Microelectronics Journal 42 (2011) 966–971
Contents lists available at ScienceDirect
Microelectronics Journal journal homepage: www.elsevier.com/locate/mejo
A high precision low dropout regulator with nested feedback loops Chua-Chin Wang , Ron-Chi Kuo, Tung-Han Tsai National Sun Yat-Sen University, Department of Electrical Engineering, 70 Lian-Hai Rd., Kaohsiung, Taiwan
a r t i c l e i n f o
a b s t r a c t
Article history: Received 15 July 2010 Received in revised form 16 March 2011 Accepted 21 March 2011 Available online 19 May 2011
A high precision low dropout regulator (LDO) with nested feedback loops is proposed in this paper. By nesting a zero-tracking compensation loop inside of the negative feedback loop comprising an error amplifier, the independence of off-chip capacitor and effective series resistance (ESR) is ensured for different load currents and operating voltages. This circuit is designed and fabricated using a 0:35 mm standard CMOS process. The die area is a 1350 mm 560 mm. The measurement results show that the total error of the output voltage caused by line and load variations is less than 7 3% in low quiescent current (Iddq) or low voltage scenarios. Besides, the smallest dropout of the LDO, 0.11 V, while the output current is 165 mA, the output load is 4:7 mF and 20 in parallel. & 2011 Elsevier Ltd. All rights reserved.
Keywords: Low dropout Low noise High precision LDO
1. Introduction The LDO (low-dropout) regulator is generally utilized to translate a voltage level or provide a stable output voltage, which has been considered as one of the most important components for the power management of wireless applications, e.g., cellular phones, hand-held computers, and particularly implanted wireless biomedical chips [1]. Traditional linear LDO regulators usually consist of a two-stage error amplifier, a power output stage and a negative feedback loop. However, since the stability and accuracy of the regulated output voltage is the most critical factor in many field applications such as biomedical chips, traditional designs suffer from poor efficiency as the difference between the input and output voltages is enlarged [2]. To attain a fast transient response and an accurate output voltage, several prior LDO regulators were proposed, which utilized high gain amplifiers with either a cascade or cascode topology. It results in the stability problem due to the appearance of inherent three poles. Besides, the varying load current also affects the stability [3–5]. Thus, several frequency compensation techniques to ensure the stability of the LDO regulator was proposed to resolve this problem [4–16]. Nevertheless, those compensation techniques usually come with disadvantages. For example, cascode compensation with a dynamic bandwidth boosting was proposed in [10,13], which guaranteed stability over the full range of alternating load current at the expense of high power consumption.
Corresponding author. Tel.: þ 886 7 5252000x4144; fax: þ 886 7 5254199.
E-mail addresses:
[email protected] (C.-C. Wang),
[email protected] (R.-C. Kuo),
[email protected] (T.-H. Tsai). 0026-2692/$ - see front matter & 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.mejo.2011.03.014
It is not welcomed in mobile devices, particularly. Besides, the off-chip capacitor and its ESR (equivalent series resistor) are designed to move the zero around to achieve a pole-zero cancelation and keep the phase margin larger than 601 to remain the feedback system stable [18]. However, the off-chip capacitor increases the area of the printed-circuit board (PCB) and the ESR value might drift owing to the variation of the temperature and different materials. Therefore, the off-chip capacitor scheme is not acceptable in many applications demanding high precision. In this work, a new compensation architecture independent of the offchip capacitor and its ESR is presented to enhance the LDO regulator’s performance regarding the transient response and the noise rejection. Besides, a bandgap reference integrated with an error amplifier in a feedback loop is proposed to achieve the high precision of LDO.
