Mobile QR Code

1. (Department of Electrical and Information Engineering, Seoul National University of Science and Technology / Seoul, Korea hkcha@seoultech.ac.kr )

Charge pump, High voltage generator, Voltage regulation, Neural stimulator

## 1. Introduction

High supply voltage (HV) is often required in implantable bio-applications, such as neural stimulators, which are used to treat neurophysiological disorders, including epilepsy, paralysis, and Parkinson's Disease [1-4]. A neural stimulator is commonly used for treatment in these applications to deliver up to a few milliamperes of current to an electrode-tissue load, which has a high impedance. Thus, the output current driver of the stimulator must operate with an HV to achieve sufficient voltage compliance.

The circuit blocks comprising the system often require multiple supply voltages for operation, so the general approach is to use a low voltage as the supply for many of the system blocks to limit the system's overall power consumption. Then, an internal HV generator boosts this low voltage to produce the higher voltage for the circuits operating at HV supply. Energy resource can be limited in an implant system environment, so good power efficiency with small output voltage ripple performance is vital for these HV generation circuits.

Furthermore, as the HV generator's load current may vary depending on the following block's operating condition, some regulation scheme should be included along with the core pump circuits to maintain the target output voltage during load transients. Many HV generation circuits using capacitive switching charge pump (CP) topologies have been proposed with several variants of regulation techniques [3-12]. Some of these architectures are briefly reviewed in the following sections.

This paper proposes a HV CP IC which employs a complementary charge transfer switch (CTS) core pumping circuit for good efficiency. A 100-kHz fixed pumping clock frequency is employed considering the trade-off between the power efficiency and output voltage ripple performance. The output voltage is regulated by employing the input voltage modulation scheme [5], which modulates the input voltage of the CP core depending on the amount of load current.

The rest of the paper is presented as follows. The architecture of the overall system is described in Section 2. Section 3 reviews the widely used CP topologies along with a description of the proposed circuit design and the regulation scheme in detail. Section 4 discusses the measurement results, and Section 5 concludes the paper.

## 2. Proposed System Architecture

Fig. 1 shows the proposed system architecture. This work targets to generate a 15 V output from a 3.2 V input voltage. The generated output voltage is intended to be used as a reliable voltage supply for the following neural stimulator's output current driver. The output current driver is specified to provide the stimulation current up to 1 mA and is assumed to have a peak transient current of up to 5 mA during switching operations in the output driver [2]. Due to this wide range as well as the large current at the output, 1 ${\mathrm{\mu}}$F on-board external capacitors are used as pumping capacitance. Likewise, for the load capacitance, a similar value is used to support large load current and help suppress the voltage ripple at the output. The overall system is comprised of the multi-stage CP core, regulation circuits, non-overlapping clock generator, and peripheral circuits.

In contrast to previous work, which used the latched topology [5] as the CP core, a multi-stage complementary CTS topology is employed in this work due to its advantage of enhanced conversion efficiency. A single switch transistor at each stage for the current path results in low voltage drop and less parasitic capacitance than the previous charge pump design. In addition, this system uses an increased pumping frequency of 100 kHz to generate a larger output voltage and to improve the output ripple performance.

For the regulation function, the input voltage modulation technique is utilized where a low-dropout regulator (LDO) is placed at the CP core input. When the CP's load current condition changes, this change is sensed by the resistor divider comprised of R$_{\mathrm{F1}}$and R$_{\mathrm{F2}}$ resistors, which then becomes the input of the error amplifier (EA) in the feedback path. The output voltage of the EA is applied to the LDO as its reference. Thus, the change in the reference voltage of the LDO due to the difference in the load condition of the CP controls the LDO's output, which as a result becomes the input of the CP core.

## 3. Circuit Design

### 3.1 Review of Existing CP Topologies

The Dickson topology is a popular and widely used CP architecture [6]. The n-stage Dickson CP is shown in Fig. 2(a), where the transistor devices act as diodes instead of switches. These diode-connected transistors are used to transfer or deliver charge from one stage to the next and has a simple architecture that can easily be combined with other standard CMOS circuits in a system. However, this topology has a drawback of degraded efficiency as more stages are added as the threshold V$_{\mathrm{TH}}$ rises due to the body effect, resulting in higher voltage drops. Several techniques to avoid the rise in threshold voltage due to the body effect have been developed to alleviate this issue [7], with added cost of complexity.

