Three Stage Amplifier With Positive Feedback

Compensation Scheme
J oiio Ramos, Michiel Steyaert
K U Leuven, ESAT-MICAS
Leuven.
Abstract- A CMOS opamp that can drive large ca-
pacitive loads is presented. The technique employs a
positive feedback compensation (PFC) t o improve fre-
quency response as compared to nested Miller compen-
sation (NMC), allowing the circuit to occupy less silicon
area and straightforward design.
At 1.5 V, the circuit dissipates 275 pW, has more than
100 dB gain, a gain bandwidth of 2. 7 MHz and 1.0 V / p
average slew rate while driving a 130 pF load.
I. I NTRODUCTI ON
The driving force behind technology development -
digital CMOS design, demands high integration to lower
fabrication costs, and lower voltage to decrease power
consumption, which in turn allows reduction of thermal
dissipation, and in portable electronics ensures a reason-
able battery lifetime. This trend, makes everyone working
with integrated circuits aware of the constrains that low
voltages poses, especially in the case of analog design.
Given current standard CMOS fabrication processes
cannot withstand supply voltages higher that 1.5 V [l],
this is a strong push to do research in circuits able to
operate at this voltages, but without decrease in perfor-
mance.
For the specific case of analog design, the basic building
block nowadays continues to be the operational amplifier.
The literature, aware of this fact continues to publish new
solutions that try to mitigate the ongoing needs, such as
operation at lower supply voltages [2-41, higher gain [3-51,
bandwidth and efficiency. Particularly at low voltages,
cascoding is no longer an advisable technique to achieve
high gain, as stacked transistors limit the available voltage
swing at the output. As a result the multistage amplifiers
is under research as a technique to overcome this limita-
tion.
For the CMOS operational amplifier presented in this
paper, a three stage (good compromise between a high
voltage gain, complexity of the circuit and power dissipa-
tion) amplifier is developed that offers simultaneously op-
eration at low voltages (1.5 V), rail-to-rail output swing,
high gain (>l o0 dB) and good bandwidth to power dis-
sipation efficiency. This is a result of the new proposed
compensation technique.
The goal of this work is to realize a three stage amplifier
with a better bandwidth to power efficiency and suitable
Belgium
Fig. 1. Block diagram of a typical NMC amplifier.
Fig. 2. Block diagram of the newly proposed PFC amplifier.
for driving high capacitive loads such as high accuracy EA
modulators, pipeline A/D converters, linear regulators,
etc.
I n the next section we will briefly review the NMC
topology, the starting point for multistage amplifiers. The
PFC amplifier and some of its characteristics are then in-
troduced, before presenting, in section 111, the measured
results from a test chip that was fabricated and measured
to show the effectiveness of the compensation scheme. Fi-
nally, some conclusions are draw.
11. PROPOSED COMPENSATION
With the NMC compensation (Figure 1) for every new
gain stage added, a new Miller capacitor is added for com-
pensation purposes. However this new capacitor reduces
the bandwidth by a factor of two [6]. Moreover both ca-
pacitors load the output, and thus, increasing the power
requirements for the last stage in order to fulfill the band-
333
width and slew requirements. If capacitor Cm2 is not
present, a higher bandwidth is possible, but not without
having an instable amplifier as now a pair of complex
poles with sinall damping factor appear close to the unity
gain frequency. If we can conceive a circuit that allows
to: control the damping factor, does not load the output
and does not increase the circuit consumption, we would
have eliminated the peaking while having a wider band-
width. The proposed solution for frequency Compensation
is depicted in Figure 2.
The feed-forward transconductance gmf bypasses all
but the first stage at high frequencies to provide a direct
path to the output, consequently, this boosts the band-
width of the PFC amplifier. This block, is implemented
using a single MOS transistor (gmf in Figure 3) , driven
by the output of the first stage and connected to the out-
put node. By using this approach to implement g mf , it
is ensured that there is no increase in power consump-
tion and silicon area, when comparing it to the NMC [6]
counterpart.
A positive feedback around gm2 allows to effectively
control the damping factor of the complex poles, and ca-
pacitor Cm2 fulfill this condition. This capacitor is pre-
dominantly smaller than Cm, , and thus, not limiting the
slew rate of the first stage.
