(Quickly) translated from French.
|1) Story, goal and hypotesis
While going to the cinema to see our saturday night movie, Scruffy
and I walked by the waiting line where employees were giving spectators
their 3D glasses for Avatar (which we didn't went to see). The sight
of the obviously battery-operated glasses immediately made us wonder
how they could work and how we could annoy the audience by partialy
controling them ourselves.
From an old visit to the Futuroscope, I remembered that they used
very heavy and ugly glasses that had an infrared sensor in between
the two LCD panels which made them go out of sync if it was obstructed.
The glasses they're using for Avatar are a lot smaller, but the same
kind of sensor was exactly in the same place. A little black square, looking
like an IR filter.
To sum up how active 3D glasses work:
There is at least two times more images per second than in a normal film,
that means if it was shot using two 24fps cameras, the reel will be projected
at 48fps with one image for each eye. To make each eye see the pictures
it should see, glasses using LCD panels are used to alternatively hide
one eye then the other, in perfect synchronisation with the movie. Therefor,
the glasses need to know which eye to hide at what exact moment.
For now, there's 3 main types of transmission systems on the market:
radio, infrared or white light (DLP-Link).
Luckily for us that night, the cardboard boxes in which the glasses were
delivered were visible and showed a nice "XpanD" logo
on each side.
At first, XpanD was an European company. They moved to the USA in Los
Angeles to produce their 3D cinema systems. They bought NuVision and their
3D glasses in 2008.
Their X101 glasses are synched by infrared and go for $110 a pair.
The other model, X102 are synched by infrared or the picture itself
(white flashes used by DLP projectors) and are sold for $50 more.
The major problem is that they can't be differenciated by their
appearence. X6D Ltd., the legal name behind the technology, patented
the same design for the two types (US #D603445).
They all have a serial number on a sticker, but I doubt it can be
used to guess the model type.
Since the technology is proprietary, there's almost no chance to find
useful details in a patent. There's no patents filed under the X6D, XpanD
or NuVision name for the transmission protocol. There's no need to know
the exact timing of the infrared signal since we only want to make the
glasses glitch for now.
|2) First screening: searching for clues
The local cinema was using a rented digital
projector exclusively for Avatar, a Christie CP2000-XB. Its specifications
doesn't say anything about DLP-Link (cool, we don't want that.)
To minimize the risks of getting caught, and to be able to see
the possible infrared projectors, I took my old
modified Canon Powershot to take pictures at the end of the
movie. This explains the black and white pictures and the bad focus.
Here are the activated glasses, right before the end of the credits
The XpanD infrared projectors
official documentation, says this:
"It contains 280 high-power, narrow-angle (10°) IR diodes and
consumes 5 watts of power.", so if they are here, we should
see them easily.
First reflex after sitting my ass on an appropriate seat: look
behind to see if there aren't any purple-ish light spots with my
camera. There was two, right under the projector's window !
The narrow angle of the projector's LED can be confusing at first,
but that's because the projectors are made to be aimed at the screen
so the light can be reflected towards the audience.
No need to explain much, here we can clearly see the 7 by 40 IR
The ridiculous 5W of power consumption comes from the fact that
the transmitted signal's duty cycle is very low. That means the
LEDs are all blinking on and off very fast, but are on for only
a very short period of time.
The dead column shows that the LEDs are certainly 40 strips of
7 wired in series.
IR Investigator is one of the only software I use on my GameBoy
which isn't a useless gadget. It can take up to 20 recordings of
IR signals with various selectable sample rates ranging from a few
µs to hundreds of ms and save them in RAM. I used my 256Kb
flash cart this time and it didn't even fail.
Thank you Ken Kaarvik, thank you Nintendo. We have all we wanted
It was quite fun trying to look serious while holding the thing
aimed at the projector, in front of like 30 people...
The recordings reveals that the sync signal is rather simple (they
say "secured"... nice marketing move...).
Avatar is known to be screened at 144 fps (3 frames for each eye
This gives a 6.94ms period for the 100µs pulses sent to tell
the glasses to swich eyes.
Every 20 pulses, there is a 750µs pulse 150µs after
the small one to tell the glasses that the left eye has to
This allows to only use one pulse to sync 19 frames, and make sure
the 20th frame is for the right eye. So if they lose sync, you'll
see at most 131ms of film with the bad pictures.
|3) Second screening: action !
After $70 worth of components, some perfboard and two days of building,
we were ready to put our discoveries to the test.
The details about the infrared projector I built are on this
page (French only, sorry).
It worked, and we got a nice bunch of people on the first row of seats
complaining about how much the picture sucked.
Mission complete !