A High-Performance Charge Pump with Improved Static and
Dynamic Matching Characteristic
Haibin Shao, Ke Lin, Bo Wang*, Chen Chen, Fang Gao, Feng Huang, Xin’an Wang
The Key Laboratory of Integrated Microsystems, Peking University Shenzhen Graduate School, Shenzhen, China
* Email: wangbo@ pkusz.edu.cn
Abstract
A high-performance Charge Pump (CP) is introduced in
this paper with both perfect static and dynamic matching
properties. Pseudo-cascode structure and bulk-biasing
technique are adopted to obtain a high equivalent output
impedance over a wide voltage range as well as a fast
transient behavior. Besides, by analyzing turn-on and
turn-off mechanisms, a MOS capacitor is inserted to
weaken the coupling from current switch to reduce the
conduction time and an extra discharging path is added
to accelerate the turn-off speed. A prototype is
implemented in SMIC 130nm CMOS process. The
simulation results show that the maximum static
mismatch is only 0.37% and the largest ripple on VCO
control voltage is 1.97mV over process, temperature and
supply voltage variable. The amount of reference spur is
-73.1dBc in the PLL output signal while the
conventional one’s level is -52dBc.
1. Introduction
A charge pump sinks or sources current for a limited
period of time. The charge pump combined with the PFD
acts like an integrator and provides an open-loop pole to
form a type-II PLL system which can achieve zero static
phase error. But non-ideal effects such as leakage current,
sink & source current mismatch, dynamic mismatch and
etc. will lead to phase offset and reference spur. In this
work, we consider all these issues and then give out the
equations of phase offset and spurs which are listed in
Section 2. Some techniques are adopted like
pseudo-cascode structure, bulk biasing and etc. to give
an improved static and dynamic matching characteristic.
The proposed charge pump is demonstrated in Section 3,
and its post-simulation results are described in Section 4.
2. Non-Ideal effects of Charge Pump
For practical PFD/CP circuit, several imperfections
always exit and it will lead to static phase error and high
ripple on the control voltage even when the loop is
locked. The ripple will be modulated to the PLL output
and cause the reference spur. In frequency synthesis, the
spur caused by CP is dominant source.
The leakage current, sink & source current mismatch and
switch delay difference are the major considerations.
Other than these static behaviors above, the transient
dynamic mismatch will also degrade the spur feature
which is always overlooked.
By taking all these into account, the phase error and
reference spur can be approximated by following
equations [1, 2]:
2
leakage mismatch switch
leakage
on d
CP ref CP ref CP ref
I
ti t q
ITITIT
H
II I I
S
§·
''' '
¨¸
©¹
(1)
2
22
1
2(2)
1
20log
2
(2 ) (2 )
20log
leakage
on
CP ref CP
BW
spur
ref
on d on
ref ref ref CP ref
ref
p
I
ti
ITI
f
PN
f
tt t q
TT TIT
f
f
SS
SS
½
ªº
''
°°
«»
°°
«»
®¾
«»
'' ' '
°°
«»
°°
«»
¬¼
¯¿
(2)
In the equation,
CP
I
is the CP current,
ref
T
and
ref
f
is the
period and frequency of reference clock,
BW
f
is the loop
band width of PLL,
1q
f
is the pole of the low-pass filter
and N is the divider ratio.
leakage
i
,
on
t'
,
i'
,
d
t'
refer to
leakage current, turn-on time of PFD which is used to
avoid the dead-zone, current mismatch and switch delay
difference respectively.
q'
is the dynamic unequal
charge and can be calculated by integration.
In the CPPLL application, the phase error due to the
undesirable effects is usually negligible but the reference
spur is possibly substantial. From equation (2), some
methods to decrease spur can be proposed.
a) Halve the divider ratio and the loop band width can
both lead to 3dB improvement of reference spur;
b) A minimum turn-on time is important to reduce
in-band noise contribution for it appears in three
formulas. But it’s limited by the dead-zone time;
c) Reduce the mismatch and the leakage current, so a
large CP current is always a nice choose as long as
the power match the requirements;
d) Reduce the switch delay mismatch which can be
implemented by adding extra delay cells to get a
better feature;
e) Reduce the dynamic charge mismatch. The current
on and off time, the clock feed-through, the charge
sharing etc. should be considered carefully.
Above all the approaches, some belong to system level
and the others are circuit level. The working frequency
band, divider ratio and loop band width are decided by
the over-all PLL system and are listed in Table 1. This
article will focus on the circuit level to optimize spur
property.
978-1-4799-8485-5/15/$31.00 ©2015 IEEE