See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/270658762

A low-cost, open-source, wireless

electrophysiology system

Article · August 2014

DOI: 10.1109/EMBC.2014.6944288 · Source: PubMed

CITATIONS READS

3 168

6 authors, including:

Najib J Majaj
New York University
42 PUBLICATIONS 1,405 CITATIONS

SEE PROFILE

All in-text references underlined in blue are linked to publications on ResearchGate, Available from: Ali Ghomashchi
letting you access and read them immediately. Retrieved on: 21 October 2016
 A Low-Cost, Open-Source, Wireless Electrophysiology System
A. Ghomashchi, Z. Zheng, N. Majaj, M. Trumpis, L. Kiorpes and J. Viventi, IEEE Member

Abstract— Many experiments in neuroscience require or We have combined these two technologies with a small
would benefit tremendously from a wireless neural recording lithium-ion coin-cell battery enabling an entire 8 channel
system. However, commercially available wireless systems are wireless biosignal recording system that is 25 mm in
expensive, have moderate to high noise and are often not diameter and 9 mm high. By utilizing the built-in Bluetooth
customizable. Academic wireless systems present impressive transceivers in modern laptops, smartphones and tablets, our
capabilities [1]–[4], but are not available for other labs to use.
wireless system eliminated the need for expensive
To overcome these limitations, we have developed an ultra-low
noise 8 channel wireless electrophysiological data acquisition proprietary neural data acquisition systems. If an external
system using standard, commercially available components. transceiver is desirable, or if the wireless system needs to be
The system is capable of recording many types of neurological used with an older computer, an inexpensive USB Bluetooth
signals, including EEG, ECoG, LFP and unit activity. With a 4.0 transceiver can be used (Fig. 2).
diameter of just 25 mm and height of 9 mm, including a
CR2032 Lithium coin cell battery, it is designed to fit into a Compared to currently marketed wireless
small recording chamber while minimizing the overall implant electrophysiology systems, our system had higher
height (Fig. 1 and 3). Using widely available parts we were able performance neural signal acquisition with lower noise and
to keep the material cost of our system under $100 dollars. The larger dynamic range, due to the use of a 24-bit sigma-delta
complete design, including schematic, PCB layout, bill of analog to digital converter. It was similar in size and weight,
materials and source code, will be released through an open but dramatically lower in cost (Table 1). The core
source license, allowing other labs to modify the design to fit technologies were developed for the mobile phone industry,
their needs. We have also developed a driver to acquire data making them widely available at low cost. The bill of
using the BCI2000 software system. Feedback from the
materials cost is approximately $100 for the 8-ch system.
community will allow us to improve the design and create a
more useful neuroscience research tool. Similar non-commercial (academic) wireless recording
systems offer improved capabilities, such as higher channel
I. INTRODUCTION counts or lower power consumption [1]–[4]. However, these
systems rely on custom application specific integrated
New commercial medical analog front end integrated circuits (ASICs), which are exceptionally expensive to
circuits from Texas Instruments [5] and others have develop and are produced in limited quantities, restricting
combined entire high performance, multichannel biosignal their availability to only the group that developed them.
acquisition systems into a single small package. A single, To overcome these limitations, we have developed a
small integrated circuit provides a direct “electrode to custom wireless data acquisition that utilizes entirely off the
digital” interface that includes programmable amplification, shelf components. By leveraging innovative technology, we
filtering and high-resolution (24-bit) analog to digital were able to design a low cost, small form factor device that
conversion. Similar innovations in integration have obtained high quality neural recordings. To accelerate the
produced new Bluetooth 4.0 Low Energy (LE) system on a adoption of wireless electrophysiology methods, we are
chip (SoC) devices [6] that combine a complete low-power releasing the entire project under an open source license, such
wireless transceiver and microcontroller into a small that other groups may freely modify and improve the design.
package. Bluetooth 4.0 LE, while not originally intended for
continuous data streaming, provides moderate data rates (up II. METHODS
to 2 Mbps), with one of the lowest levels of energy A. Hardware - Component Selection
consumption per bit in commercially available wireless Tradeoffs are always required when designing a
technologies [7]. miniaturized, battery powered system. We prioritized
minimizing the size of the design, with a special emphasis on
minimizing the overall height, in order to reduce the torque
Manuscript received April 7, 2014. This work was supported by a that can be exerted on the animal via the recording chamber.
Taking Flight Award from Citizens United for Research in Epilepsy. A minimum battery life of two hours was required for the
A. Ghomashchi, Z. Zheng, and M. Trumpis are with the Department of planned experiments.
Electrical and Computer Engineering at New York University, Brooklyn,
NY 11201 USA and NYU WIRELESS At the core of our system (Fig. 2) was a medical analog
N. Majaj, and L. Kiorpes are with Center for Neural Science at New frontend integrated circuit (ADS1298, Texas Instruments)[5].
York University, New York, NY 10003 USA. The integrated circuit provided very low noise signal
J. Viventi is with the Department of Electrical and Computer acquisition, 0.4 µV rms at 250 S/s, increasing to 1.8 µV rms
Engineering at New York University, Brooklyn, NY 11201 USA, the
Center for Neural Science at New York University, New York, NY 10003
at 8 kS/s. High common mode rejection (-115 dB) combined
USA and NYU WIRELESS (phone: 917-727-2464; fax 718-260-3906; e-
mail:[email protected])
 changes automatically with the selected sampling rate (-3dB
bandwidth at 0.262 of selected data rate).
For wireless communication we utilized the PAN1721
(Panasonic Corporation of North America), which is a
prefabricated wireless module that includes the CC2541 SoC
(Texas Instruments) and a chip antenna. The CC2541
combines a 2.4 GHz RF transceiver with an enhanced version
of the 8051 microcontroller in a compact 6x6 mm package.
The SoC is optimized for low power usage and was capable
of operating using the Bluetooth 4.0 Low Energy protocol
(BLE) or a proprietary communication mode. The effective
range of the wireless link depends on the transmission power
setting. To balance power consumption and wireless range,
we selected 0dBm output power and achieved a wireless
transmission range of ~10-15m.
Although the BLE protocol was designed for low
throughput applications, we used this mode in our system to
enable compatibility with the many BLE enabled devices
Figure 1: 8-Channel wireless electrophysiological data acquisition hardware, currently available on the market. For applications requiring
25mm in diameter and 9mm in height not including the electrodes mating
connector. On the left is the analog front end acquisition chip ADS1298 and higher data throughput, a proprietary communication mode
on the right is the Bluetooth wireless module PAN1721. At the top is a 14- along with an accompanying custom receiver module can be
pin electrode connector, with 9 connections used. The power regulator chip used to achieve raw data rates up to 2 Mbps.
is located at the bottom.
To provide regulated 3.0V power for the ADS1298 and
with right leg drive (RLD) enabled low level signal PAN1721, we chose a high efficiency (94%) switching
recordings even in high noise environments. The chip power supply (TLV61224, Texas Instruments). The power
contained eight 24-bit analog to digital converters, capable of supply had a wide input voltage range (0.7 to 3.0V) to
sampling all 8 channels at rates up to 32 kS/s. The sampling accommodate a variety of batteries with different chemistries,
rate and amplifier gain were all software programmable, including single-cell alkaline, NiMH or Li-primary batteries.
allowing simple adjustments for recording different types of No noise from the switching power supply was observed.
neurological signals. Even at the highest gain setting (12), a To enable simple and safe connectivity to potentially
large input range of +/- 100 mV allowed electrodes to be DC- delicate implanted recording devices, we designed a separate
coupled to sample low-frequency signals. This chip also printed circuit board (PCB) to interface our flexible electrode
includes a digital low pass third order sinc filter to attenuate array to the wireless system. We designed a custom electrode
high frequency noise. The frequency response of this filter

