International Journal of Engineering and Applied Sciences (IJEAS)

ISSN: 2394-3661, Volume-2, Issue-8, August 2015

Design and Optimization of a Low DC Offset in

Implanted System for ENG Recording Based on
Velocity Selectivity Method
Assad I. K. Al-Shueli

[6-9]. All these approaches are employed in bio-amplifier

Abstract The major target of this paper is the design of design to achieve DC offset removal. But, these methods
advance signal processing system based on minimized length of have some weakness points such as they are required a large
bits required for digital-to-analogy converter (DAC) for capacitors in off-chip or on-chip. In addition, these designs
velocity selectivity recording (VSR) approach. The main increase the CMOS switches number that certainly obtained
application of this device is peripheral nerves recording
switching noise and thermal noise, also increase the power
(electroneurogram-ENG) by exploring a spectral analysis for
the propagation of neural activities in the velocity domain consumption. Indeed, these designs were improved the low
recording using VSR in implantable application. This research flicker noise, however they caused higher bandwidth and
adapted a flexible, compact, and energy efficient dc offset worse thermal noise performance [10]. Moreover, these
removal circuit. An optimization design has been used based on approach either on-chip or off-chip application increased the
best possible process involving linearity and area is thus hardware size.
suggested. The system process acquired using this approach Consequently, we turned to minimize the dc offset in the
were characterized as having a 10-bit signal processing for digital signal processing (DSP) design using online
DAC resolution, with 1.4 mA rms output current, with technique for high optimum design and low-hardware cost
minimum size around 0.02 mm2 of chip area, using FPGA
solution. In this study, a DSP system proposed real time
board as prototype design. This paper also explores the design
temperature vibration in online recording; minimization the signal processing with low-offset voltage and low-power
output DC offset decrease the heat emission which is consumption to meet the essential properties of
significantly for long term implementation applications. This bio-amplifiers design. The proposed DSP was designed
study proposed an analysis circuit configuration demonstrate using FPGA board technology, achieved less than 100 V
that this approach could achieve a small DC offset error, with offset voltage and consumed only 3.5 W under a 5 V
small size required. supply. It is a significantly results for ENG signal recording
applications.
Index Terms Implantable device, neural recording,
biomedical integrated circuit, CMOS amplifier, low-offset and
low-power, dc offset rejection, velocity selectivity recording
II. SYSTEM AND CIRCUIT ARCHITECTURE OF VSR METHOD
(VSR), FPGAs. The method currently in use to estimate the velocity
spectrum of ENG data captured by an MEC-based system is
based on linear signal processing principles and is referred to
I. INTRODUCTION as delay and add as shown in Fig.1[11-14].
The amplitude of the ENG signal is most important for the
recording application especially with electrode cuff because
the amplitude recorded by this interface technology is very
small about 1-5 microvolts, with frequency range from
0.3-5kHz [1-2]. Indeed, this challenge became more difficult
with the multielectrode cuff (MEC) approach due to the
increment in the tripolar amplifiers numbers required in the
same area compared with individual tripolar amplifier cuff
[3-5]. Furthermore, there are many other challenges facing
this approach such as the noise source resistance occurring
between the electrodes, the noise generated by the amplifiers
itself and dc offset. As a result, a high gain, low noise,
minimum dc offset and low power consumption within
constricted size limit are the most important boundary
conditions[15].

Different methods have been used to overcome DC offset

challenges such as, auto-zeroing (AZ), correlated double
sampling (CDS) and chopper stabilization techniques (CHS)
Fig.1. The signal processing unit (SPU) for selecting one
velocity (taken from [11]).

