- 1 Orb Illuminator Design Files
- 2 Orb Illuminator-Final Design Synopsis
- 3 Orb Illuminator Acivity post Burning Man 2007
- 4 Design Notes
Orb Illuminator Design Files
With the addition of a second, revised set of illuminators to each Orb for Burning Man 2008, the illuminator hardware design is complete.
All Illuminator design files and source code are checked into and available from our Subversion system: http://svn.lee.org/swarm
Orb Illuminator-Final Design Synopsis
The Orb illuminators are based on the Atmel Mega8 microcontroller; board power is supplied by a 78L05 regulator. Hardware PWM is used to drive the discrete red, green, and blue LEDs. The LEDs are arranged in complimentary triangles to provide better color mixing. In addition, these LEDs have a 30 degree radiation pattern which assists with color mixing. A clever technique devised by Jon allows the LEDs to operate at 24v or 12v depending upon dropping resistor value and placement. Each Orb illuminator is connected to the serial bus via a MAX232 chip. This serial line is set up so that serial data can pass through to each illuminator and any other device in the serial chain. All the software for the illuminators was written in C by Jon.
Important Technical Notes
The new Orb Illuminators need to be powered if you are going to use a USB ISP programming wand to program them. The Universal Serial Bus does not have enough power to get the Mega8 into a programmable state. You have been warned :)
Fuse settings for the Illuminators (as of the latest rev)
- Ext. Crystal/Res. High Freq.;Start-up time: 16CK +4 ms [CKSEL=1111 SUT=10]
- BODEN ungprogrammed
- Brown-out detection level at VCC=2.7V; [BODLEVEL=1]
- Boot Reset Vector unprogrammed
- Boot Flash section size= 1024 (default value)
- Preserve EEPROM unprogrammed
- CKOPT unprogrammed
- Serial Program downloading enabled [SPIEN=0]
- WDT unprogrammed
- RESET unprogrammed
Orb Illuminator Acivity post Burning Man 2007
The illuminators worked very well on the playa. The light patterns generated by the light exiting the shells onto the playa were amazing! Two more illuminators are being added to each orb. They may replace the illuminators that are presently at the top of each orb. The old illuminators will be moved to the bottom of the orbs; these will be located on top of the motor controller at the front and back of each orb.
Note These are design notes taken during the development of the illuminators for SWARM. They may be outdated!
LED driver Designs
The most efficient way to do this might be to drive a buck regulator directly off the 24V bus. Now all we need to do is find or build one that is affordable.
Something like the IRS2540 is possible but the circuit looks darn complicated: http://www.irf.com/product-info/datasheets/data/irs2540pbf.pdf
The "buck puck" things look pretty ideal, but at $20 per channel that's too spendy.
Here's a simple hack but looks like it wastes a lot of power http://www.instructables.com/id/ERVLPDU5XVEWOF31U3/?ALLSTEPS
Note on another alternate method, posted by mark alexander: 1. It seems to be feasible to use a constant voltage DC-DC converter as a power source for high-current LEDs. I don't know if there are some runaway issues I just haven't hit yet, but when I've powered large banks of LEDs (100+ super ultra-bright), they run fine off a TI PTH08T240. I usually use a pot to adjust the output voltage until I've got the brightness level right, and then after experimentation I'll replace that with a fixed resistor.
Jon adds: this will work, but you need to be careful. Diode current goes exponentially with voltage, and temperature(!) and the voltage difference between optimal current and toast may be very small. Here are other reasons why it's generally good to use a current source: http://www.lunaraccents.com/technology-LED-driver.html
The LM2675-ADJ can be used in constant current configuration to drive one or more high powered LEDs. The on/off pin is PWM'd to control the brightness of the LED. Pros: I already have the boards for this design--originally a full RGB fiber optic illuminator. Uses an Atmel Mega8 and three of the LM2675 driver chips. Cons: Loss of some power through the current sense resistors. Board is 3.5" in diameter which takes up a lot of space. Can only drive two Atlas RGB LEDs at reduced output. The LM2675 is twice the price of the Supertex HV9910.
