Hardware components | ||||||
 |
| × | 1 | |||
WORK IN PROGRESS
Hey everyone seeing this -
This page is still very much under development and is subject to change & frequent updates. I wasn't expecting the site to be publicized yet and haven't gone through any sort of revision process yet. At this point, it's more of a brain dump than any sort of artist statement.
Also, feel free to suggest a better name. I'm still trying a few out. [email protected]
-Eva
Motivation
This project began when I was trying to explain what a panic attack felt like to someone who had never experienced one before. Realizing how little of the intense feelings could be described through words, I wanted to find a medium more suited for this type of ‘emotional explanation.’ My first thought was an audio/visual display that could simulate different emotions in a first-person perspective. I then realized that there was no way to monitor the engagement of the audience without some sort of feedback from the person experiencing the display - they would have to be somehow actively engaged with the performance in order to experience it properly.
This necessitated two additions to the project: biometric sensors and a feedback system. Feedback is when a system tries to match a control input with a sensed input though changing some sort of actuators. In this case, the control input is the desired emotion, the sensed input is some sort of biometric sensor, and the actuators are audio and video.
This biometric feedback system could easily be extended to cover emotions other than anxiety and is only limited by the data the sensors can convey.
Implementation
To create an intimate, sensory-overwhelming experience, I decided that the participant in the performance should be isolated within a relatively small space. A box of 1m wide, 1m long, and 2m tall was designed to accommodate heights up to 6’6”, which is the vast majority of people.
The walls of the box are adorned with 8192 individually controllable LEDs which essentially creates a gigantic video display. While the resolution of the display is only 1 pixel per inch, a diffused panel of acrylic ensures that there are no gaps in the light.
The audio portion of the experience is implemented through sub-bass speakers positioned beneath and above the participant. These speakers lend themselves more to feeling rather than hearing, which further helps with immersion.
The simplest sensory feedback mechanism is a heart rate sensor. While not the most precise sensor, it is relatively unintrusive and gives a good idea whether the subject is calm or not. It also creates a framework from which a more complex sensor such as an EEG can be developed.
I use the Max/MSP software package to manage all inputs and outputs. The sensor data is fed through an Arduino Due microcontroller and the display information is sent to an array of Teensy 3.1 microcontrollers running the OctoWS2811 library.
Mechanically, the box is constructed out of 2cm X 2cm extruded aluminum. Eight frames make up the four walls of the box. Each frame has 16 LED circuit boards that in turn have 64 LEDs each. This modular approach facilitates a flexible platform that can be easily reconfigured for other display formats (e.g. a 1m X 8m short wall).
Block diagram
Circuit board art
PCB quotes
Vendor Price Stencil Smart-Prototyping $1881.00 Yes
Itead Studio $2243.00 Yes
PCBcart $2524.35 No
Project-PCB $2050.13 No
PCBmain $2148.34 No
RayMing $1759.00 No
Keije $1419.00 No
I decided to go with Smart-Prototyping as they have the lowest price of all well-reviewed PCB fabricators.
Eagle files + Gerbers
Frame design
The frame of each 1m x 1m section was constructed from extruded aluminum. This material was chosen for its easy reconfigurability and the ability to purchase custom-cut sections for relatively cheap. 2cm x 2cm cross-sections offered the best tradeoff between structural stability and price. Four 2cm x 2cm x 980cm extrusions form the outside frame and three 2cm x 4cm x 960cm extrusions provide rails to mount PCBs and acrylic. Slide-in nuts are used to fasten PCBs in place and support the acrylic via standoffs. The 4x4 PCB pattern was a compromise between PCB quantity and size - both of which have a large impact on the overall cost.