Figure 2: 8-channel wireless data acquisition system overall block diagram. The left block is the implantable wireless hardware. It contains the electrode and
connector assembly along with the detachable wireless board. The analog signals are amplified and digitized with the analog frontend IC, ADS1298. The
digital signal is then transferred to the Bluetooth module, CC2541, and packaged for wireless data transfer. The right block is the control and receiving
system. It includes a Bluetooth receiver module and computer software system for acquiring the streamed data. Note that a custom Bluetooth receiver is not
required for the receiving system. Devices such as laptops, smart-phones and tablets with built in Bluetooth 4.0 radio can be used.
adapter board with a female connector (Mezzapede, power consumption of the device was ~15mA at 3V.
Advanced Interconnections). The flexible array was affixed Powered by a standard Duracell CR2032 coin cell battery, the
to the adapter board using anisotropic conducting film device lasted approximately 30-60 minutes, significantly
(Elform). Using the rigid board as a structural support, we shorter than our initial calculations.
ensured plugging and unplugging the wireless device does
We tested our wireless system in vivo in an anesthetized
not cause damage to the thin flexible electrodes (Fig. 3). This
and paralyzed nonhuman primate (Macaca nemestrina) using
design allows the electrode adapter board to be permanently
8 of 12 available electrodes in a custom µECoG electrode
implanted on the subject, while the wireless data acquisition
array (Fig. 3). The electrode array was implanted subdurally
system can be connected and disconnected at will.
over primary visual cortex. Full screen stimuli with
The completed system also included an infrared sinusoidally modulated luminance at 1, 2 and 4 Hz were
phototransistor that can be used with an external light source sequentially presented. The corresponding evoked potentials
to record timing pulses and embed them in the wireless data. were recorded. The power spectral density showed strong
The recordings can then be precisely synchronized to eternal modulation at frequencies driven by the luminance
stimulation sources, such as visual or auditory stimuli. modulation of the stimulus (Fig. 4).
Moreover, a recording chamber was designed and 3D
printed to house the system during the acute experiments.
This chamber includes an easily removable cap to replace the
battery. The overall system was designed to be light and
small enough to be used on rodents and robust enough to be
used on primates.
B. Hardware – Fabrication and Assembly
Miniaturized PCB fabrication and assembly can be
expensive and challenging, especially for designs that utilize
small features or advanced manufacturing techniques. To
reduce cost, we have designed our wireless system using only
moderately sized features, enabling most PCB manufacturing
companies to successfully fabricate our design. We used a 4
layer PCB with 1 oz. copper for all layers. All traces were a
minimum of 5 mils wide, spaced 5 mils from other traces.
Via holes and pads were 10 and 20 mils, respectively.
C. Software
We developed embedded software for the on chip 8051
microcontroller to receive the data from the ADS1298 Figure 3: Electrode adapter board and wireless board assembly. Connection
frontend and run the Bluetooth stack. The embedded was made using a 14-pin, 1 mm pitch Mezzapede connector. The assembly
application was developed around the operating system can be chronically implanted and enclosed in a 1 inch diameter chamber.
abstraction layer which is a control loop that handled the (Inset) Size comparison with a US quarter.
Bluetooth stack, hardware abstraction layer (HAL), tasks
IV. DISSCUSION
and events. Sampled data from the analog front-end were
received by the microcontroller through the SPI port, We have successfully demonstrated the design and
buffered and then sent via low energy wireless link in short fabrication of a small wireless data acquisition system
packets. capable of recording neurological signals. The system was
validated in an acute experiment in a primate using a
The open-source firmware for the on-board
subdurally implanted µECoG electrode array.