Supertex makes the HV9910 which is a very capable high power LED driver chip. It has a dedicated PWM pin for brightness control. Pros: Cheaper than the LM2675; runs about 90 cents each. Can drive LEDs with currents over one amp. Circuitry takes up much less space on board. Minimal power loss through current sense resistors. Easily drives two Lamina Atlas LEDs at full output (525ma) Cons: Some don't like that it uses an external MOSFET. I don't have a working prototype of this design yet. An application note for the HV9910 can be found here: AN-H50. I have a spreadsheet with all equations listed in the application note to aid in working with the HV9910. My new fiber optic illuminator design uses six HV9910s. This illuminator drives four Lamina Atlas LEDs (two per RGB channel). Update 4/19/07: This design is turning out to be very large; the board is now 5x6, I think we need to consider smaller alternatives.-Rick
I have a small RGB LED controller that is based on the Atmel ATTiny2313. It can drive six RGB LEDs per channel for a total of twelve RGB LEDs. It uses a very small buck switching supply to provide voltage to the LEDs and CPU. The board is already done and is presently sitting on a breadboard with some RGB LEDs, I can bring it to the next meeting. The one issue with this design is that it uses current limiting resistors and these waste power. Calculations show that the if we use twelve RGB LEDs in each module, we will be losing about 2 watts per module via the resistors. A big plus of this design is that it can be made quite small by deleting the switching power supply and replacing it with a linear regulator. The power/control components could go on one side and the LEDs on the other; just a thought. The Tiny2313 supports four hardware PWM channels that can be used to control the LEDs; this chip also has a bi-directional USART. We'll have to power the LEDs directly from the 24v rail on the orbs. 4/20 The board appears to have stopped working due to a problem with my lab supply. URGHHHH!
Here is a DIY RGB LED Controller based on a PIC. The latest design can be addressed via a serial line which is cool. This is sort of a 'food for thought' project that's close to the pod idea mentioned below.
RGB LED/LED Driver Chips
Jon suggested that we could also use dedicated RGB LED driver chips,a microcontroller, and RGB LEDs to make RGB LED pods. These pods would be distributed about the orb. The pods would need only power lines and serial links. Here are some RGB driver chips to consider; the links take you directly to the PDF datasheet:
STP08DP05 8Bit LED Sink Driver I have a nascent design based on this chip that I forgot about :).
Microcontroller buck regulators
Jon asks: if we already have a micro on board, why not use it as a buck regulator controller? Save parts and money.
Here's some app notes describing that approach, including a 3W lumiled driver (microchip): http://www.freescale.com/webapp/sps/site/overview.jsp?nodeId=01624684490DEC1BBD http://www.microchip.com/stellent/idcplg?IdcService=SS_GET_PAGE&nodeId=1824&appnote=en012124
Where are the LEDs? How many channels, x 3 for RGB?
- At least one under the battery ballast
- Uplight the interior of the sphere? How without glare?
- Central diffuser "heart"?
- External "headlights" -- running lights on ears, tail lights? Bucket handle?
- Point source inside the sphere should make REALLY NICE SHADOWS. Mount neart top of orb facing down to avoid glare?
- To save $$$, make illumination recognfigurable? E.g. one panel of ultrabright RGBs. Mount at top of sphere facing down, or move to axle underneath spherical "heart" diffuser.
- Discrete Red, Green, Blue LED boards to allow for color mixed light to be placed in multiple locations in the sphere.I have access to a design for these boards-Rick
- LEDs to indicate system status: battery charge, heat, radio link status, brain health, self-test. Any way to make these elegant, or are we left with Terminator eyes?
