DIY Touchpanels - Maker Faire 2006

This is a talk that I gave at the 2006 MAKE Magazine Maker's Faire in San Mateo. I leave it for reference. It is pretty rough, so feel free to email me with questions, recommendations, rants, etc.

Sasha Harris-Cronin

BASICALLY...
A Touchpanel is a sensor system that returns:
  • (x,y) as analog
  • z (presence) as digital
optionally returns:
  • z as analog
  • multitouch
  • ID
When creating a touchpanel, consider:
  • Application (singletouch vs. multitouch, finger vs. object)
  • Environment (lighting, space, display type)
  • Cost
Some good references:
http://www.research.ibm.com/journal/sj/393/part3/paradiso.html
http://www.protechservices.ie/touch-panel.htm
http://tangible.media.mit.edu/
http://www.pixelsumo.com/tags/tangible/

TYPES OF COMMERCIAL TOUCHPANELS
http://www.protechservices.ie/touch-panel.htm
  • Resistive (any object - smart board, AV tps - not terribly sensitive, wears easily)
  • Capacitive (finger only - computer touch pads, very robust)
  • SAW (Surface Acoustic Wave - any object - gaining popularity, expensive)
  • DST (also sound waves. bad. stay away.)
  • IR (beam arrays. Any object. very robust, long lasting, expensive, doesn't scale)

CAMERA/MACHINE VISION
most flexible (in some ways), but necessitates limiting environment
  lighting (!!)
  placement

machine vision looks for:
  change over time
  edge detection
  blob
  shape within Region O' Interest (ROI)

Note: Projectors generally have little to no IR, so generally good to use IR lighting

Particulars:

  • Requires lots of software interpretation
  • generally need 2 computers - one for machine vision, one for display
  • Allow lots of R & D - make physical model
  • try and think of "trick" to make finger visible: ie. active IR LED
  • Sort of scalable. Need some depth for camera
Benefits: Drawbacks:
  • Unreliable - lighting huge factor
  • software/CPU intensive (go through lots of data)
  • Generally setup time is tricky and difficult
1. Side "Squinting"
need to use (and evaluate) 2 cams - for instance 1 does x and 1 does y
Need wide angle lens
make sure no comes between cam and touch surface

Application:
  multitouch
  analog z

Environment:
  can use vert or horiz, fnt or rear
  Need controlled space around installation
  can use array of IR leds to provide reflection.

2. Rear Camera
1 cam with coordinate system mapped to touch surface
Usually with rear projection
usually IR lighting (760 nm and about)
Easy to evaluate, but becomes a lighting problem.
You need really smooth lighting.
In sunlight less IR than visible light, but plenty of it.
Projectors generally have almost no IR
Flourescent good. Halogen bad.

Application:
  *sometimes* analog z
  multitouch

Environment:
  dark room or no IR lighting
  lots of rear space
  better if vertical

Danny Rozin - "Easel" - 1998
Active IR LED in Paintbrush

Jeff Han - 2005
Multi-Touch Sensing through Frustrated Total Internal Reflection
IR LED Side Lighting

Alvaro Casinelli - Khronos Projector - 2005
Rear Lit with Flexible Sreen

Graham Plumb/Gyroscope Inc - "Conservation Lab" - 2005
Bishop Museum, Honolulu
Technology by Sasha Harris-Cronin/BBI Engineering
Rear Lit by reflected bounce

3. Stereoscopic Vision
2 cams with coordinate systems mapped to each other
requires the most intensive software/math
Once working, pretty robust

Application:
  coarse analog z
  multitouch

Environment:
  front or rear
  better if horizontal (easier to have cams above)
  controlled lighting (but *can* do by measuring change in image)

West Office Exhibition Design - "Datastream" - 2006
Cadence Design Systems
Technology by BBI Engineering

RFID
like Wacom Tablet

Application:

  • object only, not finger
  • multitouch
  • limited Z and tilt/direction.
  • provides ID (coded objects)
Environment:
  • Can use in any ambient conditions.
  • Only front projection
  • surface needs to be powered
Particulars:
  • requires a large amount of hardware/DSP knowledge.
  • Can't really triangulate at that size. Has to be full underlayer.
Benefits:
  • if you have the hardware knowledge, relatively cheap.
  • Very accurate.
Downsides:
  • If you can't build your own hardware, expensive.
  • Lots of RF knowledge reauired.
  • not scalable
Off the shelf: wacom

James Patten, et al - Sensetable - 2001

Sound
active - send out ultrasonic "ping" or standing wave and measure results
passive - look for a tap on a resonant surface

Application:

  • Single touch
  • limited z only in active
  • object or finger
Environment:
  • Must use resonant surface
  • Have to have transducers on front of surface
  • Noise is sometimes a problem
Benefits:
  • Very cheap hardware
  • lighting insensitive
Downsides:
  • difficult software.
  • difficult to hide hardware

Josh Paradiso, et al - Tap Tracker - 2000

BEAM ARRAY aka. LIGHT CURTAIN
Use two set of beam curtains to create X and Y.
Use narrow beam lensed IR LEDs and phototransitors.

Application:

  • can be object or finger
  • limited multitouch (due to occlusion)
Environment:
  • in front of the surface
  • pretty light resistant
Benefits:
  • simple idea. tricky, but doable.
  • reliable once set up
Downsides:
  • Not really scalable.
  • Tedious to make.
off the shelf: banner engineering light curtains

West Office Exhibition Design - "Waterfall" - 2002
San Francisco Federal Reserve Bank
Technology by Sasha Harris-Cronin/BBI Engineering

CAPACITIVE SENSORS
Either use individual capacitive sensors per "hot spot"
or an array of cap sensors.

Application:

  • only fingers
  • multitouch
Environment:
  • completely environment unaware
  • can be front or rear screen
Benefits:
  • discreet.
  • analog z
  • relatively reliable
Downsides:
  • Not terribly scalable.
  • can be expensive (though getting cheaper)
off the shelf:
Lemur
QProx Sensors
Cypress capSense PSoCs

DECEASED!

OTHER STUFF
Scanning Laser Rangefinder
expensive, bad for your eyes, hard to find one with the right range

Jeff Han's LED multitouch pad

interesting, but not very practical

power your use and use them as a capacitor