An Implantable Charge Balanced Neurostimulator IC Using Standard CMOS Process
(Tuan Vo)
1
(Hyouk-Kyu Cha)
1,*
-
(Department of Electrical and Information Engineering, Seoul National University of
Science and Technology, Seoul, South Korea. hkcha@seoultech.ac.kr)
Copyright © The Institute of Electronics and Information Engineers(IEIE)
Keywords
Neural stimulation, Active charge balancing, Transistor stacking, Dynamic biasing
1. Introduction
Neural stimulators have been applied to treat various neural disorders such as in
retinal implants to treat sight impairment and in deep brain stimulation devices to
suppress epileptic seizures [1-
3]. A stimulator circuit should deliver and recover electric charge in a precisely controlled
fashion between the electrical device and the target tissue so that it is able to
initiate an action potential leading to events such as sensory perception without
causing damage to the tissue.
Among several important requirements for the stimulator design, the stimulator should
achieve high voltage compliance so that the stimulator is able to deliver wide range
of desired current to the high impedance electrode-tissue interface. The stimulator
is thus usually designed with high-voltage (HV) CMOS or DMOS transistors as HV supply
is needed to meet the voltage compliance requirement [1]. Low voltage (LV) standard CMOS implementations have previously been presented with
low supply voltages, but these works have limitations in maximum deliverable current
[3,
4].
Another key requirement is safety as maintaining charge balance during the stimulation
process is important to prevent tissue damage. Thus, biphasic current stimulation
is usually used in which the charge is delivered to the tissue in the first phase
and is recovered back in the next phase. Ideally, there should be no charge remaining
in the tissue after each stimulation cycle. However, mismatch in the biphasic pulse
due to process variations result in a non-ideal stimulation and may cause some residual
charge to remain in the tissue. This net charge builds a voltage difference between
the electrodes, and ultimately can lead to tissue damage. Several charge balancing
methods have been proposed in previous works such as DC block capacitor [5], electrode shorting [1] and active charge balancing [6,
7] to alleviate this issue. DC blocking is not appropriate for multi-channel applications
as it requires a large external capacitor per electrode. Electrode shorting is a simple
method but takes a long time to sufficiently short out the residual charge. Active
charge balancing in [6] utilizes a complex feedback circuitry with HV comparator which requires a HV process.
In this paper, in comparison to previous similar works [1,
4,
6,
7], a current-mode bipolar neural stimulator IC is proposed whose output current driver
is composed of stacked 3.3-V LV standard CMOS transistors with dynamic gate biasing
circuit in order to reliably operate at high voltage supply and deliver from 32 µA
up to 992 µA of current to the target tissue. A regulated cascode output stage for
boosting the output impedance of the current driver is used in order to improve the
stimulation accuracy. Additionally, the proposed stimulator is equipped with a simple
feedback-based active charge balancing circuit in order to achieve minimal residual
charge after stimulation for safety requirement. All the block level circuits comprising
the overall stimulator are designed using LV transistors. Section II describes the
architecture and general design considerations. The circuit design details are presented
in Section III. The simulation results are reported in Section IV, followed by the
conclusions in Section V.
2. Design Considerations and System Architecture
There are several specifications that need to be defined when designing the stimulator.
Some of these parameters include stimulation waveform, pulse width, repetition rate,
and current magnitude range. A biphasic rectangular waveform is considered for stimulation
with a 100-µs stimulation pulse width, 20-µs interphasic delay between cathodic and
anodic pulses, and repetition rate of 400 µs. After the initial stimulation phase,
charge balancing phase takes place within the 50-µs short timing window. This work
targets current magnitudes in the range of 32 µA to 992 µA which is sufficient to
support various applications such as retinal implants and for treating epilepsy. Fig. 1 shows the equivalent electrical model [2] of the electrode and tissue interface used in the design. In order to accurately
deliver up to around 1 mA of current to a 10-kΩ load, a high supply voltage is required
in the output driver circuit. In this work, 12.8-V supply is chosen with margin considering
the voltage drop in the switches and the current sink transistor in the output driver
[8].
Fig. 1. Equivalent electrical model of electrode-tissue interface used in the design.
Fig. 2 shows the overall block diagram of the proposed neural stimulator architecture which
is comprised of the following building blocks; 5-bit digital-to-analog converter (DAC),
digital control logic, level-shifters, output current driver, and charge balancing
circuits. The DAC provides current amplitude values from 4 to 124 µA, while the digital
