experimental horology

[gfish]

I apologize for not updating about projects recently. The one big one that I've finished (fireprop) isn't officially public until this weekend. I should get a post up about it next week.

So what have I been obsessing over otherwise? Well, this.



It may look like the back of a roadsign being used for target practice, but it is, in fact, a sundial. A digital sundial. The idea is simple enough. It's obvious enough that you can make a series of baffles which will block the sun during some hours and allow it to shine a dot through during others. Well, you can get some surprising resolution doing that, down to 5-10 minutes if you take care. And if you put a whole bunch of those dots together, you can create any pattern you want which changes over the day however you want. So, theoretically, you can do something like this...



This animation isn't just a mock-up, it's an actual simulation done in a ray tracer. The shadows being cast are due to a (very complex!) set of parallel plates like the one above, and the light source is following the same arc that the sun will at Black Rock City on September 3, 2010. Because, well, where else would I be installing something like this? It runs from about 8:00 to 17:00, updating every 10 minutes for that entire range. And since I'm using the constructive solid geometry functions, working out the design of the plates is actually pretty simple. I just have to print out the final plates, glue them onto some sheet metal, and cut them out. (Or maybe send them off for CNC water-jet cutting.)

Proof of concept! Well, kind of. As the first rendering shows, there isn't much left of the plate in places. The design will need to be tweaked so bits of metal don't need to be floating in midair anymore. This is also being rendered in an ideal ray tracer world. I need to do a lot more testing to see how it will react to real world imperfections. But worst of all, the current design isn't tracing out complete arcs, but differencing out cylinders based on the sun position every 10 minutes. This reduces shape complexity so rendering each frame of that animation above only took 1.5 hours. But it leads to the following result if I place the sun at 11:53.



I really love this idea, but there remains a lot of design work to be done.

Comments

[bigbumble.livejournal.com]

Incredibly cool idea. Since you only plan on actually using it for one day, may I suggest that you use heavy black card stock instead of sheet metal? There is a standard .050" thick card stock that is regularly CNC laser cut for model rocket fins. The material is relatively cheap, which allows for testing and mistakes(or multiple days if you are so inclined). It can be treated with epoxy for strength and weather protection.

[gfish.livejournal.com]

Well, the idea is to have it for the entire week of Burning Man, I just chose the median day. I haven't run the simulation yet to prove it, but I'm pretty confident it will still work quite well on the other days.

[catsittingstill.livejournal.com]

Very cool!

What if you make the "light-blocking-array" longer, so that the projection for, say, 12:10 falls in a different place than the projection for, say 12:20? And alternate back and forth, so 12:30 falls in the first spot, and 12:40 in the second, and 12:50 in the first and so on?

That way it seems to me like you could have fewer holes and thus more remaining background; maybe that would eliminate the "bits have to float in midair" problem?

[gfish.livejournal.com]

I thought about that, but if I can get it working this way I'd really prefer it. It just seems... more magical to have them all in the same place.

[tylik.livejournal.com]

That is highly awesome.

If you made the holes a little wider, could you have it so at 12:15 both 12:10 and 12:20 were visable, at different intensities? Then you could kind of interpolate...

[gfish.livejournal.com]

Wider is bad, because that lowers temporal resolution. But I suspect in reality that is more how it would work, since the sun isn't actually a point source and you'll get some diffraction through the apertures etc.

[sistawendy.livejournal.com]

Two Three things:

1. What software did you use for this?
2. Be careful about wind resistance.
3. This is awesome!

[gfish.livejournal.com]

POV-Ray (http://www.povray.org/), which I've been playing with on and off since the mid 90s.

[peteralway.livejournal.com]

And if you prop it up to tilt it, you can use it on other days as well!

Lovely, lovely scheme.

Hmm, you could probably use the same system for a digital display simulating a dial. Or some other slowly evolving image.

[memegarden.livejournal.com]

Stunning concept!

You might consider Lexan or some other clear surface, to solve the "dots floating in midair" problem.

[neuro42.livejournal.com]

In theory I have access to a 60" wide printer that can print on transparent media; you could do the entire system that way.

[neuro42.livejournal.com]

The single frame at 11:53 doesn't tell the whole story; I'd like to see an animation of every few seconds for a given 10m window to see what the transition effect is like. Might not be bad.

[anansi133.livejournal.com]

Dang, I'm glad someone is doing this!

Last I took a whack at it, the 7 stroke display was made up of plexi rods in different lengths. Then a mirrored strip would illuminate one end of each rod, and be blacked out when the rod was supposed to be dim.

15 mirrored strips, and 15 masking schemes... easier to wait for someone else to do it!

[gustavolacerda.livejournal.com]

yay!

Can't the floating bits of metal be attached via a transparent solid?
Do you have two layers of metal?

It would be interesting (from an AI perspective) to have the design problem solved by search / constraint satisfaction algorithms.

[orawnzva.livejournal.com]

I was wondering if there's a way to do this that factors out the redundancy in the faster-changing digits, which must show the same digit at many different sun angles in a repeating pattern, and I think it can be done with a lot of mirrors, but it'd be an awful pain.