DIGITAL COMMUNICATION FOR TELEMETRIC MULTI-NEURON RECORDINGS

Academic Open Internet Journal

www.acadjournal.com

Volume 16, 2005

DIGITAL COMMUNICATION FOR TELEMETRIC MULTI-NEURON RECORDINGS

Veselin Geortchev¹, Ivilin Stoianov², Rumiana Krasteva¹,
Ani Boneva¹, Ditchko Batchvarov¹, Konstantin Stanishev¹,
Roman Zahariev¹, Giorgio Vallortigara³

¹Central Laboratory of Mechatronics and Instrumentation - BAS
Acad. G. Bontchev Str. Bl.2, 1113 Sofia, BULGARIA
Phone: (+359 2) 72 13 61; E-mail: veso@clmi.bas.bg
²University of Padova, Department of General Psychology, University of Padova
³Dept. of Psychology and B.R.A.I.N., Centre for Neuroscience, University of Trieste, ITALY

This paper investigates which type of digital communication would permit telemetric multi-neuron population recordings in flying birds. We first outline the challenges that multi-neuron recordings put forward, the most important of which are small weight/size and large transmission band. Given these requirements, we propose that data should be transmitted in a digital way, using ZigBee or Bluetooth wireless protocols. We critically review these protocols and conclude that Bluetooth is suitable for recordings of large neuron populations due to its larger transmission band while the lower transmission bandwidth of ZigBee is sufficient to record a lower number of neurons. We propose basic specifications of the architecture and information processing of such a telemetric system.

Keywords: Multi-neuron population recording, Single cell recording, Radiotelemetry, wireless data transmission, Bluetooth, ZigBee

The paper reviews the state-of-art of wireless transmission methods based upon commercial hardware, aiming at the development of an enhanced telemetric system for multi-neuron population recordings in completely unconstrained birds. The usual approaches for single-cell recording have been based upon analogical signal transmission, which is advantageous due to compatibility with classical single-cell recording, simply replacing the wire-links with miniature wireless systems [19, 2]. The development of miniature low-energy single-chip FM-radio transmitters has further facilitated the implementation of such telemetry systems [8]. These systems, however, are limited to one-way transmission of up to two channels. In contrast, the latest developments in neuroscience require the bidirectional transmission of multiple neuronal signals, in order to allow the recordings and/or stimulation of neuron populations [20]. Apparently, the most feasible way to do this is by means of using digital systems that permit the encoding, decoding, compression, and transmission of multiple analogical channels, using two-way communication. These capacities, however, could be obtained at the cost of complex hardware with increased energy consumption and larger physical dimensions. Hence, in order to find an optimal solution, it’s needed a precise estimation of the currently common digital wireless protocols and corresponding hardware. For this purpose, we critically confront two digital transmission protocols that we consider as most perspective for a telemetric system: Bluetooth and ZigBee, each of which has hardware implementations on miniature low-energy chips.

Neuronal recordings of awake behaving animals is the most direct approach for the analysis of their brain organization and functioning, providing evidence that is even stronger than brain-imaging techniques in humans and optical imaging in animals because directly accounts for neural activities [21]. Neuroscience research classically uses single-cell recordings, which since 1940 is a basic instrument for the analysis of the neural functionality. Physical restraint of animal subjects is usually needed in this type of electrophysiological studies. For some neuro-ethological experiments with highly mobile animals such as birds, however, it is extremely desirable to have the animals free from the restriction of connecting wires. In these cases, wireless data transmission could prove advantageous and enables new perspectives for combined behavioural-electrophysiological approaches.

