Hacker News Reader

So many devices rely on the timing of the waves that the generating companies actually keep count and then adjust the speed to make up any they might have missed!

That's not at all why grid timing is so precise.

Multiple generation systems feed into the same grid. If they're out of phase by even a few milliseconds, that translates to huge power losses. If power generation companies didn't synchronize to high precision, they would lose a lot of power and probably break some equipment.

I have a textbook somewhere that said the adjustments were due to so many things relying on it for syntonisation. I'm not in a position to dig it up and find a proper reference unfortunately.

I've heard otherwise elsewhere: that grid frequency was initially imprecise, was made precise to enable synchronous electric clocks, and then grid interconnection became feasible. https://en.wikipedia.org/wiki/Telechron#Henry_Warren:_the_Sy... , for instance.

I can't immediately find a cite for the story that some enterprising electric clock company would give clocks to execs at power companies, then let them get annoyed at large errors and demand stable frequency, but that's surely too good to need a cite, eh?

(edit: initially imprecise)

Grid interconnection has always been possible, whether it is precise or not, however if one grid was running faster than the other doing such an interconnect would cause the grids to eventually stabilise somewhere in the middle as the faster grid caused the slower grid to speed up and the slower grid caused the faster grid to slow down.

The amount of power wasted is insane though. Bringing new turbines online for example means getting the turbine to run in sync with the current clock cycle within a certain percentage otherwise massive damage can be done as the turbine jumps backwards or forwards when connected to the grid.

Equipment being out of phase has nothing to do with the timing. The timing of the 50/60 Hz signal actually changes throughout the day. On average the signal over 24 hours will equal about 24 hours so you can keep time with it, but as the load changes on the grid it will actually slow down and then speed back up when people turn off their devices at night.

The phases of the grid as a whole actually get synched up automatically. For example when they spin up a new turbine at a power plant if it is not within a certain percentage of the phase cycle it can actually cause massive amounts of damage as it gets pulled into the phase automatically.

The electric grid companies actually want to lower the current requirements for the 50/60 Hz cycle to be accurate in a 24 hour period to longer, such as possibly a week. This could mean that the clock in your oven for example may be a minute or two slow and slowly catch up over time as the grid has moments where it is clocked faster than 50/60 Hz.

For example see this graph of the last 60 minutes: http://www.nationalgrid.com/uk/Electricity/Data/Realtime/Fre... and here it is in real time for the whole US: http://fnetpublic.utk.edu/gradientmap.html

And here is the experiment that the power companies wanted to run/still are running regarding the relaxing of keeping the grid at 60 Hz: http://www.nbcnews.com/id/43532031/ns/technology_and_science...

> as the load changes on the grid it will actually slow down and then speed back up when people turn off their devices at night.

When I sailed on ships we knew not to use alarm clocks that plug into an outlet, because the load on the ship's generators is always changing, thus making plug-in alarm clocks less dependable than wrist watches.

This is also the reason why a lot of clocks like that won't work on inverters that convert 12 VDC to 120 VAC since any load and most of them fall below spec.

For reference, the specification for the European power grid is in the "Continental Europe Operation Handbook", Part "P1, Load-Frequency Control and Performance".

https://www.entsoe.eu/publications/system-operations-reports...

It's actually a pretty elaborate system (who would have guessed?) with phase-stiffness between subsystems specified, because this basically defines the amount of power drawn/sourced. Timing is just the "outermost" regulation loop.

> On average the signal over 24 hours will equal about 24 hours so you can keep time with it, but as the load changes on the grid it will actually slow down and then speed back up when people turn off their devices at night.

Yes, the load changes the frequency over the course of a day. In Great-Britain there's even a special peak after the afternoon TV shows finishes and people start turning on their electric kettles to prepare tea.

http://www.bbc.co.uk/britainfromabove/stories/people/teatime...

In fact, the exact frequency of the power grid can be used to identify precisely the time of the day!

http://en.wikipedia.org/wiki/Electrical_network_frequency_an...

That is really cool that they can use it as a sort of fingerprint!

Yes, they do synchronize and yes, the number of cycles averaged over a day cycle is kept close to constant but: the frequency still floats. In fact, it floats enough to produce a unique "background hum fingerprint" which can be used to verify authenticity of audio recordings: http://www.bbc.co.uk/news/science-environment-20629671

It is also worth noting that most alarm clock chips not only keep time by counting AC cycles (often with fallback to 9V battery and suitably divided 32768Hz crystal) but also derive timing of all other internal functions from it. For example when LED display is multiplexed into two sections (as it often is) there are no common electrode drivers and instead the display is directly connected to two 1-way rectified taps of mains transformer. Economies of scale in AC-powered radio clocks are such that center taped (or even multi-taped as required for things like LM8562, which seems common in European clocks) transformer is significantly cheaper than two additional transistors.

And LM8562 and this timing of everything off mains is the reason why many modern alarm/radio clocks have that convoluted interface for setting time and alarm with two buttons that decrement setting by one minute and increment it really fast (ie. 50/60 minutes per second).