control logic is responsible for controlling the output current amplitude, timing
of the waveforms, and the charge balancer circuit. The level-shifters are required
to convert the 0-3.2 V trigger signal to 9.6-12.8 V signal to drive the PMOS transistor
switch of the output current driver. The output of the DAC is applied to the current
driver, in which the current is ultimately amplified by 8 times through current mirroring
in order to provide up to a maximum of 992 µA stimulation current. The final component
is the simple active charge balancing circuit that ensures minimal net charge remaining
in the tissue. The charge balancer first senses the voltage difference between the
two electrodes. If the difference exceeds the safety range of -50 mV to +50 mV [6], the charge balancing circuit sends a feedback signal to the digital control logic
so that the current driver is controlled to generate additional cathodic or anodic
current in order to reduce the voltage difference between the electrodes. The charge
balancer keeps sending feedback to the control logic until the voltage difference
between the two electrodes is lower than 50 mV. Thus, the residual charge would be
sufficiently small enough to prevent tissue damage. In addition to the active charge
balancer function, the electrode shorting option can be employed to support residual
charge reduction after each stimulation cycle. The shorting technique is also controlled
by the digital control logic, which shunts two electrodes in order to force the voltage
between them to be equal to zero.
Fig. 2. Overall architecture of neural stimulator IC.
3. Circuit Design
Fig. 3 presents the simplified block diagram of the output current driver. A bipolar H-bridge
topology is employed in which two switches diagonally opposite to each other are turned
ON in each phase to produce the biphasic current waveform. Due to HV operation of
the output current driver, a HV transistor switch is commonly utilized. In this work,
each HV switch transistor is replaced by a stack of four standard CMOS transistors
as depicted in Fig. 4. Each standard transistor has the nominal operation voltage of 3.3 V, and the voltage
difference between any of its terminals is designed so that it would not exceed the
nominal voltage to avoid device breakdown. Thus, the dynamic gate biasing topology
[9] is utilized in order to ensure all the transistors work within the nominal voltage
during transient operation. Since the PMOS transistor switch requires 9.6–12.8 V control
signal, level-shifters are provided in order to produce this signal.
Fig. 3. Block diagram of biphasic bipolar H-bridge output current driver.
Fig. 4. Circuit schematic of proposed output current driver.
The basic operation of the output driver in Fig. 4 is described as follows; INP1,2 and INN1,2 are the control signals which control the ON and OFF of the switch transistors. 3.2
V, 6.4 V, and 9.6 V are fixed DC voltages which can be generated internally [8] or applied externally. In order to turn OFF the upper switch (HSW1) and turn ON the
lower switch (HSW2), the INP1 must be set to 12.8 V, while the INN1 must be 3.2 V. This turns the transistor MN1 ON. As a result, the source of MN2 is shorted to the current sink, and the transistor MN2 would turn ON as its gate is constantly biased to 3.2 V. When MN2 is ON, the node n1 is connected to the current sink, which means this node goes ‘LOW’
causing MBP1 to turn ON. Thus, the gate of MN3 is shorted to 3.2 V allowing the transistor MN3 to turn ON. Similar sequence of events occurs, and results in transistor MN4 to turn ON. Finally, the switch HSW2 is ON since every NMOS transistor of its stack
is active. On the other hand, INP1 is set to be 12.8 V which turns OFF MP1. MP2 is also OFF because its gate is constantly biased at 9.6 V. Moreover, when the gate
of MN4 is shorted to 3.2 V, transistor MP4 is OFF as its drain is connected to a ‘LOW’ voltage. At the same time, MBN3 is OFF and MBP3 is ON which makes the gate voltage of MP3 to be 6.4 V. This turns OFF MP3. As a result, the switch HSW1 is OFF since every PMOS transistor of its stack is
OFF. Similarly, the same mechanism would occur to turn ON all the transistors of the
upper stack and turn OFF those of the lower one in case when the upper switch is turned
ON and the lower switch is turned OFF.
The current sink part of the current driver in Fig. 4 is designed using a regulated cascode topology [10] in order to enhance the output impedance as depicted in Fig. 5. A differential amplifier ADA is applied to equalize the drain-source voltage values of VDS1 and VDS2. This makes VGS and VDS of the two mirror transistors M1 and M2 equal to each other. The differential amplifier is composed of a differential pair
that consists of M5 and M6, followed by a DC-level shifter M7. When transistors M2 and M4 are in saturation mode, the output resistance of the current sink is determined by
where ADA is the gain of the differential amplifier, and AV-M4 denotes the gain of the cascode transistor M4 which is computed as AV-M4 = gm4r04. Therefore, the output impedance is boosted to a higher value in comparison to a
simple current mirror circuit.
Fig. 5. Circuit schematic of current sink in the output driver.
