Stephen P. DeWeerth

Biologically Inspired Joint Stiffness Control

Biological systems are able to perform movements in unpredictable environments more elegantly than traditionally engineered robotic systems. A current limitation of robotic systems is their inability to simultaneously and independently control both joint angle and joint stiffness without electromechanical feedback loops, which can reduce system stability. In this paper, we describe the development and physical implementation of a servo-actuated robotic joint that uses antagonistic, series-elastic actuation with novel nonlinear spring mechanisms. These mechanisms form a real-time mechanical feedback loop that provides the joint with angle and stiffness control through differential and common-mode actuation of the servos, respectively. This approach to joint control emulates the mechanics of antagonistic muscle groups used by animals, and we experimentally show that it is capable of independently controlling both joint angle and joint stiffness using a simple open-loop control algorithm.

A CMOS programmable analog memory-cell array using floating-gate circuits

The complexity of analog VLSI systems is often limited by the number of pins on a chip rather than by the die area. Currently, many analog parameters and biases are stored off-chip. Moving parameter storage on-chip could save pins and allow us to create complex programmable analog systems. In this paper, we present a design for an on-chip nonvolatile analog memory cell that can be configured in addressable arrays and programmed easily. We use floating-gate MOS transistors to store charge, and we use the processes of tunneling and hot-electron injection to program values. We have fabricated two versions of this design: one with an nFET injection mechanism and one with a pFET injection mechanism. With these designs, we achieve greater than 13-bit output precision with a 39-dB power-supply rejection ratio and no crosstalk between memory cells.

A multiconductance silicon neuron with biologically matched dynamics

We have designed, fabricated, and tested an analog integrated-circuit architecture to implement the conductance-based dynamics that model the electrical activity of neurons. The dynamics of this architecture are in accordance with the Hodgkin-Huxley formalism, a widely exploited, biophysically plausible model of the dynamics of living neurons. Furthermore the architecture is modular and compact in size so that we can implement networks of silicon neurons, each of desired complexity, on a single integrated circuit. We present in this paper a six-conductance silicon-neuron implementation, and characterize it in relation to the Hodgkin-Huxley formalism. This silicon neuron incorporates both fast and slow ionic conductances, which are required to model complex oscillatory behaviors (spiking, bursting, subthreshold oscillations).

Analog VLSI circuits for stimulus localization and centroid computation

An analog aggregation network that extracts the position of a stimulus in a sensory field is presented. This network is integrated with photodiodes in a VLSI circuit that performs stimulus localization through the computation of the centroid of a visual image. In this implementation, bipolar transistors and global subtraction are used to produce a high-precision centroid implementation. Theory for the localization of a bright visual stimulus is developed, and the theoretical predictions are compared to experimental data taken from the 160×160-pixel centroid circuit. Finally, the applications of these circuits to more complex feature extraction and to sensorimotor feedback systems are discussed.

Stimulus-Artifact Elimination in a Multi-Electrode System

To fully exploit the recording capabilities provided by current and future generations of multi-electrode arrays, some means to eliminate the residual charge and subsequent artifacts generated by stimulation protocols is required. Custom electronics can be used to achieve such goals, and by making them scalable, a large number of electrodes can be accessed in an experiment. In this work, we present a system built around a custom 16-channel IC that can stimulate and record, within 3 ms of the stimulus, on the stimulating channel, and within 500 mus on adjacent channels. This effectiveness is achieved by directly discharging the electrode through a novel feedback scheme, and by shaping such feedback to optimize electrode behavior. We characterize the different features of the system that makes such performance possible and present biological data that show the system in operation. To enable this characterization, we present a framework for measuring, classifying, and understanding the multiple sources of stimulus artifacts. This framework facilitates comparisons between artifact elimination methodologies and enables future artifact studies.

Using a Hybrid Neural System to Reveal Regulation of Neuronal Network Activity by an Intrinsic Current

The generation of rhythmic patterns by neuronal networks is a complex phenomenon, relying on the interaction of numerous intrinsic and synaptic currents, as well as modulatory agents. To investigate the functional contribution of an individual ionic current to rhythmic pattern generation in a network, we constructed a hybrid system composed of a silicon model neuron and a heart interneuron from the heartbeat timing network of the medicinal leech. When the model neuron and a heart interneuron are connected by inhibitory synapses, they produce rhythmic activity similar to that observed in the heartbeat network. We focused our studies on investigating the functional role of the hyperpolarization-activated inward current (Ih) on the rhythmic bursts produced by the network. By introducing changes in both the model and the heart interneuron, we showed that Ih determines both the period of rhythmic bursts and the balance of activity between the two sides of the network, because the amount and the activation/deactivation time constant of Ih determines the length of time that a neuron spends in the inhibited phase of its burst cycle. Moreover, we demonstrated that the model neuron is an effective replacement for a heart interneuron and that changes made in the model can accurately mimic similar changes made in the living system. Finally, we used a previously developed mathematical model (Hill et al. 2001) of two mutually inhibitory interneurons to corroborate these findings. Our results demonstrated that this hybrid system technique is advantageous for investigating neuronal properties that are inaccessible with traditional techniques.