Assad I. K. Al-Shueli, Biomedical Engineering Department, University

The resulting profile of relative signal output as a function of
College of Engineering, Thi-Qar University, City Thi-Qar, Iraq, velocity is called the intrinsic velocity spectrum (IVS) [12].
009647827241232

1 www.ijeas.org
Design and Optimization of a Low DC Offset in Implanted System for ENG Recording Based on Velocity Selectivity
Method
An Implementation of hardware systems using VSR has been
presented in [15], this system implemented in Austria
Microsystems 0.8 m process . Fig.2 shows that the general
structure of the system which is consists from two parts, the
first part called the electrode unit (EU) which is a mixed
analogue/digital signal acquisition system. The second part
connected with first part by wires which is called the
monitoring unit (MU), both of them are fabricated in 0.8 m
CMOS. Finally, the MU provides the wireless data to
external signal processor.

Fig. III: Block diagram of the action potential simulation

system with white Gaussian noise[21].

The simulated AP generation system consists of an FPGA

based digital controller with a USB interface to a desktop PC.
The controller drives a multi-channel low pass filter with a
PWM sequence that follows the desired AP waveform. Each
channel is skewed in time from the previous channel to
simulate the effect of the AP travelling along the nerve
bundle as it goes through the multi-electrode cuff. The
controller makes use of an off-the-shelf Digilent Basys 2
board which includes a USB interface. Uncommitted digital
outputs from the Basys 2 are used as the PWM channel
outputs. Each PWM output is connected to an analogue low
Fig. 2. System construction overview (taken from [15]). pass filter that also acts as an attenuator to reduce signal
levels and switching noise reaching the system under test on
In this research a field Programmable Gate Array (FPGA) the outputs of the signal simulator [22-24].
have used as a prototype for the design and implementation The overall plan of the complete system is shown in Fig. 4
of signal processing approach for DC offset removal. Since and consists of three separate stages. The first stage on the
FPGAs offer a flexible and parallel processing supplies and left (block 1) is the programmable AP and noise signal
proper for online and high-performance demands. The ability generator (synthesiser) whose output is eleven unipolar
of FPGAs to reconfigure present a significant advantages for signals delayed as described in [22-24]. For a typical MEC of
online multichannel recording systems, for example length 3 cm the propagation velocity is variable range form
dimension and word length of neuronal data, and that make 10 100 m/s in the A fibre types. To these signals controlled
the system changeable and adaptable for various amounts of white Gaussian noise (AWGN) were added in the
requirements [16]. manner described in chapter three. Blocks 2 and 3 are the two
The FPGAs synthesis and implementation provide high principal sections of a neural recording system intended for
performance, impressive scalability and reconfigurable at the ultimate implantation and described in outline in [15]. Block
same time compare to other alternative appreoach such as, 2 is referred to as the electrode unit (EU) and is intended to
system-on-chip (SoC) [17]. Consequently, FPGAs have be mounted on the MEC for optimum performance. It
been used for implementation the signal processing system consists of a set of specially designed low noise differential
for neural recording application in several studies [18-24]. amplifiers whose 10 bipolar (or single differential) outputs
More attentions have been spent to optimize the circuit are digitised (10 bits resolution) and multiplexed into one
design include reduction in the amount of logic gates channel for transmission along an implantable cable [15]. In
required and power consumption. Finally, the hardware the architecture described in [15] several EUs are placed at
systems supplying particular circuits for digital signal different sites in the peripheral nervous system and
processing that could considerably replicate the simulation connected by implanted cables to a single monitoring unit
process into real time hardware process. In addition, the (MU). The MU is a demultiplexer (DMUX) and digital
hardware solutions are presenting the necessary signal processing system that interfaces with an RF telemetry
bio-implantability feature for real-time signal recordings and system (not described in this study). In addition to processing
analysis where these are essential. and transmitting the recorded data for external logging, the
MU processes the system commands and provides power
III. METHODS supplies [22-23].
Although a single MU can process the data from and control
A. Description of General Plan of the System and power several EUs [15], in the experiments described in
The block diagram in Fig.3 shows the system which is used this paper only one of each type of unit is used as shown in
to generate an eleven channel simulated AP signal from a Fig. 4. In addition, the dc offset removal is a complete
multi-electrode cuff with additive noise [21]. prototype is realised in FPGA. Although for ultimate
implantation the MU will be a second CMOS ASIC, the
comparison of different VSR signal processing strategies

2 www.ijeas.org
 International Journal of Engineering and Applied Sciences (IJEAS)
ISSN: 2394-3661, Volume-2, Issue-8, August 2015
that lies at the heart of this investigation requires easy
flexibility of design suggesting the use of FPGA.

