A Giant DIY LED Display

smf28 writes "Dheera Venkatraman has created a giant DIY LED display featuring 36 blue Luxeons in a 6x6 array on the windows of Simmons Hall, an undergraduate dormitory at MIT famous (or infamous, if you wish) for its design. Recent uses included welcoming students in September, Pirate Day, and others."

Microchip PICs

[bhima]

Another fine use for PICs. I love these things...

Re:Microchip PICs

[Anonymous Coward]

Fine use for any uC.

Check out Atmel AVRs. They're especially good if you've ever needed to do I2C and really didn't want to write a bit-banger system. They also have on-die oscillators, which could have removed the external crystal in their schematic.

~AVR Fan Boy

Re:

[Alioth]

I think some PICs also have internal oscillators - not that I've yet built anything that uses a uC (but I've been looking around for my next project).

Re:

[dheera]

Many PICs have internal oscillators, but they are RC which isn't very stable, and def. not good enough for RS232 or any other communication - especially when the temperature of the room fluctuates a lot...

Re:

[juniorkindergarten]

Actually, I work with Microchip's 16f819, and that has a built in crystal running at 8 Mhz, and is extremely stable with temperature (automotive).

Re:

[stevenm86]

On the contrary. The RC oscillators work quite well with RS232. Besides, some of the 18F series have internal RC and crystal oscillators, and a built in USB peripheral. The 30F I don't think have USB, but they have a ton of memory and an FPU.

Re:

[marcansoft]

RC causes trouble with RS232 if it isn't calibrated right. Forget about using USB with it, it won't work.

I2C, OTOH, is perfectly fine. RC oscillators are perfectly fine for synchronous comms, since they don't depend on a precise clock rate. And most newer PICs have I2C built-in, so no need to bitbang.

Re:

[SageMusings]

Nope. That's why you resynch constantly on the pre-amble err...start bit...
RC osc's are just fine.

Cheers.

Re:

[Eternauta3k]

Well, you could read the last byte before flashing the pic. It contains a calibration number for the rc oscillator, if I remember right.

Re:

[Tyge]

Interesting post date. Oct 21 2006. Edison's bulb was invented Oct. 21 1879 :)

Slashdotted

[Anonymous Coward]

Trust slashdot to put the lights out! Someone should build a mirror array.

Re:

[suv4x4]

Trust slashdot to put the lights out! Someone should build a mirror array.

You know, people like you who keep reminding us of the "slashdot effect" always get modded up since it flatters the slashdot readers that theya re part of something big, the colective power of which can bring a server down.

Truth is the slashdot effect is nothing like what it used to be when blog had decent articles (versus unchecked factas, duped, misinformation etc. in massive quantities).

And second, you can bring a small server, or

Re:Slashdotted

[dheera]

I looked at my logs. It's probably the initial rush of visits in the first couple of minutes. A potential "slashdotting-protection" system that could be implemented by large sites is to select different IP ranges at random and serve up the site with a 0, 30, 60, 90, 120, 150, or 180 second delay based on the incoming IP range (i.e. if you're in the IP range corresponding to the 120 second delay, the article doesn't even show up for you until 120 seconds after it has been posted. I don't know if this has other moral issues, or if it's Slashdot's responsibility to care, but it's just a thought about a hypothetical solution.

Re:

[suv4x4]

I looked at my logs. It's probably the initial rush of visits in the first couple of minutes. A potential "slashdotting-protection" system that could be implemented by large sites is to select different IP ranges at random and serve up the site with a 0, 30, 60, 90, 120, 150, or 180 second delay based on the incoming IP range (i.e. if you're in the IP range corresponding to the 120 second delay, the article doesn't even show up for you until 120 seconds after it has been posted. I don't know if this has oth

Re:

[Dirtside]

Truth is the slashdot effect is nothing like what it used to be when blog had decent articles (versus unchecked factas, duped, misinformation etc. in massive quantities).

Go back and browse slashdot articles from two, four, six, or even eight years ago. The quality hasn't gotten worse. Or really changed much at all, for that matter (Roland Piquepaille notwithstanding).

Hackers

[UnderDark]

Reminds me of whay 0cool(Crash Override) did in the movie "Hackers" with the building lights: spelt out "Crash and Burn"

Re:

[Anonymous Coward]

Looks like a project out of the book Hardware Hacking Projects for Geeks.

Nothing to see here, move along.

Re:

[Joe The Dragon]

Building do that for real like saying go bears in chicago.