A PDMS-Based Integrated Stretchable Microelectrode Array (isMEA) for Neural and Muscular Surface Interfacing

Numerous applications in neuroscience research and neural prosthetics, such as electrocorticogram (ECoG) recording and retinal prosthesis, involve electrical interactions with soft excitable tissues using a surface recording and/or stimulation approach. These applications require an interface that is capable of setting up high-throughput communications between the electrical circuit and the excitable tissue and that can dynamically conform to the shape of the soft tissue. Being a compliant material with mechanical impedance close to that of soft tissues, polydimethylsiloxane (PDMS) offers excellent potential as a substrate material for such neural interfaces. This paper describes an integrated technology for fabrication of PDMS-based stretchable microelectrode arrays (MEAs). Specifically, as an integral part of the fabrication process, a stretchable MEA is directly fabricated with a rigid substrate, such as a thin printed circuit board (PCB), through an innovative bonding technology-via-bonding-for integrated packaging. This integrated strategy overcomes the conventional challenge of high-density packaging for this type of stretchable electronics. Combined with a high-density interconnect technology developed previously, this stretchable MEA technology facilitates a high-resolution, high-density integrated system solution for neural and muscular surface interfacing. In this paper, this PDMS-based integrated stretchable MEA (isMEA) technology is demonstrated by an example design that packages a stretchable MEA with a small PCB. The resulting isMEA is assessed for its biocompatibility, surface conformability, electrode impedance spectrum, and capability to record muscle fiber activity when applied epimysially.

An Integrated System for Simultaneous, Multichannel Neuronal Stimulation and Recording

Precision electronics that provide multi-electrode stimulation and recording capabilities are an important tool for the experimental study of neuronal development and plasticity. Towards this end, we present a custom analog integrated circuit (IC), fabricated in a 0.35-mum process, incorporating stimulation buffers and recording preamplifiers for multiple electrodes onto a single die. The architecture of the IC allows for arbitrary, independent configuration of electrodes for stimulation or recording, and the IC includes artifact-elimination circuitry that returns the stimulation electrode to its previous voltage following stimulation, minimizing the interference with recording. We analyze the thermal noise levels in the recording preamplifiers and experimentally measure input-referred noise as low as 4.77 muVrms in the frequency range of 30 Hz-3 kHz at a power consumption of 100 muW from a total power supply of 3.8 V. We also consider the temporal response and stability of the artifact elimination circuitry. We demonstrate that the use of the artifact-elimination circuitry with a 30-mum diameter stimulation electrode permits a return to recording mode in les 2 ms after stimulation, facilitating near-simultaneous stimulation and recording of neuronal signals. (Patent applied for, U.S. No. 2007/0178579.)

Object-based selection within an analog VLSI visual attention system

An object-based analog very large-scale integration (VLSI) model of selective attentional processing has been implemented using a standard 2.0-/spl mu/m CMOS process. This chip extends previous work on modeling a saliency-map-based selection and scanning mechanism to incorporate the ability to group pixels into objects. This grouping, or segmentation, couples the circuitry of the objects pixels to act as a single, larger pixel. The grouping of pixels is dynamic, driven solely by the segmentation criterion at the input. In this demonstration circuit, image intensity has been chosen for the input saliency map and the segmentation is based on spatial low-pass filtering followed by an intensity threshold. We present experimental results from a one-dimensional implementation of the object-based analog VLSI selective-attention system.

Nonlinear antenna technology

Nonlinear antennas combine advances in nonlinear dynamics, active antenna design, and analog microelectronics to generate beam steering and beam forming across an array of nonlinear oscillators. Nonlinear antennas exploit two phenomena typically shunned in traditional designs: nonlinear unit cells and interelement coupling. The design stems from nonlinear coupled differential equation analysis that by virtue of the dynamic control is far less complex than the linear counterparts by eliminating the need for phase shifters and beam forming computers. These advantages arise from incorporating nonlinear dynamics rather than limiting the system to linear quasisteady state operation. A theoretical framework describing beam shaping and beam forming by exploiting the phase, amplitude, and coupling dynamics of nonlinear oscillator arrays is presented. Experimental demonstration of nonlinear beam steering is realized using analog microelectronics.