2. LDO circuit design Fig. 1 shows the block diagram of the proposed LDO, which is composed of a reference voltage bandgap, an error amplifier, a source follower, a power MOS (MP), a voltage divider composed of R1 and R2, a variable resistance Rc. 2.1. Bandgap reference To generate a stable voltage against process and temperature variations, a bandgap reference circuit is adopted [3]. To enhance the capability against the supply voltage variation and the noise coupling from the substrate, a feedback loop with an error amplifier (EAB) is used as shown in Fig. 2. Assuming the gain of
C.-C. Wang et al. / Microelectronics Journal 42 (2011) 966–971
967
VIN 3.45 V ~ 5 V
Reference Voltage Bandgap ( 1.2 V )
EA OUT
Acts as a resistor Rc MA Cc
EAF
MP Source Follower
VOUT 3.3 V
1 F
R2
Mc
Quiescent current = VREF / R1
C1 VFB R1
Fig. 1. Architecture of the proposed low noise low dropout regulator.
VIN
VIN 3.45 V ~ 5 V
M12 M11
M13 VB4
M14
M9
M8
M10 MP
M6
M5
EAOUT
M7 VREF
R1
R2
Source Follower
R4
Fig. 3. The schematic of source follower.
I3 M3
EAB
M4 Vy
Vx
M1 I1
M2
VB3
VB4
R3
I2 Q1
VDD
Q2
VP
Q3
VN
VB3
VB2 EAOUT Fig. 2. The schematic of bandgap reference.
VB2 the EAB is very large, Vx and Vy are approximately equal, where Vy and Vx are the voltages at the inverting and noninverting input terminals of EAB, respectively. In Fig. 2, transistors M1 and M2 are both operating in saturation region. Taking channel-length modulation into consideration, the drain currents of M1 and M2, I1 and I2, can be written as I1 ¼
b1
I2 ¼
b2
2
2
2
ðVGS1 VTH1 Þ ð1 þ lVDS1 Þ
ð1Þ
ðVGS2 VTH2 Þ2 ð1 þ lVDS2 Þ
ð2Þ
where VGS1 and VGS2 are the gate to source voltage of M1 and M2, respectively, VTH1 and VTH2 are the threshold voltage of M1 and M2, respectively, l is the channel-length modulation coefficient,
VB1 VB1
Fig. 4. Error amplifier (EAF).
VDS1 and VDS2 are the drain to source voltage of M1 and M2, respectively, b1 and b2 represent the device parameters of M1 and M2, respectively. Given that I1 ¼ I2, VGS1 ¼ VGS2 , VDS1 ¼ VDS2 , we attain the following equation: VGS1 þVBE1 ¼ VGS2 þI1 R3 þVBE2
ð3Þ
968
C.-C. Wang et al. / Microelectronics Journal 42 (2011) 966–971
Then, the sensitivity of I3 versus temperature can be derived as
Eq. (3) can be rewritten as I2 ¼
VBE1 VBE2 þVGS1 VGS2 VBE1 VBE2 ¼ ¼ I3 R3 R3
ð4Þ
Since VBE ¼ VT lnðIc=IsÞ, and the size of Q2 and Q3 are both the n times of that of Q1, I3 can be expressed as I3 ¼
VT lnðnÞ R3
ð5Þ
@I3 K lnðnÞ ¼ @T q R3
ð6Þ
Since Is ¼ cT 4 þ m expðEq=KTÞ, where c and m are constants, the sensitivity of VBE versus temperature can be derived as follows: @VBE VBE ð4þ mÞVT Eq=q ¼ 1:5 ðmV=KÞ ¼ T @T
ð7Þ
where T¼300 K, VBE ¼ 0.75 V, in Eq. (7). In Fig. 2, VREF ¼ I3 R4þVBE3. Hence, the sensitivity of VREF versus temperature can be derived as Series Pass Element
@VREF R4 K VBE ð4 þ mÞVT Eq=q ¼ lnðnÞ þ @T R3 T T
S
G
Cgd
Vgsgp
Vgs
Rds
ð8Þ
Thus, the sensitivity of VREF can be eliminated by adjusting the value of R4 to generate a stable reference voltage.
D
2.2. Source follower Zo
As shown in Fig. 3, source follower is used to drive the large capacitive load and the small resistive load. The EAOUT is fed into the source follower to drive MP in the saturation region, where VB4 is a bias voltage connected to EAF, as shown in Fig. 4.
gp R2 RESR ga
Vin Rpar
Cpar
R1 Vref
Vo
Cb Co
Fig. 5. The small signal equivalent model of LDO.
