Hacker News Reader

>depend on inputs that are merely convenient side-effects of third-party systems provided free of charge and without obligation.

I would disagree. You don't get it for free, you pay for that. The Grid frequency is regulated and the powergrid providers have, IIRC, even legal obligations on how far they are allowed to deviate the time.

Using AC as clock signal is good enough if being wrong by a few minutes is not mission critical. If you absolutely do need accurate time use DCF77 or GPS.

> As a community of mostly engineering-minded folks, I hope we can take this as a warning to not make our mission-critical systems depend on inputs that are merely convenient side-effects of third-party systems provided free of charge and without obligation.

That is not entirely true. The TSO is committed to keeping the average frequency of the power grid at 50Hz and has always communicated that. It tries to keep grid time within 20 seconds of UTC, and has historically achieved this. So it is reasonable to use the grid as a time source where high accuracy is not required. The current situation is highly irregular. Using the grid frequency for time keeping is a secondary function of the electricity grid, but it is not merely an unintended side-effect.

I agree with the rest of your points about using a different time source where high accuracy is required.

Good point about this service, I almost forgot that it still exists (although reception is somewhat spotty indoors). The basic advice is sane ("do not rely on side effects"), yet "third-party systems provided free of charge and without obligation" sounds exactly like that, except now it's putting all your eggs in a basket that's controlled by one entity (be it DCF77, or the GPS temporal component). I would think that the AC-as-a-clock is usually chosen for convenience rather than high reliability - "the juice is needed anyway, why bother with an extra receiver?" - and if you actually need to tell the time, checking multiple independent sources would be encouraged :)

DCF77 surely cannot be that bad indoors on the Continent?

When living at what was then often quoted as the extreme fringe of the coverage area - Trondheim, Norway, on 63,5 degrees of latitude - I had reliable coverage as long as the clocks were kept in window sills (hence, effectively outdoors)

However, just moving down to my current home (on 62 degrees, or -roughly speaking- 11% closer to the transmitter site), I now have reliable coverage everywhere (granted, in a wooden house - but the DCF77 alarm clock in the basement synchs every hour, too)

I've found that antenna orientation matters, even in the Netherlands which is much closer to DCF77 (concrete house).

Yes, it needs to be perpendicular to the direction of frankfurt for best results

You can find more details in the second page of this datasheet

https://www.gemischtwaren-haendler.de/shopdateien/3942_1.pdf

More importantly, you would need to know how the antenna is oriented within the enclosure of your clock. I'm pretty sure that that wasn't mentioned in the manual of the one I bought.

Most clocks are thin, so it's easier to guess the placement of the antenna.

Usually the antenna is parallel to the board - and the board is positioned in the largest dimension, which tends to be also parallel with the display.

Citizen, the Japanese watches maker, offered a passive (without batteries) "enhancer" for their AT series, IIRC was a ferrite block where to put the watch, I think I've seen a photo.

Or one can use a PC to transmit the DCF77 signal locally (it could be illegal) to a clock using a pair of headphones, I've seen at least a couple of examples but never tried them.

Edit: found the photo http://forums.watchuseek.com/f17/possible-method-improve-ato...

Beats me - I'm within reasonable driving distance to the antenna (one border away), yet clocks fail to acquire signal unless outdoors/directly at a window. Just a bad coverage area, I guess.

>11% closer to the transmitter site

This is completely outside of my area of expertise, but doesn't the radio signal strength decrease exponentially? So that last 11% could be a big deal?

Inverse square, so you're in the vicinity of being correct: https://en.wikipedia.org/wiki/Inverse-square_law

Quadratically.

Line-of-sight attenuation of an EM wave in the empty space is quadratical, whilst - for example - attenuation of an electrical signal in a wire is exponential.

Obviously this is a super simplification...

Oh, that makes sense. Thanks for your explanation, good enough for now :)

Another advantage of DCF77 and similar is that they also broadcast the current time, while the 50Hz or 60Hz AC is just a frequency. If you have only a frequency, you have to set the initial time manually every time you lose power.

DCF77 also tells you what time it is, AC can only keep whatever time's been set in sync. The equivalent would be to include a 50-cent crystal oscillator instead.

On a only barely related note, I've been playing with the idea of implementing a NTP synced clock on an ESP8266. It'd wake up, join wifi and sync NTP every few hours, then keep time on its internal crystal in between. It should be a good deal cheaper, work anywhere in the world, and have better indoor coverage (I have a DCF77 clock in an interior bathroom in London, and it never syncs. Kind-of defeats the purpose that I have to move it to a south-facing room for the day to get it to sync).

The UI for entering the Wifi password is always the biggest problem with these things. The Dash buttons have an interesting solution involving audio.

One solution would be qr scanner...

You could try a USB SATA drive with a FAT16 filesystem containing "/wifipass.txt". USB-1.0 low speed is 1.5Mb/s and while I can't find documentation on ESP8266's GPIO switch speed, it's CPU is 40MHz, so bitbanging should be feasible.

Or just reporpose an old phone.

I was thinking about this in terms of building a product.

Even with that, a clock generally runs for months, even years, on a single battery. Not exactly phone territory, but might be feasible on the ESP8266, depending on how the oscillator behaves in deep sleep mode.

>how the oscillator behaves in deep sleep mode.

horribly

There are many dedicated real time clock chips, pre-designed for backup time-keeping and low power usage. Interfacing with one is going to be way faster/easier/better than trying to roll your own on a general purpose microcontroller, IMO.

TI's bq32000 is $0.55@1K, is 3.3V for easy interface to the ESP, and takes just over 1 microamp in backup power mode. That's just one that I quickly poked at.

(Side note: at this point, I'd think I'd want to be using an ESP32 in any new designs.)

I imagine the difference between DCF77 and a 50Hz clock is similar to the difference between a $40 clock and a $10 clock. :P

Radio-controlled is pretty much the default around here (and the cheap ones cost perhaps 5 €); I think I've yet to see a grid-powered alarm clock.

Here's an IC used in some alarm clocks that uses the power grid's frequency, along with a list of products that use it: https://www.radiomuseum.org/tubes/tube_lm8560.html

It was mostly used in the common radio alarm clocks, the ones with a 4digit red 7 segment display that were all based on an IC from the 70s that handles everything (minus the radio which is a completely indipendent module).

GPS is also an extremely accurate source of time, although since it's on a much shorter wavelength your reciever needs a clear view of the sky and is much more complicated.