Fig. III. Block diagram of the complete system.

The dc offset removal signal processing circuitry is
contained in the unit on the far right of block 3 in Fig. 4 Fig.6. The DC offset value for the delay and add channel
(SPU) and is preceded by stages of de-multiplexing into 10, before removing the DC offset.
parallel bipolar channels and conversion into 9 tripolar (or
Removal of the DC offset before signal processing results in
double differential) channels (tripolar signals are generally
a significant reduction in the size of memory which is
employed in ENG recording applications since they have
enhanced immunity to common mode signal contamination required to save samples from the signal processing data
(such as electromyogram-EMG) compared to monopolar or channel.
Table (1) and Fig.6 illustrate the DC offset value for delay
bipolar ones). Finally the captured and recorded signals are
and add channel related to temperature and power supply
processed for dc removal process. Verilog HDL has been
voltage. The result shows a slight variation in the DC offset
used to syntheses the FPGA board (XC3S1200E-5FG320C)
for both approaches under Xilinx ISE 9.2i platform. and this amount of variation cannot be affect on bit number
or memory size we have used in digital signal processing
system. So if we are supposed the DC offset constant value,
B. The Design and Implementation of DC Offset Removal the output will has very small DC offset depends on
In the amplifiers and ADC unit in Fig.4, the DC offset have variation of temperature and voltage supply. The rest of DC
been measured for ten channels, the all channels have range offset can be easily removed with band pass filter in next
of DC offset from 0.8-1.1mV. As a result, the delay and add stage of the system.
channels have a large amplitude DC offset due to the
integration of the DC offset for ten channels, consequently, Table (1). The DC offset value for delay
this increased the hardware size for the digital system
implementation, because the signal with high DC Offset will
require a large number of bits to represent it. Fig. 5 illustrates
the DC offset value for each channel; it is very large
compared with the target signal.

Fig.5. The DC offset value for 9 channels before the DC

offset removal. Fig.6.The DC offset variation curve for delay and add
channel
Each input channel sample from the ADC is 10 bits and the
Fig.7 shows the procedure which has been followed to
ten channels sample values are combined, so the delay and
reduce number of bit for each sample and for each channel.
add output must be 14 bits (see Fig. 6). From fig.6 , we can
In the DC offset remover unit the DC offset value is
see clearly the amount of the DC offset value for the delay
subtracted from the output from each channel which is
and add channel ( about 1.4mV) compared with peak to peak
represented by 9 bit (Fig.7(b)) .This leads to the production
signal value (about 1.1mV).
an output signal from the signal processing system with very

3 www.ijeas.org
Design and Optimization of a Low DC Offset in Implanted System for ENG Recording Based on Velocity Selectivity
Method
small (ineffective) value of DC offset within the range of a data samples should be supplied in a serial processing;
10 bit signed number (Fig .7 (c)). The DC offset remover will therefore the parallel processing could be transformed to
satisfy our goal of optimizing the memory size for each serial processing necessary.
channel to 6 bits for our application. Clearly, the benefit may
be more or less for applications.
IV. RESULTS AND DISCUSSION
Serial processing DC offset remover could be easy
implemented with multiplexer and subtraction; the serial
implementation needs only one multiplexer and one
subtraction to implement 10-bit serial DC offset removal
circuit for our application.
This technique demonstrates the strategy of signal
processing techniques which could be used effectively in a
practical application, it demonstrates possibly to create a
high-performance circuit that is able to combine samples
directly from ADC using low complexity hardware.
Moreover, the results show how the DC offset value can
affected hardware size. The DC offset remover requires a 10
bits subtraction and multiplexer and even the smallest FPGA
(Field-programmable gate array) can support more than 10
channels. The size of memories required to store the sample
values could easily be supported by block RAMs
(Random-Access Memory) on board the FPGA.