Re: Slashdotted

[dheera]

oh god. mirror: http://web.mit.edu/dheera/www/simdisplay.php.html [mit.edu]

Re:

[gaber1187]

speaking of "oh god"... oh god that building is ugly... I guess at night its not so ugly, but man... anybody ever seen that thing? jeez... the eecs building is kinda funky, but at least its not "ugly"...

Blinkenlights?

[Animaether]

http://www.blinkenlights.de/index.en.html [blinkenlights.de]

And I *know* there was a Dutch team that did much the same as well, and a Dutch commercial venture (was it KPN - Dutch telecom?) has one still up and running, I think.

But I guess they didn't use the Ooh! Shiny! blue LEDs :)

BlinkenLights and some other examples

[xmas2003]

I have some other examples of projects like this listed here. [komar.org] Very cool job by the MIT guys ... now they just need to add some more of 'em in all of the windows and provide the ability to generate alphanumber (or image) messages that can be uploaded from the Internet - heh, heh! ;-)

Re:

[account_deleted]

Comment removed based on user account deletion

Re:

[Usquebaugh]

Then set up RGB in each window :-)

Re:

[PhilDEE]

... now they just need to add some more of 'em in all of the windows and provide the ability to generate alphanumber (or image) messages that can be uploaded from the Internet - heh, heh! ;-)

But thing is, will it support ClearType?

Potential Income Opportunity

[Critical Facilities]

If they're really clever, they could sell advertising and offset the cost of adding to the array. I'm betting companies like Intel, AMD, IBM, Dell, etc would be willing to part with a little cash for some display time on a building/billboard at MIT.

Re:Potential Income Opportunity

[JoshJ]

I'd much rather they not corrupt the project by doing that. It's awesome as is, if it turned into yet another kind of lame advertising, that'd suck.

Re:

[Critical Facilities]

From TFA:
Acknowledgements Thanks to everyone on the Simmons Rush team that helped with the installation including soldering, duct taping, and cutting. Thanks to Simmons Hall for providing the funding for this project.

While I agree that "overcommercialization" of this would indeed suck, I was thinking of it more in a sponsorship type of income opportunity. You know, getting a PC vendor/LED manufacturer to provide parts at cost, leasing equipment at preferred rates, etc.

Re:

[fredcai]

The array cost less than 150 and was sponsored. Evil advertising unnecessary.

Too bad

[fuckingsound]

"We need babes now!" doesn't fit on a 6x6 array.

Re:Too bad

[smf28]

It is animated and scrolls text or any 6x6 animation; see the video [mit.edu] on the link.

Re:

[suv4x4]

It is animated and scrolls text or any 6x6 animation; see the video on the link.

As you see in the video is pretty hard at times to follow a scrolling text on a 6x6 display. Now make it 12x6 or 10x6 and it might work.

Of course, this is not meant to be useful, it's meant to be cool :) and as such, it's quite cool already.

Something like it has been done, and it was cooler

[BeeBeard]

Make the array bigger and put Tetris on that thing, kind of like these guys [com.com] or even these guys [engadget.com].

Summary lacking something...

[gumpish]

You really, I mean, you really couldn't just say how big this "giant" display is?

Take some notice.

[nazera]

Being a crontrols system desiger, I hope some of the goofs of industry look at stuff like this. I am constantly yelling at my vendors that I don't need a super screwed up version of RS232/485/422 etc to network sensors around a machine......binary and ASCII protocols WORK GREAT for stuff like this. KISS (Keep it Simple and Stupid). A few micros some twisted pairs and your basic switcher......bingo big network of fun.....if you need some more bandwidth, throw Ethernet at it. I've been saying this for over 10 years and still all the big players want to sell you a "Field Bus". I hope some of the MIT guys move in with the big guys...and slap some sense into them.....rant off.

Re:

[maxume]

Good luck. The problem with problems is that they often get 'solved' by someone who thinks it is interesting, while the guy who has the right solution never even knew there was a problem because it seemed so obvious to him.

Re:

[CPMO]

It's exactly the other way around. The dude who thinks the problem is uninteresting thinks his first idea is right because he never thinks beyond that first idea. When things screw up later (imagine the system growing to 100+ sensors, each with their own different binary protocol) the excuse is that noone knew any better at the time. The dude who finds the stuff interesting figured out all these problems long before the project even started because he enjoys figuring out that sort of thing. But he can *nev

Re:

[maxume]

It's at least debatable. For example:

http://hardware.slashdot.org/comments.pl?sid=20192 5&cid=16528159 [slashdot.org]

Which is the better solution? The first guy thinks the problem is interesting.