Iout
2.3. Error amplifier Speed and gain always determine the performance of the LDO, and the speed of the LDO are mainly determined by the error amplifier, i.e., EAF, in Fig. 1. In order to reduce the the noise from VIN, EAF is implemented with a folded cascode structure, as
Fig. 6. Bandgap reference simulations given different process corners.
C.-C. Wang et al. / Microelectronics Journal 42 (2011) 966–971
969
Power Supply
Digital Oscilloscope Function Generator Current/ Meter VOUT LDO VIN Fig. 7. VREF is stable given different VIN from 3 to 5 V.
1350 µm VOUT
VOUT resister
VREF
GND (core) GND (core)
BandGap
Fig. 9. Measurement environment for the proposed LDO.
560 µm
PWM (W/L = 80,000/0.35 um)
GND
VIN VREF Fig. 8. Die photo of the proposed LDO.
shown in Fig. 4, where VB4 is a voltage bias coupled to the following source follower. Moreover, the cascode structure can enlarge the gain to provide a precise output voltage of LDO. 2.4. Frequency compensation In a negative feedback system, the stability is always a critical issue. To ensure the stability of the system, the phase margin of the open loop gain should be carefully designed. Traditionally, the small signal equivalent model of LDO is shown in Fig. 5 [17], where ga is the transconductance of error amplifier, Cpar and Rpar are the parasitic capacitor and resistor of error amplifier, respectively, gp is the transconductance of power MOS, Co is the load capacitor, RESR is the load effective series resistance, Cb is the bypass capacitor, and Iout is the load current. In most cases, Co is much larger than Cb , the output impedance, Z0 , can be simplified as Z0
Rds ð1 þSRESR Co Þ ½1 þ SðRds þRESR ÞCo ½1 þ SðRds JRESR ÞCb
ð9Þ
By Eq. (9), three poles and one zero could be attained. However, the phase margin is not high enough for a LDO and the output voltage may be easily oscillated. In other words, the phase margin of the traditional LDO is dominated by the load. Therefore, two frequency compensation methods are adopted to keep our LDO apart from load capacitor and its ESR, which are given in detail with following two sections. 2.4.1. Zero-tracking compensation The main pole frequency of the LDO will varied with its load. When the load current decreases and the equivalent output resistance increases, the frequency of main pole determined by the output capacitor and resistor will move forward left to reduce the LDO’s bandwidth. The circuit may then become unstable
when the bandwidth is reduced. Therefore, a mobile zero varying with load current is needed for LDO [12]. As shown in Fig. 1, the PMOS, MA , is biased in the triode region to act as a variable resistance. Its resistance is varied with the voltage at VOUT, which is the load voltage. Therefore, the zero is varied with the dominant pole to avoid the bandwidth reduction of the LDO. Meanwhile, the phase margin is also compensated. 2.4.2. Nested Miller compensation A small Miller capacitor, Mc, is coupled to the gate of MA on the feedback path, as shown in Fig. 1. Besides, the feedback voltage divider composed of R1 nd R2 is connected with C1 in parallel to generate a extra zero and pole. When the pole’s frequency equals (1þR2/R1) k times of the zero’s frequency, the noise of the system can be eliminated, where k is an experimental constant. According to our measurement in Section 3, the frequency of the pole is about 2.75 times of the zero’s frequency.