I've seen a chinese place on the web that looked really really cheap. I'll try to find it again
...This is probably it Chinese RGB LEDs. Part number 540R2GBC-CA (common anode)
Note on lumens/$ posted by Mark Alexander: Getting a significant amount of light out of LEDs is EXPENSIVE. I've been trying to find different ways of skinning this cat for the last year or two. Lumileds <http://www.lumileds.com> are very bright and very efficient, but they cost about $7 each for the 1W and $25 for the 5W (retail). Superflux LED modules <http://www.litewave.co.uk/superflux_led_modules.asp> seem like a good deal (and they contain their own driver), but they make less light than the Luxeon Lumileds. Dropping back to a more labor-intensive solution, I started building modules of 10-20 regular ultrabright LEDs ($.50 ea), and I think in lumens per dollar, it's cheaper than lumileds or superflux. I haven’t checked this with a lightmeter, but intend to.
The Lamina Atlas series of RGB LEDs provides an amazing amount of light at standard drive currents of 350 and 525ma. So bright that you cannot look at them without sunglasses! $33/each but well worth the price. When you want a good point source, these high powered RGB LEDs are the way to go. Inexpensive focusing optics and heatsinks are also available.
Piranha High Flux RGB LEDs These are brighter than the Chinese RGB LEDs. They are square and mount with four pins and are common anode. About $1.20 each in quantities of 25 depending upon exchange rates
Kids, LSDiodes has some seriously bright S-Flux square, four pin LEDs and the prices are very good.
5mm LED light reflectors LEDs will appear brighter with these
A nice Candela to Lumen Calculator. Lot's of other LED resources here,too.
Group Meetin' Notes
- A very cool color palette tool is available on-line from Adobe: Kuler
- Mark's lumens per dollar calculations showed that the Lamina Atlas high powered LED costs 19 cents per lumen; the inexpensive Chinese RGB LEDs cost 22 cents per lumen.
- The Chinese RGB LEDs are part # 599R2GBC-CA and are available from from HB, Ltd
- We propose that a single point source for illuminating the spheres would work best. We are going to set up a Lamina Atlas RGB LED and make a 70 RGB LED matrix to test which works better.
- The 70 RGB LED matrix will be roughly equivalent to the power of one Atlas RGB LED; we realize that there may be power issues with this configuration.
- The light source will be placed near the top of the sphere.
- The final controller/driver board will be based on an Atmel Mega88 and will use the HV9910 LED driver chip.
RGB vs. Discrete LEDs
4/22-I had an informative discussion with Jeremy Lutes creator of the Lilypond and many other amazing illuminated works. I have been working under the assumption that we'd be using RGB LEDs in arrays to do the illumination of the orbs. It turns out that it is much easier to use discrete LEDs when they are going to be clustered. RGB LEDs are much harder to lay out due to the need for resistors. One could use a constant current design, but this really makes things more complex. Jeremy suggests using discrete LEDs in a circular pattern and has lent us a sample board with this configuration. RGB LEDs are best used when you have lots of space between LEDs. When the LEDs need to be close together, discrete LEDs are fine as color mixing is not as much of an issue. Jeremy also lent us some discrete LED strips to play with. I will be bringing a Lamina BL-4000 (obsolete lil bro to the Atlas)to the meeting so that we can see a VERY bright point source in action
The round, discrete LED board is 2" in diameter and has interconnect points for Red, Green, and Blue. Jeremy suggested putting the control electronics on another board that is attached to the back of the LED board. This is what he did for Splash! at Burning Man 2001. This makes for a very compact little illuminator. Jeremy noted that we need to design the LED board so that it still work once the batteries begin to dip below 24 volts. When using discrete LEDs, each color of LED should have a similar radiation angle.
4-29 Making discrete RGB arrays:
- The LEDs should be as close together as possible to avoid color banding.
- A linear array of discrete LEDs would look like: R G B R G B R G B etc.
- A circular array of discrete LEDs uses triangles of LEDs with one color at each apex; the triangles mirror each other.