Integrated low noise, low power amplifiers, and control for the recording of electrocorticograms

Brain machine interfaces allow one to use one's thoughts to control the actions of a machine. To accomplish this, the interface must record and send brain signals to a signal processor to be decoded. The recording of brain signals such as the firing of individual nerves and electroencephalograms has been used for some time now to control machines. Recently it has been discovered that electrocorticograms are also a viable brain signal to control machines. This thesis covers the design and testing of the amplifiers and digital control for an integrated circuit that can record electrocorticograms and broadcast such data from multiple electrodes. The chip is powered wirelessly by an inductive link operating at 2.765 MHz. This same link uses amplitude modulation to send commands to the chip. Data can be collected from 100 electrodes. Each electrode is capacitively coupled to an amplifier. Each amplifier's output can be multiplexed to an ADC, digitized and then broadcast off chip via an RF transmitter. The chip was fabricated in a commercially available 0.6 (im, 2 poly, 3 metal BiCMOS process. The chip was tested, and all functions of the chip performed within their respective design tolerances. Specifically, the amplifier's bandwidth ranges from 0.05 Hz to a programmable high cut-off frequency of 79 Hz to 240 Hz. The amplifier has an electrode-referred noise of 3.5 jiW and requires 4.5 fiW of power. The transmission of data from multiple electrodes was also tested and it was found that individual electrode data could be reconstructed.