This is an electronic reprint of the original article.
This reprint may differ from the original in pagination and typographic detail.
Powered by TCPDF (www.tcpdf.org)
This material is protected by copyright and other intellectual property rights, and duplication or sale of all or
part of any of the repository collections is not permitted, except that material may be duplicated by you for
your research use or educational purposes in electronic or print form. You must obtain permission for any
other use. Electronic or print copies may not be offered, whether for sale or otherwise to anyone who is not
an authorised user.
Olabode, Olaitan; Kosunen, Marko; Halonen, Kari
A current controlled oscillator based readout front-end for neurochemical sensing in 65nm
CMOS technology
Published in:
ISCAS 2016 - IEEE International Symposium on Circuits and Systems
DOI:
10.1109/ISCAS.2016.7527290
Published: 11/08/2016
Document Version
Peer reviewed version
Please cite the original version:
Olabode, O., Kosunen, M., & Halonen, K. (2016). A current controlled oscillator based readout front-end for
neurochemical sensing in 65nm CMOS technology. In ISCAS 2016 - IEEE International Symposium on Circuits
and Systems (pp. 514-517). [7527290] (IEEE International Symposium on Circuits and Systems). IEEE.
https://doi.org/10.1109/ISCAS.2016.7527290
A Current Controlled Oscillator Based Readout
Front-end for Neurochemical Sensing in 65nm
CMOS Technology
Olaitan Olabode, Marko Kosunen and Kari Halonen
Department of Micro and Nanosciences
Aalto University School of Electrical Engineering, Espoo, Finland
Email: olaitan.olamilehin@aalto.fi
Abstract—This paper presents the design of an integrated
current-controlled oscillator (CCO) based readout front-end for
neurochemical sensing applications. The readout front-end chip
is implemented in 65 nm CMOS technology and occupies an area
of 0.059 mm
2
. The proposed design supports an input current
range of 1.2 µA (±600 nA) and can also be configured to support
wider current range. The CCO-based structure utilized in this
design results in noise averaging of the detected neurochemical
input signal due to its inherent ∆Σ first-order noise shaping
and anti-alias filtering characteristics. Thus, the prototype chip
achieves a current resolution of 1 0 0 pA and can detect dopamine
concentrations as small as 10 µMol based on measured data from
novel diamond-like carbon electrodes. In addition, the digital
codes obtained from the readout front-end attain a signal-to-noise
(SNR) of 82 dB and linearity limited effective-number-of-bits
(ENOB) of 8 at full current range input, without employing any
calibration or linearization techniques. The proposed read-out
front-end consumes 33.7 µW of power in continuous operation.
I. INTRODUCTION
Sensing and real-time monitoring of neural activities within
the central nervous system (CNS) has become a fast-growing
area of research due to the need to understand more about how
neurons communicate as well as the emerging needs related to
personalized healthcare and brain machine interfaces. In addi-
tion, further knowledge on how neurons transmit information
within the CNS is of significant value to researchers in the
field of neuroscience for improving treatment of neurological
disorders and neurodegenerative diseases.
Neurons in the CNS are connected by synapses and
communicate through electrical and chemical impulses or
signals. Transmission of neurochemical signals occurs over
short distances in the order of (20 − 30) nm, across chemical
synapses; as a result of discharge and absorption of bio-
chemical molecules also known as neurotransmitters [1]. The
region or gap between chemical synapses forms a chemical
synaptic junction also known as synaptic cleft. The synaptic
cleft is filled with an extracellular fluid which aids the chemical
reactions that occur during neurochemical transmission. Neu-
rochemical signals are responsible for controlling cognitive,
learning and memory functions in the brain. Thus, several neu-
rological disorders such as Parkinson’s disease, Schizophrenia,
Alzeihmers and Epilepsy have been reported to be associated
with abnormal concentration levels of neurotransmitters such
as glutamate and dopamine [2]. Hence, the readout and anal-
ysis of neurotransmitter concentration levels from the brain
provides insight into neurochemical signalling and plays a
Fig. 1. System block diagram showing the interface between the proposed
readout front-end (DORSI) and the neurochemical sensor electrodes which are
designed to be implanted at chemical synaptic junctions in the brain.
vital role in the development of more effective treatments for
patients suffering from neurological disorders.
