IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMSII: EXPRESS BRIEFS, VOL. 54, NO.

9, SEPTEMBER 2007

755

A High Slew-Rate PushPull Output Amplifier for

Low-Quiescent Current Low-Dropout Regulators
With Transient-Response Improvement
Tsz Yin Man, Student Member, IEEE, Philip K. T. Mok, Senior Member, IEEE, and
Mansun Chan, Senior Member, IEEE

AbstractA high slew-rate amplifier with pushpull output

driving capability is proposed to enable an ultra-low quiescent
1 A) low-dropout (LDO) regulator with imcurrent (
proved transient responses. The proposed amplifier eliminates the
and large slew-rate that is imposed
tradeoff between small
by the tail-current in conventional amplifier design. Pushpull
output stage is introduced to enhance the output driving ability.
Small dropout voltage ( DO ) with large-size pass transistor and
can thus be used to minimize power loss of LDO
ultra-low
regulator without transient-response degradation. The proposed
amplifier helps to improve stability of LDO regulators without
using any on-chip and off-chip compensation capacitors. This is
beneficial to chip-level power management requiring high-area
efficiency. An LDO regulator with the proposed amplifier has been
implemented in a 0.18- m standard CMOS process and occupies
0.09 mm2 . The LDO regulator can deliver 50-mA load current at
1-V input and 100-mV DO . It only consumes 1.2 A
and is
able to recover within 4 s even under the worst case scenario.
Index TermsHigh slew rate, low-dropout (LDO) regulator,
low-quiescent current, tail current.

I. INTRODUCTION
OW-DROPOUT (LDO) regulators are often used in
battery-powered mobile systems requiring small size
and clean supply voltage. To prolong the battery life, power
efficiency of LDO regulators can be improved by reducing the
). Fig. 1(a)
quiescent current ( ) and dropout voltage (
shows the regulator efficiency as a function of load current
(
) with different
and under a particular output voltage
) and
. It shows that the use of ultra-low
effec(
tively improves the regulator efficiency at small
, which
often happens when systems are operating in standby mode
to extend the battery life. From the efficiency expression in
also improves the regulator efficiency and
Fig. 1(a), small
this improvement is more significant in the small
case.
However, reduction of
and
unavoidably slows down
the transient responses of an LDO regulator. This can be explained with a generic LDO regulator shown in Fig. 1(b). The
tail-current of amplifier is equal to
of the LDO regulator.

Manuscript received January 15, 2007; revised April 3, 2007. This work was
supported by the Research Grant Council of Hong Kong SAR Government
under Project 617705. This paper was recommended by Associate Editor
T. C. Carusone.
The authors are with the Department of Electronic and Computer Engineering, The Hong Kong University of Science and Technology, Hong Kong.
(e-mail: eesm@ece.ust.hk; eemok@ece.ust.hk; mchan@ece.ust.hk).
Digital Object Identifier 10.1109/TCSII.2007.900347

Fig. 1. (a) Plot of regulator efficiency. (b) Schematic of LDO regulator with
conventional amplifier. (c) Plot of regulator dropout voltage.

Since this
sets the maximum current to charge and discharge
the gate capacitance (
) of the pass transistor, requirement
of ultra-low
inevitably reduces the slew-rate (
) at
the gate of the pass transistor. Transient responses are therefore
degraded.
reduction also degrades transient responses of an LDO
regulator. As shown in Fig. 1(c),
can be reduced by
enlarging the width of the pass transistor (
). To achieve
the same
at smaller input voltage (
),
needs to
be greatly enlarged to compensate for the limited gate-drive
voltage. As
is proportional to
, time required
to charge or discharge
is greatly increased when both
ultra-low
and small
at low
are required. Transient
responses of an LDO regulator are hence significantly degraded.
of existing LDO regulators [1][6] is normally ranged
from several tens of to few hundreds of microamperes. For applications required ultra-fast responses [7], the
is up to several thousands of microamperes. Efficiency of those LDO regulators at small
is degraded.
In this work, a high slew-rate amplifier with pushpull output
stage is thus proposed to enable an ultra-low
LDO regulator

