Biomedical Engineering Reference
In-Depth Information
whereas
V
ref
and diode-connected resistors provide the DC operational bias. The OTA is a typical
two-stage OTA with a P-type input differential pair loaded with Wilson current mirrors followed
by a class A output stage.
Table
7.2
shows the performance of our current designs per electrode using 0.5-µm technol-
ogy. Better performance can be obtained with newer technologies. Dies can also be wired bound to
the circuit board to minimize the physical size and avoid noise.
7.3 ThE PICo SySTEM
We have developed and tested a data acquisition/preprocessor/wireless system to implement mode
2. We called it the PICO system, and its goal is to acquire neural recordings from the headstage,
amplify the signals, perform spike detection (feature extraction/compression), spike sorting, and
wirelessly transmit binned data (rate codes) to a remote station. Alternatively, this system can sim-
ply amplify the raw neural data, do spike detection with a simple threshold, and send just these
windows to a remote system. As is, it already plays an important role in BMIs because it implements
a first-generation backpack. However, it does not allow the neurophysiologist to analyze and store
the continuous raw voltage traces being collected because of bandwidth limitations (mode 1), nor
it is capable of performing the high speed computations of mode 3. This is where a neural signal
processor (NSP) (local decoding algorithm implementation) will add capabilities in the future by
performing data reduction at the frontend such that the data can be reconstructed in the backend
for visualization and archiving purposes. In the particular path described here, the previously re-
ported Pico Recording system [
52
] is capable of replacing a multiconductor commutator cable used
in animal experimental paradigms with a high-bandwidth, low-power wireless interface that allows
the subject more mobility and a more natural response to a given experiment. This new wireless
interface presents a compact architecture as well as new challenges related to the power consump-
tion of the remote device.
To optimize the trade-off between power consumption and processing power, the low-
power, low-noise, high-gain amplifier shown in Section
7.2
was designed in conjunction with
off-the-shelf components, specifically MSP430F1611 as the microcontroller and the Nordic
nRF2401 wireless transceiver. The processor consumes less than 2 mW of power with 8 million
instructions per second of processing and has the added benefit of conserving space by integrating
two universal asynchronous receiver/transmitters, three direct memory access channels, and an
eight-channel, 12-bit A/D converter all in one small package. In this work, we demonstrate the
integration of the low-power amplifier with the Pico system and collect data from a live subject.
The Pico Neural Data Collection System consists of the Pico Remote (battery-powered neural
amplifiers, a neural data processor, and a wireless transceiver), the Pico Receiver (a wireless-transceiver
to USB converter), and PC-based software (for spike sorting and firing rate quantification). Figure
7.8