The DAC, shown in Fig. 6, is designed to generate current whose amplitude varies from 4 to 124 µA. A cascode
current-steering topology is utilized to construct the DAC in order to achieve high
linearity [11].
Fig. 6. Circuit schematic of DAC.
After every stimulation cycle, there is some residual charge accumulated in the electrodes
due to the mismatch in the circuit. This residual charge leads to the increment of
the voltage difference between the two electrodes. Therefore, the designed active
charge balancer operates based on that output differential voltage. At first, the
charge balancer circuit initially determines the voltage difference between the two
electrodes. If the voltage difference is smaller than 50 mV, no feedback signal is
sent back to the digital controller. In case the voltage difference exceeds 50 mV,
the charge balancer delivers a feedback signal to the digital controller to produce
additional cathodic or anodic current in order to minimize the voltage difference
between the electrodes.
Fig. 7. Simplified block diagram of active charge balancer.
Fig. 7 shows the simplified block diagram of the active charge balancer, and explains its
principle. If the voltage of the electrode 1 is 50 mV greater than that of the electrode
2 (EL1 > EL2), the feedback signal would be sent to make the digital controller generate more
current whose direction would be from electrode 2 to electrode 1 in order to lessen
the accumulated charge. Similarly, if voltage of the electrode 1 is 50-mV smaller
(EL1 < EL2), an additional opposite-direction current would be generated. The charge balancer
continues sending feedback until the voltage difference becomes lower than 50 mV.
4. Simulation Results
The proposed stimulator IC is designed using 0.18-µm standard CMOS process. The layout
capture of the overall IC is shown in Fig. 8. The core layout area excluding the metal routing paths is 0.12 mm2. Fig. 9 shows the stimulation transient voltage waveform. Some mismatch between the cathodic
and anodic waveforms is added on purpose to the pulse width to observe the operation
of the charge balancing circuit. It can be observed that a large initial voltage difference
is reduced below 50 mV after charge balancing. Fig. 10 shows the simulated stimulus output current versus the output voltage for the four
different DAC input values. The voltage compliance at the maximum reachable current
is approximately 12.4 V, which is 96.8% of the supply voltage.
Fig. 8. Layout capture of designed neurostimulator IC.
Fig. 9. Stimulation output voltage waveform with charge balancer in operation.
Fig. 10. Simulated output current for selected DAC codes.
Fig. 11 (a) depicts the cathodic and anodic currents of the proposed stimulator with respect
to the DAC input codes. The output current varies with a step of 32 µA (1 LSB current
of the DAC) to 992 µA (full scale current of the DAC). According to the result, the
dissimilarity between the cathodic and anodic currents is negligibly small, as shown
in Fig. 11 (b).
Fig. 11. (a) Cathodic and anodic current versus DAC input codes. (b) Mismatch between
cathodic and anodic current versus DAC input codes.
Regarding the power consumption, the total static power consumed by the IC is 0.7
mW, which is due to the operation of the current DAC, current mirror, and charge balancing
circuits.
Table 1 summarizes the performance of the proposed stimulator, and compares them with those
from other previous works. Despite using a low-voltage standard CMOS process, the
stimulator in this work is able to accurately deliver a wide range of current to a
high impedance load.
Table 1. Performance summary and comparison.
|
Parameters
|
[1] |
[4] |
[12] |
This work*
|
|
Stimulation current (A)
|
32-1000µ
|
0.5-465µ
|
2µ-10m
|
32-992µ
|
|
Supply voltage (V)
|
15
|
3
|
-
|
12.8
|
|
Tissue load model (Ω)
|
10 k
|
10 k
|
1.8 k
|
10 k
|
|
DAC INL/DNL (LSB)
|
0.1/0.2
|
0.7/0.35
|
0.71/0.84
|
0.66/0.02
|
|
Voltage compliance (V)
|
-
|
-
|
3.3-49
|
12.4
|
|
Current mismatch (%)
|
0.3
|
0.5
|
-
|
0.01
|
|
Charge balancing
|
Shorting
|
Shorting
|
Shorting/Active
|
Active
|
|
Area (mm2)
|
0.11
|
8 (256 ch.)
|
15.6 (6 ch.)
|
0.12
|
|
CMOS Technology (µm)
|
0.35 HV
|
0.065 LV
|
0.35 HV
|
0.18 LV
|
*Post-layout simulation results
5. Conclusion
A charge balanced biphasic neurostimulator circuit which has the ability to deliver
up to 1-mA current through 10-kΩ load using 12.8 V supply voltage is designed using
180-nm standard CMOS process. The proposed stimulator employs a transistor-stacked
output current driver architecture for reliable high voltage operation and achieves
a voltage compliance of 12.4 V by using a regulated cascode current mirror topology.
For future work, benchtop testing and in-vitro measurements will be carried out to
verify the performance of the IC.
Acknowledgement
This study was supported by the Research Program funded by Seoultech (Seoul National
University of Science and Technology). The authors thank IC Design Education Center
(IDEC) for providing EDA tool.
References
H. Chun , Y. Yang , T. Lehmann , Safety ensuring retinal prosthesis with precise
charge balance and low power consumption, IEEE Transactions on Biomedical Circuits
and Systems, Vol. 8, No. 1, pp. 108-118, 2014