A. Theoretical Analysis of Frequency Response
It can be shown that, for the arrangement of Figure 2,
the open loop gain is given by
With the purpose of having symmetrical current capa-
bility in the output stage of Figure 3 and noting that gm3
and gmJ are in the same branch, both transconductances
are set to have equal values: gmf =gm3. After simplifica-
tion, the DC gain is given by
with a dominant pole at
The determination of the values for the second-order
system in the denominator of H( s ) follows the assumption
that the amplifier has third order Butterworth response
with unity gain feedback [6]. This assumption does not
take into account the presence of the zeros in the transfer
function and implies that the poles will be complex with
the damping ratio I equal to 1/ d, leading tc a phase
margin of 60.
The overall transfer function includes two zeros: one
LHP zero and one RHP zero. Given the constrains of
(4) applied in (1): the RHP zero will have a value one
magnitude order higher than the GBW minimizing the
degradation in the phase margin; a value smaller but close
to WO for the zero in the LHP enhances the overs11 phase
margin. The location of the zeros in the PFC amplifier
is translated into a gain in stability.
From (1) to ( 5 ) , the unity gain frequency of the PFC
amplifier can be found to be
GBW =0.875 x w,
( 6)
Solving the previous system with the result from (4) in
order to obtain GBW as function of circuit small signal
parameters, and using the approach presented in [4]
GBW( ~FC) = (e) x 1.46
where
I n the case P being greater than one, the proposed tech-
nique achieves higher bandwidth than the NMC for the
same power consumption. However, ,Bcan be made larger,
which is the case when loaded by large capacitances, as it
is proportional to a. Maximizing gm2 with respect to
gm3 and minimizing parasitic capacitors at the output of
the second stage also makes this approach more ejficient.
I n the case /3 is greater than 4, the bandwidth of the PFC
amplifier may be even wider than a single stage amplifier
141.
334 19-3-2
AVDD
AVSS
Fig. 3. Schematic diagram of the PFC amplifier.
B. Circuit Implementation
The circuit of Figure 3, intended to demonstrate the
proposed compensation scheme of Figure 2, has been fab-
ricated and tested. The typical threshold voltage for the
NMOS and PMOS transistors is 0.59 V and -0.57 V, re-
spect ively.
With the purpose of maximizing the efficiency of the
amplifier, a small signal equivalent of the circuit from Fig-
ure 3 was created in order to wrap around it an optimiza-
tion routine in Matlab. All parameters were optimized
with the exception of capacitor Cml r that had its value
set to be one order of magnitude higher than the parasitic
capacitors of the nodes it connects to. For the optimiza-
tion algorithm, a simple brute force approach was chosen,
being easy to implement and still fast to run, giving the
results in less than one minute.
111. EXPERIMENTAL RESULTS
A microphotograph of the PFC amplifier designed in
a 0.35 pm CMOS technology with five metal layers and
double poly capacitors is shown in Figure 6.
A. Measured Results
The bandwidth of the amplifier was chosen to be within
the range of the testset for automatic characterization of
opamps [ 7] . The frequency response is shown in Figure 4
and the measurement of the slew rate is shown in Figure 5
and was done with a Tektronix TDS680B oscilloscope.
The performances are summarized in Table I.
B. Performance Comparison
Considering that each amplifier has its own character-
istics, comparing them requires a figure of merit (FOM)
that weights the trade-off between the bandwidth, load
capacitance, slew rate and power consumption. Two of
them will be used: one for small signal [3] and one for
large signal performance [4].
4 0 -qi__L F,q" e" q
Fig. 4 Measured frequency response of the PFC amplifier.
-0 1 *I_ T l W .to-
Fig. 5. Measured transient response (130 pF//24 kR)
(10)
G B W [ M H z ] . Cr . [ p F]
FOMs =
power [mw]
The units are shown between brackets and the average
value of the SR is used for the calculations. From (10)
and (11) the higher the FOM, the better is the amplifier.