Radiotelemetry – wireless transmission of data via radiowaves – has fascinated researchers in biomedical sciences for several decades (for review, [4,3]) and is widely applied to convey bioelectric potentials like electrocardiograms, electromyograms or electroencephalograms. Compared to the transmission of such potentials, telemetry systems carried by birds and broadcasting activity of nerve cells have to meet specific demands. The transmission systems should be miniature (max 10-15 cm²), light (max 20-30 gr.) and allow an autonomy of few hours or days. Neuronal signal consists of action potentials (spikes) and noise. A typical spike is shown on Figure 1. It’s composed of a fast negative- and a positive impulses with a typical amplitude range of 50 - 200 uVpp (peak-to-peak voltage between the two extremes) when recorded extracellularly with microelectrodes of impedance several MOm. The spike’s complete waveform is circa 2 ms long. The noise has an amplitude range of 10 - 15 uVpp.

Several transmitters suitable for broadcasting the neuronal activity of single units have been previously designed in pioneering studies, all of them based upon transmission of analogical signals [1, 2, 5, 6, 7, 9, 10]. Continuous improvements of electronic components and the application of highly-integrated circuits offered the possibility for construction of miniature systems with more than one transmission channel. [8], in particular, has developed a miniature FM-stereo radio transmitter that permitted simultaneous radio-telemetric transmission of neuronal activity from two electrodes, which device was successfully used to record from freely-moving behaving barn owls.

The current neuroscience research is shifting the focus on studying whole neuronal populations by means of the simultaneous recording of multiple neurons [20, 21]. Neural processing is commonly organized in topographic maps that could be more accurately understood by means of simultaneous recordings of multiple neurons with arrays of microelectrodes (e.g., 8x1, 6x6). Registering the activation of neuronal populations using sequential single-cell recordings produces results with markedly increased variance due to variability of the whole brain activity because of processes specific to behaving animals, such as attention or intention [22]. Instead, multi-neuron recordings simultaneously register the neural activities and produce results that could be considered as invariant to that brain “noise”.

Fully active neurons typically produce action potentials at a frequency up to 300 spikes per second. A spike could sufficiently accurate be discretized by 20 samples, 50 levels per sample, which means that 6-bit sampling with frequency of 6000 samples per second could be sufficient for the digitalization of the activity of a single neuron (see Figure 1). However, each microelectrode typically registers more than one neuron, usually 3.4 per electrode, even up to 5, which could be then separated by means of PCA-analysis of the shape of the spike waveform [21]. Hence, in order fully to represent such a compound neuronal signal, the neural recording system should digitalize it with even larger frequency. Yet, typical analogical single-cell registration systems clear the raw electrode signal with a filter of a bandpass 300 – 7000 Hz [2, 8, 23], which means that an upper bound of 7000 samples per second should be sufficient to represent the collective neuronal activity registered by a single electrode. If we consider 16-microelectrode recording, it would then produce digital signal with a data-rate of circa 650 Kbps, while 64 sensors would require circa 2.5 Mbps.

Analogical transmission apparently can not permit such multi-neuron recording: linear scaling of analogical telemetry systems (e.g., by means of applying multiple analogical devices) is impractical because it leads to linearly increased size, weight, and energy consumption. Moreover, multiple-frequency radio transmission would interfere in each other. In contrast, digital transmission proves advantageous here because of the current miniaturization of a wide variety of digital elements, also available for commercial use.

In the race of eliminating any form of wiring between products by adopting wireless RF (Radio Frequency) based technologies, the designers are faced with an ever growing number of communications protocols. Hence, the decision as to which one is best for a given application has become anything but trivial. Nowadays there are various wireless RF communications technologies – a breadth of choice that makes difficult identifying the optimum technology for a given application.

One group of wireless technologies that is particularly appropriate for the current problem is unified under the so-called Personal Area Networks (PAN) – is the interconnection of information technology devices within short range (typically 10 meters) [11]. PAN-technology allows to interconnect, for example, a laptop, a personal digital assistant (PDA), and a portable printer without having anything to plug in, using wireless communication. The currently appropriate for our purpose PAN-technologies are shown in Table 1.

According to the characteristics of the above shown protocols, we consider that Bluetooth and ZigBee are most relevant for the specific requirements of a telemetry system: large transmission data-rate and low energy consumption. Those protocols will be further reviewed in the following sections. The IrDA technology needs direct visibility [11] between the two nodes and has bandwidth and energy consumption less advantageous than ZigBee.

