Project Hardware

The project has involved the use of a lot of hardware. Programming and Matlab simulation for the chip was done on two wintel machines: my PII-400 desktop and my P90 laptop running Windows 95/8; some work was also done in Bruce Land's lab in Mudd Hall. The actual code of the neuron resides in an Atmel AT90S1200 8-bit MCU, a very simple and cheap (about $1.80) microcontroller. The chip consists of 1k of flash program memory, 256 bytes of EEPROM, and 16 i/o ports. The chip has no other memory or computing power. Although very simple, when clocked at 4 or 8MHz, it is more than powerful enough to run the neuron simulation code, which needs around .8ms with 4MHz and .4ms on the 8MHz clock. The chip is programmed on an Atmel STK200 development board, which gets hooked up via parallel cable to the computer, and downloaded with the Atmel ISP software.

Once the 1200 has been programmed on the STK200, I took them off and placed them onto a R.S.R. Electronics Analog/Digital Trainer Pad, which is a glorified prototype board with fancy switches and LEDs, which was wonderful for debugging a small neural network made out of only five neurons (used to control the swimming of the robotic fish). Analog and frequency data was observed using a HungChang PC Scope Card 200, an oscilloscope for the PC.

The servos for the robot fish movement are Acroname standard servos which are small, light, and cheap--perfect for this project!