Fig. 2(b) shows the widely used cross-coupled latched CP circuit [5, 8, 9], in a two-stage configuration as an example. The cross-coupled latched CP has the advantage of achieving a reliable conversion operation at high voltage with reasonable efficiency. One drawback of this structure is that it uses two pumping switches in series at each stage, which leads to the conduction loss due to the larger switch on-resistance (R$_{\mathrm{ON}}$) than the topology that uses only one pumping switch in the current path.

### 3.2 Proposed CP Core

Fig. 3 shows the proposed CP core circuit which employs the complementary CTS topology. The advantage of this circuit is that it employs a single transistor in each stage which leads to less conduction loss in comparison to the latched topology, which ultimately results in enhanced efficiency. Four stages are used to convert 3.2 V input to 15 V output. NMOS devices are used for the input stage switches and PMOS switches are used in the following stages in the core. To produce a large voltage output despite using 180 nm CMOS process, 3.3 V transistors are employed in the design. The switching transistors are sized with dimensions as shown in Fig. 3, which is decided considering in-rush current, R$_{\mathrm{ON}}$ value, and area. Inverters are utilized to control the ON-OFF of the PMOS switches. The source of the NMOS transistor in the inverter is connected to the previous stage node while the drain of the PMOS in the inverter is connected to the following stage node. Thus, the lower voltage from the previous stage is used to turn on the CTS switch, while the higher voltage in the following stage is used to turn it off.

The body terminal of the switches is connected to the dynamic body biasing (DBB) circuit for reliability purpose. This is used to connect the highest voltage between its source and drain for the PMOS transistor and the lowest potential for the NMOS transistor. As the input stage NMOS devices utilize deep n-well transistors, additional pairs of DBB circuits are used to ensure the n-well and the substrate are connected to the appropriate voltage.

The basic operation of the CP is described as follows. In the first stage, the charge passes through NMOS transistors SWN$_{1}$ and SWN$_{2}$ in the upper half and lower half branches of the cross-coupled structure, respectively. In the first cycle, when CLK$_{1}$ is low and CLK$_{2}$ is high, the voltage at node A$_{1}$ is close to V$_{\mathrm{IN}}$ and the voltage at node B$_{1}$ is about 2V$_{\mathrm{IN}}$. This causes the transistor SWN$_{1}$ to turn on and the charge gets transferred to capacitor C$_{1}$ from the input node. The voltage drop occurring across SWN$_{1}$ (V$_{\mathrm{DS}}$) is smaller than the threshold voltage. When CLK$_{1}$ is high and CLK$_{2}$ is low in the next clock cycle, the voltage at node A$_{1}$ becomes twice the V$_{\mathrm{IN}}$ while node B$_{1}$ decreases to around V$_{\mathrm{IN}}$, thus turning SWN$_{1}$ off and blocks the current path. SWN$_{2}$ operates similarly in the opposite way. After that, the charge will be passed through transistors SWP$_{1-6}$, where these switches are controlled using inverters. As the charge transfer operation of switch transistors SWP$_{1}$ to SWP$_{6}$ is similar, the second stage operation is described as an example. At this point, the voltage at the source node of SWP$_{1}$ is around 2V$_{\mathrm{DD}}$. To turn on and off SWP$_{1}$ using the lower voltage from the previous stage, both the NMOS and PMOS gates of the inverter INV$_{1}$ are connected to node A$_{1}$, the NMOS source of INV$_{1}$ is connected to B$_{1}$, and PMOS source of INV$_{1}$ is connected to node A$_{2}$. In contrast, the drain nodes of NMOS and PMOS devices are connected to the gate of SWP$_{1}$. Likewise for SWP$_{2}$, both NMOS and PMOS gates of inverter INV$_{2}$ are coupled to B$_{1}$, the NMOS source of INV$_{2}$ is connected to node A$_{1}$, and the PMOS drain of INV$_{2}$ is connected to node B$_{1}$, while the drain nodes of NMOS and PMOS devices are coupled to the gate of SWP$_{2}$. When CLK$_{1}$ is high, the SWP$_{1}$ switch turns off and cuts off the current path to the next stage. When the CLK$_{1}$is low, the SWP$_{1}$ turns on. The transistor SWP$_{2}$ works in a similar manner with CLK$_{2}$.

