Hacker News Reader

Comment by codesnik | Hacker News Reader

using AC frequency for timekeeping seems ridiculous. Does quartz generator cost anything?

Depends if you're making an analogue clock or a digital one.

If you look inside a cheap mechanical timeswitch [1], you'll find a synchronous motor the size of a sugar cube, turning 50 or 60 times a second, then half a dozen plastic gears reducing the rotation to 1 turn per day. AC wall clocks follow the same principle, but with a slightly different gear ratio.

This means you don't need a printed circuit board, or any integrated circuits, or anything like that. Such efficiency is how you make a profit, if you're selling timers on Aliexpress for $3.50.

If you're making a digital timeswitch, a quartz crystal might be cheaper, or might not - it depends on the insulation/isolation you think your product requires.

[1] http://swarfer.co.za/rc/03-timer.php

You'd expect even a fairly good quartz to deviate by more than a minute a year from temperature fluctuations. Relying on AC frequency should allow your clock to operate for many years without gaining or losing time as long as the AC frequency is compensated, which it needs to be or someone is stealing energy.

In ordinary circumstances, the grid frequency is far more accurate than a cheap quartz crystal oscillator.

What for, if you already have a clock signal that has been reliable until now?

Because its reliability is happenstance.

No it's standardized. Outside of accidents the power providers take extreme care to keep the frequency synchronized to UTC. Exactly.

A clock using the power grid as reference might be off by 4 minutes over the day but on average it will be perfectly on time.

It used to be in the US that some operators would have a plain off-the-shelf alarm clock in the control room, and, in case the evening load made the frequency dip, operators would adjust things so that the clock would read 6am (for example) exactly the next day.

Nope. There's man-years of effort per day in keeping it reliable; as evidenced by the point that it takes major political meddling over a prolonged period to throw it off, and an article on HN for people to even notice "it has been acting sort of weird, now that you mention it".

"I didn't notice until now" is quite the opposite of happenstance, on par with the PHB mantra "what do we need devops for, anyway, the network seems to be up!" Well it's up because devops is keeping it up. Likewise, the grid doesn't just magically self-balance around 50 Hz.

I mean, in software engineering we try to avoid unnecessary dependencies right? This is exactly what's happening, why would those clocks ever depend on the power grid??

EDIT: Nevermind, sounds like there are precision concerns that justify using the power grid for timekeeping.

Indeed. There's a tradeoff in both directions - either you trust the cloud^W grid, or you trust your local clock, and both of them have failure modes; there's no one clearly superior option here. You could combine them, but then it gets complex in short order.

> I mean, in software engineering we try to avoid unnecessary dependencies right?

I/you wish.

> in software engineering we try to avoid unnecessary dependencies

leftpad would like to have a word

Putting Lego blocks together is engineering now?

Edit: NPM has engineering challenges, and Node as well...but I would be extremely hesitant to call a library of ~ten LoC "engineering."

i'm not a fan of what the JS ecosystem looks like, but you can and will get bitten by transitive dependencies if you don't store them all in a local mirror that doesn't respect upstream deletions...

Of course. But that's an engineering problem worthy of the name (cache-invalidation and whatnot).

I definitely noticed this in the last few weeks already, since my alarm clock is set to switch the radio on in the half minute before the news on the top of the hour. I was already searching for reasons, and I would have expected an announcement about this a lot sooner.

Grids ensure the cycle count is correct over the day. There will be some drift from spikes and peak loads, but they adjust the frequency to compensate. They have atomic clocks, for the EU grid it's in Switzerland, to track the cycle count.

UK and EU grids have websites letting you look at frequency history.

There's still an awful lot of infrastructure keyed off having an accurate synchronous time, including billing systems, traffic lights, street lights...

I think it is really old and cheap clocks (such as old radio alarm clocks) which use AC frequency for the clock signal. Their power consumption on AC is also rediculous compared to what they do (displaying the time).

I think the only clock I have that does it that way is the one in the stove/oven as I recently noticed the deviation after having set it correctly just a few weeks ago.

The real challenge is to have all clocks within a building display the same time. Getting them powered and synchronized through the electric grid is a straightforward solution.

No but things like heating systems are presumably designed for 20 year + lifespans, so there is probably a large number of legacy systems out there and will be for some time.

AC frequency makes the entire process cheaper.

You can use a resistor to lower the voltage and plug in a stepper motor and there is your clock.

When the idea was initially proposed quartz generators were also a bit more expensive so using the grid to synchronize was not a bad idea at all.

Even today it's not the worst idea since it's being kept very close to 50hz. Close enough that you can safely operate an alarm clock from it without being off to far by the end of the day.

You wouldn't use a resistor to lower the voltage, because then you have most of the power being dissipated in the resistor rather than the stepper motor. Much easier to just use a transformer or a 240V-rated stepper motor.

Well, normally you wouldn't draw power over that resistor, I was thinking more in the direction of using the power line as a "sense" pin for the stepper and power was provided properly via a transformer.

Before being relegated to cheapest chinese stuff, elecromechanical timers were the norm.

There's probably thousands of electrical panels with old DIN rail timers that don't get replaced until they die.

I was thinking the exact same thing. I've always assumed that AC frequency wasn't exactly 50Hz and was allowed to fluctuate slightly.

It's supposed to average 50Hz over time, so using it for clocks make perfect sense.

That assumption is right. But until now it was always compensated, so that in the mean it would be ~50hz. This correction mechanism apparently isn't enough anymore for the current fluctuations in europe.

Not mean, average. That's a very important distinction because if it was mean, then it might never correct, but it does.

It can be allowed to fluctuate, but still regulated to keep pretty much exactly 50Hz mean value over time.

Reply on news.ycombinator.com