Introduced in 1998, Bluetooth is a standard that allows various types of electronic equipment to connect between them without wires within a short range (10/100 m). It is developed by Ericsson, Intel, Toshiba, Nokia & IBM, but currently there is a large number of companies that further improve the Bluetooth concept. Bluetooth targets low-cost, low-power, secure and robust short-range connectivity. The technology has been designed for ease of use, simultaneous voice and data, and multi-point communications.

A Bluetooth-system has four basic parts [12, 16]: (i) a radio (RF) that receives and transmits data and voice; (ii) a baseband or link control unit that processes the transmitted or received data; (iii) link management software that manages the transmission; and (iv) supporting application software.

The Bluetooth radio (i) is a short-distance [16], low-power radio that operates in the unlicensed spectrum of 2.4 GHz. This spectrum is shared by other types of equipment (e.g. cordless telephones, microwave ovens). In order to avoid interference, the Bluetooth specification employs Frequency Hopping Spread Spectrum (FHSS) techniques, which divides the frequency band into a number of channels (2.402 - 2.480 GHz yields 79 channels). The radio-transceivers hop from one channel to another in a pseudo-random fashion, determined by the master. Using a nominal antenna power of 0 dBm, the range is 10 meters. Optionally, a range of 100 meters may be achieved by using an antenna power of 20 dBm. Data is transmitted at a maximum gross rate of up to 1 Mbps. Protocol overhead limits the practical data rate to a little over 721 kbps.

Baseband (ii), or link controller, is the hardware that turns received radio signals into a digital form, which can be processed by a host application. It also converts digital or voice data into a form that can be transmitted using a radio signal [12]. The baseband unit takes care of converting data from one form to another (such as voice to digital data), compressing it, putting it into packets, taking it out from packets, assigning identifiers and error correction information and then reversing the entire process for data that is received.

The link manager software (iii) runs on a microprocessor and manages the communication between Bluetooth devices. Each Bluetooth device has its own link manager, which discovers other remote link managers, and communicates with them to handle link setup, negotiate features, authenticate QoS and to encrypt and adjust data rate on link, dynamically. The application software (iv) is embedded in the device that operates an application over the Bluetooth protocol stack. This software allows the PDA, mobile phone, or keyboard to do its job. All Bluetooth devices must have compatible sections in their Bluetooth stack, so that all Bluetooth devices would be able to interoperate with each other.

Bluetooth used two topological structures. [11,16]. The simpler one, piconet, is built of up to 8 basic devices (1 master and 7 slaves). Larger scatternet structures are built of piconets. The piconect topology is a collection of devices connected in an ad hoc fashion. One unit acts as a master and the others as slaves for the duration of the piconet connection. The master-unit sets the clock and hopping pattern and can connect up to 7 simultaneously active slave-units or 200+ inactive (parked) units [16]. Each piconet has a unique hopping pattern/ID. The scatternet topology, in turn, links multiple co-located piconets by means of sharing common master- or slave-devices. One device can be both a master-unit in one piconet and a slave-unit in another piconet. The scaternet topology preserves the full bandwith, each piconet having the maximum capacity (720 Kbps).

Bluetooth targets mid-low power consumption, with expected battery life of few days. A standard Bluetooth implementation would consume 150 mW during transmission and 10 mW in a sleep-mode [18].

Zigbee is a home-area network designed specifically to replace the proliferation of individual remote controls. It has been created to satisfy the needs for a cost-effective, standards-based wireless network that supports low data rates, low power consumption, security, and reliability in application areas such as remote monitoring and control. The protocol is managed by the so-called Zigbee-alliance [15] - a global consortium of more than 50 companies - which also develops standardized application software.

Following the standard Open Systems Interconnection reference model, ZigBee's protocol stack is structured in layers. The first two layers, physical and media access, are defined by the IEEE 802.15.4 (ZigBee) standard [13]. The physical layer contains the radio frequency (RF) transceiver along with its low-level control mechanism. The media-access layer provides access to the physical channel for all types of transfer.

