RGel is conceived not as a device, but a prototype material. As evidenced by databases like Materia, institutions like Material ConneXion[1], and the work of designers such as Linnaea Tillett and Neri Oxman, an industry and a practice focused on generating and using 'new' materials with novel properties is blooming. As further testament, Erica Robles and Mikael Wiberg from New York University and Umeå University will be presenting a paper entitled "Texturing the 'Material Turn' in Interaction Design" at the TEI 10 conference this year. RGel envisions such a "material turn", a future where new materials exhibit more than novel material properties, but controllable phenomenal properties as well. Call it phenomenological material design. As electronic components grow more advanced, smaller, cheaper, and more efficient, the possibility of designing materials with computationally-enabled active or responsive properties approaches reality. RGel invites us to bracket the harsh unchangeable materiality of material itself[2] and think of matter as something programmable, to imagine new class of hybrid materials with responsive capabilities enabled by distributed sensors and actuators. These new active materials could be thought of and used as raw material with which to build responsive interfaces, tools, spaces, and buildings.
For the first prototype of RGel, we established the hypothesis that we could synthesize the sensation of breaking a brittle material such as glass or a cracker. Read more about our initial vision in the proposal. Find more technical details and follow our progress at the progress blog.
1. "Every idea has a material solution" reads their slogan
2. As a designer from the future might put it, 'by its current reductive definition'
Technical
The first RGel prototype consists of an 8"x10"x2" slab of silicone gel laced with electronics. It detects contortions with a grid of four fiber optic bend sensors, which are used to drive software sound synthesis. The sounds generated by the synthesizers are then used to control an array of vibrotactile actuators in the gel.
Fiber optic bend sensors were chosen over traditional bend sensors and paper sensors mainly for aesthetic and educational reasons. They were constructed using an IF-E96 LED (emitter) and IF-D91 photodiode (receiver) from Industrial Fiber Optics with 1mm plastic fiber optic cable purchased from Abra Electronics. To create a bend sensor, the jacket was stripped completely and the cladding scraped off on one side of the fiber optic cable to allow a small amount of light to escape. Once this is done, when the cable is bent in the direction away from the removed cladding, more light is allowed to escape, while when the cable is bent toward the removed cladding, more light is retained. This allows for bi-directional sensing, unlike traditional bend sensors. By reading the voltage difference generated by the photodiode in response to the light fluctuation, a measure of the deformation of the fiber optic cable can be acquired. The voltage difference was read by an Arduino, an ATmega-based microcontroller which streamed the data to a laptop over USB.
Max was used to process the sensor data. Because the amount of light leaked while bending away from the scraped cladding is greater than the amount of light retained while bending toward the scraped cladding, an algorithm for measuring and compensating for this effect was devised. In addition to typical smoothing and scaling operations, the signal was rectified to give identical values for bending "up" and "down". Future iterations will take advantage of the bi-directionality of the sensors. The cooked sensor data was then mapped to a function resembling f(x)=(log(x)+2)/2 using the Max "object" [function] to yield a sharp onset.
Synthesis was implemented in Max/MSP. The first presentation of the device used a signal file playback buffer and an instance of IRCAM's Modalys running a plate simulation. Sound from the file playback buffer was used to excite the plate and the output consisted of a mix of both signals. The physical modeling was removed for the second presentation. We plan to use multiple playback buffers and a physical model with an array of spatialized exciters in the future. By placing the exciters at different locations on the plate corresponding to the locations of the bend sensors in the gel, we can create an accurately spatialized multi-channel output signal. Audio from the single playback buffer was split into four channels and filtered appropriately for the destination actuators. Audio sent to the speakers was highpass filtered and DC offset was removed while audio sent to the inertial motors was lowpass filtered. A fifth channel intended for the DC vibration motor consisted of a comparator and a lowpass filter. We found that large pager-type vibration motors respond best to DC pulses.
To retain the DC and sub-sonic frequencies in the output signal, we bypassed the "DC blockers" (highpass filter capacitors) in a Presonus Firebox firewire audio interface. This was done easily by soldering the leads together on the PCB. The DC blockers are the 10 identical capacitors on the top PCB of the device, off to one side and grouped in fours.
Choosing vibrotactile actuators (vibration motors) was the most difficult portion of the project. We conducted in depth research into the available options and acquired as many as we could for testing. Unfortunately due to shipping delays, we were not able to test some actuators before making the mould and we had to make do with what we had. Among the pager motors that we ordered were 12mm and 20mm motors from Precision Microdrives and a generic 20mm motor purchased from All Electronics.