The last stage comprised of switches SWP$_{7}$ and SWP$_{8}$operates in a similar working principle as the first stage but uses the PMOS transistor instead of NMOS. At this point, the voltage level at the source node of SWP$_{7}$ and SWP$_{8}$ is around 5V$_{\mathrm{DD}}$. The gate terminal of SWP$_{7}$ is connected to B$_{\mathrm{4,}}$ while the gate of SWP$_{8}$ is connected to A$_{4}$. When CLK$_{1}$ is low, SWP$_{7}$is turned on while SWP$_{8}$ is turned off. The on-state of SWP$_{7}$ and SWP$_{8}$ allows transfer of charge from C$_{7}$ and C$_{8}$ to the output capacitor C$_{\mathrm{out}}$.

To generate the two-phase clock for driving the complementary branches, a non-overlapping clock signal generation circuit, shown in Fig. 4, is realized by combining NAND gate and inverters. A small delay is included to minimize the pumping transistors turning on and off simultaneously.

### 3.3 Regulation Scheme

The output voltage of a complementary CP circuit can be defined as:

##### (1)
$V_{o}=\left(N+1\right)V_{I{N_{-}}CP}-N\left(\frac{I_{L}}{2f_{osc}C_{p}}\right)$

where $V_{I{N_{-}}CP}$ represents the input of the CP, N is the number of stages, f$_{\mathrm{OSC}}$ is the pumping frequency, C$_{\mathrm{P}}$ is the pumping capacitor, and I$_{\mathrm{L}}$ is the load current. It can be observed from this equation that either $V_{I{N_{-}}CP}$ or f$_{\mathrm{OSC}}$ should be controlled for regulating the output V$_{\mathrm{O}}$ with a change in load current [4,10]. The issue of tuning the pumping frequency is that regulation may not be achieved at very light load condition as I$_{\mathrm{L}}$ ${\approx}$ 0. At heavy load condition where the pumping frequency is dynamically increased to pump more current to the output, there can be excessive switching loss due to the increased switching frequency which may degrade the overall efficiency of the CP. On the other hand, modulating the input voltage can be done with no limitations at both light and heavy loads with a fixed frequency for switching. The pumping frequency affects the efficiency as well as the output ripple voltage. A trade-off exists between these two parameters where a high clock frequency degrades the efficiency while it is favorable for the ripple performance. This work employs the input voltage modulation regulation method while using a fixed 100 kHz pumping frequency for minimizing the loss due to fast switching and achieve a reasonable ripple performance.

While employing the 1 ${\mathrm{\mu}}$F pumping capacitor, the appropriate $V_{I{N_{-}}CP}$ value can be calculated using equation (1). Two load conditions are considered where I$_{\mathrm{L}}$= 0 mA and I$_{\mathrm{L}}$ = 5 mA, which are minimum and maximum load conditions, respectively:

##### (2)
$V_{I{N_{-}}CP}=\frac{V_{O}}{N+1}=\frac{15}{5}=3V,\text{ when }I_{L}=0$
##### (3)
$V_{I{N_{-}}CP}=\frac{V_{O}+N\cdot \frac{I_{L}}{2f_{osc}\cdot C_{p}}}{N+1}=\frac{15.1}{5}=3.02V,\text{ when }I_{L}=5mA$

Now, the relationship between $V_{I{N_{-}}CP}$ and $V_{REF\_ LDO}$can be decided by the following equation:

##### (4)
$V_{IN\_ CP}\cdot \frac{R_{L2}}{R_{L1}+R_{L2}}=V_{REF\_ LDO}$

where R$_{\mathrm{L1}}$ and R$_{\mathrm{L2}}$ are assumed to be the feedback voltage divider resistors in the output of the LDO circuit. Assuming R$_{\mathrm{L1}}$ = R$_{\mathrm{L2}}$, when $V_{I{N_{-}}CP}$varies from 3 V to 3.02 V from Eqs. (2) and (3), the $V_{REF\_ LDO}$value should fall between 1.5 V and 1.51 V in order to support the load current values between 0 to 5 mA. From this data, the reference voltage of EA ($V_{REF\_ EA}$) is set to 1.5 V.

