xDev: a mixed-signal, software-defined neurotechnology interface platform for accelerated system development

Abstract

Objective.Advances in electronics and materials science have led to the development of sophisticated components for clinical and research neurotechnology systems. However, instrumentation to easily evaluate how these components function in a complete system does not yet exist. In this work, we set out to design and validate a software-defined mixed-signal routing fabric, 'xDev', that enables neurotechnology system designers to rapidly iterate, evaluate, and deploy advanced multi-component systems.Approach.We developed a set of system requirements for xDev, and implemented a design based on a 16 × 16 analog crosspoint multiplexer. We then tested the impedance and switching characteristics of the design, assessed signal gain and crosstalk attenuation across biological and high-speed digital signaling frequencies, and evaluated the ability of xDev to flexibly reroute microvolt-scale amplitude and high-speed signals. Finally, we conducted an intraoperativein vivodeployment of xDev to rapidly conduct neuromodulation experiments using diverse neurotechnology submodules.Main results.The xDev system impedance matching, crosstalk attenuation, and frequency response characteristics accurately transmitted signals over a broad range of frequencies, encapsulating features typical of biosignals and extending into high-speed digital ranges. Microvolt-scale biosignals and 600 Mbps Ethernet connections were accurately routed through the fabric. These performance characteristics culminated in anin vivodemonstration of the flexibility of the system via implanted spinal electrode arrays in an ovine model.Significance.xDev represents a first-of-its-kind, low-cost, software-defined neurotechnology development accelerator platform. Through the public, open-source distribution of our designs, we lower the obstacles facing the development of future neurotechnology systems.

Keywords: neural interfaces; neurotechnology; system development; system integration.

Figure 1.

A conceptual overview of the present implanted device design process, and the flexible process using xDev. (a) Vendor-supplied evaluation boards or benchtop devices are subsystems connected using a proprietary backplane, which provides real-time configuration and monitoring to a host computer (PC). (b) Once a system is designed, the component subsystem integrated circuits (ICs) are laid out on a new printed circuit board (PCB) for evaluation. This is a rigid processing step, as changes in the subsystems, layout, or routing are complex. (c) In order to iterate system designs, or correct mistakes, entirely new circuit boards are required, before arriving at the final goal of an implanted device in panel (d). (e) Instead, using xDev, subsystems can either be vendor-supplied evaluation boards, benchtop systems, or custom modules that test a particular layout implementation. The modules are connected using a switch matrix, defined in software. Real-time oversight is provided by a microprocessor. (f) To iterate the system design, modules can be rapidly introduced or removed by updating the software definition for the switch matrix. Here, the microprocessor oversight is disabled, to simulate the modules being on their own PCB.

Figure 2.

Design and implementation of the initial xDev system. (a) A schematic block diagram of the xDev system. An Arduino Nano facilitates communication to the host computer (PC) over a universal serial bus (USB) interface using a universal asynchronous receiver/transmitter (UART) protocol. Programming of the AD75019 multiplexer is conducted by delivering the desired bitstream over general purpose input output (GPIO) pins, and power is supplied by the Arduino’s 5 V rail. The AD75019 uses digital logic components to configure the 16 × 16 crosspoint switch matrix, which enables arbitrary connection between two banks of 16 pins. (b) A photograph of the initial xDev system as tested here. (c) A photograph of the xDev system in an example deployment to deliver stimulation and record responses using an Arduino Nano 33 IoT and an Intan RHS2116 chip. (d) A photograph of the xDev system in an example deployment to control an application specific integrated circuit (ASIC).

Figure 3.

The xDev system demonstrates low direct current (DC) and alternating current (AC) impedance. For (a)–(d): Note, the multimeter uses a 5 V test voltage to determine DC impedance. (a) The distribution of DC line resistances observed for 3 boards. No significant difference in resistance distribution was found between the 3 boards examined (p > 0.05, Kruskal–Wallis test, Tukey–Kramer correction). (b) The histograms of DC line resistances observed for 3 boards. (c) The distribution of DC impedances mismatches computed using a bootstrapping method. No significant difference in resistance mismatch distribution was found between the 3 boards examined (p > 0.05, Kruskal–Wallis test, Tukey–Kramer correction). (d) The histograms of DC impedance mismatches between randomly selected lines. (e) An impedance spectrogram showing the distribution of impedance magnitude and phase mismatch between randomly selected lines (n = 10 000).

Figure 4.

DC impedance flatness performance and off-state resistance. (a) A circuit diagram showing the test circuit used to measure the resistance flatness for an xDev trace. When measuring in the reverse direction, the voltage source was connected to the Y gate and the resistor was connected to the X gate. (b) (top) Resistance-flatness curves for 3 boards. (bottom) Resistance flatness curves for a single connection on a single board, measured in the forward and reverse settings. Random samples are drawn by randomly establishing single connections on the board. (c) Off-state impedance measurements for an xDev trace.

Figure 5.

Configuration and switching time were minimized using xDev. (a) A histogram of the time taken to configure the switch matrix using the serial programming interface. (b) A circuit diagram showing the test circuit used to measure the switching time (that is, the time from a configuration being finalized to the connection being made). (c) A time-domain trace showing the configuration being finalized (falling edge of PCLK trace) and the connection being made (falling edge of V_I trace). (d) A histogram showing the connection time. The limit of time resolution is 0.2 ns at 5 Gsps.

Figure 6.