J.-Y. Son , H.-K. Cha , An implantable neural stimulator IC with anodic current
pulse modulation based active charge balancing, IEEE Access, Vol. 8, pp. 136449-136458,
2020

H.-G. Rhew , J. Jeong , J. A. Fredenburg , S. Dodani , P. G. Patil , M.
P. Flynn , A fully self-contained logarithmic closed-loop deep brain stimulation
SoC with wireless telemetry and wireless power management, IEEE Journal of Solid-State
Circuits, Vol. 49, No. 10, pp. 2213-2227, 2014

N. Tran , S. Bai , J. Yang , H. Chun , O. Kavehei , Y. Yang , V. Muktamath
, D. Ng , H. Meffin , M. Halpern , E. Skafidas , A complete 256-electrode
retinal prosthesis chip, IEEE Journal of Solid-State Circuits, Vol. 49, No. 3, pp.
751-765, 2014

K. N. Hageman , Z. K. Kalayjian , F. Tejada , B. Chiang , M. A. Rahman ,
G. Y. Fridman , C. Dai , P. O. Pouliquen , J. Georgiou , C. C. Della Santina
, A. G. Andreou , A CMOS neural interface for a multichannel vestibular prosthesis,
IEEE Transactions on Biomedical Circuits and Systems, Vol. 10, No. 2, pp. 269-279,
2016

M. Ortmanns , A. Rocke , M. Gehrke , H.-J. Tiedtke , A 232-channel epiretinal
stimulator ASIC, IEEE Journal of Solid-State Circuits, Vol. 42, No. 12, pp. 2946-2959,
2007

R. Ranjandish , A. Schmid , An active charge balancing method based on anodic
current variation monitoring, Proc. of 2017 IEEE Biomedical Circuits and Systems Conference
(BioCAS), pp. 1-4, 2017

A. Abdi , H. S. Kim , H.-K. Cha , A high-voltage generation charge-pump IC
using input voltage modulated regulation for neural implant devices, IEEE Transactions
on Circuits and Systems II: Express Briefs, Vol. 66, No. 3, pp. 342-346, 2019

A. Banuaji , H.-K. Cha , A 15-V bidirectional ultrasound interface analog front-end
IC for medical imaging using standard CMOS technology, IEEE Transactions on Circuits
and Systems II: Express Briefs, Vol. 61, No. 8, pp. 604-608, 2014

J. Ramírez-Angulo , R. G. Carvajal , G. O. Ducoudray , A. J. López-Martín
, A. Torralba , New compact CMOS continuous-time low-voltage analog rank-order
filter architecture, IEEE Transactions on Circuits and Systems II: Express Briefs,
Vol. 51, No. 5, pp. 257-261, 2004

M. Sivaprakasam , W. Liu , M. S. Humayun , J. D. Weiland , A variable range
bi-phasic current stimulus driver circuitry for an implantable retinal prosthetic
device, IEEE Journal of Solid-State Circuits, Vol. 40, No. 3, pp. 763-771, 2005

A. Taschwer , N. Butz , M. Köhler , D. Rossbach , Y. Manoli , A charge
balanced neural stimulator with 3.3 V to 49 V supply compliance and arbitrary programmable
current pulse shapes, Proc. of 2018 IEEE Biomedical Circuits and Systems Conference
(BioCAS), pp. 1-4, 2018

Tuan Vo received his B.S. degree in biomedical engineering at International University,
Vietnam National University in 2015, Ho Chi Minh City, Vietnam, and an M.S. degree
in electrical and information engineering at Seoul National University of Science
and Technology, Seoul, Korea in 2019. Since 2019, he has been working toward a Ph.D
degree in the Department of Electrical Engineering and Computer Science at Florida
Atlantic University in the United States. His research interests include neuroscience,
brain-computer interfaces, and analog IC design for biomedical devices.
Hyouk-Kyu Cha received his B.S. and Ph.D degrees in electrical engineering at Korea
Advanced Institute of Science and Technology (KAIST), Daejeon, Korea, in 2003 and
2009, respectively. From 2009 to 2012, he was a Scientist with the Institute of Microelectronics,
(IME), Agency for Science, Technology, and Research (A*STAR), Singapore, where he
was involved in the research and development of analog/RF ICs for biomedical applications.
Since 2012, he has been with the Department of Electrical and Information Engineering,
Seoul National University of Science and Technology, Seoul, Korea, where he is now
a Professor. His research interests include low-power CMOS analog/RF IC and system
design for biomedical devices.