Fig. 5 shows the simplified circuit schematic of the EA in the feedback regulation path, which performs the error comparison between voltage-divided output voltage V$_{\mathrm{FB}}$and the reference voltage. The voltage error is first converted to current $I_{Fb}$ using the operational transconductance amplifier (OTA) in the EA and then re-converted back to voltage output $V_{REF\_ LDO}$. The $V_{REF\_ LDO}$can be expressed as follows:

##### (5)
$V_{RE{F_{-}}LDO}=\left(I_{Fb}+I_{C}\right)\cdot R_{C}$

The general idea is that when the V$_{\mathrm{OUT}}$ of the CP does not exceed 15 V, then V$_{\mathrm{FB}}$ < $V_{REF\_ EA}$and OTA delivers $I_{Fb}$ to R$_{\mathrm{C}}$, increasing $V_{REF\_ LDO}$, which leads to the increase of $V_{I{N_{-}}CP}$and thus the increase of V$_{\mathrm{OUT}}$. When V$_{\mathrm{OUT}}$ exceeds 15 V, the OTA will sink $I_{Fb}$ from R$_{\mathrm{C}}$, thus decreasing $V_{I{N_{-}}CP}$and V$_{\mathrm{OUT}}$. The resulting output voltage generated is applied to the LDO in the feedforward input path as the LDO reference voltage. The output of the EA circuit also features a limiter (upper and lower limiters) to limit the voltage applied to the LDO at start-up of the CP when the error between the feedback V$_{\mathrm{FB}}$ and the reference $V_{REF\_ EA}$ is too large. This forces $V_{I{N_{-}}CP}$ to be within the range of 2.8 V to 3.1 V.

Fig. 6 shows the LDO used in this system. A conventional LDO circuit [5] is used with adjustments to enable a larger output voltage and load current. As shown, a PMOS pass transistor, a telescopic error amplifier, a buffer, and two feedback resistors are used in the LDO circuit. A similar limiter block used for the EA is also utilized in the LDO to limit the current in the pass transistor at start-up.

## 4. Experimental Results

The proposed HV CP IC is implemented using 180 nm CMOS process. Fig. 7 shows the chip micrograph where its core area is 968 ${\mathrm{\mu}}$m ${\times}$ 1012 ${\mathrm{\mu}}$m. The die is assembled on an FR4 PCB while using the chip-on-board (COB) package for measurement. External on-board voltage regulators generate bias voltages for the IC, while the 100~kHz clock signal is generated and applied from a waveform generator. The pumping and load capacitors with values of 1-${\mathrm{\mu}}$F are implemented by using surface mount device components.

The measured startup transient waveform is shown in Fig. 8, which closely follows the simulated startup transient results. Fig. 9 shows the measured power efficiency and output voltage against the change in load current. Thanks to the regulation function, the output voltage is kept at around 15 V with changing load current. At a light load of 1 mA, the efficiency performance is around 25%. This efficiency improves to over 88% as the load increases up to 5 mA. Fig. 10 shows the measured output voltage while the load current is switched between values of 0 and 5 mA. A 79 mV of output voltage ripple exists during the no load condition of 0 mA while this increases to 150 mV at 5 mA, which is around 1% of ripple voltage (${\Delta}$V/V$_{\mathrm{OUT}}$).

Table 1 compares the performance to that of previously published CP ICs that utilize similar off-chip pumping capacitors and generate over 10-V output voltages from an input voltage less than 5 V. In comparison to previous CPs, this work achieves improved power efficiency, has good ripple performance, and supports wide range of load current.

##### Table 1. Performance summary and comparison.
 Parameters [3] [5] [10] [11] This work Output Voltage 17 V 12.8 V 20 V 16 V 15 V Input Voltage 3.3 V 2.8 V 4.6 V 3.3 V 3.2 No. of Stages 10 4 4 6 4 Pump. Cap. 50 pF 1 µF 1 µF 1 µF 1 µF Load Cap. 20 nF 1 µF 1 µF 100 nF 1 µF Freq. 2.5 MHz 10 kHz 3 kHz-1 MHz 6.67 MHz 100 kHz Ripple N/A 31.6 mV @ 10 µA 120 mV @ 0.6 mA 160 mV @ 7 mA 150 mV @ 5 mA Load Current 0-0.31 mA 10 µA-1 mA 0.6 mA 0.1-7 mA 0-5 mA Max. Eff. 10%@N/A 84.7% @ 2 mA 82% @ 0.6 mA 70.8% @ 4 mA 88.3% @ 5 mA CMOS Tech. 350 nm HV 180 nm 180 nm HV 160 nm 180 nm

## 5. Conclusion

A highly efficient charge pump circuit was designed and implemented using 180-nm CMOS technology. The proposed circuit utilizes a four-stage complementary CTS core with an effective regulation technique. The circuit generates 15-V output from an input voltage of 3.2 V and achieves a maximum power efficiency of 88.3% at 5 mA current load with 1% output ripple voltage.