1549-7747/$25.00 2007 IEEE

756

IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMSII: EXPRESS BRIEFS, VOL. 54, NO. 9, SEPTEMBER 2007

Fig. 2. Conceptual schematic of an LDO regulator with the proposed high slewrate amplifier with pushpull output configuration.

with improved transient responses. Concept of the proposed amplifier is discussed in Section II. Circuit design and implementation are shown in Section III. Experimental results and conclusions are given in Sections IV and V, respectively.
II. CONCEPT OF PROPOSED AMPLIFIER
Fig. 2 shows the conceptual schematic of an LDO regulator with the proposed amplifier, which is constructed by
two common-gate differential-input transconductance (
)
cells,
and
, and a current-summation circuit. Each
-cell is made by a constant current source ( ) and a pair of
matched transistors,
and
, in the form of current mirror.
The maximum output current of
-cell (
), and therefore the maximum amplifier-output current ( ), are no longer
limited by the constant-current source
as in the case of
conventional amplifier with a tail-current. As shown in Fig. 2,
the common-gate configuration enables
depending on
its input-voltage difference, which is the difference between
the reference voltage (
) and the
. Since all the transistors of the
-cell operate in saturation region
has a
quadratic dependence on its input-voltage difference according
to the square-law characteristic of MOS transistor [8]. Even if
an ultra-low
is used, the amplifier-output current ( ) and
also the amplifier slew-rate is no longer limited.
High slew-rate in both positive and negative direction is
achieved with the current-summation circuit and the cross-coupled connection scheme. As shown in Fig. 2, the inputs of the
-cells,
and
, are connected in a cross-coutwo
pled manner. The outputs of
-cells are connected by a
current-summation circuit. Regardless of whether the
becomes larger or smaller than the
during transient
responses, the large
from either one of the
-cells is
able to charge or discharge the gate capacitance
quickly.
As a result, both positive and negative direction slew-rates can
be successfully enhanced even if an ultra-low
is used.

-cell helps to improve

The small input-resistance of the
stability of LDO regulator without using any off-chip and
on-chip compensation capacitors. As shown in Fig. 2, the
source terminal of transistor
insider the
-cell,
,
is directly connected to the output of the LDO regulator. The
open-loop output resistance of LDO regulator is now modified
from its original value
to
. This
provides the regulator with a small open-loop output resistance
from several tens of milliamperes
under a wide range of
to several A. Even the output-parasitic capacitance (
)
due to the metal lines used to perform chip-level power distribution is as large as several hundreds of pF (e.g.,
), the
regulator-output pole
is still
located at high frequency and much far away from the low-frequency amplifier-output pole (
). Stability
of an LDO regulator with the proposed amplifier is greatly
improved without using any on-chip and off-chip compensation
capacitors. Area efficiency of such LDO regulator is highly
improved. This is particular suitable for chip-level power
management requiring multiple LDO regulators with high
chip-area efficiency. In case a large off-chip output capacitor is
used to improve transient responses under fast changing load
current, stability of proposed LDO regulator can be achieved
like conventional LDO regulators by the pole-zero cancellation
[3], [5] where the zero is generated from the output capacitor
and its equivalent-series resistor.
III. CIRCUIT DESIGN AND IMPLEMENTATION
Fig. 3 shows the schematic of the proposed LDO regulator,
which consists of a pMOS pass transistor (
), the proposed high slew-rate pushpull output amplifier and a reference buffer. As mentioned before, the
reduction, especially
under a very-low
(e.g., 1 V in our case), is very important to
the efficiency improvement and is usually achieved by enlarging
the pass transistor width
. Hence, to provide 50-mA load
current at 100-mV dropout under 1-V input, the aspect ratio of
(
), where
is the channel length of
, is chosen to be 5940 m/0.18 m in a 0.18- m standard CMOS process where the threshold voltage of
is
.
To ensure the proposed amplifier has fast-transient responses
and large voltage-gain for input-offset reduction, the channel
lengths of all transistors except the
are designed to be
two and half times of the minimum feature size of 0.18- m
process, which balances the tradeoff between the need for smallgate capacitance and large-drain resistance.
To make sure the proposed regulator can be tested under sev(e.g., 1.2 A in our case), gate-to-source voltage
eral- A
of
,
, and
, (e.g.,
) is generated by injecting
a well-defined DC current to the drain terminal of diode-connected transistor (not shown in Fig. 3) that has the same size as
and is closely integrated on the same chip with,
,
,
and
. Impact of process variation on the bias current (e.g.,
100 nA in our case) generated by those transistors is greatly minimized. It is important to note that conventional supply-insensitive biasing circuits [8] can be used to define the
and be integrated with the proposed LDO regulator without any problems.
The proposed amplifier is constructed by two common-gate
-cells,
and
, and a current-summation circuit. As