3. Simulation and measurement 3.1. Simulation of the bandgap reference The low noise LDO is designed and fabricated using a 0:35 mm standard CMOS process. The bandgap reference with a compensated error amplifier has been simulated given the temperature from 40 to 120 1C. As shown in Fig. 6, VREF has a variation of 71 mV from the mean value 1.204 V given different process corners. A stable VREF is obtained given different VIN from 3 to 5 V, as shown in Fig. 7. 3.2. Measurement results for the proposed LDO Fig. 8 shows the die photo of the proposed circuit. The measured environment is set up as Fig. 9, which is composed of an infinium Oscilloscope, a function generator, and a power supply. The input port of the proposed LDO, VIN, is connected to the power supply. Fig. 10 presents the smallest dropout of the LDO, 0.018 V, while the output current is 165 mA, the output load is 4:7 mF and 20 O in parallel. Figs. 11 and 12 show the transient response of LDO in the line regulation measurement. Fig. 12 is the detailed illustration of Fig. 11 with VIN switching from 3.45 to 5 V. It can be observed that VOUT is stable with a settling time
970
C.-C. Wang et al. / Microelectronics Journal 42 (2011) 966–971
Power supply
VIN
LDO
VOUT M0 R5
Function gererator
R6
CLK R5 = 1K Ohm R6 = 14 Ohm
Fig. 13. Measurement configuration for LDO load regulation.
Fig. 10. The smallest dropout with output current ¼ 165 mA.
Fig. 14. Measured load regulation of proposed LDO.
Fig. 11. Measured line regulation of proposed LDO.
Fig. 15. Measured settling time of LDO load regulation.
Fig. 12. Measured settling time of LDO line regulation.
less than 36:55 ms and a line regulation of 0.012%. When load regulation is measured, the test configuration with R5 ¼ 1 KO and R6 ¼ 14 O in Fig. 13 is used, where R5 and R6 are load resistances. Fig. 14 shows the measured transient response of LDO at
VIN¼3.3 V with the output current switching from 120 to 0 mA. Fig. 15 is the detailed illustration of Fig. 14, where CLK represents the transient when M0 is turned off. VOUT is stabilized with a settling time less than 42:54 ms and a load regulation of 0.005%. Figs. 16 and 17 shows the measured PSRR when VIN is at 30.08 and 3.08 kHz, respectively. The quiescent current, Iddq, is 50 mA with a load current ¼ 134.35 mA. The performance comparison of LDOs is shown in Table 1. Our design has the edge of minimal line regulation, load regulation and dropout voltage. 4. Conclusion In this paper, a bandgap reference with a compensated error amplifier for high precision LDO is disclosed. This design
C.-C. Wang et al. / Microelectronics Journal 42 (2011) 966–971
971
PSRR stays over 50 dB even if the operating frequency is up to 3.08 kHz.
Acknowledgments This investigation is partially supported by Ministry of Economic Affairs, Taiwan, under Grant 99-EC-17-A-01-S1-104 and under Grant 99-EC-17-A-19-S1-133. It is also partially supported by National Science Council, Taiwan, under Grant NSC99-2221E-110-082-MY3, NSC99-2923-E-110-002-MY2, NSC-99-2221E-110-081-MY3, NSC-99-2220-E-110-001, and EZ-10-09-44-98. The authors would like to express their deepest gratefulness to CIC (Chip Implementation Center) of NARL (National Applied Research Laboratories), Taiwan, for their thoughtful chip fabrication service.
References Fig. 16. Measured PSRR at VIN ¼ 30.08 kHz.
Fig. 17. Measured PSRR at VIN ¼ 3.08 kHz.
Table 1 Performance comparison of LDO.