### ACKNOWLEDGMENTS

This research was supported by the Research Program funded by Seoultech (Seoul National University of Science and Technology).

### REFERENCES

1
Noorsal E., Sooksood K., Xu H., Hornig R., Becker J., Ortmanns M., Jan. 2012, A neural stimulator frontend with high-voltage compliance and programmable pulse shape for epiretinal implants, IEEE J. of Solid-State Circuits, Vol. 47, No. 1, pp. 244-256
2
Abdi A., Cha H.-K., 2016, A bidirectional neural interface CMOS analog front-end IC with embedded isolation switch for implantable devices, Microelectronics J., Vol. 58, pp. 70-75
3
Bisoni L., Carboni C., Raffo L., Carta N., Barbaro M., Feb 2015, An HV-CMOS integrated circuit for neural stimulation in prosthetic applications, IEEE Trans. on Circuits and Systems II: Express Briefs, Vol. 62, No. 2, pp. 184-188
4
Chen W., et al. , Jan 2014, A fully integrated 8-channel closed-loop neural-prosthetic CMOS SoC for real-time epileptic seizure control, IEEE J. Solid-State Circuits, Vol. 49, No. 1, pp. 232-247
5
Abdi A., Kim H. S., Cha H.-K., March 2019, A high-voltage generation charge-pump IC using input voltage modulated regulation for neural implant devices, IEEE Trans. on Circuits and Systems II: Express Briefs, Vol. 66, No. 3, pp. 342-346
6
Dickson J. F., Jun. 1976, On-chip high-voltage generation in MNOS integrated circuits using an improved voltage multiplier technique, IEEE J. Solid-State Circuits, Vol. sc-11, No. 3, pp. 374-378
7
Jiang X., Yu X., Moez K., Elliott D. G., Chen J., Mar. 2018, High efficiency charge pumps for low-power on-chip applications, IEEE Trans. on Circuits and Systems I: Reg. Papers, Vol. 65, No. 3, pp. 1143-1153
8
Ker M.-D, Chen S.-L., Tsai C.-S., May 2006, Design of charge pump circuit with consideration of gate-oxide reliability in low-voltage CMOS processes, IEEE J. Solid-State Circuits, Vol. 41, No. 5, pp. 1100-1107
9
Pelliconi R., Iezzi D., Baroni A., Pasotti M., Rolandi P. L., Jun. 2003, Power efficient charge pump in deep submicron standard CMOS technology, IEEE J. Solid-State Circuits, Vol. 38, No. 6, pp. 1068-1071
10
Zhao J., Yao L., Xue R., Li P., Je M., Xu Y. P., Feb 2016, An integrated wireless power management and data telemetry IC for high-compliance-voltage electrical stimulation applications, IEEE Trans. on Biomedical Circuits and Systems, Vol. 10, No. 1, pp. 113-124
11
Karadi R., Pique G. V., Feb 2014, 3-phase 6/1 switched-capacitor dc-dc boost converter providing 16 V at 7 mA and 70.3% efficiency in 1.1 mm$^{3}$, in 2014 IEEE Intl. Solid-State Circuits Conf. Dig. of Tech. Papers (ISSCC), pp. 92-93

## Author

##### Geri Paksi

Geri Paksi received the B.S. degree in electrical engineering at Universitas Indonesia, Indonesia, in 2018, and the M.S. degree in electrical and information engineering at Seoul National University of Science and Technology, Seoul, Korea in 2021. In 2021, he joined Accenture, Indonesia, as an Application Development Analyst. His research interest includes the design of highly efficient charge pumps for biomedical applications.

##### Hyouk-Kyu Cha

Hyouk-Kyu Cha received the 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 an Associate Professor. His research interests include low-power CMOS analog/RF IC and system design for implantable and wearable biomedical devices.