Professional Documents
Culture Documents
Abstract—A charge pump circuit to minimize current (M14, M15). This can be solved using long transistors in the
mismatch and current variation over a wide voltage compliance output stage to increase the output impedance. However, long
range is proposed. A feedback loop is used to cancel both transistors can’t completely eliminate the effect and large
deterministic and random mismatches between charging and parasitic capacitance limits the speed and operation. A current-
discharging current to minimize PLL reference spurs and static compensation circuit can be used to extend the flat range of
phase offset. A current compensation circuit is used to minimize output voltage significantly to minimize the current variation
current variation to avoid bandwidth variation and loop [5,6]. Gain-boosted charge pumps use regulated cascodes to
instability. The circuit can operate at low supply voltage. The increase the output resistance [7] for minimizing current
power overhead due to added circuitry is 3.36% making it
variation keeping the charging and discharging currents
suitable for low-power and low-voltage applications. The
proposed current-steering charge pump circuit achieves
approximately constant with the control voltage. However, this
mismatch lower than 0.44% over the output voltage range from lowers the voltage compliance range and a higher supply
0.06 V to 0.85 V. Also the current variation is reduced to less than voltage is needed for proper operation. This solution also
1.19% when the output voltage varies from 0.06 V to 0.85 V in increases the power consumption since the op amps used
the 65 nm CMOS process with 1 V supply. should have very large GBW.
I. INTRODUCTION
Conventional charge pump (CP) circuit non-idealities, such
as charge injection, charge sharing, clock feedthrough, and
charging and discharging current variation, greatly degrade the
performance in phase locked loops (PLLs). These non-
idealities worsen PLL stability, phase noise and increase
reference spurs. Also, random and systematic mismatches
between charging and discharging currents develop a phase Fig. 1. Conventional Current Steering CP
offset in the PLL resulting in increasing the PLL reference
spurs, phase noise, and jitter [1]. This paper presents a new technique for current-mismatch
cancellation and current-variation reduction in CSCP over a
To overcome charge sharing, charge injection and clock wide voltage range. The paper is organized as follows: Section
feedthrough, current steering charge pump (CSCP) shown in II presents the proposed circuit with mismatch-cancellation and
Fig. 1 can be used [2]. The UP and DN switches are far from current-compensation techniques. Detailed design is presented
the output node. Also, the current sources (M2, M7) are always in Section III. Section IV reports the simulation results. Finally,
on, and this current is steered to the output or to other branches Section V draws the conclusions.
depending on UP and DN signals. CSCP eliminates current
glitches and enhances switching speed among other topologies II. PROPOSED CHARGE-PUMP CIRCUIT
reported in literature [3,4] since we do not need more time for
activating the current sources. This in turns enhances PLL The proposed CP is shown in Fig. 2. It utilizes a mismatch-
phase noise performance and reduces reference spurs. cancellation circuit to reduce current mismatch and a current-
Moreover, it is compatible with CML based PFD with compensation circuit to compensate for current variations. The
differential operation for inherent common mode rejection, current-compensation circuit senses the output voltage and
better noise immunity, and reduce sensitivity to power supply controls part of the charging current, so that the total charging
noise. current Iup is nearly constant over a wide output voltage range.
On the other hand, the mismatch-cancellation circuit senses any
However, the conventional CSCP suffers from current mismatch between charging and discharging current. Then
variation and current mismatch between the charging and according to this mismatch, the biasing of Idn is automatically
discharging currents; Iup and Idn respectively. This is because of adjusted so that discharging current Idn follows any change in
the channel-length modulation effect of the output transistors the charging current Iup. At steady state, there is a very low
1667
−gm gm −gm MHz which is higher than 10 times the PLL BW (600 kHz).
Aβ = A ∗ ∗ ∗ ∗ (−gm (r //r ))
gm gm gm Also, the DC open loop gain varies from 57.6 dB to 84.12 dB
which is high enough to reduce the current mismatch
Aβ (@DC) ≅ −A ∗ gm ∗ (r //r ) (2) significantly.
100
∗ ∗( )
( )
BW = ≅ (3)
∗ ∗
50
Gain(dB)
The opamp A0 plays an important role in the current
matching. It should have high enough gain to achieve very low 0
gain error and hence perfect matching between charging and
discharging currents. Also, it should be designed to support
rail-to-rail input common mode voltage. Moreover, it is -50
Phase(degree)
while consuming only 4.3 μW.
0
The current-compensation circuit should account for ICP
variation with Vout. Without current compensation, ICP -100
decreases when Vout increases according to (4)
I =I α 1+V α (1 + (V −V )) (4) -200 2
10 10
4 6
10
8
10
Freq(Hz)
where is the channel-length-modulation parameter. If ICP Fig. 5. Loop Stability Analysis
changes from (I + ∆ ) to (I − ∆ ) in the voltage
compliance range, then Icomp, I , and I ranges can be
expressed in (5) Fig. 6(a) shows the charging current Iup and discharging
current Idn variation with the output control voltage Vout of the
∆1 ∆2 conventional CSCP. Fig. 6(b) shows current variation in
− ≤(I =I −I )≤
M M proposed CSCP when the current compensation circuit is not
used. As shown, Iup decreases when Vout increases and Idn
∆2 follows Iup due to the negative feedback loop used. Fig. 6(c)
0≤I ≤ &I = 0, when 0.5 V ≤ V ≤V
M shows the current variation of the proposed CSCP when the
∆ current compensation technique is used under nominal
0≤I ≤ &I = 0, when V ≤V ≤ 0.5 V (5) conditions. Fig. 6(d) shows current variation under process and
temperature variations. As shown, the current variation has
where M is the mirroring ratio and equals to . been reduced significantly. Iup and Idn variation is from 150.78
μΑ to 148.22 μΑ, which is less than 1.19% when Vout varies
The proposed technique can be used for low-voltage low- from 0.06 V to 0.85 V at nominal conditions, whereas the
power applications. It can operate at supply voltage down to variation is less than 7% across process and temperature
0.8 volts since there are no more than three transistors in the corners. Moreover, the current mismatch ratio of the proposed
stack. Also, the current flowing in the added branches can be CSCP varies from 0.435% to –0.023% while the percentage
reduced to a very small value such that Iup is a multiplier from mismatch in the conventional CSCP varies from 46.55% to
I1 and the same for I2 and Idn. However, if the supply voltage is –16.2%, when Vout varies from 0.06 V to 0.85 V.
