
(Pb) that is exempt under RoHS requirements. The lead finish
is JESD97 category e3, and is compatible with both lead-based
and lead-free soldering processes. A “#” anywhere on the top
mark denotes an RoHS-compliant device.
Mystery solved!
-
That would be a fantastic message!
Some time back there were discussions it is not available anymore, it seems to me at that time I had not seen the # in production. It could be, based on many letters Maxim's CEO had received from the folk, they put it in an active status again.. :)
-
.. Even my car, which cost over $40,000 uses a cheap clock chip that goes out of time by at least 5 minutes every time I check it.
It could be the 1min/month clock is an option N.345 for $999.99 you simply forgot to put on your car's configuration list :)
PS: I've seen somewhere this watch is going to be produced as mechanical clock version for your car, option N.346 :)
https://www.youtube.com/watch?v=UDx_d2Xln3M (https://www.youtube.com/watch?v=UDx_d2Xln3M)
-
The older style mains synchronized clocks, the ones that were based on a synchronous motor, had major advantages over newer clocks of all sorts. First, they weren't sensitive to the high frequency noise that seems to get some of the newer clocks and thus were quite accurate over the long term. And second, when the power went out they gave a measure of how long it had been off. When you returned home at the end of the day and found power off you knew when it went off. With today's clocks all you know is that you need to reset them.
The mechanism to adjust clock rate based on the adjustment is actually quite simple on the older mechanical movements. All that was done was to move the tie down point of the hairspring that controlled the mechanical oscillator. The big problem is that most people don't reset on any kind of precise interval, so the correction factor was at best a guess (was that ten minute in a day or a week). By reducing the "gain" of the adjustment (during the design phase) you could get a fairly good likelihood of eventually converging on a good value.
-
The older style mains synchronized clocks, the ones that were based on a synchronous motor, had major advantages over newer clocks of all sorts. First, they weren't sensitive to the high frequency noise that seems to get some of the newer clocks and thus were quite accurate over the long term. And second, when the power went out they gave a measure of how long it had been off. When you returned home at the end of the day and found power off you knew when it went off. With today's clocks all you know is that you need to reset them.
that could also be problem if you don't notice and now you are an hour late
-
The mechanism to adjust clock rate based on the adjustment is actually quite simple on the older mechanical movements. All that was done was to move the tie down point of the hairspring that controlled the mechanical oscillator.
I always liked the idea of shining a laser through the wheel connected to the hairspring and have it saving a counter value that is driven from a GPSDO. Then have some form of servo that gradually moves the hairspring back and forth to settle on the right point so that the average counter value is dead on.
-
I always liked the idea of shining a laser through the wheel connected to the hairspring and have it saving a counter value that is driven from a GPSDO. Then have some form of servo that gradually moves the hairspring back and forth to settle on the right point so that the average counter value is dead on.
Photo sensors sensing the gears position are "common" today:
https://www.seikowatches.com/global-en/customerservice/knowledge/kinetic-perpetual-knowledge (https://www.seikowatches.com/global-en/customerservice/knowledge/kinetic-perpetual-knowledge)
GPSDO and mechanical watch :palm:
-
that could also be problem if you don't notice and now you are an hour late
I have been late to work more than once because of this. They cut off the power in the middle of the night to do work because that is when they cause least disruption. Then the alarm is late by the same amount of time the power was off.
-
that could also be problem if you don't notice and now you are an hour late
I have been late to work more than once because of this. They cut off the power in the middle of the night to do work because that is when they cause least disruption. Then the alarm is late by the same amount of time the power was off.
I've been using my smartphone as a wake-up alarm for several years now.
I never have to worry about power outages.
I can still remember using a mechanical alarm clock. That never let me down either.
-
A cheap mechanical clock powered by a battery, will in theory be as good or as bad as a digital clock, powered by a battery, because it will also use a quartz crystal oscillator to keep the time. Some really old battery powered clocks did use a flywheel with magnets and a coil, driven by a transistor oscillator, but any clock made in the last 30 or so years will be quartz, because it's cheaper.
-
The first mechanical watch with electronics inside was Bulova Accutron, 1960.
It had a tuning fork with coils inside, and a transistor creating the oscillator.
No crystal and no chips inside..
https://www.youtube.com/watch?v=UILl6PFbOzA (https://www.youtube.com/watch?v=UILl6PFbOzA)
-
'Atomic' Clocks, are basically 'Chrystal' regulated clocks, but with a periodic
correction factored in due to another element.
-
I've been using my smartphone as a wake-up alarm for several years now.
I never have to worry about power outages.
I can still remember using a mechanical alarm clock. That never let me down either.
Now you have to worry about software bugs causing skipped or silenced alarms. :)
This has happened multiple times on Apple’s iOS.
-
Now you have to worry about software bugs causing skipped or silenced alarms. :)
This has happened multiple times on Apple’s iOS.
This exact thing has happened to my wife several times on her iphone.
-
Now you have to worry about software bugs causing skipped or silenced alarms. :)
This has happened multiple times on Apple’s iOS.
This exact thing has happened to my wife several times on her iphone.
I can't say its ever happened to me with Android.
-
Now you have to worry about software bugs causing skipped or silenced alarms. :)
This has happened multiple times on Apple’s iOS.
This exact thing has happened to my wife several times on her iphone.
I can't say its ever happened to me with Android.
The only time I have ever had a missed alarm on my iPhone was when I accidentally set the alarm to 7:00 PM instead of 7:00 AM.
I guess that is an advantage of 24 hour time...
Otherwise I have been using my phone as an alarm clock for years with perfect reliability.
-
Both Android and iOS haven't been reliable. Usable, but far from perfect. They're consumer grade computers with associated reliability and it pays to plan accordingly. Other types of alarm clock have proven to be more reliable.