MAN et al.: HIGH SLEW-RATE PUSHPULL OUTPUT AMPLIFIER

757

Fig. 3. Schematic of an LDO regulator with the proposed high slew-rate pushpull output amplifier and a reference buffer.
TABLE I
DIMENSIONS OF TRANSISTORS

shown in Fig. 3, the

(
) consists of two pMOS transistors,
and
(
and
), and one nMOS transistor,
(
), acting as a constant-current source. Mismatch between the input transistors (e.g.,
and
) can be reduced
with methods commonly used in conventional amplifier design
such as placing these input transistors with common-centroid
method and designing them in large width. The current-summation circuit is made by three current mirrors,
and
,
and
, and
and
, respectively.
To enable the proposed amplifier with sink and source outputdriving ability, the three current mirrors are required to construct the pushpull output stage. As shown in Fig. 3, the curand
, redirects
rent mirror constructed with transistors,
the output current of
to the amplifier output. The large
is now able to discharge the
transient-output current of
large gate capacitance of
, from which sink ability is
achieved. Source ability is achieved using current mirrors made
by transistors,
and
, and
and
. The large transient-output current of
is redirected to the output of amplifier and is now able to charge the large gate capacitance of
. To further enhance this sink and source output-driving
ability, the size of transistors,
and
, is set to three times
of the size of transistors,
and
, respectively. Ratios of
different transistors shown in Fig. 3 are summarized in Table I.
Moreover, the pushpull output stage constructed with tranand
enables the proposed amplifier with large
sistors
output-voltage swing, which facilitates an LDO regulator using
only moderate size
to provide a wide range of load currents under small
. The amplifier-output voltage is able to go
of the regulator when compared
from nearly zero to almost
with the amplifiers using telescopic output stage. In other words,

can operate in cut-off region for small load current, or

operate in saturation region or even linear region to handle heavy
load current. Hence, very-large size
used to compensate
the limited amplifier-output swing is no longer required in the
proposed LDO regulator. The use of moderate size
also
reduces its gate capacitance, and with the proposed slew-rate enhanced amplifier, the transient responses of an LDO regulator is
further improved.
Since most of voltage reference circuits do not have
output-current-driving ability, a reference buffer is introduced
here to transfer the supply- and temperature-independent referto both the inverting input of
and the
ence voltage
non-inverting input of
. As shown in Fig. 3, the reference
buffer is a two-stage amplifier in unity-gain feedback. Since
will not be changed quickly with time, high slew-rate
is not required in the design of reference buffer. Therefore, a
conventional common-source differential-input amplifier with
small tail-current is chosen as the first gain stage, which is
constructed with transistors,
,
,
and
, and
acted as a tail-current. Transistors,
,
,
and
, make up the second gain stage.
Moreover, it is important to note that the output of reference buffer is connected to a low-resistance node, which is the
inside
. The pole at the
source terminal of transistor
output of reference buffer is therefore pushed to high frequency
regime, where is much far away from the low-frequency pole
at the output of first gain stage. Frequency compensation is
hence not required in the reference buffer when compared with
a stand-alone two-stage amplifier in unity-gain feedback.
IV. EXPERIMENTAL RESULTS
The proposed LDO regulator has been implemented in
TSMC 0.18- m CMOS process. The chip micrograph is shown
in Fig. 4. Since the proposed regulator does not require any
area-consuming on-chip compensation capacitors, the LDO
regulator, including all the pads, only occupies 0.09 mm only.
The total quiescent current
of the proposed LDO regulator
is 1.2 A only. Even when the input voltage
is changed from
1.8 V (the maximum voltage allowed in the 0.18- m standard
CMOS process) down to 1 V, the regulator is still able to provide
and 0.9-V output voltage
.
a 50-mA load current