Table I1 presents an overview of multistage amplifiers
present in the open literature. Based on measured data,
both figures of merit are also presented. As can be seen,
TABLE I
MEASURED PERFORMANCE
Parameter
Technology
Low Frequency Gain
Unity Gain Requency
Phase Margin
Positive Slew Rate
Negative Slew Rate
Power Consumption
Power Supply
Load Condition
Area
Measured
0.35 pm CMOS
>l o0 dB
2. 7 MHz
52"
1.0 v/ps
1.0 v/ps
275 pW
3r0.75 V
130 pF / /24 kR
0.03 mm2
19-3-3
335
(dB) (MHz) (mW@Vdd)
Fig. 6. Microphotograph of the realized amplifier.
( V/ p s ) (pF) (MEyF) (w)
the figure of merit varies by more than one order of
magnitude. Different topologies and how much empha-
sis has been devoted to the optimization can justify this
effect. Especially note that the different load capacitors
are within a close range, so their weight in the calculations
is minimized.
From the results presented in Table 11, it can be con-
clude that the PFC amplifier is two times more efficient
in the case of small signal. In the case of large signal, the
best result is obtained. As indicated by equations (7) and
(9), a higher load capacitor will allow a higher FOMs [4].
I V. CONCLUSI ON
A new compensation scheme for a three stage ampli-
fier has been presented. The compensation capacitors
are small which allows the circuit to occupy less silicon.
The poles and zeros depend on ratios of currents and ca-
pacitors, making the stability less sensitive to matching
and thus making it suitable for integration in commercial
CMOS processes.
Measured results shows that this work outperforms the
other referenced amplifiers for a similar load capacitor.
This can be explained by the newly proposed topology
and a proper optimization of the overall circuit.
ACKNOWLEDGMENTS
J . Ramos gratefully acknowledges the financial support
from Fundapio para a Cicncia e a Tecnologiii and the
computer facilities provided by Ajuda a Igreja que Sofre
in the beginning of this work, both from Portugal.
REFERENCES
111 Overall Roadmap Technology Characteristics (ORTC), http:
//public.itrs.net/Files/2000UpdateFinal/ORTC2OOOfinal.
pdf.
[2] J . Fonderie and J . H. Huijsing, Operational Amplifier with 1-V
Rail-to-Rail Multipath-Driven Output Stage, IEEE J. Solid-
State Circuits, vol. 26, no. 12, pp. 1817-1824, Dec. 1.991.
131 H. T. Ng, R. M. Ziazadeh, and D. J. Allstot, A Muhistage Am-
plifier Technique with Embedded Frequency Compensation,
IEEE J . Solid-State Circuits, vol. 34, no. 3, pp. 339-347, Mar.
1999.
[4] K. N. Leung, P. K. T. Mok, W. 1. Ki, and J . K. 0. Sin, Three-
Stage Large Capacitive Load Amplifier with Damping-Factor-
Control Frequency Compensation, IEEE J . Solid-State Cir-
cuits, vol. 35, no. 2, pp. 221-229, Feb. 2000.
[5] K. Bult and G. J . G. M. Geeleri, A Fast-Settling CMOS Op
Amp for SC Circuits with 90-dB .DC Gain, IEEE J . Solid-State
Circuits, vol. 25, no. 6, pp. 1379--1384, Dec. 1990.
[6] R. G. H. Eschauzier, L. P. T. Kerklaan, and J . H. Huijsing,
100-MHz 100-dB Operational Amplifier with Multipath Nested
Miller Compensation Structure, IEEE J . Solid-Stai!e Circuits,
vol. 27, no. 12, pp. 1709-1717, Dec. 1992.
[7] C. Van Grieken and W. Sansen, A testset for automatic char-
acterisation of opamps in the frequency domain, in IEEE
International Conference on Microelectronic Test Structures,
Barcelona, Spain, Mar. 1993, vol. 6, pp. 83-88.
A CPdOS Low-
Distortion Fully Differential Power Amplifier with doc ble Nested
Miller Compensation, IEEE J . Solid-State Circuits, vol. 28, no.
7, pp. 758-763, J uly 1993.
[8] S . Pernici, S. Nicollini, and R. Castello,
336
19-3-4