On your board there's going to be a 32.768KHz crystal for the real time clock. It's definitely possible for the frequency to drift as they get old. Or if they get mechanically damaged by something like a shock or ultrasonic cleaner. The tuning fork kind of crystal (the cylindrical type) are especially sensitive to this.

If you have a scope, you can check the frequency. Use your probe/scope in x10 mode so you don't load the crystal.

They're super cheap and common. You can always just swap it out and see what happens.

Appliances often use the zero crossing point of your AC mains power as a timer. It's cheaper than a dedicated RTC IC/circuit. Which is not as accurate as a crystal oscillator. So it's common for appliances to drift.

In general computer clocks have never ever been particularly good, not in the 80s (when they were competing with people bothering, or maybe not, to input it correctly on boot) and certainly not today (the internet + NTP age) 😀

Quartzes (?) also change in frequency with temperature, some more some less but the serious clocks use active heating to keep it constant; then you have a capacitor in series with it (that may be integrated in the RTC) which could have shifted; then you have different chips with or without temperature compensation (look at the Raspberry accessories market for a wide range of current Dallas I2C products), which in any case is not measured inside the quartz, so… 😀

Windows ME For Dummies wrote:

Most computers nowadays have an internal clock […] That said, they are not very reliable, especially on laptops with "power saving". Do compare, every now and then, your 20 € wristwatch and your 2000 € computer to ensure time is advancing correctly!

The power line frequency (probably last used in a mainstream computer for the Amiga, and then not for the main wall-clock and only because it recycled a C64 chip) is indeed generally better: http://www.repairfaq.org/sam/audiofaq.htm#audwm2lc

Ryccardo wrote on 2024-03-04, 11:44:

In general computer clocks have never ever been particularly good, not in the 80s (when they were competing with people botherin […]
Show full quote

In general computer clocks have never ever been particularly good, not in the 80s (when they were competing with people bothering, or maybe not, to input it correctly on boot) and certainly not today (the internet + NTP age) 😀

Quartzes (?) also change in frequency with temperature, some more some less but the serious clocks use active heating to keep it constant; then you have a capacitor in series with it (that may be integrated in the RTC) which could have shifted; then you have different chips with or without temperature compensation (look at the Raspberry accessories market for a wide range of current Dallas I2C products), which in any case is not measured inside the quartz, so… 😀

Windows ME For Dummies wrote:

Most computers nowadays have an internal clock […] That said, they are not very reliable, especially on laptops with "power saving". Do compare, every now and then, your 20 € wristwatch and your 2000 € computer to ensure time is advancing correctly!

The power line frequency (probably last used in a mainstream computer for the Amiga, and then not for the main wall-clock and only because it recycled a C64 chip) is indeed generally better: http://www.repairfaq.org/sam/audiofaq.htm#audwm2lc

A $20 wristwatch also uses a crystal. No idea what you are going on about.

As kingcake says its probably just a bad crystal, Necroware just had a repair video where the crystal on his motherboard was up to like 44KHz and the clock was off. He replaced it and it kept time fine it looked like.

The crystals have external load capacitors in a typical RTC oscillator and they have to be tuned exactly right for the time to pass at a correct speed, of course at a given temperature. When I built a clock, using crystal from a PC mobo, it took some effort to tune it just right so that over the course of weeks there was only a second worth of offnes. I have seen such trimming capability only on one PC mobo... so in the end there will be some amount of clock offness due to not exact tuning and probably small amount of drift too from any temperature and/or aging effects.

Tiido wrote on 2024-03-04, 16:46:

The crystals have external load capacitors in a typical RTC oscillator and they have to be tuned exactly right for the time to pass at a correct speed, of course at a given temperature. When I built a clock, using crystal from a PC mobo, it took some effort to tune it just right so that over the course of weeks there was only a second worth of offnes. I have seen such trimming capability only on one PC mobo... so in the end there will be some amount of clock offness due to not exact tuning and probably small amount of drift too from any temperature and/or aging effects.