758

IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMSII: EXPRESS BRIEFS, VOL. 54, NO. 9, SEPTEMBER 2007

Fig. 4. Chip micrograph of the proposed LDO regulator.

Fig. 6. Simulated line regulation with different levels of channel length mismatch between transistors M and M .

Fig. 5. Measured line regulation under 50 A and 50-mA load current.

Fig. 5 shows the measured line regulation with different loadcurrent levels (e.g., 50 A and 50 mA). The line regulation is
measured to be 4.75 mV/V at 50- A
and 3.625 mV/V
at 50 mA
. The load regulation at 1.0 V
is measured
to be 148
mA. From the measurement results, satisfactory
line and load regulation are successfully achieved using only
1.2 A .
In Fig. 5, it is found that the
is decreasing but not
increasing with an increasing
. With extensive simulations
and detailed analyses, the origin of this interesting phenomenon
is found to be caused by the process-variation induced drain-resistance mismatch between
and
. Since the channel
and
are in deep sub-micron regime,
lengths of
channel length modulation (CLM) and short-channel effect
such as drain-induced barrier lowering (DIBL) [9] make the
drain resistance of these transistors to be very sensitive to the
absolute value of channel length. Even though these transistors
are matched in the layout using common-centroid method,
any process variation induced channel-length mismatch still
inevitably introduces mismatch between the drain resistance of
transistors
and
.
Fig. 6 shows the simulated line regulation with different levels
of channel-length mismatch between
and
. When the
(
) is smaller than that of
(
),
channel length of
the drain resistance of
becomes smaller than that of
.
is increased, the drain current of
becomes larger
Once
than that of
. To match the drain currents of transistor
and
,
is forced to be decreased by the negative-feedback action of the amplifier. The negative slope of the measured
line regulation (Fig. 5) is resulted from the channel-length mismatch between
and
(e.g.,
).

Fig. 7. Measured load-transient responses of the proposed LDO regulator

without an off-chip output capacitor.

Load and line transient responses are measured to evaluate

the transient performance of LDO regulator with the proposed
amplifier. Fig. 7 shows the measured load-transient responses
varies from
without an off-chip output capacitor. The
50 A to 50 mA, and
is 1.2 V. Measurement result shows
that the proposed LDO regulator responds within
ns and
is fully recovered within
s at 1.2 A .
Fig. 8 shows the measured load-transient responses with a
100 pF off-chip output capacitor. It is used to model the regulator-output-parasitic capacitance from the metal lines used
to implement chip-level power distribution. The
again
varies from 50 A to 50 mA at 1.2 V
. Measurement result shows the proposed LDO regulator is fully recovered within
3 s at 1.2 A .
The measured line-transient responses without an off-chip
output capacitor and with a 100 pF off-chip output capacitor are
varies from 1.2 to
shown in Figs. 8 and 9, respectively. The
1.7 V and the load current is 50 mA. Measurement results show
that the output voltage of the LDO regulator can be recovered
within
s.
The measurement results shown in Figs. 710 confirm that
stability of LDO regulator with the proposed amplifier is successfully achieved without using any on-chip and off-chip compensation capacitors. Even a parasitic capacitor with several
hundreds of pF existed at the regulator output, stability is also
guaranteed.

MAN et al.: HIGH SLEW-RATE PUSHPULL OUTPUT AMPLIFIER

759

TABLE II
PERFORMANCE COMPARISON

Fig. 8. Measured load-transient responses of the proposed LDO regulator with

a 100 pF off-chip output capacitor.