Circuit design error

[Flying pig]

The diagram shows a simple, basic mistake in circuit design. The Vcc of the microcontroller should not ever be directly connected to the Vcc of the power bus. Adding extra capacitance to the supply is not smart, it is quite unnecessary.

The microcontroller should have a separate supply, and as the consumption of the PIC is so low this could be derived simply by passing the LED supply through a small low voltage drop diode (Schottky diode) and preferably a suitable inductor, and then decoupling it with electrolytic and ceramic capacitors (say 1000uF and 100nF) in parallel as close to the Vcc pin as possible. With this arrangement, the LED Vcc can even momentarily drop to zero and the microcontroller will just keep running.

(In fact, if you are thinking of doing this from scratch, you do not need an expensive supply at all. Rectify the output of a transformer directly to provide pulsating DC (100Hz Europe, 120Hz US.) This is the LED bus. You can do that with a 35A 50V bridge rectifier bolted to a nice big alumin(i)um strip. Then pass the output through a diode, a suitable resistor, and stabilise it with a 5.1V Zener. Assuming a peak of about 8V from your transformer, a 1A Schottky, a 10 ohm 3W wirewound resistor and a 5W 5.1V Zener will do just fine, with maybe a 1000uF electrolytic and a 100nF ceramic to stabilise the voltage at the PIC and provide enough surge capacity to drive the MOSFET gates. That way, you avoid the major disadvantage of switching power supplies, which is that they do not like rapidly varying loads.

Oh, another thing. Do not put a resistor between the PIC and the MOSFET gate. Use a driver chip to translate the current levels. Cheap insurance.

Re:Circuit design error

[dheera]

Thanks for the suggestions. The PIC should actually be able to drive the MOSFET, it's just that there ends up being some small amount of ringing, which then [in some way that I didn't bother to analyze] caused the entire thing to hang. Adding the resistors killed that and made it work especially for cases where a large fraction of the entire array is suddenly turned on. But yeah, redoing it with better power supplies and power circuitry would definitely be something to do when we get time. At the time it was built we were trying to finish as quickly as possible and used ATX power supplies which are essentially free at MIT (you find them lying around in dumped computers everywhere).

Re:

[tuxicle]

I'm guessing without the gate resistors, the rise time of the load current was high enough to cause the voltage to sag at the load (inductance of the supply wiring), making your PICs brown-out. You should ideally have a separate wire carrying the supply voltage for the PICs from the ATX supply. This way, load transients won't cause the PIC to brown out. Perhaps a ferrite on the load wire would help.

OOO man I partially understood that

[technoextreme]

Rectify the output of a transformer directly to provide pulsating DC (100Hz Europe, 120Hz US.) This is the LED bus.

Now why are you powering the LEDs with the pulsating DC? I know no one is going to notice the flashing because the frequency is too fast. Im a engineering student who only just recently figured out how a walwart works and why its so heavy.

Re:

[Com2Kid]

I am CS, but I like LEDs, so here is what I know about pulsing them...

LEDs are pulsed for a variety of reasons. Altering brightness is one of them, it is simple to vary the brightness of an LED by pulsing it at different frequencies.

Heat is an other one, if an LED is being over driven, pulsing it can help lengthen the LEDs life span.

Pulsing also reduces power consumption (only an issue if running from batteries really...)

Re:Circuit design error

[dpaton.net]

Oh, another thing. Do not put a resistor between the PIC and the MOSFET gate. Use a driver chip to translate the current levels. Cheap insurance.
You're kidding right?

A resistor for gate isolation is just fine, especially for a low side FET drive. A driver chip would cost as much as the FETs, and is overkill to the extreme. In a perfect world, where money and time are infinite for design, it's easy to make anything better. For something like this, a little realism is in order.

My $0.02 on the design:

I've done something similar as a proof of concept for a customer...256 RGB LEDs (50mA/color, ~38A at full bright/full white) with 64 custom processors controlled by a big Atmel. It ran off a standard 600W ATX supply, and it worked just fine, no voltage dropouts at all. I don't think the ATX supply itself was the problem, rather the layout of the circuit. A normal ATX supply has rather good transient reacitve capabilities. Using a single power supply for an entire floor is likely the culprit. It looks like the run on each floor was about 60', and I highly doubt that he used the right sized wire for that run (25A @ 60'-> #8). The accumulated coltage drop would be pretty extreme, making the PICs low voltage brownout inevitable. Combine that with an improper power supply arangement at each processor location and bam, crashes. The 6600uF caps are a band-aid, I agree. A fat wire feeding the high sides of the LEDs, and a secondary wire feeding the PICs would be my choice. Yes, they can safely be tied together, but ONLY AT THE SOURCE. That long run of wire will be all the isolation they need. Standard long distance bypassing at the PICs will keep them happy (10uF/1uF/0.1uF) and a nice fat ground return keeps it all under control. There were a few mistakes, but by no means is it fatally flawed.