[Edit: 50/60 minutes per second, not 25/30, datasheet says that this rate is not dependent on rate of pulses on the setting pin, but I think that nobody really tested what will happen if the button is connected to anything other than mains frequency]

So at 7PM in the UK all the alarm clocks will be a little slow ?

http://www.bbc.co.uk/britainfromabove/stories/people/teatime...

Yes, as the load changes on the grid the frequency of the grid will slow down/speed up. So yes alarm clocks and other time keeping devices will be a little slow during those times the frequency itself is slow.

I think the author might be referring to clocks not counting in binary, rather than clocks not counting off power cycles. Full text:

> (6) "I figured it was actually 512 seconds (2^9)," one informant speculated. "Or maybe, since the clock is counting (typically) the power cycles from the wall socket, it's because nine minutes is 32,400 cycles, very close to 2^15 (32,768)." Engineer's comment: Nice try, bub, but clocks don't count that way.

Yes, some alarm clocks do count the cycles "from the wall".

Source: I helped a friend hack his clock to change his alarm clock from 60Hz to 50Hz.

I'm fairly curious about how you did that, and if the same is doable for larger load devices (60Hz -> 50Hz or 50Hz -> 60Hz).

It's not difficult, the IC has a pin to select between 50Hz and 60Hz so you just need to solder one or two resistors to make it work.

Something like this: http://www.bucek.name/pdf/lm8560.pdf

edit: It sounds like the first alarm clock with a snooze might have been synchronous so maybe it's worth thinking more about that :)

http://en.wikipedia.org/wiki/Telechron http://blog.onlineclock.net/evil-snooze-clock-history/

It's a bummer the Wikipedia article for Alarm clocks cites the straight dope article which links to some suggestive information that is totally missing now. It's too bad that Wikipedia doesn't store source information that is important for verifying the statements in articles in case the linked source disappears (or does it?)

originally, and probably wrong:

Yeah, some clocks used this, but I don't think it was most clocks and unless the first clock with a snooze functionality was this type, it doesn't give any more insight into why 9 minutes was chosen.

>It's too bad that Wikipedia doesn't store source information that is important for verifying the statements in articles in case the linked source disappears (or does it?)

Good reason to donate to www.archive.org :) It's probably accessible there, and the citation should include a date you can rewind to. Not that that's ideal, but it's at least verifiable, and probably available now if not forever.

Wikipedia encourage editors to use archive.org.

http://en.wikipedia.org/wiki/Wikipedia:Using_the_Wayback_Mac...

This all began in the 1970's when National Semiconductor brought out a series of integrated circuits for alarm clocks, clock radios, clocks for cars and so on. These chips included the LED display driver and the timing clock usually came from the mains.

Here is an example with the page of the data sheet regarding how the snooze function works:

http://html.alldatasheet.com/html-pdf/9252/NSC/MM5387AA/168/...

When the alarm goes there is a 'latch' that is operational for 59 minutes. Pressing snooze does not turn this 'latch' off, it suppresses the alarm output for 8-9 minutes. Turning the alarm off is what resets the 'latch'.

Chips of this era worked on BCD - BCD ruled once upon a time, when people did things in FORTRAN...

I don't think that National Semiconductor deliberately sought to emulate the snooze time of analog alarms from a previous era, it just was do-able to do 8-9 minutes in available silicon.

BCD was my thought as well. For the uninformed: http://en.wikipedia.org/wiki/Binary-coded_decimal

> What? Some clocks do count that way.

Lots of stuff used to... The realtime clock in the C64 (almost nevery used) worked in that way too. (it was one of a handful of things that depended on AC; you could power the machine off a 9V battery if you were ok with certain minor functionality not working)

So instead of measuring vibrations from a crystal, they are measuring cycles from the wall? Why?

In the past, the mains AC frequency was a cheap and accurate way to keep track of time. A lot of devices relied on the mains frequency.

Why does your modern day LCD monitor have a 60 Hz refresh rate? It's legacy from 1950's vacuum tube televisions that synced their beam with the mains voltage frequency.

The AC frequency used to be very inaccurate, it was only through a lot of work and legislation that it was made accurate, specifically so that electronic devices could use it to keep time.

Nowadays the electric companies want more leeway in how they run their grids and would love to have the frequency be less accurate, trying to keep it accurate actually costs a lot of money, since now as load changes on the grid they have to make sure to spin up new capacity. As the load increases the frequency goes down, as load decreases the frequency goes up. It is a careful balancing act that is required for the power companies to keep it within spec.

If it was allowed to drift more the power companies wouldn't be required to spin up more capacity that then goes unused.

Creating a pure enough crystal and cutting it in the right format was once very expensive.

There was a time when all the clocks relied on mechanical timers, and a MW sized electrical generator is more stable than anything you can put at your living room.

> What? Some clocks do count that way. That's why you can buy converters.

And it's also why your microwave clock is always off :)

Microwave clock, oven clock, your VCR/DVD players clock, and your alarm clock (if it is the plug-in kind that doesn't use a crystal to keep time).