There's no need to trim the load caps. The motherboard is designed for the correct caps in the first place. This is well documented in the chipset datasheets. No need to experiment here. Replace the crystal.

Wristwatches can usually rely on being stabilised near body temperature, cheap quartz wall clocks usually stay near room temperatures. I have had wristwatches go plus or minus a noticeable amount when out in subzero winter temperatures, or sitting with my arm in the sun in hot summers. Also I had a wallclock in my garage, and figured out after a couple of years that even if I put the duracell pro alkaline in it, it was still only gonna be a plus or minus 5 minute guess at the time because the crystal wasn't temperature stabilised and it would get up to 50C maybe with the doors closed in summer and hit -20c in some parts of winter.

Also what is gonna make a difference is where on the board your RTC is, if it is up between the first PCI/ISA slot and the CPU, even with airflow it's not gonna cope too well with the swings. If it's down on bottom right of board, where the speaker and front panel connectors usually are, that's a relatively cooler = less swing spot.

Back in the day before ubiquitous internet connectivity and NTP correction of the clock frequently, there were numerous ways to get time corrected, or continuous "right time" some appliances also used them. There were time signals in some FM broadcasts that can keep radio alarms honest, there used to be some in PBS TV broadcasts that some VCRs set themselves by (A royal PITA post digital transition because many of them never had a manual timeset option). In europe there were teletext services going out over broadcast, and it was more simple to grab the timesignal out of those than fully decode the text so there was a couple of computer cards that just did that.... Then there were weather service and navigational broadcasts on VHF channels that had a time signal. Then there was/still is the atomic clock time signals which go out on a low frequency broadcast. Later in the 90s GPS dongles just started getting attention as a time source. There were multiple small production boards for catching time signals from these sources for various commercial/scientific needs. Some projects in the likes of Elektor and Electronics Today etc as well.

Edit: Maplin in the UK used to do an atomic clock receiver that connected to a serial port. UniversalSolder.ca is a modern source for atomic clock receivers but not sure if any of theirs connect to a PC like the Maplin one would, but it's probably do-able.

EditII: Heh, or you can get your very own atomic clock not just a receiver of the signal from elsewhere https://shop.stepglobal.com/XHTF1021/Compact- … dash_232/pd.php

kingcake wrote on 2024-03-04, 18:15:

There's no need to trim the load caps. The motherboard is designed for the correct caps in the first place. This is well documented in the chipset datasheets. No need to experiment here. Replace the crystal.

That really depends on how accurate is accurate. What I was talking about is a difference between 2pF ws 2.1pF, there are ~32768 oscillations a second in a typical RTC and any deviation, hovever slight, will add up over a timescale. The things you refer to are very much just a ballpark stuff and absolutely nothing with exactness, physics do get in the way of that. Of course if the clock goes wildly off in a day or week something is actually wrong and one should start seeing about replacing parts. But assuming 10ppm quartz with low tolerance C0G capacitors and ignoring everything else you're still going to be seconds off over the course of some months without tuning, the tolerances are big enough even though you're using essentially premium parts. Really cheap parts may have such a bad tempcoeff that there's no point trying to tune anything, it'll go off as soon as there's any temperature change, even one from just the presence of one's fingers near the tuning capacitor 🤣 (I very much speak from hands on experience on this matter).

Thanks for the replies. It's not something I'll likely mess with in the near term, but I think I know where the oscillator is that you're talking about. It's a BX6r2 and I seem to recall a cylindrical oscillator near the battery. Maybe someday I'll buy a cheap scope and test it.

A proper NTP daemon will slow or speed the system clock to allow for any drift in the hardware.

I think what was called SNTP only does periodic time sets, with I presume time jumps when it does.

No doubt there is a Windows version of ntp.org's NTP daemon, or OpenNTP from OpenBSD, or another implementation, and that could well allow for an effective software work-around for the aging hardware.