Hacker News Reader

Hacker News

Comment by xdissent | Hacker News Reader

Full thread

xdissent

> The unit is powered by an external 12VDC adaptor, connected on the back of the frame.

But later...

> The clock reference, in other words the heart beat of this clock comes from the AC outlet.

Maybe it's really a 12VAC adapter?

View on HN

Does anyone know if this 'experiment' became permanent?:

Planned U.S. Power System Experiment Means Some Clocks Will Speed Up

http://spectrum.ieee.org/tech-talk/energy/the-smarter-grid/p...

Because if it did, then at least in the US, the 60Hz grid frequency is no longer a 'good' stable time-of-day clock reference.

Anecdotal evidence (my own old 60Hz line-synced alarm clock) prior to this experiment kept very good time (drift of 1-2 minutes over a year or more). Ever since this announcement, it has drifted fast such that I need to reset it about once a month for it to be reasonably accurate. And the drift now after a month or two is on the order of 5+ minutes.

If you use a tuned quartz crystal oscillator you shouldn't have any problems.

The signal can still be pulled from the output side of a simple wall-wart type power adapter. Here's a Maxim app note that describes how you might do that:

http://www.maximintegrated.com/en/app-notes/index.mvp/id/199...

From the note:

> Toward the end of the day the power company speeds up/slows down the frequency as needed so the total cycles in a given day is 5,184,000

That's incredible! I had no idea it was that accurate.

I'd love to see a citation for that level of accuracy, I can't believe it's that good, maybe to within +/- 4 cycles? If you watch this:

https://www.youtube.com/watch?v=slDAvewWfrA

This is about how the UK National Grid manage power availability at peak times. Over here we're running 50Hz rather than 60, but one of the key goals is to keep the grid frequency as close to 50Hz as possible.

At around 2m15s into the clip they lose power from a French on-demand provider and the frequency takes a "nose-dive" to 49.6Hz causing the operator to call on backup from a Welsh hydro-station to keep up with demand.

Theoretically in the UK we should get 4,320,000 cycles per day, but I suspect it's probably + or - 3 cycles. Unless of course it eventually balances out over the course of several days.

Happy to be corrected though.

I agree with you that "always accurate" is an optimistic statement but I wouldn't say any level of accuracy from that note is implied. A well engineered solution would take the 5million figure as an average over a longer period of time as stated.

When I first discovered that this was how grid demand was managed, stood in a UK National Grid control centre next to my brother who was interning for them, I was amazed.

Yes, I was under the impression this would be automated. I'm sure many parts of demand management probably is, but the manual intervention involved to keep things under control during simultaneous mass tea breaks surprised me hugely.

In Europe (also including UK ;-) ), the joint Europe electrical grid is maintained according to the mechanism outlined in this technical document.

https://www.entsoe.eu/fileadmin/user_upload/_library/publica...

    Page P1-29, Definition D-D1, Tolerated Range of Discrepancy

    A discrepancy between SYNCHRONOUS TIME and UTC is tolerated
    within a range of ±20 seconds (without need for time control
    actions).

And corrective action is to change the global-grid setpoint for frequency (of 50Hz) up or down by 0,01 Hz (i.e. 1sec difference in time shown on a clock / 5000 seconds of real time). So the system allows for ±20s·50Hz=1000 cycles of deviation until corrective action starts, and allowing for some time until the system reacts, that means that the clock might be off by ±30s in reality (the last one is a guess by me).

But it's part of a much more elaborate scheme with several control-loops taking care of different things, the timing is only the global, outermost, regulation. All this makes much more sense, once one has realized that phase between two parts in the grid is the main controlled variable to determine flow of energy, so this document may appear to be quite obscure and strange to most people.

EDIT: corrected my math.

ADD: And, obviously, "Time Nut" Tim van Baak has a plot on the topic. For the US ;-). http://leapsecond.com/pages/mains/

Thanks for the links.

I always wondered what the importance of maintaining such an accurate frequency is. Are there reasons other than keeping parts of the system in phase? (From what you're saying it sounds like they take additional measures to ensure that too.)

I suspect he uses both 120VAC and 12VDC - the first to provide a clock reference, and the second to power the circuitry. It wouldn't be hard to wire in the 120VAC connection in addition to having a 12VDC adapter.

The C64 did this (pretty much). It had a rarely used realtime clock function in the CIA 6526 (Complex Interface Adapter) chips. You could power the C64 off just 5VDC for most purposes (or a voltage roughly in that region - I experimented with running mine off both a 4.5V and 9V batteries; it was surprisely robust against abuse from people like me who preferred physical experimentation to actually reading up on what'd be advisable to subject it to), but if you didn't provide an AC source on the right pin too, the realtime clock and the user port wouldn't work as they should.

Theres a transformer 120 vac to 12 vac and i use the signal of the 12 vac. Theres 3 wires that goes from the power supply unit to the clock, Ground, +12 DC and the 12 vac 60hz signal.

Yeah, I'm no electronics expert (far from it), but that part also confused me. Can someone explain?

Depending on the supply requirements, he might use voltage ripple as a reference.

With an unsmoothed rectified supply you get a nice "pulsed" input anyway:

http://falstad.com/circuit/#%24+1+5.0E-6+1.8479586061009856+...

True but you'd expect that to be 120Hz rather than 60Hz.

So you could: a) use a half-rectified supply b) halve the count.

A) would be easier. :)

True ... not so good for efficiency and you'd need a bigger smoothing capacitor. (I'm assuming since this is all discrete, there's no need for a fully regulated power supply).

You'd want a fully regulated power supply for this, /especially/ because it's discrete. You want the LEDs to remain constant brightness, but, more importantly, you don't want an unregulated power supply to overvolt and destroy all this hard work.

Use a diode so you only get half of the sin waves, that's 60hz which makes it very easy to build a clock off since you don't need to do any multiplications on the base clock frequency it self.

Reply on news.ycombinator.com