Inspired by the maddening timepiece from Discworld, this clock keeps time, but anyone watching the seconds tick by may be mentally unstable for it. [Renaud Schleck] built the stuttering clock using very few components. He undertook the build after being inspired by the version which [Simon Inns] built.
The clock itself is a run-of-the-mill item which uses one battery to keep time. We’re always impressed by how these dirt-cheap things remain so accurate over the long haul — but we digress. The method of attack uses coil injection to drive the hands. [Renaud] used one of the microcontrollers from the MSP430 Launchpad, along with the clock crystal which also shipped with the kit, to gain control of the mechanism. The crystal triggers an interrupt which does the actual time-keeping. The seconds hand is driven rather sporadically based on an algorithm explained in his write-up.
You can watch the uneven ticking in the video after the break. Despite that visually disturbing functionality, the short and long ticks balance each other and the correct time continues to be displayed.
[via Reddit]
I think I might try a version that uses a synchronous AC motor and varies the frequency.
That would be far worse. It would be like fluorescent lights that continuously pulse over time. Makes me ill just thinking about it.
Combine it with the Harry Potter clock and you will always know where your cat is.
It looks like the coil might be getting a bit too much current, judging from the swinging motion the second hand makes after each tick. While it does make the clock extra loud (and annoying, which is great in this particular case), it wears down the cogs inside the clock pretty fast. Not that replacing the clockwork would cost you an arm and a leg, but it will cost you significantly more than adding some current limiting resistors…
Indeed the pulses do seem a bit long, though I found the biggest detrimental effect of this is wearing out the battery sooner. The solution for me was shorter pulses. The optimal pulse length has to be tuned on a per-clock basis. I had a project to do the exact same thing also using the same devices (both the old launchpads and the new fancier ones) at https://github.com/bnahill/clock. Off 2 AA batteries it seems to get at least a few months of run time. To improve power, I think this design could afford to wake up less frequently than 64Hz, but I think most of the drain is in the clock driving. While mine only uses a consistent probability distribution for tick length (and hard bounds at +/- 30s), it does seed with a random process using a pair of oscillators as described in TI’s SLAA338.
That said, it is amazingly annoying to have this in the room: http://www.youtube.com/watch?v=nS-rX8b5SzI
“the short and long ticks balance each other and the correct time continues to be displayed.”
I’m assuming you mean accurate to within 10 seconds. For a few seconds, the time will be inaccurate until a series of “short ticks” catch it up, or “long ticks” allow time to catch up.
I love the idea, but both this and the other one have a couple of fatal flaws. First, they’re obviously driving the coil at 5v with far too much current – the ticking motion is far too abrupt. No doubt this is the cause of the shortened battery life.
More significantly, I think these are missing out on some of the spirit of the original. Vetinari’s clock was _subtle_. Ticks shouldn’t be twice as long or half as long as regular ones, they should be fractionally shorter or longer. Just enough to unsettle, but not so that you notice the discrepancy immediately.
Honestly, I think this would be far more disturbing if the hands where independent. Minutes and hours always in time, and seconds always slower. A little bit slower. 1.3 seconds, for example.
That would be hardly noticeable (as they are always equally wrong) and horrible for someone looking at it.
As arachnidster said this version is missing the subtlety of the “real” thing, If it were possible to have it tick backwards once in a ten minute block and only drop one tick in a sixty second block then you would have a much more authentic Vetanari clock.
The description of the original isn’t even that it misses a tick now and then, but that now and then a tick is just maybe noticeably delayed. Maybe delay a tick by about 30% every 20s, then make the subsequent seconds 1.5% faster to keep up.
Seems like that would be the ticket. Enough regular ticks to establish the auditory pattern, then once a person would begin to anticipate the ticks, toss in a slightly delayed one. Would be even better to randomize when the delayed tick comes in so a person wouldn’t begin to anticipate them as well.
I really want to do this and hang it at work to experiment on my co-workers.
I’d go for an attiny 85 with external oscillator.
My issue with this example is that it is no where near subtle enough.
I think getting it just right is a real challenge, as real testing would require a steady stream of test subjects unaware of the modification.
anyone experimented with this and gotten some figures they perceive as subtle but nerve wreaking?
From my experience with the MSP430 watch from TI, you dont have to do anything to make an inaccurate clock. The dang thing loses 2 hours a month on it’s own when you are trying to be accurate.
I’ve wanted to make one of these for along time.
I agree with those who already mentioned that this model isn’t suble enough.
One tick backwards with a lot of time in between, now and again two seconds at once and some slight inconsistencies in the length of a second would be much more unsettling. With the model here everyone can see it’s the clock which is just strange, but the goal of Lord Vetinari’s clock is to make people think they are mad, not the clock.
from “Feet of Clay” by Terry Pratchett
“But somehow, and against all usual horological practice, the tick and the tock were irregular. Tick tock tick … and then the merest fraction of a second longer before … tock tick tock … and then a tick a fraction of a second earlier than the minds ear was prepared for.”