has anyone ever made this?

scubasteve_911

Joined Dec 27, 2007
1,203
Never made one, and if I did, I'd make something a bit nicer than what they show. There are too few light sensors and the response is a bit cheesy, but I guess they need boundaries to make it at a reasonable cost.

If you're going to do something like this, get the LEDs from ebay from a chinese source. You can get hundreds for cheap, unlike from a US distributor.

Steve
 

Salgat

Joined Dec 23, 2006
218
If you don't mind the repetitive nature of this project, then go for it. I saw recently a project made of plexiglass that was much smaller and acted as a night light. If you don't have the patience to do this one, then do that instead.
 

beenthere

Joined Apr 20, 2004
15,819
Can anybody tell how the table "interacts"? If it's just a matter of humongous numbers of LED's that get brighter when something shadows a phototransistor, then making a dupe might be pretty easy.
 

scubasteve_911

Joined Dec 27, 2007
1,203
There's a bit more to it, as you notice, it pulses and fades. It does seem very easy to copy and improve while you are at it.

I think the best way to do this would be using surface mount parts, since you don't need to drill a ton of holes. You'd definitely have to make it modular though...

steve
 

SgtWookie

Joined Jul 17, 2007
22,230
SMT would be one approach, but I'm afraid that would take away from the "organic appeal" of having lights appearing in pseudo-random 3-d space, along with the pulsing/fading and its reaction to external input.

It's not that highly regular 2-d and 3-d lit objects are uninteresting; quite the contrary. However, they have a more mechanical and artificial "feel" to them.
 

beenthere

Joined Apr 20, 2004
15,819
In every case I've seen, the smt LED's cost more than thru-hole, especially if you go grab-bag hunting. Too bad it's mostly green in the bags.

If one has over a thousand feet of 26 ga wire wrap wire and a gun, there is a better way than pcb's. Lots of holes to drill, but only one per LED.

The drawback to something like that is if it isn't really neat, it will get old pretty quick. Something on the wall instead of a table might be another way to go...
 

scubasteve_911

Joined Dec 27, 2007
1,203
It would be fun to use the larger 1206/1210 LEDs to go to town with conductive epoxy. You can get artistic and have it done in a day of 'painting'

Steve
 

SgtWookie

Joined Jul 17, 2007
22,230
It would be fun to use the larger 1206/1210 LEDs to go to town with conductive epoxy. You can get artistic and have it done in a day of 'painting'
Gee Steve, while an attractive idea, I don't know of conductive epoxies available with low enough resistance to attempt something like that - not that I've looked recently.

In order to keep a blue LED supplied with 20mA @ 3.8v (just tested one earlier) via a foot-long epoxy conductor, the epoxy would have to have a resistance of 410 Ohms (or roughly 34 Ohms/inch) in order to be supplied via 12VDC - and that's not counting Vdrops from other components in the circuit.

Keeping the dimensions of such a conductive epoxy track would be problematic; that would figure in heavily unless it was nearly a perfect conductor.
 

scubasteve_911

Joined Dec 27, 2007
1,203
Imagine the beauty of using the epoxy to create current limit resistors?

http://www.epoxies.com/tech/40-3900R.pdf

They have a resistance of 0.0001ohms/cm, which should work for something like this. The problem with it is, they use silver in the epoxy, which means $$$. They have a nickel filled epoxy, I wonder if that is a bit cheaper?

You're right about keeping the resistance constant. Maybe the uneven light will fit right in with the whole "imperfection is beauty in art" notion.

Steve
 

beenthere

Joined Apr 20, 2004
15,819
The operating principle is not quite established. The sensors seem to detect changes in both visible illumination and also in the near IR. How this translates to lighting how many LED's and how to fade them out isn't clear. The small IC's visible on the pcb's are probably ATtiny microprocessors, although some sites say the LED control is all analog.
 

SgtWookie

Joined Jul 17, 2007
22,230
Hmm, yes - silver-filled epoxy would be prohibitively expensive. Now that my brain is beginning to work again, I remember they were using conductive epoxy to assemble planar array X and KU band antennas.

One could likely mix finely powdered charcoal (ie: carbon) into a liquid epoxy to create "on the fly" resistive traces, which would be much more economical than the silver-filled epoxy. It would take some experimentation to determine optimal mixes.
 

SgtWookie

Joined Jul 17, 2007
22,230
Well, here's a big clue as to how it works:
"Each circuit board panel has four photosensor nodes, each of which drives 20 LEDs, for a total of 80 LEDs per board. So, the 6-panel board has 24 sensors and 480 LEDs, while the 8-panel version has 32 sensor nodes and 640 LEDs."

From the original link, on the 9th page, it showed a process of matching LEDs for brightness @ the same voltage. Sounds like they're using a bunch of LEDs in parallel with one load R, which doesn't sit well with me.

Anyway, I think what they did is for each zone, simply compare the output from that zones' photosensor with the average output from the other three zones' photosensor. If the light level in the zone is less than the average of the other three, more current is fed to the LEDs to raise the intensity. If the light level in the zone is more than the other three, less current is fed to the LEDs.

There must be a low-pass filter of about 1Hz on the output of the compare circuit to prevent rapid oscillations, but this also gives a "pulsating" or "undulating" feel to the circuit. There must be a biasing network to set an average ambient light level; that would of course change with room lighting conditions.

I happen to have a slew of CdS photocells, op amps and comparators around here, along with 80 blue diffused LEDs. Might have to do some tinkering ;)
 

beenthere

Joined Apr 20, 2004
15,819
In case you get a bit tired of funky CdS cells, look at TAOS light-to-voltage converters. Mouser carries them in either TO-92 or smt. They have a voltage output and faster and wider response than CdS. And they're even cheaper! Page 164 of the current catalog.
 

SgtWookie

Joined Jul 17, 2007
22,230
Interesting, Beenthere :) Hadn't seen those before.

However, I have over a dozen CdS photocells here to fiddle with that I'd picked up a number of years ago. Might as well use 'em for something like this ;)

I have a sneaking suspicion they used an OPA2674 dual high current op amp in their design; after letting my subconcious chew on this for a while and looking at what was available along with what their design looks like, that IC is the only one I've found so far that makes sense. It could compare the signals from the other photosensors, and has enough output current to drive 10 series pairs of LEDs at up to 50mA/pair.
 

beenthere

Joined Apr 20, 2004
15,819
Unless there are smt transistors on the other side of the board. Sure like to see a nice detailed close-up of that ic.

Forget the smt trannies - this is available as a kit, so it should all be thru-hole parts. I was thinking something more like a dual comparator doing the flashing action.
 

SgtWookie

Joined Jul 17, 2007
22,230
Yes, I don't think they'd be putting SMT's in there, as everything else appears to be thru-hole.

After some more thinking and looking, an LM1877N audio amp is looking even more likely than the OPA2674; it's a 14 pin DIP dual amp with output up to an amp; double the OPA2674, and the OPA2674 is SOIC, not a DIP. But, I've found some other interesting ICs that could be substitutes ;)
 
Top