reduced to be lower than 0.8 V, the folded-cascode opamp
should be replaced with another opamp that can work at lower A conventional CMOS PFD circuit [1] followed by a
supplies. CMOS to CML conversion stage were designed to derive the
CSCP (a conventional CML PFD [2] can be used instead, but
IV. SIMULATION RESULTS this consumes more power). Noise analysis has been done to
The proposed charge pump circuit was implemented in a the PFD/CP circuits (PSS & PNOISE) where the input signals
65-nm CMOS technology to be used in a PLL of 600 kHz to PFD are in phase (PLL Locking at steady state), and the
bandwidth. Stability analysis across temperature, process, and output current noise versus frequency at 150 μA is shown in
supply corners has been done to the negative feedback loop to Fig. 7. The proposed technique for CSCP increased the current
ensure that loop is stable and the proposed technique works noise by less than 0.5 dB. Monte Carlo simulation has been
properly under different conditions. Fig. 5 shows the open loop done to estimate the effect of random mismatch on the
gain and phase versus frequency under process variations when conventional and proposed CSCP. As shown in Fig. 8, the
the supply varies from 0.9 V to 1.1 V and temperature varies mean of current mismatch (Iup –Idn) has been reduced from 16.8
from 0o to 85o. The phase margin of the loop varies from 45.83o uA in the conventional CSCP to -46 nA in the proposed CSCP,
to 57.91o, whereas, the loop BW varies from 6.8 MHz to 37.2 also the standard deviation is reduced from 1.9 uA to 0.9 uA.
1668
The power consumption of the proposed CSCP is 0.332
mW at 1 V supply including the 150 μΑ current source. The
power overhead due to added circuitry for cancelling mismatch
and reducing current variation is 3.36%. The proposed
technique is one of the best solutions for mismatch cancellation
and current variation reduction without affecting the power,
noise, or PLL settling and stability. The layout of the proposed
CSCP is shown in Fig. 9, whereas the active area is 82 * 53
μm2. Table I summarizes the performance of the proposed
CSCP compared to other charge pump circuits reported in Fig. 9. Layout of the proposed CSCP
literature.
TABLE I. Performance Summary
REF [4] [6] [7] This Work
150 150
Tech. (nm) 65 130 180 65
Iup,Idn(uA)
Iup,Idn(uA)
100
Supply (Volt) 1 1.2 1.8 1
100
ICP (μA) 150 100 100 150
50 50 Iup
Compliance range 0.1-0.85V 0.2 - 1 V 0.4 – 1.4 V 0.1 – 0.85 V
Iup
Idn Idn I-Mismatch <0.03% 3.2 % NA < 0.03%
0 0 I-Variation <1.2 % 1.7 % <1 % < 1.2 %
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Vcontrol(volt) Vcontrol(volt)
-210
Proposed REFERENCES
-220
[1] Behzad Razavi. RF microelectronics. Vol. 1. New Jersey: Prentice Hall,
1998.
-230
[2] John Rogers, Calvin Plett, and Foster Dai. Integrated circuit design for
high-speed frequency synthesis. Boston, London: Artech House, 2006.
-240
[3] Jae Shin Lee, Woo Kang Jin, Done Myung Choi. "A wide range PLL for
-250 3 64X speed CD-ROMs & 10X speed DVD-ROMs." Consumer
4 5 6 7 8
10 10 10 10 10 10 Electronics, IEEE Transactions on 46.3 (2000): 487-493.
Freq(Hz) [4] Peng Liu, P. Sun, J. Jung and D. Heo. "PLL charge pump with adaptive
Fig. 7. Output current noise of conventional and proposed CSCP body-bias compensation for minimum current variation." Electronics
letters 48.1 (2012): 16-18.
60 [5] Sheng Chen, Zhiqun Li, and Qin Li. "An improved high speed charge
mu = - 46 nA Conventional mu = 16.8 uA pump in 90 nm CMOS technology." Communication Technology
sd = 0.89 uA Proposed sd = 1.86 uA (ICCT), 2011 IEEE 13th International Conference on. IEEE, 2011.
40 [6] M-S. Hwang, Jung-Ho Kim, and Deog-Kyoon Jeong. "Reduction of
pump current mismatch in charge-pump PLL." Electronics letters 45.3
(2009): 135-136.
20 [7] Jae Hyung Noh, and Hang Geun Jeong. "Charge pump with a regulated
cascode circuit for reducing current mismatch in PLLs." International
journal of electrical and computer engineering 3.9 (2008): 576-578
[8] Wakayama, Myles H. "Low offset and low glitch energy charge pump
0 for PLL-based timing recovery systems." U.S. Patent No. 6,897,733. 24
-5 0 5 10 15 20
I mismatch(uA) May 2005.
Fig. 8. Monte Carlo Simulation Results
1669