In addition, the delay and add channel is considerably

reduced by approximate 30% of bits size. Fig.9 shows the
delay and add signal after the DC offset removal within the
range of a 10 bit signed number.

Fig. 7. DSP unit with and without DC offset Remover

A simple circuit can be represented in the digital domain to
create a very efficient and practical function of DC remover.
The subtraction and multiplexers are all able to operate in 50
MHz which is provided in FPGA board. Therefore, it is also
possible to construct a high-performance circuit that is able Fig.9. The delay and add channel after removing the DC
to capture samples straight from the ADCs. offset.
In order to test and execute the hardware design, we have
employed three methods to implement the system. The first
technique has been used in the system which is without DC
remover unit and this approach has been employed the BPF
filter as DC offset remover as well as its function. Fig.10
shows technique procedure which have been used to
implement the system. The second approach it is our method
which is describe in detail above (see Fig.11).The third
Method have been implement in different structure from
previous approaches by setting the DC offset remover
between Delay and add channel and BPF (see Fig.12). In this
section we evaluate and compare the benefit for our
Fig.8 shows how serial processing can assist to remove DC technique against the other techniques. Consequently, we
offset thus considerably decreasing the size of hardware. can decide the capability and efficiency of our approach in
Serial processing can reduce hardware implementation for practical application.
lower sample-rate applications. In this type of function, the

4 www.ijeas.org
 International Journal of Engineering and Applied Sciences (IJEAS)
ISSN: 2394-3661, Volume-2, Issue-8, August 2015
board. Indeed, the flexibility of the FPGA allowed
reconfiguring and generating multi- prototype
implementations designs using only one FPGA board. This
advantage of choosing FPGA for implementation target is
saving the efforts, time and cost during prototype design.
Also, generation multi- prototypes for both approaches
provide a wide selection options to have fair comparison
among these implementation methods without needing to
Fig.10. DSP Unit without DC offset removal redesign the system each time because it is easy to upload
them and running at any time. In fact, this FPGA future
cannot be easy offer with alternative technologies.
Moreover, the cost itself is very reasonable and efficient for
producing a prototype task.

Fig.11. DSP Unit with DC offset remover before channels

Fig.12. DSP Unit with DC offset remover Delay and add

channel.