Neurochemicals such as dopamine, histamine, nore-
pinephrine, and serotonin; are primarily monitored with the
help of potentiostats which operate based on electrochemical
transduction principle [3]. Electrochemical transduction prin-
ciple is the process of applying an electrical potential (V
cell
)
across an electrochemical cell and measuring the induced
reduction-oxidation (redox) current within the cell (I
cell
) as
illustrated in Fig. 1. Thus, this paper proposes a readout front-
end for the measurement of neurochemical induced currents
based on redox reactions of neurotransmitters within the
synaptic cleft. The detected current profiles from the readout
chip (DORSI) represents change in concentration levels of
neurotransmitters at the neurochemical sensor interface as
depicted in Fig. 1. As a result, the detected oxidation and
reduction peak potentials help to regulate the voltage applied
by neurostimulation electrodes when used in deep-brain stim-
ulation of patients suffering from dopamine-deficient disorders
such as Parkinson’s disease [4].
This paper is organized as follows; Section II describes the
system level design of the proposed readout front-end. Section
III presents post-layout simulation results based on measured
data from the neurochemical sensor electrodes. Finally, perfor-
mance of the proposed design is summarized in Section IV.
II. PROPOSED DESIGN
The main challenge in the design of readout front-ends for
neurochemical sensing is the required support of a wide range
of input currents while achieving current resolution in pA or
less range. Hence, the proposed design of the readout front-
end is based on a mixed-signal architecture for minimizing
Fig. 2. Neurochemical signal processing stages in DORSI.
the effect of noise and achieving high resolution of detected
current signals. In addition, sensitivity and selectivity of the
sensor electrodes to neurochemicals play an important role
in achieving good current resolution from the readout front-
end. Thus, the novel diamond-like carbon electrodes used in
this work ensures more stable detection of neurochemicals and
provides lower background current (I
bg
). The input signal is
processed within the readout front-end along three main stages,
across analog and digital domains as shown in Fig. 2. First,
the neurochemical induced current is acquired in the IA stage.
Then, the acquired current from the IA stage is converted to
frequency in the I-F stage. Lastly, the pulses generated from
the I-F stage are quantized in the ID stage such that the digital
output from the readout front-end can be further processed by
neurostimulation circuitry or externally outside the brain.
A. System Architecture
This section describes the internal structure of the proposed
design and noise averaging technique of the current-controlled
oscillator (CCO) based architecture. The induced cell current
(I
cell
) from the neurochemical sensor is processed in the IA
stage in order to generate the control current (I
ctrl
) for the
I-F stage. Then, a simple RC-filter is applied to I
ctrl
in
order to limit the noise bandwidth. The low-pass filtered I
ctrl
signal is conveyed to the CCO which converts the current
signal to frequency domain with I-F conversion gain (K
cco
)
as illustrated in Fig. 3. Thus, the frequency of the oscillator
(F
cco
) is modulated by changes in the detected input current.
The frequency-domain information at the output of the I-F
gain stage is integrated to generate changes in phase domain
φ
cco
(t). Hence, the continuous change in the phase of signal
x(t) is quantized to the amplitude domain in the ID stage with
an integrate-and-dump algorithm.
The quantizer is implemented as an up-counter which is
triggered at the rising edge of the generated pulses from the
CCO. In addition, the accumulated counter codes (Σ
(φ)
) at
the output of the quantizer are sampled to generate discrete
representation of the accumulated phase change φ[n]. As a
result, the ID stage is controlled by two clocks namely; F
cco
and F
s
, which divides the ID block into increment and readout
clock domains respectively. The increment clock domain is
controlled by an asynchronous clock from the CCO since
F
cco
varies with changes in I
cell
. On the other hand, the
readout clock domain is controlled by a fixed sampling clock
F
s
which synchronizes subsequent processing by the discrete-
time derivator and determines the effective data output rate of
the system. Hence, the counter utilizes gray-coding in order to
mitigate the effect of possible timing violations that may occur
at the clock domain crossing (CDC) between the two clock
domains as depicted in Fig. 3. In addition, the use of gray-
code counters ensures that the minimum error due to possible
metastability in the digital circuitry is limited to 1 LSB.
Furthermore, the discrete-time phase sampler defines the
Fig. 3. Electrochemically induced cell current life-cycle within DORSI.