The ZigBee radio uses the following unlicensed bands: 2.4 GHz (global), 902-928 MHz (Americas), and 868 MHz (Europe). It targets bandwidth of: (a) 40 Kbps at 915 MHz (10 channels), (b) 20 Kbps at 868 MHz
(1 channel), and (c) 250Kbps at 2.4GHz (16 channels). Transmission distance ranges from 10 to 75m, depending on power output and environmental characteristics [13]. In the 2.4 GHz band, direct-sequence spread spectrum is used, with offset-quadrature phase-shift keying modulation. In this case, channel width is 2 MHz with 5 MHz channel spacing. The 868 and 900 MHz bands also use direct-sequence spread spectrum but with binary-phase-shift keying modulation.

ZigBee networks consist of three types of devices [13]: (i) A network coordinator that maintains overall network knowledge. It's the most sophisticated of the three types and requires the most memory and computing power. (ii) A full function device (FFD) that supports all ZigBee functions and features. It can function as a network coordinator. Additional memory and computing power make it ideal for network router functions, but it could also be used in network-edge devices, where the network touches the real world; (iii) A reduced function device (RFD) carries limited functionality to lower cost and complexity. It's generally used as network-edge devices. The communication is based upon fully handshaked protocol for transfer reliability

The Network layer supports multiple network topologies, including star, cluster tree, and mesh [11]. In a star topology, one of the FFD-type devices assumes the role of a network coordinator and is responsible for initiating and maintaining the devices on the network. All other devices are end-devices and directly communicate with the coordinator. In a mesh topology, a ZigBee-coordinator is responsible for starting the network and for choosing key network parameters. The network might further be extended through the use of ZigBee routers. The routing algorithm uses a request-response protocol to eliminate sub-optimal routing. Ultimate network size can reach up to 264 nodes [13]. Using local addressing, one can configure simple networks of more than 65,000 (216) nodes, thereby reducing address overhead.

ZigBee targes low power consumption, with expected battery life of months to years. This is achieved using a very low duty-cycle (<0.1%). A standard ZigBee implementation should consume 50 mW during transmission and 0.003 mW in a sleep-mode [15].

ZigBee is advantageous for a telemetric system due to low energy consumption and battery life, and wider range. However, it has a relatively limited bandwidth (20 / 250 kbps), which would only be sufficient for a telemetric system with up to eight recording electrodes. In contrast, the bandwidth of Bluetooth (720 kbps) is about three times as much as that of Zigbee. This is achieved, however, at the cost of an increased energy consumption: a Bluetooth-based system is expected to have an autonomy of up to several days. Yet, this is not an obstacle for a telemetry system because a single experiment would probably have the duration of several hours, maximum few days. The ZigBee protocol is designed for a greater autonomy (several months) that, however, could be obtained only in case of low-rate data transmission. This is not the case in multi-neuron recording, where neuronal signals should continuously be transmitted, which would drastically increase the energy consumption, hence, reduce the autonomy to several days.

Bluetooth networks have a more limited range than ZigBee networks (10m. vs. 100m), which could be problematic for Bluetooth-based systems in cases of experiments in a wide natural environment. However, laboratory test-scenes or cages are usually limited to several meters, which is within the range of the basic low-consumption Bluetooth devices.

In sum, the comparison between the above two wireless technologies showed that the Bluetooth standard is more advantageous for a digital telemetry system. It provides the largest bandwidth; is relatively simple to use and enters within the information-constraints put to a basic multi-neuron telemetry-system. It remains to be seen whether the physical characteristics of the commercially available electronic components needed to build a Bluetooth device also meet the physical constraints of such a telemetry system.

This section proposes a prototype for the communication part of a telemetry system for multi-neuron recordings, with a general block-scheme shown in Figure 2 (remote site). We first suggest hardware equipment that could be used to build a miniature low-energy Bluetooth communication module for the remote site. Then, we propose pre-processing methods that would allow increasing the amount of electrodes in the recording matrix. We also discuss the network structure and information processing in the telemetry system.