microcontroller allows researchers to create and download
custom algorithms such as spike detection and data In future work we will optimize the embedded software to
compression. The Bluetooth wireless radio module can reduce power consumption and investigate different battery
directly pair with most modern smartphones, tablets and technologies to extend the system battery life. Additionally,
laptops for data display and logging, eliminating the need for operating the wireless system in Bluetooth 4.0 BLE mode
an expensive, dedicated neurophysiology system. We limited the overall bandwidth to 40-50Kbps, which restricted
additionally developed a driver for the BCI2000 [8] brain- the sampling rates available for use. Increasing the data rate
computer interface system to provide real-time display and will allow higher channel counts and higher sampling rates.
analysis of recorded neurological signals. The ADS1298 can also be replaced with one of the
RHD2000 series digital electrophysiology interface chips
III. RESULTS
(Intan Technologies), yielding a 32 or 64 channel wireless
Our wireless data acquisition system was fabricated and system without increasing the physical size of the system.
assembled (Advanced Circuits). We tested the completed However, the cost of the system will dramatically increase
system using test signals recorded at different gain settings ($400 - $700). We utilized a prefabricated wireless module
and sampling rates to confirm functionality. The input for our first prototype to speed development. Future designs
referred noise of the system was 0.4 µV RMS when sampling could integrate the CC2541 directly onto the main PCB to
at 250 S/s and using a gain of 12, as expected. The average reduce the size of the wireless system.
 REFERENCES
[1] R. Bashirullah, J. G. Harris, J. C. Sanchez, T. Nishida, and J. C.
Principe, “Florida Wireless Implantable Recording Electrodes
(FWIRE) for Brain Machine Interfaces,” in 2007 IEEE
International Symposium on Circuits and Systems, 2007, no. 4,
pp. 2084–2087.
[2] R. Harrison and P. Watkins, “A low-power integrated circuit for a
wireless 100-electrode neural recording system,” Solid-State
Circuits, …, vol. 42, no. 1, pp. 123–133, 2007.
[3] M. S. Chae, Z. Yang, M. R. Yuce, L. Hoang, and W. Liu, “A 128-
channel 6 mW wireless neural recording IC with spike feature
extraction and UWB transmitter.,” IEEE Trans. Neural Syst.
Rehabil. Eng., vol. 17, no. 4, pp. 312–21, Aug. 2009.
[4] H. Gao, R. M. Walker, P. Nuyujukian, K. A. A. Makinwa, K. V
Shenoy, B. Murmann, and T. H. Meng, “HermesE: A 96-Channel
Full Data Rate Direct Neural Interface in 0.13 mCMOS,” IEEE J.
Solid-State Circuits, vol. 47, no. 4, pp. 1043–1055, Apr. 2012.
[5] Texas Instruments, “ADS1298: Low-Power , 8-Channel , 24-Bit
Analog Front-End for Biopotential Measurements,” no. January
2010. 2012.
[6] Panasonic, “PAN1721: Ultra Low Power, Bluetooth Low Energy
Module.” Panasonic Corporation of North America, 2012.
[7] P. Smith, “Comparisons between Low Power Wireless
Technologies,” 2011.
[8] G. Schalk, D. J. McFarland, T. Hinterberger, N. Birbaumer, and J.
R. Wolpaw, “BCI2000: a general-purpose brain-computer
interface (BCI) system.,” IEEE Trans. Biomed. Eng., vol. 51, no.
6, pp. 1034–43, Jun. 2004.
[9] J. a Gregory, A. Borna, S. Roy, X. Wang, B. Lewandowski, M.
Schmidt, and K. Najafi, “Low-cost wireless neural recording
Figure 4: Wireless recording is validated in primary visual cortex of an system and software.,” Conf. Proc. IEEE Eng. Med. Biol. Soc.,
animal viewing sinusoidal luminance modulation. (A) The black trace shows vol. 2009, pp. 3833–6, Jan. 2009.
µECoG recorded in one channel. Visual cortex activity is driven at 4 Hz, [10] M. Yin and M. Ghovanloo, “A low-noise clockless simultaneous
responding to both the bright and dark contrast phases of the 4 Hz luminance 32-channel wireless neural recording system with adjustable
modulation. The best-fit 4 Hz sinusoid (plotted in red) illustrates a tight resolution,” Analog Integr. Circuits Signal Process., vol. 66, no.
phase locking with the visual stimulation. (B) The power spectral density of 3, pp. 417–431, Nov. 2010.
this channel indicates high SNR in the driven component of the µECoG
compared to broadband power.