The all above approaches have been implemented using

FPGA board with Xilinx ISE 9.2i software for a 32.5 KHz Fig. 13: The hardware system constructions for the complete
sample rate and serial process and 50 MHz clock rates. The design
results show the amount of interest design that have been
obtained after the implementation of the two methods, it is
clear in Table(2) the amount of gates and slices that are save V. CONCLUSION
in our technique of comparison with conventional method This study proposed an analysis circuit configuration
where the provision of 35819 gates and 113 slices, in demonstrate that this approach could achieve a small DC
addition we save 1 MB from internal FPGA memory. This is offset error (less than 100V), with small size required. A
significant indicator of the effectiveness and efficiency of low-offset low-power and low hardware cost with online DC
our approach, including qualifies it to be a practical and offset removal design technique for VSR method in nerve
applicable method. activities recording applications is presented in this paper,
without hardware cost to enhance the circuitry ability over
Table (2) Device utilization summary for three methods power supply and temperature. The system process obtained
using this approach having a 10-bit signal processing for
DAC resolution, with 1.4 mA rms output current, with
minimum size around 0.02 mm2 of chip area. The FPGA
board has been used as a prototype for the design and
implementation the dc offset signal processing approach.
The whole circuit configured in FPGA board and consumed
only 495 Slices from the FPGA board hardware area and the
completed system consumed 665,864 gate while the system
without dc offset removal consumed 701,683 gate. In
The ADC and preamplifier unit which has been used in our addition, the dc offset removal reduces the dc offset value to
design has operating DC power supply range from 4.5V to be less than 100V, which was 1.4mV in conventional
5.5V DC. In order to determine the temperature effect on the method.
DC offset value we are set some of experiments to calculate
the DC offset value at different temperature value. REFERENCES
[1] M. K. Haugland and T. Sinkjar, Cutaneous Whole Nerve Recordings
The full hardware system after implementation and Used for Correction of Footdrop in Hemiplegic Man, IEEE
Transactions on Rehabilitation Engineering, vol. 3, no. 4, 1995,pp.
verification is shown in Fig. 13 which includes electrode 307317.
simulation unit, amplifiers and ADC unit and the DSP unit. [2] M. Haugland, A. Lickel, J. Haase and T. Sinkjaer, Control of FES
The DSP unit is used for processing the dc offset removal for thumb force using slip information obtained from the cutaneous
the data in both conventional and TDNN approach which electroneurogram in quadriplegic man, Rehabilitation Engineering,
IEEE Transactions on, vol. 7, no. 2, Jun. 1999,pp. 21527.
have been synthesised and implemented on the same FPGA