A multi-channel digital telemetry system needs to digitalize and process a number of analogical signals, which nowadays are implemented in a one-chip microcontroller that integrates: an analog-digital converter (ADC); digital signal processing and temporal storage in working memory (RAM), and a digital interface to the transceiver device. Table 2 provides a list of few microcontrollers that we found perspective for this purpose: AT90LS8535, ATMega128L (produced by Atmel), PIC16F8X (by Microchip), and MSP430F149 (by Texas Instruments).

Some of the key characteristics that were important for choosing the most appropriate solution were: (a) physical dimensions; (b) active power; (c) number of ADC; (d) RAM. The first two parameters determine the total physical dimension and mass of the telemetric device. The latter two determine its working capacity: we need to convert multiple analogical signals and probably to compress the digital signal. Hence, in order to maximize the number of channels transmitted and reduce the energy needed for transmission, we need more ADCs and RAM. One could also use analogue multiplexers, at the cost of increased total physical dimension of the system. It is also possible to use implanted amplifiers and multiplexers [18], but this is a matter of further research and commercial availability.

As far as the Bluetooth communication is concerned, it could be implemented with the one-chip solution by Ericsson: ROK 101 008 [17] - a short-range module that implements full Bluetooth functionality. It operates at
5 V and consumes 26 mA during data-transfer mode. It is 3.2x1,6x0,275 cm. large and weights less than 3 gr.

Analysing the characteristics of the microcontrollers (MCUs) shown above, we concluded that ATmega128L is most suitable for the telemetry system. It is a 8-bit MCU clocked at up to 8 MHz. When clocked at 4 MHz, it consumes only 5.5 mA in active mode. This MCU has 8 anaog input ports on the board, which allows sampling from 8 micro-sensors without external ADC. The ADC subsystem is extendable to 32 channels using a multiplexer. The analog inputs could be measured with a total frequency of up to 500 KS/s with a resolution of 8 bits per channels. Data could be temporally stored on the 4 KB on-chip memory, but an external memory of 64 KB could also be added. Control software and supporting data could be stored in the on-board 128 KB flash-memory. There are 2 UARTs onboard: one for digital communication with the external Bluetooth module and another for debugging [24].

In sum, the proposed hardware solution has the following characteristics:

MCU:	Atmel ATmega128L
Memory (Built in):	4K SRAM, 128K FLASH, 4K EEPROM
External memory:	64 K RAM (optionally)
ADC:	8 Channel 10-bit A/D-converter
Serial communication:	2 programmable serial UARTs

A complete telemetry system built upon MCU Atmel ATmega128L and Bluetooth module Ericsson ROK 101 008 would expectedly have the following energy consumption [17, 24]:
• ATmega128L MCU, at 3 V, 4MHz clock

In the hardware/software telemetric system, all measurement and processed activities of the microprocessor remote unit will be controlled by control commands sent by the base station (PC) via the Bluetooth communication cannel. The scheduling of the microprocessor tasks will be controlled by a flash-resident Micro Operating System (MOS).

The device will have the ability of scanning up to 32 ADC-channels, connected to micro-probes, with a sampling frequency of up to 16 KHz per channel (16.000 measurements per second for each channel). We remind that 6-bits, 7.000 measurements per channel are sufficient to represent and analyse the neuronal signal. The data values (8 significant bits) could be filtered with on-line programmable software filters and buffered. The base-station (PC) could change the state of each of the analogue channels, the length of their filter, the length
of the data-buffer and the process mode.

Each of the analogue channels will be buffered in the MCU working memory, which continuously will be sent to the base station using the Bluetooth interface (under of the control of the MOS), depending of the traffic time scheme set by the base station (PC). In case of 32 ADC channels, 4 KB (32 kbit) working memory provides
1 kbit of buffer length per channel, i.e., 150 samples per channel could be stored with 6-bit sample length. If each channel is scanned 7.000 times per second, then about 20 ms of raw data signal could temporally be stored (circa 10 spikes) [24]. This is sufficient for certain digital filtering and eventually simple compressing.

