Those LED cubes are cool, but also very directional because of the general shape of LEDs, their lenses and diffusers. Does anyone make a spherical, omnidirectional-view LED with axial leads, which would make construction easier?
I can envision how to build an LED globe using an arc of LED's and a micro-controller. I'd use two pins to control the motor and the rest to control the LED's. By pretending the LED's are in rows and columns you use one pin to select a row and another pin to select a column. So 36 LED's in the arc would only need 12 pins, plus 2 for the motor for 14 pins.
But LED cubes mystify me. Even an 8 x 8 x 8 cube would seem to require an enormous number of pins. You could use one of eight pins to select the level of the cube, then then you would need 64 pins to address all the LED's within a level. Unless you could again select a row and column, but that's the part that mystifies me.
I know charlieplexing can allow many more LED's to be addressed, but it seems unreliable.
How are you generating that third state? How will it be detected?
Hi Leon. Any method to get 50%, such as PWM or an external resistor divider circuit will work. Detection can happen with LEDs showing states, as in the example to read numbers, or as in states read by the chip in input mode. With so many states it seems like the method should have numerous applications.
I've seen a similar technique used in robotics - with high and low outputs used for motor direction, and the pin switched to input to stop the motor. The input state was detected with an op-amp.
For light loads (e.g. LEDs), it's even easier. The micro's output structure is already one half of an H-bridge. All you have to provide is the other half:
An artificial neuron could be created with a simple training mode and a using mode would drive it. It would have many inputs and one output. A neuron can be trained to fire or not fire based on input patterns. When a learned pattern is confirmed, associated output becomes the current output. If the input pattern does not belong in the taught list of input patterns, the firing rule is used to determine whether to fire or not. Perhaps inputs could be trinary so neurons could be multiplied using the state method described. After that, it's very complicated.
Comments
-Phil
http://www.coolstuffexpress.com/store/p/227-3D-Pin-Impression-Art.html
And where is the heat going to...
http://wn.com/3D_led_display_globe
http://www.seekway.com.cn/ledsys9.htm
But LED cubes mystify me. Even an 8 x 8 x 8 cube would seem to require an enormous number of pins. You could use one of eight pins to select the level of the cube, then then you would need 64 pins to address all the LED's within a level. Unless you could again select a row and column, but that's the part that mystifies me.
I know charlieplexing can allow many more LED's to be addressed, but it seems unreliable.
approximately millions and millions of pin states
The idea is to find practical applications for 32 pins which have 1x10^15 pin states. (The state of one pin can range from on, off, or 50% biased.)
This is all I can think of (it seems too vague)
- Signal an external device (Robot)
- Set up a giant matrix (LED Sculpture)
- Display large numbers on 32 LEDs
- Output Directions of a large map
Thank you for your replies.http://forums.parallax.com/showpost.php?p=942590&postcount=4
-Phil
Are the neural net experts going to chime in too?
An artificial neuron could be created with a simple training mode and a using mode would drive it. It would have many inputs and one output. A neuron can be trained to fire or not fire based on input patterns. When a learned pattern is confirmed, associated output becomes the current output. If the input pattern does not belong in the taught list of input patterns, the firing rule is used to determine whether to fire or not. Perhaps inputs could be trinary so neurons could be multiplied using the state method described. After that, it's very complicated.