Fig. 9. Measured line-transient responses of the proposed LDO regulator

without an off-chip output capacitor.

V. CONCLUSION
A high slew-rate amplifier with pushpull output stage is introduced to improve transient responses of an LDO regulator
to improve light-load efficiency and
which uses ultra-low
large-size pass transistor to reduce dropout loss. The proposed
amplifier is able to provide a pushpull transient-output current,
which is much larger than its quiescent current. Time required
to charge or discharge the large pass transistor gate capacitance
is greatly reduced. Both load and line transient responses of an
LDO regulator are thus improved.
Experimental results show that the proposed LDO regulator
with only 1.2 A
is able to recover within
s under
the worst case scenario. It also proves that the proposed unique
structure enables a stable LDO regulator without using any
on-chip and off-chip compensation capacitors. Chip-area efficiency is thus greatly improved and multiple proposed LDO
regulators can be applied in chip-level power management.
ACKNOWLEDGMENT
The authors would like to thank A. Ngs technical support.
REFERENCES

Fig. 10. Measured line-transient responses of the proposed LDO regulator with
a 100 pF off-chip output capacitor.

Table II provides comparison between performance of the

proposed LDO regulator and other published designs that are
targeted for chip-level power management. A figure of merit
(
) used in [7] is adopted here to
evaluate the effect of
to the load transient response time ( )
in different designs. A lower FOM implies a better slewing performance. The proposed regulator has very comparable FOM
(0.067 ns) to the design in [7]. However, the proposed regulator
only requires 1.2 A compared with 6 mA in [7]. It makes the
proposed regulator with superior current efficiency at light-load
condition. This feature is very important and attractive to any
battery-powered mobile systems.

[1] H. Lee, P. K. T. Mok, and K. N. Leung, Design of low-power

analog drivers based on slew-rate enhancement circuits for CMOS
low-dropout regulators, IEEE Trans. Circuits Syst. II, Exp. Briefs,
vol. 52, no. 9, pp. 563567, Sep. 2005.
[2] G. A. Rincon-More and P. E. Allen, Optimized frequency-shaping circuit topologies for LDOs, IEEE Trans. Circuits Syst. II, Analog Digit.
Signal Process., vol. 45, no. 6, pp. 703708, Jun. 1998.
[3] C. K. Chava and J. Silva-Martinez, A frequency compensation scheme
for LDO voltage regulators, IEEE Trans. Circuits Syst. I, Reg. Papers,
vol. 51, no. 6, pp. 10411050, Jun. 2004.
[4] S. K. Lau, P. K. T. Mok, and K. N. Leung, A low-dropout regulator
for SoC with Q-reduction, IEEE J. Solid-State Circuits, vol. 42, no. 4,
pp. 658664, Mar. 2007.
[5] K. N. Leung and P. K. T. Mok, A capacitor-free CMOS low-dropout
regulator with damping-factor-control frequency compensation, IEEE
J. Solid-State Circuits, vol. 38, no. 10, pp. 16911702, Oct. 2003.
[6] G. A. Rincon-Mora and P. E. Allen, A low-voltage, low quiescent
current, low drop-out regulator, IEEE J. Solid-State Circuits, vol. 33,
no. 1, pp. 3644, Jan. 1998.
[7] P. Hazucha, T. Karnik, B. A. Bloechel, C. Parsons, D. Finan, and S.
Borkar, Area-efficient linear regulator with ultra-fast load regulation,
IEEE J. Solid-State Circuits, vol. 40, no. 4, pp. 933940, Apr. 2005.
[8] P. R. Gray, P. H. Hurst, S. H. Lewis, and R. G. Meyer, Analysis and
Design of Analog Integrated Circuits, 4th ed. New York: Wiley, 2001,
ch. 1 and 4.
[9] S. M. Sze, Physics of Semiconductor Devices, 2nd ed. New York:
Wiley, 1981, ch. 8.