The bandwidth of the Bluetooth protocol is currently limited to some 720 kbps, which is insufficient for the transmission of more than 18 recording units (18 units x 7.000sps x 6bit). Hence, some compression would be needed in order to allow the transmission of a larger recording matrix, e.g., 64.

We first note that if the microelectrodes are arranged in a 2D-matrix, not all electrodes could detect neuronal signal due to the folded shape of the neuronal tissue. In [21] have found in experiments with a 2D-matrix microelectrode that typically only 58% (40-70%) of the electrodes register neuronal spikes. Therefore, one could use a recording matrix that has a number of microelectrodes that is about twice the number of channels the communication system could transmit and send with the available transmission bandrate only the signal from a limited number of electrodes: those that exhibit high Signal-to-Noise-Ratio (SNR). This means that our system could use a matrix with some 32 microelectrodes (e.g., 4x8 ).

Then, since neurons are frequently inactive, one could also compress the filtered signal (at least 1:3 compression rate is expected). More sophisticated compression methods that study the dynamics of the signal could further improve the compression rate and maintain it even during sustained neuronal activity. In sum, the proposed communication system could allow the transmission of a 2D-matrix with as much as 100 microelectrodes.

A network topology that is appropriate for the projected telemetry system is piconet, where the master-unit is the base station (a PC) and a slave unit is the remote measurement unit. The base station inquires the measurement unit and after initiating a connection, maintains it permanently, keeping the unit always in a slave-mode. This topology represents a directed piconet, with two nodes only. Data transmission will be provided by means of command-events. When the Bluetooth module of the remote unit accepts a packet from the base, it would return a packet containing a confirmative signal and data encoding the measurements of the channels. After a successful transmission/receipt of the data-packets, the base-station would send back an acknowledging message.

This article presented our preliminary investigations for the realization of the communication module of an improved telemetric system that would allow multi-neuron recordings in completely unconstrained animals. Radiotelemetry has been used in bio-medical sciences for wireless transmission of data [8], but previous systems have been based upon analogical transmission only that allowed the transmission of up to two channels. We aimed at extending this limit, to allow telemetric recordings from arrays of microelectrodes.

For this purpose, we proposed that the neural signals should be transmitted in a digital way, after appropriate pre-processing: digitalization and compression. For data transmission, we suggested that the so-called Personal Area Networks wireless technology-devices could be appropriate to the current problem because they allow low-energy, large-band interconnection within the range of 10-100 meters. We confronted two of the most promising wireless PAN-protocols that could be used for the implementation of such a system: Zigbee and Bluetooth. We were particularly interested in deciding which protocol could fulfil the tight requirements for a device that could be transported by a flying bird (such as a pigeon): small size, low energy consumption, and large transmission bandwidth necessary for the transmission of multiple analogue signals. We concluded that Bluetooth, with a 720 kbps transmission bandwidth and modest energy consumption of the needed hardware elements is the most suitable protocol for such a telemetry system. The competitive wireless interfaces were impractical for the current purpose due to limited band-rate (ZigBee) or high energy consumption (Ultra Wide Band).

We also investigated prospective microcontrollers that would pre-process the data and communicate it to the Bluetooth device. Size, energy consumption, and number of on-board, and allowed external ADC were among the main criteria. We found that currently the Atmel ATmega128L processor was most appropriate for this purpose. We also proposed the basic principles of data preprocessing, encoding and transmission that we plan to implement in our future telemetric system.

Finally, we conclude that the implementation of a multi-neuron telemetry system portable by flying birds with currently available electronic elements is a realistic project, with sufficiently small size and energy consumption of the hardware needed, and sufficiently large transmission data-rate.