5 www.ijeas.org
Design and Optimization of a Low DC Offset in Implanted System for ENG Recording Based on Velocity Selectivity
Method
[3] J. Taylor, N. Donaldson and J. Winter, Multiple-electrode nerve cuffs
for low-velocity and velocity-selective neural recording, Med Biol
Eng Comput, vol. 42, no. 5, 2004, pp. 634643.
Assad I. K. Al-Shueli Received BSc. degree in
[4] N. Donaldson, R. Rieger, M. Schuettler and J. Taylor, Noise and
Electrical Engineering in 1998, MSc. degree in
selectivity of velocity-selective multi-electrode nerve cuffs, Medical
Electronics and Communication in 2001 from
& biological engineering & computing, vol. 46, no. 10, Oct. 2008,pp.
Al-Mustansiriya University in Iraq and PhD
100518.
degree in Electronics in 2013 from Bath
[5] R. Rieger, M. Schuettler, D. Pal, C. Clarke, P. Langlois, J. Taylor and
University in UK.
N. Donaldson, Very low-noise ENG amplifier system using CMOS
In July 2009, he started his research with the
technology., Neural Systems and Rehabilitation Engineering, IEEE
Microelectronics and Optoelectronics research
Transactions on, vol. 14, no. 4, Dec. 2006,pp. 42737.
group as PhD student in the Department of
[6] C. H. Chan, J. Wills, J. LaCoss, J. J. Granacki and J. Choma, A novel
Electronic and Electrical Engineering at the
variable-gain micro-power band-pass auto-zeroing CMOS Amplifier,
University of Bath. In 2001 2015. He worked at University of Thi-Qar in
IEEE Interna- tional Symposium Circuits and Systems, 2007,pp.
Iraq as a lecturer. In. He has published 4 papers in conferences and journals.
337340.
[7] B. Permanod, N. Silawan and C. Toumazou, Drift reduction in
Ion-sensitive FETs using correlated double sampling, IET
Electronics Letter, vol. 43,no.16 ,2007, pp.857 859.
[8] N. Panitantum and N. Wongkomet, A low-offset high-CMRR
switched-capacitor instrumentation amplifier for data acquisition
applications, Circuits and Systems, no.1, 2002,pp. 339342.
[9] G. A. Hanasusanto and Y. J. Zheng, A chopper stabilized
pre-amplifier for biomedical signal acquisition, International
Symposium on Integrated Circuits, 2007,pp. 200203.
[10] L. Li, High gain low power operational amplifier design and
compensation techniques, A Dissertation of department of Electrical
and Computer Engineering, Brigham Young University,2007.
[11] C. T. Clarke, X. Xu, R. Rieger, J. Taylor and N. Donaldson, An
implanted system for multi-site nerve cuff-based ENG recording using
velocity selectivity, Analog Integrated Circuits and Signal
Processing, vol. 58, no. 2, Nov. 2008, pp. 91104,
[12] J. Taylor, M. Schuettler, C. Clarke and N. Donaldson, The theory of
velocity selective neural recording: a study based on simulation,
University of Bath Online Publication Store, vol. 50, no. 3, 2012, pp.
309318,
[13] K. Yoshida, G. a M. Kurstjens and K. Hennings, Experimental
validation of the nerve conduction velocity selective recording
technique using a multi-contact cuff electrode., Medical engineering
& physics, vol. 31, no. 10, 2009,pp. 126170, Dec.
[14] M. Schuettler, V. Seetohul, J. Taylor and N. Donaldson,
Velocity-selective recording from frog nerve using a multi-contact
cuff electrode., in Engineering in Medicine and Biology Society,
EMBS ,06. 28th Annual International Conference of the IEEE, vol. 1,
2006,pp. 29625.
[15] C. T. Clarke, X. Xu, R. Rieger, J. Taylor and N. Donaldson, An
implanted system for multi-site nerve cuff-based ENG recording using
velocity selectivity, Analog Integrated Circuits and Signal
Processing, vol. 58, no. 2, Nov. 2008, pp. 91104.
[16] B. Yu, T. Mak, X. Li, F. Xia, A. Yakovlev and S. Member, Real-Time
FPGA-Based Multi-Channel Spike Sorting Using Hebbian
Eigenfilters, Emerging and Selected Topics in Circuits and Systems,
IEEE Journal on, vol. 1, no. 4, 2011 pp. 502515.
[17] K. Imfeld, S. Neukom, A. Maccione, Y. Bornat, S. Martinoia, P.-A.
Farine, M. Koudelka-Hep and L. Berdondini, Large-scale,
high-resolution data acquisition system for extracellular recording of
electrophysiological activity., IEEE transactions on bio-medical
engineering, vol. 55, no. 8, Aug. 2008,pp. 206473.
[18] U. Frey, U. Egert, F. Heer, S. Hafizovic and A. Hierlemann,
Microelectronic System for High-Resolution Mapping of
Extracellular Electric Fields Applied to Brain Slices, Biosensors &
bioelectronics, vol. 24, no. 7, Mar. 2009,pp. 21918.
[19] R. Rieger, J. Taylor and C. Clarke, Signal processing for velocity
selective recording systems using analogue delay lines, 2012 IEEE
International Symposium on Circuits and Systems, May 2012, pp.
21952198.
[20] A.I. Al-Shueli, C.T. Clarke and J.T. Taylor, "Simulated Nerve Signal
Generation for Multi-electrode Cuff System Testing," Biomedical
Engineering and Biotechnology (iCBEB), 2012 International
Conference on , vol., no., pp.892,896, 28-30 May 2012.
[21] A.I. Al-Shueli, C.T. Clarke, N. Donaldson and J.T. Taylor, Improved
Signal Processing Methods for Velocity Selective Neural Recording
Using Multi-Electrode Cuffs, Biomedical Circuits and Systems,
IEEE Transactions on , vol.8, no.3, pp.401,410, June 2014.
[22] A.I. Al-Shueli, Signal Processing for Advanced Neural Recording
Systems, PhD Thesis, University of Bath, Bath, UK, 2013.
[23] B. Yu, T. Mak, X. Li, F. Xia, A. Yakovlev and S. Member, Real-Time
FPGA-Based Multi-Channel Spike Sorting Using Hebbian
Eigenfilters, Emerging and Selected Topics in Circuits and Systems,
IEEE Journal on, vol. 1, no. 4, pp. 502515, 2011.

6 www.ijeas.org