Our System, Triangle BioSystems University of Michigan, Georgia Inst. of Tech,

Parameter
ADS1298 analog frontend Intl., W5 FMTv2 [9] WINeR-5 [10]
25mm diameter ×
Device Dimension 24.4 × 18.6 × 10 mm 36 × 40 × 10 mm
9 mm height
# of Recording
8 5 15 32
Channels
0.2 – 2.4 V 4mV
Analog Input Range
(maximum input Vp-p) (maximum input Vp-p)
1 – 12
Gain 800 51 77.1 dB
(programmable)
Resolution 24-bits 16-bits 8-bits
250 S/s – 32 kS/s 22 kS/s 58-680 kHz
Sampling Rate 50 kS/s
(programmable) (programmable) (programmable)
Adjustable
Bandwidth 0.8 Hz – 7 kHz 0.1 Hz – 7 kHz 0.1 Hz – 10 kHz
(DC Coupled)
0.5 µVrms 5.5 µVrms
25 µVrms 4.9 µVrms
(at 500 S/s, G=12) (for 500Hz – 5 kHz)
Input Referred Noise
1.8 µVrms
(at 8 kS/s, G=12)
112 dB
SNR
(at fin = 10Hz, G=6)
CMRR 115 dB 139 dB

Input Impedance 1 GΩ 12 MΩ

Battery Life 1 Hour 4 Hours 24 Hours

Wireles Range 10-15 m 4m 3m 1m

Table 1: Wireless acquisition systems performance comparison