Charge injection performance achieved at safe levels. (a) A circuit diagram showing the test circuit used to assess the amount of charge injected during switching. The gate was opened and closed rapidly, allowing for the charge to accumulate in the 1 nF capacitor, which could be measured as a voltage. (b) The histogram of injected charge during 58 gate cycles. (c) A log-scale showing the measured injected charge, with safe limits for several electrode architectures overlaid in dashed lines (Nordhausen et al , Cogan et al , De uso n.d.). (d) A circuit diagram showing the test circuit used to measure the charge injection artifact during switching. The gate was closed, and the injected current flowed through the load resistor R, representing an electrode impedance. The resultant voltage was amplified, and recorded. (e) The waveforms of the injected currents for each load resistance. Solid lines are the mean, and the shaded region is ±1 standard deviation. The PCLK pulse closed the switch at t = 0, indicated by a dashed grey line.

Figure 7.

Frequency response of xDev demonstrates minimal attenuation and distortion. (a) A diagram showing the test setup to measure the frequency response. (b) A circuit diagram of the test circuit used to measure the frequency response. (c) A magnitude and phase plot of a single xDev trace, with randomly selected traces overlaid as a scatter plot. The frequency range common for biosignals is shaded, and common digital communication fundamental frequencies are overlaid as dashed lines. Several time-domain examples are inset, to illustrate the change in signal magnitude and phase observed at different frequencies.

Figure 8.

Ethernet network connection delivers high-bandwidth data at sufficient speed. (a) A diagram showing the experimental setup to measure the Ethernet performance of xDev. (b) A diagram showing the different trace configurations tested. (c) A box plot showing the distribution of Ethernet speeds achieved through the three configurations. No significant difference in Ethernet speed was found between the 3 boards examined (p > 0.05, Kruskal–Wallis test, Tukey–Kramer correction), however including xDev in the communication path reduced the observed connection speed compared to the ‘no xDev’ control configuration (p < 0.0001, Kruskal–Wallis test, Tukey–Kramer correction). (d) Histograms showing the distribution of Ethernet speeds achieved through the three configurations.

Figure 9.

xDev crosstalk was minimal at biologically relevant frequencies. (a) A circuit diagram showing the test circuit used to measure the crosstalk between traces. The aggressor trace is shown in purple, and the victim trace is shown in green. (b) A time-domain trace showing the induced crosstalk at 10 kHz. (c) A magnitude plot showing the crosstalk attenuation between xDev traces vs frequency. Several time-domain examples are inset to demonstrate the changes in magnitude. (d) A time domain trace showing the induced crosstalk from a digital aggressor signal at 100 kHz. (e) A time domain trace showing the effective reduction of crosstalk from a digital aggressor by a 15 kHz antialiasing filter. The crosstalk waveform is reduced to less than 1 μV in amplitude.

Figure 10.

The xDev system faithfully conveys microvolt-scale biosignals. (a) A diagram showing the experimental setup to play back prerecorded biosignals. (b) Example traces for the three biosignals tested. The black trace is measured from the output of the NeuroDAC and the blue trace is measured from the output of xDev. (c) Correlation coefficients between the NeuroDAC and xDev output signals for 300 μV sine waves of various frequencies. (d) Correlation coefficients between the biosignals recorded from the output of NeuroDAC and xDev. (e) Band powers for ovine hindlimb electromyography (EMG) and spinal local field potential (LFP) biosignals with xDev included and excluded from the signal chain. No significant difference in band powers was observed between the ‘xDev’ and ‘no xDev’ conditions for either biosignal (p > 0.05, Mann–Whittney U test). (f) A raster plot showing the spike times with xDev included and excluded from the signal chain. Each condition was repeated 3 times. (g) A box plot showing the difference in spike counts in 1 s windows with xDev included and excluded from the signal chain. No significant difference in spike count error was found between the ‘xDev’ and ‘No xDev’ conditions (p > 0.05, Mann–Whittney U test).

Figure 11.

Overview of the in-vivo demonstration of system prototyping using xDev. (a) A diagram showing the experimental setup. The epidural paddle array (HD64) was implanted epidurally on the dorsal aspect of the spinal cord, and exposed percutaneously for connection to xDev. The implanted hindlimb electromyography (EMG) telemetry system recorded intramuscular EMG from three lower extremity muscles. The BF is the biceps femoris, the Gas. is the gastrocnemius, and the EDL is the extensor digitorum longus. (b) A radiograph showing the HD64 implanted on the dorsal spinal cord of the sheep at the L5 vertebral level. (c) Photographs of the HD64, showing the 60-contact array (left), and the onboard hermetically sealed multiplexing ASIC (right). (d) A timeline showing the two xDev configurations used in this experiment.

Figure 12.

xDev enables rapid system reconfiguration during in vivo deployment. (a) The xDev configuration for testing a commercial stimulator device and active epidural electrical stimulation (EES) paddle. Electrophysiology signals and HD64 control signals were routed through the xDev using simple modules. (b) The xDev configuration used to test a custom implementation of the Intan RHS2116 stimulator chip and active EES paddle. A complex module was created for the Intan chip. The chip’s communication signals, electrophysiology signals, and HD64 control signals were all routed through xDev. (c) A fundamental demonstration of recorded stimulation evoked compound action potentials (ECAPs) using xDev. Stimulation was provided using the orange electrode, and the recording was taken from the blue bipolar electrodes. (d) A fundamental demonstration of using xDev to deliver stimulation using a custom stimulator module. In the top panel, stimulation was delivered to the purple electrode at 50 Hz and 2 mA. In the bottom panel, stimulation was delivered to the green electrode at 25 Hz and 2.5 mA.