A Low-Offset Dynamic Comparator with Input Offset-Cancellation
Ruihan Pei
1
, Jia Liu
1
, Xian Tang
2
, Fule Li
1
* and Zhihua Wang
1
1
Institute of Microelectronics, Tsinghua University, Beijing 100084, China
2
Graduate School at Shenzhen, Tsinghua University, Shenzhen, China
*Email: lifule@mail.tsinghua.edu.cn
Abstract
This paper presents a low-offset dynamic comparator
using an input offset-cancellation technique. The offset of
a dynamic comparator is mainly determined by the
dynamic preamplifier. The proposed technique achieves
input offset-cancellation under the assistance of the
dynamic preamplifier and input-series capacitors, without
quiescent current. The offset resulting from both threshold
voltage mismatch and sizing factor mismatch can be
cancelled. A prototype comparator is implemented by 65-
nm CMOS technology with an area of 127μm
2
. The
simulation results show it achieves 1.16-mV offset at 1-
sigma(σ), while its counterpart without offset-
cancellation achieves 8.91-mV offset at 1-sigma(σ). After
post-layout simulation, the offset at 1-sigma(σ) only
slightly increases to 1.56-mV. The power dissipation is
135-μW from a 1.2V supply, at 500MHz operating.
1. Introduction
With the data processing speed of modern electronic
system rising, high-performance analog-to-digital
converters (ADC) have become an inevitable trend. As an
important component in ADCs, the comparator and its
performance optimization are very vital.
Because of its low power consumption, high speed,
high input impedance and good compatibility with deep
sub-micron processes, the dynamic comparator with a
dynamic preamplifier followed by a regenerative latch is
very attractive in high speed and resolution applications.
It can amplify a small input voltage difference to a large
voltage to overcome the backend latch’s offset and
prevents the kickback noise [1]. Fig. 1 redraws one
conventional double–tail latch-type dynamic comparator
with a preamplifier [2]. The preamplifier’s mechanism is
that when clock ϕ is low, the output nodes are precharged
to power supply. Then ϕ goes high, and the output are
discharged by the transistors M1, M2 and M3. The
following latch is used to shape and detect the different
discharging speed of the two nodes owning to different
input voltage. However, the dynamic comparator suffers
from large input offset due to the transistor’s mismatch,
especially when CMOS processes scale down. Offset
cancellation or calibration techniques are highly required.
ϕ
9
LQ
9
LQ
9
RXW
9
RXW
9
1
9
3
9
''
9
''
ϕ
ϕ
Figure 1. Conventional latched dynamic comparator with
a preamplifier [2]
The method reported in [3] does not need static power
dissipation and refresh process, but requires configurable
capacitor arrays, memories and logic controllers. This
results in lower speed and larger area for a high
cancellation resolution. The current cancellation
technique introduced in [4] exploits additional one pair of
compensation input transistors and a charge pump. Its
weakness lies in a long calibrating time due to the
successive approximation operation and a large area for
the auxiliary circuits. A dynamic offset cancellation
technique, by sampling the mismatches at the drain
terminal of the input transistors, is proposed in [5]. A large
bridging capacitor is required to store the mismatches and
cancel the offset finally. This capacitor increases the
differential load, thus reducing the gain of the dynamic
preamplifier. Another dynamic offset cancellation
technique, by storing offset at the source terminal of the
input transistors, is reported in [6]. In this method,
however only the threshold mismatch is suppressed and
the bias voltage
had better to be set low to get a low
offset voltage, leading to a long comparison time.
In this paper, by storing offset at the gate terminal of
the input transistors, a low-offset and small-area dynamic
preamplifier with input offset-cancellation technique, is
proposed. The offset resulting from both threshold voltage
978-1-5090-6625-4/17/$31.00 ©2017 IEEE
132