Process (mm) Year Line regulation (%) Load regulation (%) Dropout (V) ESR zero required Iddq (mA) ILmax (mA)
[10]
[4]
[5]
[8]
Ours
0.18 2009 0.024 0.7 0.3 Yes 0.028 N/A
0.18 2008 N/A N/A N/A No 0.027 150
0.35 2007 N/A N/A 0.2 No 0.02 200
0.18 2009 0.231 0.101 0.541 Yes 0.035 150
0.35 2010 0.012 0.005 0.11 No 0.05 240
is independent of the off-chip capacitor and its ESR. The measurement results justify its high precision with line regulation¼ 0.012% and load regulation¼0.005%. Moreover, the
[1] R.J. Baker, H.W. Li, D.E. Boyce, CMOS circuit design, layout, and simulation, in: Proceedings of the 15th International Conference on Electronics, Circuits and Systems, August 2008, pp. 1047–1050. [2] K.N. Leung, P.K.T. Mok, A capacitor-free CMOS low-dropout regulator with damping-factor-control frequency compensation, IEEE Journal of Solid-State Circuits 38 (10) (2003) 1697–1702. [3] S. Hui, W. Xiaobo, Y. Xiaolang, A precise bandgap reference with high PSRR, in: Proceedings of the Seventh IEEE International Conference on Electron Devices and Solid-State Circuits, December 2005, pp. 267–270. [4] H.-C. Lin, H.-H. Wu, T.-Y. Chang, An active-frequency compensation scheme for CMOS low-dropout regulators with transient-response improvement, IEEE Transactions on Circuits and Systems II: Express Briefs 55 (9) (2008) 853–857. [5] M. Al-Shyoukh, H. Lee, R. Perez, A transient-enhanced low-quiescent current low-dropout regulator with buffer impedance attenuation, IEEE Journal of Solid-State Circuits 42 (8) (2007) 1732–1742. [6] W. Oh, B. Bakkaloglu, C. Wang, S.K. Hoon, A CMOS low noise, chopper stabilized low-dropout regulator with current-mode feedback error amplifier, IEEE Transactions on Circuits and Systems I: Regulator Papers 55 (10) (2008) 3006–3015. [7] M. Kayal, F. Vaucher, P. Deval, New error amplifier topology for low dropout voltage regulators using compound OTA-OPAMP, in: Proceedings of the 32nd European Solid-State Circuits Conference, September 2006, pp. 536–539. [8] S. Heng, C.-K. Pham, Improvement of LDO’s PSRR deteriorated by reducing power consumption: implementation and experimental results, in: Proceedings of the IEEE International Conference on IC Design and Technology, May 2009, pp. 11–15. [9] W.-J. Huang, S.-I. Liu, A sub-1 V low-dropout regulator with an on-chip voltage reference, in: Proceedings of the IEEE Asian Solid-state Circuits Conference, November 2008, pp. 165–168. [10] V. Majidzadeh, A. Schmid, Y. Leblebici, A fully on-chip LDO voltage regulator for remotely powered cortical implants, in: Proceedings of the IEEE ESSCIRC, September 2009, pp. 424–427. [11] T. Akin, K. Najafi, R.M. Bradley, A wireless implantable multichannel digital neural recording system for a micromachined sieve electrode, IEEE Journal of Solid-State Circuits 33 (1) (1998) 109–118. [12] K.C. Kwok, P.K.T. Mok, Pole-zero tracking frequency compensation for low dropout regulator, in: Proceedings of the IEEE International Symposium on Circuits and Systems, vol. 4, May 2002, pp. 735–738. [13] G.K. Balachandran, R.E. Barnett, A 110 nA voltage regulator system with dynamic bandwidth boosting for FRIO systems, IEEE Journal of Solid State Circuits 41 (9) (2006) 2019–2028. [14] V. Gupta, G.A. Rincon-Mora, A 5 mA 0:6 mm CMOS miller-compensated LDO regulator with 27 dB worst-case power-supply rejection using 60 pF of on-chip capacitance, in: Digest of Technical Papers IEEE International SolidState Circuits Conference, 2007, pp. 520–521. [15] R.K. Dokania, G.A. Rincon-Mora, Cancellation of load regulation in low dropout regulators, Electronics Letters (2002) 1300–1302. [16] C.-C. Wang, C.-C. Huang, T.-J. Lee, U.-F. Chio, A linear LDO regulator with modified NMCF frequency compensation independent of off-chip capictor and ESR, in: Digest of Technical Papers IEEE Asia Pacific Conference on Circuits and Systems, November 2006, pp. 880–883. [17] Technical review of low dropout voltage regulator operation and performance, Texas Instruments, Texas, 1999. [18] B. Razavi, Design of Analog CMOS Integrated Circuits, McGraw-Hill, 2001.