[1]. Eichenbaum H, Pettijohn D, Delucia AM, Chorover SL. Compact miniature microelectrode-telemetry system. Physiol Behav (1977), v.18:1175–8.
[2]. Grohrock P, Hausler U, Jurgens U. (1997). Dual-channel telemetry system for recording vocalization-correlated neuronal activity in freely moving squirrel monkeys. Journal of Neurosc. Methods, 76, 7–13.
[3]. Kimmich HP, Vos JA. (1972) Biotelemetry. Leiden:Meander.
[4]. Mackay RS. Bio-Medical Telemetry. Sensing and Transmitting Biological Information from Animals and Man. New York: Wiley, 1968.
[5]. McElligott JG. A telemetry system for the transmission of single and multiple channel data from individual neurons in the brain. In: Phillips MI, editor. Brain Unit Activity during Behavior. Springfield, IL: Thomas, 1973.
[6]. Nieder A, Klump GM. Adjustable frequency selectivity of auditory forebrain neurons recorded in a freely-moving songbird radiotelemetry. Hear Res 1999a;127:41–54.
[7]. Nieder A, Klump GM. Time course of simultaneous masking in starling‘s auditory forebrain. Exp Brain Res 1999b;124:311.
[8]. Nieder A, Miniature stereo radio transmitter for simultaneous recording of multiple single-neuron signals from behaving owls. Journ. of Neuroscience Methods 2000, v.101, 157-164.
[9]. Pinkwart C, Borchers H-W. Miniature three-function transmitting system for single neuron recording, wireless brain stimulation and marking. J Neurosci Methods 1987;20:341–52.
[10]. Skutt HR, Beschle RG, Moulton DG, Koella W. New subminiature amplifier-transmitters for telemetering biopotentials. Electroenceph Clin. Neurophysiol. 1967; v. 22: 275–7.
[11]. Eric Glaenzer, SIG WirelessOne, PAN - Personal Area Network, (2004) www.siliconfrench.com/workshops/ presentations/PanSiliconFrench.ppt
[12]. Atmel Corporation, Blootooth General Information,
http://www.atmel.com/dyn/products/product_card.asp?part_id=2205, 2000
[13]. IEEE 802.15.4 standard (ZigBee) http://standards.ieee.org/getieee802/download/802.15.4-2003.pdf
[14]. Atmel Corporation, ZigBee™-compliant RF Transceivers and Controllers, http://www.atmel.com/products/zigbee/, 2004.
[15]. ZigBee Alliance, http://www.zigbee.org/
[16]. IEEE 802.15.1 standard (Bluetooth): http://standards.ieee.org/getieee802/download/802.15.1-2002_sectionone.pdf
[17]. Ericsson Microelectronics AB, Ericsson Bluetooth module: ROK 101 008, Datasheet, 2000, pp.1- 12
[18]. Jan Beutel, Oliver Kasten, A Minimal Bluetooth-Based Computing and Communication Platform, IT Papers, June 2001, http://www.itpapers.com/abstract.aspx?compid=16237&docid=91473
[19]. Ando N, Shimoyama I, Kanzaki R. (2002). A dual-channel FM transmitter for acquisition of flight muscle activities from the freely flying hawkmoth, Agrius convolvuli. Journal of Neuroscience Methods, 115, p.181-187.
[20]. Chapin JK (2004). Using multi-neuron population recordings for neural prosthetics. Nature Neuroscience, 7, 5, 452-455.
[21]. Nordhausen CT, Maynard EM, and Normann RA, (1996). Single unit recording capabilities of a 100-microelectrode array, Brain Research, 726, 129–140.
[22]. Chapin JK & Woodward DJ (1981). Modulation of sensory responsiveness of single somatosensory cortical cells during movement and arousal behaviors. Exp. Neurol. 72, 164–178.
[23]. Harrison RR & Charles C, (2003). A low-power low-noise CMOS amplifier for neural recording applications. IEEE Journal of Solid-State Circuits, 38/6, 958-965.
[24]. Atmel Corporation, ATmega128(L) - Datasheet Complete, 11/2004 - http://atmel.com/dyn/general/tech_doc.asp?doc_id=7236