Driver chip costs

[Flying pig]

For information, an AHC244 (8 way driver) with a +/-25mA output on each driven line, easily capable of switching 5V drive MOSFETS up to several amps, costs all of $0.90US in one off. I consider that cheap insurance. But then in my formative years, before mobile phones and readily available Internet, I had to design stuff for installation in remote sites. And to make sure I got the point, I had to do some of the fixing. When you have traveled from the UK to the US to do a repair under warranty, and found the

Re:

[dpaton.net]

I've spent more than my fair share of time in the field doing repairs due to blown transistors, failed drivers, and the rest. Every single failure (in an outdoor, harsh environment, intrinsically safe, and heavily abused embedded system for agricultire, which I'm using as an example here) was the result of amazing abuse of the output devices. Load dump spikes hundreds of times over the design limits, connecting 120VAC to a terminal marked 12VDC, lightning (which it survived with only one blown transistor ou

That's nothing compared to the Cubatron

[neuro.slug]

The Big Round Cubatron [nw.com] is a much bigger, much cooler DIY LED display. It was the cool thing at this year's Burning Man.

Videos here [youtube.com], here [youtube.com], here [youtube.com], ...well, you get the idea.

Why is this a story?

[Bender_]

Sorry, I dont see why this is a story. The technical difficulty of this is something like "my first microcontroller project" from "toying with electronics 101". The implementation is not even that interesting. (Multiplexing anyone?). The novelty is almost zero (giant LED display.. uhhh). Yes, it is at this geek university, but that is the only real point about this story i can see.

Re:

[randcv]

It's not even breaking news, this was on hackaday weeks ago. This site needs some serious quality control before it loses its entire fanbase. If I want shitty news, I'll read US Weekly.

Delft University of Technology was first

[Thijs van As]

In 1995, Electrical Engineering students of Delft University of Technology did this, with playing Tetris on their 100m high building.
Link: http://www.etv.tudelft.nl/vereeniging/archief/lust rum/90/english.html [tudelft.nl]

In 2001, they used their building as a big SMS display.
Link: http://www.etv.tudelft.nl/vereeniging/archief/lust rum/95/english.php [tudelft.nl]

In 2006, a huge 8x4x2m LED MatriXX was created.
Link: http://www.etv.tudelft.nl/vereeniging/commissies/e lco/matrixx/ [tudelft.nl]

Re:

[belg4mit]

Sorry, no. Thank you for playing [mit.edu].

Those just happen to be the incidents that someone got photographs of,
there have been many more, including Tetris.

Tetris

[SnoopJeDi]

It's not even large enough to play large-building tetris on, but I still think it's really cool.

OT: I have an interview with an MIT Education Counselor (EC) for early action this Monday. Between this and the recent protein gel discovery [slashdot.org], I should have a lot to talk about with him. Neat.

Re:

[dheera]

in case i didn't mention it anywhere on the site, the pure reason why it's limited to 6x6 was the fact that it's the largest continuous available public space on the face of the building. if you go left, there's nothing, to the right, the windows stop being regular, and above/below are student rooms that we don't want to bother. we would have definitely liked to build a bigger one, though, if we could.

Re:

[SnoopJeDi]

Yeah, I did see that on the site. The fitness room is the best spot for contiguous windows available, and 6x6 is the absolute best.

Not to mention that 6x6 works just fine for displaying text, as you guys are doing. More LEDs would just mean more money and more potential trouble spots, anyway. Good stuff all the same.

Not do it yourself, but bigger yet

[geekzer]

OK, it was done by pro's and cost a bit more than $150, about $17 million more! But at 12.5 million LED's I think the Fremont Street Viva Vision is Las Vegas wins the "biggest" contest. http://www.vegasexperience.com/viva.cfm [vegasexperience.com] To provide some scale on the photos and minivid clip, the peak of the ceiling is at least 6 stories above you and the small segment you see is just a fraction of the 1,500+ feet long display. Oh, and from the looks of those photoshop-ed pictures, I wouldn't be surprised if this was

RGB

[tygerstripes]

So, your resolution is hugely limited... try RGB. Works for TV. Tellys have a pathetic resolution, but a decent colour range, so they can produce much more convincing images than would be expected (this is why B&W TV always looks shoddier, even with modern footage).

Can't imagine it'd be that hard or expensive to introduce some colour to the scheme.