On Thu, Mar 8, 2018 at 9:16 PM, Barry Roberts <b...@robertsr.us> wrote: > It's not Linux-related, but I'm curious what other nerds think. I was > listening to KUER on the way home from work tonight, and they were > talking about this: > https://www.theverge.com/2018/3/8/17095440/europe-clocks-running-slow-electricity-frequency-kosovo-serbia > > I don't know if I've ever owned a clock that would act like that. > Maybe some really old clock with an AC motor. But they talk about the > clock on a microwave, and show a digital clock on a stove. Those all > have crystal oscillators that run on DC and I'm pretty sure would not > be affected at all by a .1% change in the AC frequency (that number > was from the radio --.1% or 344 seconds in about 6 months, I think). > > Am I crazy? Are there really a bunch of clocks in Europe that > actually depend on the 50hz alternating current frequency for > timekeeping? The whole thing sounds like some non-technical reporter > grossly misunderstanding electronics to me. > > Barry
All microcontrollers have some kind of oscillator for their high-frequency clock, but they don't all have a crystal oscillator. Low-cost ones might have an internal RC oscillator (i.e. one that oscillates based on a resistor/capacitor time constant) that typically isn't nearly as accurate as a crystal oscillator and is sensitive to voltage level changes, but gets the job done for simple control tasks like monitoring a keypad and turning on relays. It *is* very common to put a quartz watch crystal (32768Hz resonant frequency) in microcontroller devices, but those are typically only accurate to 20 or 50 parts-per-million. That sounds good, but small errors accumulate pretty quickly when you're oscillating at tens of thousands of cycles per second. With a cheap one, it's easy to accumulate minutes of error per month. Wristwatches typically use some sort of calibration/compensation mechanism to get things to around 5ppm, plus they stay pretty temperature-stable thanks to being worn next to a human body, which has a pretty nice internal temperature stabilization mechanism. That kind of compensation isn't typically built into microcontrollers, though, which very often don't really care about long-term timing accuracy. On the other hand, if your device is plugged into the wall, you have a nice 60Hz (or 50Hz in Europe) signal with which to count the passing of time. This varies a bit during the day, but it's used for timekeeping enough that they typically try to re-synchronize it during the evening. Thus, you can get more accurate long-term timekeeping by synchronizing to the mains frequency rather than relying on a high-frequency crystal oscillator, and it's cheaper too! This is an interesting page that attempts to measure the stability of the mains 60Hz signal: http://www.leapsecond.com/pages/mains/ I haven't examined the control circuitry of any digital microwaves or ovens lately, but I did look at some microcontroller datasheets; many explicitly had RTC modules that described setup procedures for using either a 32768Hz crystal or a 50/60Hz mains zero-crossing timebase. So it must be a reasonably common thing to do, especially in cost-optimized devices that are always plugged in. In countries that have had their mains frequency regulated like this for timekeeping purposes, there are likely to be a lot of devices that do in fact rely on it for timekeeping, even if they're digital devices that largely run on DC. Getting a pulse at the zero crossing point of the mains power is not a difficult bit of circuitry, and is most likely cheaper and has better long-term timekeeping potential than a crystal if you don't already need a crystal for some other reason. --Levi /* PLUG: http://plug.org, #utah on irc.freenode.net Unsubscribe: http://plug.org/mailman/options/plug Don't fear the penguin. */