Fig. 4. Top-level schematic of the readout front-end (DORSI).
integration time (T
s
) of the changes in phase and the output
of the sampler φ[n] is differentiated in order to obtain the
digital output codes (D
OU T
). Hence, the CCO-based analog-
to-digital (A/D) conversion with integrate-and-dump digital
interface results in a cascaded-integrator-comb (CIC) filter
which has a continuous-time sinc frequency response. As a
result, the integrate-and-dump structure effectively averages
the noise over the sampling interval T
s
and prevents aliasing of
wide-band noise into the signal band [5]. Thus, the frequency
response of this system is given as follows with images on
integer multiples of F
s
[6].
[H(f)] =
K
cco
∗ sin
π ∗ f
F
s
π ∗ f
(1)
⇒ [H(0)] =
K
cco
F
s
= K
cco
∗ T
s
(2)
B. Circuit Level Design
This section describes the system level design of the
readout front-end and circuit implementations of each signal
processing stage of the proposed design. Fig. 4 presents the
top-level schematic of the readout front-end based on the
system architecture described in the previous Section II-A.
The following sub-sections elaborate on the design of each
processing block in the system.
1) Current Acquisition (IA): This block represents the
analog front-end of the system and controls the cell voltage
(V
cell
) that is applied across the working electrode (WE) and
Fig. 5. Relationship between V
in
, V
ref
, V
cell
, I
cell
and I
bg
.
reference electrode (RE) of the neurochemical sensor as shown
in Fig. 1. The applied V
cell
is controlled based on fast-scan
cyclic voltammetry (FSCV) sweep of the input voltage (V
in
)
in order to acquire the induced cell current (I
cell
) flowing
between the counter electrode (CE) and WE. Thus, this block
is designed to provide stable V
cell
between −0.7 V to 0.8 V
in order to detect oxidation (I
ox
) and reduction (I
red
) current
peaks of neurochemicals during the forward and reverse sweep
of the input voltage as depicted in Fig. 5. As a result, the FSCV
requirements for V
in
range of 1.5 V, limits the minimum sup-
ply voltage and defines the input common-mode range (ICMR)
requirements of the operational transconductance amplifiers
(OTAs). The OTAs are designed using the conventional Miller
OTA architecture and can operate with supply voltage between
1.8 V and 2.5 V. In addition, the OTA is designed to provide
high gain of 80 dB in order to minimize DC offsets and gain
errors, which in turn ensures stable and accurate control of
V
cell
. Furthermore, the structure of the IA stage is based on a
transimpedance topology where OTA
1
and OTA
2
are operating
in a voltage follower configuration for setting the voltage at
RE and WE respectively, which in turn sets V
cell
as follows.
V
cell
= V
W E
− V
RE
= V
ref
− V
in
(3)
⇒ I
cell
=
V
cell
Z
cell
=
V
ref
− V
in
Z
cell
(4)
The neurochemical sensor is presented in Fig. 4 as an elec-
trochemical cell and can be modelled as a voltage-controlled
current source (VCCS) which exhibits non-linear resistance
based on changes in the neurochemical concentration levels.
Hence, the acquired cell current is defined as the ratio of
the applied V
cell
to the total impedance of the cell (Z
cell
)
as expressed in equation (4). The reference current (E
I
ref
)
is used to define the control current I
ctrl
of CCO such that
the acquired redox current I
cell
is added to or subtracted from
E I
ref
as described in the following equation.
I
ctrl
= E I
ref
∓ I
cell
(5)
As a result, I
ctrl
is a positive value that decreases or increases
from the defined reference current E
I
ref
. Thus, E I
ref
can
be configured to support a wide range of input current. In
addition, the reference current E I
ref
can also be reduced if
the detected current range from the neurochemical of interest
is low, in order to reduce power consumption. The IA block
occupies an area of 0.019 mm
2
and draws an average current
of 6.64 µA from 1.8 V supply.
2) Current-to-Frequency (I-F): This block crosses both
analog and digital domains as depicted in Fig. 2, since it
performs the initial A/D conversion of the varying current
signal to frequency. Thereby providing as an output, phase rep-
resentation of the continuous-time analog signal which serves
as an input to the ID stage. As a result, this block performs
continuous-time sampling of the analog signal in frequency
domain and the ID block implements further processing of
the sampled signal such as quantization and discrete-time
sampling. Hence, this block and the ID block make-up a CCO-
based ADC for the readout front-end.