In the prototype, we used two hand-made "inertial" motors and one hand-made "pager" motor (DC motor with eccentric weight). The inertial motors were designed by Hsin-Yun Yao, Vincent Hayward, and Randy E. Ellis at the McGill Haptics Laboratory for the Microtactus project. The parts for the motors were generously given to us by Joseph Malloch from the Input Devices and Music Interaction Laboratory who had manufactured some for use in digital musical instruments. They allow for extremely accurate reproduction of audio signals from the subsonic (<20hz) domain to well into the sonic domain (<1khz). Unlike pager motors, they have excellent transient response which allows them to easily transduce complex signals. The larger pager motor was added to provide "kick" to the signal. By filtering the audio through a comparator, we were able to send short DC pulses that correlated to transients in the signal that provided extra impact. (Researching vibration motors, we found that some had a 'nominal vibration' rating measured in g's. We were not sure what to make of this scale–how much vibration was necessary to achieve what we wanted? After receiving a few rated motors after the prototype was completed, we discovered that even a motor rated at 4g was still insufficient to vibrate the gel in a visceral way (most pager motors are rated at less than 1.5g). Only the larger motor for which we milled a weight had the power we were looking for, but unfortunately we are not able to provide a measure of the vibrational amplitude. What would have worked best would have been large inertial motors. The inertial motors are exceptionally powerful for their size and were they about twice the size, they would combine the best of both worlds: amazingly fast response to changes in voltage and high (hopefully sufficient) vibrational amplitude for their size. Unfortunately we do not have access to the equations governing the proportions of these motors, so without help from the designers, we may not be able to produce larger ones. The options, then, would be to reduce the size of the silicone and aim for a subtler effect, use more inertial motors coupled with large pager motors, or give up on the inertial motors completely.)
In addition to the vibration motors, we used a pair of ordinary speakers (102-1549-ND) to reproduce some of the signal aurally. This was done to experiment with coupling between vibrotactile and aural phenomena. To drive the speakers and the vibration motors, we used a simple 1W, 5V amplifier designed by Mark T. Marshal using the TDA7052 IC. Due to the high current loads required, we used a MOSFET to control the large pager motor. A 5V voltage regulator was used to step down the voltage from a 12V power supply for the amplifiers.
The components were soldered together and set in a rectangular slab of silicone gel. To do this, we constructed a resize-able mould from plexiglass based on advice from the silicone dealer. After trying a number of different silicone products at Sial in Laval, QC, we settled on Smooth-On Dragon Skin 10, a high quality water-clear silicone with a shore hardness of 10A as our choice of silicone.
Originally we had planned to sew the electronics onto a piece of cloth or vinyl, but for the first prototype we experimented with laying the electronics directly into the silicone. This proved difficult as the components floated or, in the case of the bend sensors, insisted on retaining their bent shape. Nevertheless, the final product was functional.
Components
- 4x LED and photodiode mounted in barrel-mounts from Industrial Fiber Optics
- 1mm plastic fiber optic cable
- custom software written with Max/MSP/Jitter
- Presonus Firebox audio interface with DC blockers removed
- 2x custom fabricated "inertial" motors
- 1x generic ~30mm DC motor with custom milled eccentric weight (pager motor)
- 2x 38mm low-profile speakers (102-1549-ND)
- 4x TDA7052-based amplifier circuits
- 4x photodiode amplifier circuits
- 2 lbs. Smooth-On Dragon Skin 10 silicone gel
- plexiglass for display stand and adjustable mould
Conclusion
Though constructing the first RGel prototype, we learned an immense amount about vibrotactile actuators, bend sensing, silicone moulds, analog signal conditioning, and driving motors in atypical ways (i.e. with sound). In constructing the prototype, we made a number of mistakes that rendered the device only semi-functional, but the mistakes have been identified and a second prototype is being planned that will correct the mistakes. In the second prototype, we will: (1) use less silicone and/or stronger motors. The silicone turned out to be too much mass for the vibration motors to actuate sufficiently and we were not able to achieve the desired vibrotactile effects, (2) sew the components onto a piece of cloth before placing them into the silicone to ensure even distribution and the appropriate shape, (3) experiment more with physical synthesis models and a large number of actuators to achieve more convincing spatialization, (4) develop a better signal conditioning circuit for the bend sensors (we were measuring voltage fluctuations of only 10mV from the photodiode). We believe that a new prototype taking these lessons into consideration will achieve our goal of synthesizing the sensation of deforming a brittle object.
References
- Natural Interactive Walking, vibrotactile floor tiles, by Yon Visell and Jeremy Cooperstock at the Shared Reality Lab
- Semblance, art installation experimenting with full body 'just noticeable' vibrotactility by Chris Salter, Marije Baalman, Harry Smoak, etc. (in production)
- TapTap [PDF], wearable that records touches and replays them using vibration motors, by Jeff Lieberman, Oritz Zuckerman, Cati Vaucelle, Leonard Bonanni at the MIT Media Lab and Harvard
- Microtactus, a surgical instrument for amplifying texture, by Hsin-Yun Yao, Vincent Hayward and Randy E. Ellis
Acknowledgements
We would like to thank our mentors and teachers, Vincent Leclerc, Gary Scavone and Sha Xin Wei for immeasurable guidance. Elio Bidinost from the Sensor Lab contributed sage advice and patient help without which the project could not have been realized. We would also like to thank Martin Peach and Avrum Hollinger for advice on constructing fiber optic bend sensors as well as Joseph Malloch and Dr. Marcelo Wanderly for helping us build and contributing the inertial motors.
–Morgan Sutherland & Faiq Hussein
