Si & Joel's Disco Dance Floor Project

Si & Joel's Disco Dance Floor

(This is a page of notes, so will change!)

You know how you get those stupid ideas when drunk? Well, we were out together on Joel's birthday, discussing how funky an underlit dancefloor would be at this year's Eurovision party - the same sort of thing as the infamous Studio54 dance floor in New York, the sort of thing John Travolta threw his moves on in 'Saturday Night Fever'.

Oh dear.

News

Communication protocol basics worked out.
Si finished the first quadrant - photo.
Joel's done some LED mixing and fiddling with the Arduino - photos and video of the test cell.
Stop Press! First experimental matrix can be seen running in this video.

Requirements

No matter how stupid the idea seems, we need some aims:

Full RGB floor. Graded RGB not so important, but would be nice - although simple on/off gives us 8 colours (including black), how very 8-bit.
Connection to a PC to allow your own specification of colours on the light matrix.
Withstand a small hen party dancing 'I Will Survive'. Even the big girls.
Must be transportable. It won't be a permanent feature and may go to other parties.
Cost effective - Nicky will have my testicles for earrings if it's over budget or just too pricey.

Logically speaking it's easiest to do this as an 8x8 matrix; that gives a controller requirement for 192 LEDs (8x8x3) and means we can block the floor into four segments of 16 cells each.

Control Systems

First thought, could we circuit-bend a ready made kit? Sadly these are bloody expensive, and the controllers are little more than light flashers - so no.
Big Clive has some RGB modules. We'd need one of these per cell, and still wouldn't meet the 'computer controlled' requirement. That's a 'no' then, although they do look quite fun.
I'd wondered if we might be able to frig the daughterboard on this but they're not in production so getting hold of one might be a bit of a pain in the bum.
The MIT dancefloor used Maxim MAX7313 LED drivers controlled from an Atmega8.
Hmm, there's lots of Maxim drivers to choose from really.
Si seemed quite hot on using a PicAXE (or some other Pic-based solution).
I found several references (including this) which suggested we could use an Arduino Diecimilia (as used by Beeb bodybuilder Mike Cook) connected to another Maxim driver, the 7219.

After much grinding of teeth I chose the Arduino solution with the 7219s - three of those, one per colour, would be sufficient.

Physical Construction

A 5' square floor split into 8x8 cells gives us a cell size of 7.5" square. Actually, the apertures will be smaller because of the timber.
Of course there'll need to be enough light dissipation to 'mix' the RGB, so we can't have the LEDs flush against the translucent floor itself! There needs to be a space, coated with something reflective (good old tinfoil will do for that).
This will have people jumping up and down on it - so it's got to be strong. Perspex is expensive, it'll cost £200 to cover the floor, so we'll look at something else there.

The four segments can be connected together using banana clips - nice and easy - and cabinet locks on the sides to keep them together. B&Q sell timber 18mm x 70mm x 2.1M, which on its end will provide a reasonable boxing, and some hardboard or ply on the base, screwed in and countersunk to provide access for the frame to be removed if need be. Thus, we're actually building the LED sections on the ply itself with the frame over the top, with locating blocks on the base on the inside corners of each quadrant of 4 cells.

Communication

A quick note that we worked out a serial protocol last night for talking to the dancefloor live from a computer via the Arduino's in-built USB, or a MAX3323-based serial interface, although Maplin sell the MAX3232CPE which I'm going to try instead since they're local.

The idea is that the floor can run in one of two modes - computer-driven or 'automatic'. The auto mode is simple, it runs without intervention if you don't want to do any snazzy stuff. The computer-driven one however I think will work something like sending a stream of ASCII characters in [KRGYBMCW] and a control char to flip to the floor - that way you'd have a 'shadow' screen system which would eliminate flickering.

So for instance if you were to just whack it into minicom and draw out a rainbow stripe effect, assuming usual ASCII values in square brackets:

[STX]
RYGCBMRY
GCBMRYGC
BMRYGCBM
RYGCBMRY
GCBMRYGC
BMRYGCBM
RYGCBMRY
GCBMRYGC
[ETX]
OK

Some notes about the protocol (also helpful to refer to an ASCII table):

"OK" is the response from the floor itself - always 2 chars, can also be "E0..E9".
I've chosen STX (0x02) and ETX (0x03) since they seemed the most sensible.
FF (0x12) blanks the whole floor to black.
In a floor pattern packet (the bit between STX and ETX), we ignore CR, LF, and space (0x20).
In the event of not enough characters being sent for the whole floor, the remaining cells to be reset to black.
In the event of an error, the whole packet is discarded and an error code returned.
Errors: 0:Timeout 1:BadChars

I'm not quite sure yet whether to do some sort of compression but probably not.

Useful Links

Big Clive's LED Shop - featuring the 'Happy Cock' light modules.
Arduino Diecimila USB experimenter board (I got mine from Cool Components).
MAX7219 controller at Farnell.
Useful forum/blog posts here and here.

Index

This is a work-in-progress, and more information or construction info will appear here as and when we do stuff.