A Unity-Gain Fully-Differential Sample-and-Hold Amplifier in CMOS 0.18µm Technology IJARAS

©International society of academic and industrial research
International Journal of Academic Research in Applied Science
IJARAS 1(1): 1-9, 2012
A Unity-Gain Fully-Differential Sample-and-Hold Amplifier in
CMOS 0.18µm Technology
Win Toe, Samruay Thein
Assumption University, Thailand
[email protected]
Sample-and-hold circuits are the essential front-end part of an analog-to-digital converter. In
this work a fully-differential sample-and-hold circuit based on a two stage differential OpAmp, with sampling frequency of 20MHz is presented. The amplifier has a unity-gain
bandwidth of 370MHz and phase-margin of 75°, which meets the requirements for fast
settling of the output. Bootstrapped switches are used in order to reduce nonlinearity of
conventional transmission gate switches and charge injection effect. A CMOS temperature
and voltage-independent current reference is used with temperature and voltage coefficients
of 81ppm and 4393ppm respectively. Average power consumption of the system is 2.2mw.
Sample and Hold, operational amplifier, Bootstrapped switches, Analog IC design
Int. J. Acad. Res. Appl. Sci., 1(1): 1-9, 2012
1. Introduction
Sample-and-hold (S/H) is an important analog building block with many applications,
including analog-to-digital converters (ADCs) and switched-capacitor filters. The function of
the S/H circuit is to sample an analog input signal and hold this value over a certain length of
time for subsequent processing. S/H is used as front-end stage of an ADC. To obtain an
analog-to-digital converter with Nbit resolution, the sampling error must be less than 0.5
LSB. A part of this error, including finite-gain error and settling time error, is due to OpAmp's non-ideal properties. Gain error can simply be obtained of,
Gain error 
In this work, a unity-gain fully-differential S/H amplifier, with sampling frequency of
20MHz, and output voltage-swing of 2Vp-p,diff, for the front end stage of a 10Bit, 50MS/s
ADC is presented. The design is in standard TSMC 0.18µm CMOS process and its operation
in all corners of process; in temperature range of 0 ~ 85°C and input voltage of 1.8 V ±10% is
guaranteed. Maximum allowed error to obtain given specifications is,
0.5LSB 
1 V full Scale
 0.98mV
2 2n
It is essential that the response settles at the sampling time. For sampling frequency of
20MHz, settling time must be less than 24nsec. The deviation of output of its final value at
the end of sampling time is known as settling error. To obtain a settling error of less than
0.1%, the output takes more than 7τ, where τ is time constant of the output.
2. Op-Amp Design
The needed Op-Amp for the S/H application is designed in two-stage implementation. The
first stage is a folded cascade with N-type inputs, as shown in Figure 1.
Figure 1 First stage: A folded cascode with CMFB circuit
MOSFETs M15-M22 operates as the common-mode-feedback (CMFB) network of the
circuit and set the common-mode voltage of the first stage at VREF=0.9V. M15 and M16
operate in deep-triode region. Vb is determined that ID17follows ID18and VREF,
Int. J. Acad. Res. Appl. Sci., 1(1): 1-9, 2012
W 
W  W  W 
W 
        ,    
 L 19  L 15  L 16  L 17  L 18
Therefore ID17= ID18, if and only if Vo,cm= VREF. M21 and M22 are added to set
VDS17=VDS18to eliminate finite-error due to channel-length modulation. Diode-connected
M20 provides bias voltage of M5 and M6 and its size is chosen to equalize Ip and ID17; It is
essential for avoiding slewing. To increase the gain of this stage, length of M5/M6 and
M9/M10 are chosen twice bigger than Lmin, This will increase output impedance of the stage,
R out ,stage 1  ( rO 3  rO 5  g m 3 rO 3 rO 5 ) || (rO 7  rO 9  g m 7 rO 7 rO 9 )
To increase the unity gain bandwidth and hence the speed of the amplifier, transconductance
of the amplifier should be large enough. Device sizes to meet the goals explained above are
shown in tabel1. The second stage is a fully differential amplifier with active load. This
architecture is chosen for its high output swing. Because the input common-mode level of
this stage is the output common-mode level of the first stage which is fixed, the second stage
does not need tail current biasing, therefore fewer voltage headroom consumption. This stage,
with a simple structure of CMFB is shown in Fig. 2. In fact, M13/M14, that operate as
current references, are biased through output common-mode voltage. Output common-mode
voltage is in a level that M9 and M12 balance the current of M13 and M14. The size of M11
specifies the current of this stage to obtain desired gain. Therefore by changing the size of
PMOS, the output voltage could be set in proper point, here 0.9V.