Shopping list
Description Supplier Part# Quantity Price ea. Total price WS2812B Alibaba WS2812B 10000 $0.13 $1300.00 Acrylic TAP Plastics Sign lighting white 60% 50cm x 50cm 32 $19.86 $635.52 2cm x 2cm x 98 cm aluminum extrusion Misumi HFS5-2020-980 32 $5.58 $178.58 2cm x 4cm x 96 cm aluminum extrusion Misumi HFS5-2040-960 24 $10.36 $248.64 Thin brackets w/ nuts & screws Misumi HBLSS5-SET 80 $1.80 $144 Insertion nuts (M3) (pack of 100) Misumi PACK-HNTTSN5-3 6 $28 $168 Pan head screw (M3 x 6mm) Misumi HCBM3-6 512 $0.06 $30.72 Connecting metal strip w/ nuts & screws Misumi SHPTSD5-SET 16 $3.80 $60.80 5V 60A power supply w/ PFC Bravo Electro Components RSP-320-5 8 $59.30 $474.40 USB microcontroller PJRC Teensy 3.1 8 $19.30 $154.4 Buffer chip (74HCT245) Digikey 296-1208-1-ND 10 $0.433 $4.33 Standoff - 18mm x 5mm (M3) Digikey 952-1508-ND 150 $0.44 $66.00 Crimp contact Digikey A100453CT-ND 1300 $0.03384 $43.99 100nF 25V X7R 0805 cap Mouser 80-C0805C104K3R7210 10000 $0.009 $90.00 Male 2p header vertical Mouser 571-6404562 500 $0.05 $25.00 Male 2p header horizontal Mouser 571-6404572 250 $0.075 $18.75 Female 2s socket Mouser 571-13758202 650 $0.064 $41.60 6A 32V fast blow fuse Mouser 637-F1206FA6000V024 150 $0.191 $28.65 Cable 2 x 22AWG twisted/shielded 500ft Mouser 566-5500FE-U500-08 1 $52.53 $52.53 Male header 2 x 14p 2.54mm spacing Mouser 649-68000-414HLF 20 $0.403 $8.06 Female header 2 x 14s 2.54mm spacing Mouser 517-929974-01-14-RK 20 $1.38 $27.60 75 ohm 1% 0805 resistors Mouser 71-CRCW0805-75-E3 100 $0.027 $2.70 Total = If you have to ask....
Cabling
Cabling is surprisingly tricky in a project like this, both for power transmission and signal integrity.
For power transmission, it is important to use wires that are rated at a suitable ampacity to prevent fusing or arcing. It's also critical that wire with a low enough resistance is chosen to prevent voltage droop when high currents are drawn. Calculators like this one are great tools for figuring things like this out. With it I was able to calculate that 22AWG wire would be sufficient for the 3.84A maximum of each LED panel.
Although the WS2812B LEDs have a data rate of only 800kHz, it is important to preserve the integrity of the odd harmonics in the square wave signal which are in the MHz range. At those frequencies, transmission line effects such as voltage reflections come into play. From the cable's datasheet, R=16mΩ/ft, L=180nH/ft, C=33pF/ft. The conductance can be estimated to be in the micro- to nanoSiemens range and can be ignored in calculations. The characteristic impedance Zo = sqrt((R + jωL)/(G + jωC)) or approximately sqrt(L/C). This works out to almost exactly 75Ω. Ideally, the source (buffer chip) would have a resistance of 75Ω and the load would have a resistance of 75Ω as well. Unfortunately while this would be ideal from a signal integrity point of view, it would also function as voltage divider and make the LEDs non-functional :( As a compromise, only the 75Ω source resistance will be used as suggested in the OctoWS2811 library.
Power supplies
For a project of this magnitude, it is essential to carefully plan out the power supplies. Each WS2812B LED consumes around 50mA at full brightness. For 8 * 8 * 128 = 8192 LEDs, this is an incredible 410A. To minimize voltage droop, eight 5V 60A supplies are used with each one supplying 1024 LEDs. These supplies are rated for an efficiency of 85% and will draw 410A * (1 / 0.85) * (5V / 120V) = 20A from the mains outlet. It's crucial to plug them into a suitable circuit or a fuse will likely be tripped.
In addition to supply current, it's also important to consider the power factor of the system since there is such a large load. Many switch-mode power supplies without built in power factor correction (PFC) have poor power factors, though this is not usually an issue unless many are connected to the same circuit. I opted to use power supplies with active PFC despite the increased cost because it would be a tricky issue to debug with non-PFC supplies.
Costs spreadsheet/BOM
PCB Yield
A combination of relatively cheap PCB fabbing, 90s-era pick-and-place, and notorious unreliable LEDs means that a 100% yield for PCBs would be impossible. Thus far in testing, the results are surprisingly good!
Status:
1 Board half hand assembled
1 Board toasted by incorrect reflow settings
1 Board without any caps (ran out)
26 Boards in need of rework
121 Tested functioning boards
Yield = ~80%
Comments