In addition, this block is implemented as a differential
CCO based on two single-ended ring oscillators where each
oscillator has three current starved inverter stages and two
output buffer stages. The inverter stages are controlled by I
ctrl
and the bias current of the oscillator (O
I
ref
). Hence, the
operating frequency (F
cco
) range of this block can be tuned
by adjusting O I
ref
of the CCO. Furthermore, the frequency
outputs of the differential CCO are complementary to each
other and when subtracted from each other, the attained I-
F sensitivity is twice that of a single oscillator. Hence, the
differential operation of the CCO improves the linearity of the
I-F conversion and increases the conversion gain, which in
turn improves the resolution attained from the ID stage. The
I-F block occupies an area of 0.005 mm
2
and draws an average
current of 10.46 µA from 1 V supply.
3) Current Discretization (ID): This block performs digital
signal processing (DSP) functions such as noise averaging
and encoding of the digital codes (D
OU T
). Noise averaging
is achieved over a long integration time (T
s
) of the pulses
from the I-F block followed by discrete-time sampling and
derivation as discussed in Section II-A. Hence, the data rate
of the system is defined by the sampling frequency (F
s
) of this
block which is limited by the signal bandwidth. In addition,
the current-to-digital code conversion gain of this block is
defined by F
cco
and F
s
based on the number of pulses or phase
transitions (N
p
) within each sampling interval T
s
, where N
p
changes as F
cco
is modulated by I
cell
. Therefore, tunability
of the F
cco
range offers flexible control of the dynamic range
(DR) and digital code resolution (n) of the system as shown in
the following equations based on equation (2). However, the
maximum F
cco
tuning range is limited by non-linearity of the
CCO which significantly degrades the effective-number-of-bits
(ENOB) of the ADC [5].
Dynamic range (DR) ≈
F
cco
(max) − F
cco
(min)
F
s
(6)
⇒ Resolution (n) ≈ log
2
(DR) (7)
Finally, it should be noted that the derivation stage is evalu-
ated off-chip during post-processing in Matlab. The ID block
occupies an area of 0.035 mm
2
and draws an average current
of 11.2 µA from 1 V supply.
III. POST-LAYOUT SIMULATION RESULTS
The readout front-end is implemented in 65 nm CMOS
technology and Fig. 6 shows the layout of the fabricated chip.
Fig. 7 shows the performance of the I-F and ID blocks based
on acquired I
cell
from the IA block. The I-F block and the
ID block achieve current sensitivity of 13 kHz/nA and 100
pA/LSB respectively. The I-F block is optimized to provide
16 MHz frequency range from the differential CCO as shown
in Fig. 7a, but it should be noted that the current sensitivity of
the I-F block can be increased by tuning the CCO to provide
wider frequency range. Fig 7b shows that 13.3-bits digital code
resolution was obtained from the ID block which could be
further increased by reducing the sampling frequency. Fig. 8
Fig. 6. Layout of the readout front-end chip implemented in 65 nm CMOS
−6 −4 −2 0 2 4 6
x 10
−7
−8
−6
−4
−2
0
2
4
6
8
x 10
6
F
osc
vs I
cell
Cell current [A]
Frequency [Hz]
F
oscp
− F
oscm
(a) OSC OU T P − OSC OU T M
−6 −4 −2 0 2 4 6
x 10
−7
−6000
−4000
−2000
0
2000
4000
6000
Digital codes as a function of I
cell
Cell current [A]
Digital output code
(b) D IG OUT P − DIG OUT M
Fig. 7. (a) I-F sensitivity of the CCO and (b) represents corresponding digital
codes D
OU T
from the ID block based on I
cell
range of 1.2µA.
10
2
10
3
−150
−120
−90
−60
−30
0
Power spectral density (PSD)
Frequency [Hz]
Magnitude [dB]
Fig. 8. FFT analysis of the digital output codes with 1 kHz and ±0.8 V
sine wave input which corresponds to full input current range of ±600 nA.
Cell voltage [V]
-1 -0.5 0 0.5 1
Cell current [A]
×10
-8
-4
-2
0
2
4
10 µ Mol ox.
10 µ Mol red.