Table 1 Device Sizes of the First Stage of the Designed Op-Amp.
Figure 2 Second stage: A simple fully-differential amplifier with CMFB Network
Int. J. Acad. Res. Appl. Sci., 1(1): 1-9, 2012
2.1. Compensation
Compensator capacitor (Cc=0.2 pF) is placed between source of cascode MOSFETs (M7,
M8) and output node. Using the model of Fig. 3, One zero exists in frequency
of gm7Reqgm11/CC, This value is much greater than gm11/CC. If other capacitances be
neglected, dominant pole is approximately in frequency, 1/RLReqgm11CC as if Cc were
connected to the gate of M11, rather than source of M7. Also, the first non-dominant pole is
given by, gm7Reqgm11/CL. In Fig. 4, if Iss > Ip, then during slewing, M3 turns off and Vx falls to
a low level such that M1 and the tail current source enters the triode region. Thus for the
circuit to return to equilibrium after M2 turns on Vx must experience a large swing, slowing
down the settling. To alleviate this issue, two clamp transistors can be added as shown in Fig.
5. The Idea is that the difference with Iss and Ip now flows through Diode-connected
transistors, requiring only enough drops in Vx or VY to turn on one of these transistors. These
two transistors are off in normal status operation.
Figure 3 Equivalent circuit with compensator capacitor
Figure 4 Slewing situation
Int. J. Acad. Res. Appl. Sci., 1(1): 1-9, 2012
Figure 5 Slewing corrector
2.2. Op-Amp Simulation Results
In this section, H-SPICE simulation results of the designed Op-Amp are presented.
Simulation is done in 8 corners of CMOS process. At TT, a DC Differential Gain of 8200
(78.3dB), 370 MHZ unity gain bandwidth and 74° Phase margin obtained. With this value of
UGBW, the closed loop time constant (in unityfeedback) is of 2.7 nsec, and during 24 nsec,
the output passes 8.88τ. Therefore the response is fast and settling time error is so small. HSPICE Simulation results are illustrated in Fig. 6.
Notice, this frequency response is calculated where the load capacitance, CL=2pF, is brought
in to account. Table 2 shows the simulation results of the Op-Amp in 8 corners of process.
The operation of the amplifier is guaranteed in all corners.
Figure 6 Simulation results for amplitude and bode plots of the designed Op-Amp
Table 2 Gain, unity-gain bandwidth and phase margin of the Op-Amp in 8 corners of CMOS process.
Int. J. Acad. Res. Appl. Sci., 1(1): 1-9, 2012
3. Sample-And-Hold Circuit
After design of op-amp with desired properties, we are prepared to implement the sample and
hold circuit. A greatly used unity-gain differential structure is shown in Figure 7. Operation
of differential architecture is similar to single ended, but the input-independent nature of the
charge injected by the reset switch allows complete cancelation by differential operation.
In the amplification mode, this circuit operates as a unity-gain Buffer, and produces an output
voltage approximately equal to voltage stored across the capacitor. Generally, the input
capacitance of Op-Amp is assumed finite and calculates the output voltage when the circuit is
transferred from sampling mode to amplification mode. Assuming finite gain for the OpAmp, Vx ≠0 in amplification mode, a charge equal to CinVx is injected to Cin. The
conservation of charge at X requires that CinVx Come from CH, raising the charge on CH to CinVx+ CHVO. It follows that the voltage across CH equals CinVx+ CHVO/CH. We therefore
V out  (C HV O  C inV X ) / C H V X
VX 
V out
AV 1
1  C in
V out V O 1 
 1 
 AV 1  C H
According to the equations above, Cin must be minimized to reduce error.
Figure 7 unity-gain differential sample-and-hold
Int. J. Acad. Res. Appl. Sci., 1(1): 1-9, 2012
The switches S1 and S2 must work in same phase, but S2 should turn off a little sooner than
S1. When S2 switches off, the input node of Op-Amp, X, get high-impedance, therefore
injected charge due to S1 does not affect output voltage. In the on-state cycle of S1 and S2,
output voltage will be sampled across CH .Switch S3 operates in a phase against that of S1
and S2. In this phase of operation, the sampled voltage on CH, will be held in the output.
Switch Seq is used to solve the problem of different charge injection of Switches S2 and S2’
which cause nonlinearity. Seq switches off a little later than S2’, a little sooner than S1 and
S1’, to guarantee equal level on nodes X and Y.
3.2. Simulation Results
In this step, differential sinusoidal inputs with 1Vp-p and DC level equals to 1V and frequency
of 1MHz and 8MHz are used for simulation. The input/output waveforms of the 1MHz input
are illustrated in Fig 8. In table 3, sampling errors of some samples, calculated in a period of
input signal are shown. The average of sampling errors is approximately 0.58mv that is
smaller than maximum allowed error (0.98mv). In Fig. 9, output waveform, where the phase
margin is 100°, is shown. The circuit with such phase margin has little ringing in settling.
Average sampling error is 1.7mv, where the input signal frequency is 8MHz.
Figure 8 S/H Input/Output waveforms, Cc=0.2pf and phase-margin is 75°.
Int. J. Acad. Res. Appl. Sci., 1(1): 1-9, 2012
Table 3 Sampling errors of some samples
Figure 9 S/H Input/Output waveforms, where input frequency is 8MHz.
A S/H circuit was simulated through H-SPICE. A robust fully differential amplifier which is
the main part of Op-Amps highly needs a proper CMFB. To guarantee stability of the system,
Miller consumption was used in the two-stage Op-Amp presented in this work. Switches are
the essential part of S/H circuits; as a result, linear response of these blocks is necessary.
Bootstrapped switches were used here because of their high linearity. A temperature and
voltage independent current reference is designed to obtain a firm current, in spite of
temperature and voltage fluctuations. The Op-Amp is the main power consumer part of the
S/H circuits; therefore the power consumption of this block should be reduced as low as
Int. J. Acad. Res. Appl. Sci., 1(1): 1-9, 2012
1. B. Razavi, Design of Analog CMOS Integrated Circuit,Mc Graw-Hill, New York, 2003.
2. A. M. Abo and P. R. Gray, "A 1.5-V, 10-bit, 14.3-MS/s CMOS Pipeline Analog-to-Digital
3. S. Chang, G. Wang, and Q. Huang, " A Novel MOSFET-Only Current Reference with
Multiple Temperature Compensations", IEEE, 2007.
4. Bendali, A., and Audet, Y.: ‘A 1-V CMOS current reference with temperature and process
compensation’, IEEE Trans. Circuits Syst. I, 2007, 54, (2), pp. 1424–1429
5. Camacho-Galeano, E.M., Galup-Montoro, C., and Schneider, M.C.: ‘A 2-nW 1.1-V selfbiased current reference in CMOS technology’, IEEE Trans. Circuits Syst. II, 2005, 52, (2),
pp. 61–65.
6. C. Yoo and J. Park "CMOS current reference with supply and temperature compensation"
ELECTRONICS LETTERS 6th December 2007 Vol. 43 No. 25.