Ibg oxidation
Ibg reduction
(a) 10 µMol of dopamine
Cell voltage [V]
-1 -0.5 0 0.5 1
Cell current [A]
×10
-7
-1
0
1
2
3
4
1 mMol ox.
1 mMol red.
Ibg oxidation
Ibg reduction
(b) 1 mMol of dopamine
Fig. 9. Simulation of detected cell current (I
cell
) and background current
(I
bg
) based on FSCV dopamine measurements from the sensor electrodes.
presents the power spectral density of the digital codes from
the ID block where the achieved SNR is 82 dB and SFDR is 50
dB, which corresponds to an ENOB of 8-bits. The significant
difference between the SNR and SFDR is due to the known
effect of inherent frequency tuning non-linearity of oscillator
based ADCs [5]. Nonetheless, the achieved resolution provides
good sensitivity for neurochemical applications as shown in
the simulation results presented in Fig. 9 and Fig. 10 based on
dopamine measurements with FSCV scan rate of 400 V/s.
Cell voltage [V]
-1 -0.5 0 0.5 1
Digital output codes
-300
-200
-100
0
100
200
300
(a) 10 µMol of dopamine
Cell voltage [V]
-1 -0.5 0 0.5 1
Digital output codes
-1000
0
1000
2000
3000
4000
(b) 1 mMol of dopamine
Fig. 10. Corresponding digital codes from the readout front-end based on
detected cell currents for the dopamine concentrations shown in Fig. 9.
IV. CONCLUSION
The readout of concentration levels of neurochemicals from
the brain contributes to the realization of fully-implantable
closed-loop interfaces for stimulation of degenerative neurons
and control of neural activities. This paper described the design
of a CCO-based readout front-end for neurochemical sensing
applications. The proposed design was implemented in a 65
nm CMOS process. Post-layout simulation results shows that
the readout front-end provides current resolution of 100 pA
and could detect minimum dopamine concentration of 10
µMol based on measured data from novel diamond-like carbon
electrodes. Higher and lower dopamine concentration than 10
µMol can also be detected from the readout front-end due to
its support for a wide current range of 1.2 µA(±600 nA).
The digital code representation of the detected dopamine has
a resolution of 13.3-bits with RMS conversion error of 0.18
LSB which results in an SNR of 82 dB at full current range
input. The achieved resolution of the readout front-end pro-
vides good sensitivity of released neurochemicals in the brain
which is useful for further understanding of neurotransmitters
and fostering research into improved treatments of related
neurodegenerative diseases.
ACKNOWLEDGMENT
The authors would like to thank Academy of Finland and
Aalto University for funding this research work.
REFERENCES
[1] R. K. Goyal and A. Chaudhury, “Structure activity relationship of
synaptic and junctional neurotransmission,” Autonomic Neuroscience,
vol. 176, no. 1-2, pp. 11–31, 2013.
[2] J. M. Beaulieu and R. R. Gainetdinov, “The physiology, signaling, and
pharmacology of dopamine receptors,” Pharmacological reviews, vol. 63,
no. 1, pp. 182–217, Mar 2011.
[3] M. Roham, D. Covey, D. Daberkow, E. Ramsson, C. Howard, B. Hei-
denreich, P. Garris, and P. Mohseni, “A Wireless IC for Time-Share
Chemical and Electrical Neural Recording,” IEEE Journal of Solid-State
Circuits, vol. 44, no. 12, pp. 3645–3658, Dec 2009.
[4] K. H. Lee, C. D. Blaha, P. A. Garris, P. Mohseni et al., “Evolution
of Deep Brain Stimulation: Human Electrometer and Smart Devices
Supporting the Next Generation of Therapy,” Neuromodulation:
Technology at the Neural Interface, vol. 12, no. 2, pp. 85–103, 2009.
[5] S. Rao, K. Reddy, B. Young, and P. Hanumolu, “A deterministic
digital background calibration technique for VCO-based ADCs,” IEEE
J. Solid-State Circuits, vol. 49, no. 4, pp. 950–960, 2014.
[6] V. Unnikrishnan and M. Vesterbacka, “Time-mode analog-to-digital
conversion using standard cells,” IEEE Transactions on Circuits and
Systems I: Regular Papers, vol. 61, no. 12, pp. 3348–3357, Dec 2014.