You are on page 1of 4

A High-Gain, Low-Power CMOS Op Amp Using

Composite Cascode Stages


David J. Comer, Fellow IEEE, Donald T. Comer, and Rishi Pratap Singh
Department of Electrical and Computer Engineering Brigham Young University
Provo, UT 84602
Email: comer@ee.byu.edu
June 1, 2010

Abstract–This work demonstrates that the com- on the chip. In the next generation of IC op amps, appear-
posite cascode differential stage, operating in the ing in 1968, the Miller effect was used in conjunction with
subthreshold region, can form the basis of a high- the second stage to reflect a large capacitance to the output
gain (110 dB), low-power op amp (27.8 µW). The of the first stage for compensation [6].
circuit can be fabricated without adding a compen- Although this Miller capacitance eliminates the need for
sation capacitance. Advantages of this architecture an external compensating capacitance, a few negative fac-
include high voltage gain, low harmonic distortion, tors must now be considered. The first is the feedthrough
low quiescent current and power, and small chip that takes place in the second stage. The signal applied
area. These advantages suggest that this design to the input of the second stage now has two paths to the
might be well-suited for low-power instrumentation third stage; one is through the amplifier, but a second sig-
applications requiring multiple amplifiers as often nal feeds through the capacitance that bridges the input
found in biomedical applications. and output nodes of the second stage. In most applica-
tions, the feedthrough signal is negligible compared to the
I. Introduction amplified signal and can be ignored. In a small number of
In recent years, an area of increasing interest is that of applications, especially digital signals with small rise times,
biomedical instrumentation amplifiers [1], [2]. These ap- the feedthrough signal can be troublesome. An additional
plications typically require high gain, low power, low fre- resistance in series with the compensation capacitance [8]
quency, amplifiers that occupy minimal chip real estate. can be added to minimize feedthrough.
This work discusses the subthreshold operation of compos- A second factor to consider is one that involves the analy-
ite cascode stages to achieve advantages such as high volt- sis or design of the op amp. Bridging the input and output
age gain [3], low nonlinear distortion [1], [4] low noise [1], nodes of the second stage results in a phenomenon known as
low power, low chip area, and low bandwidth. Although “pole splitting”. While the objective of lowering the pole
low bandwidth is often considered a shortcoming, in this or 3-dB frequency of the first stage is achieved, the pole
case, it can eliminate the need for a compensation capaci- of the second stage is moved to a higher frequency. This
tor to achieve stable operation in the presence of feedback. effect is more pronounced in BJT circuits than in CMOS
Required chip area is reduced by deleting this capacitance. circuits, but must often be considered in the design with
II. The Classical Op Amp Architecture either device [7]. Pole splitting is advantageous in achiev-
The first generation, bipolar junction transistor (BJT), ing a higher frequency design, but requires a capacitance
general purpose, integrated circuit (IC) op amp of 1964 and a resistance for proper compensation and, thus, adds
was designed to have a high voltage gain differential input complexity to the design.
stage, a moderately high voltage gain second stage, and a A third factor that can be significant in single-chip ampli-
low voltage gain/high current gain third stage that acts as fier arrays is the need for additional chip area required by
a buffer [5]. The first two stages develop the high overall the compensation capacitance and resistance. Each com-
voltage gain needed for the op amp circuit. This config- pensating capacitor may have a value of tens of pF and
uration, sometimes referred to as the Widlar architecture, occupy a significant chip area. The resistance is often im-
has continued in the design of op amps, for both BJT and plemented by a MOSFET in the triode region and requires
CMOS devices. little extra area.
These early IC op amps required an external capacitance As CMOS circuits grew in popularity, the BJT architec-
or an external RC combination to compensate the op amp ture was applied to CMOS designs. The low transconduc-
[5]. This compensating network was placed between the tance of the MOS device compared to that of the BJT has
output of the first stage and ground. The capacitance resulted in several departures from the classical architec-
needed was considerably larger than could be integrated ture for comparable performance of the CMOS op amp. If

978-1-4244-7773-9/10/$26.00 ©2010 IEEE 600


VDD VDD = 1V

21/1
M9

M3
Vbias M10 103/1
93/1 M13

M3 5/1 M4
M4 V- V+

Vout Vout
M1 120/1 M2

3.62uA Ibias
M2 50/1
Cc 46/1
M11 28/1
M5 M6 M14

M12
M1 M7 M8 1/1
Vin 1/1

VGG VSS = −1V

Figure 2: Composite cascode op amp.


Figure 1: Single-ended composite cascode gain stage.
apportioned almost equally between the first and second
stage. Both of these stages have high gain (≈ 1000 V/V)
only two stages of voltage amplification are used, voltage and moderate bandwidth. The circuit for this op amp is
gain of the op amp is relatively low [8]. Other implementa- shown in Fig. 2. Fabrication of the op amp [3] demon-
tions add more gain stages to increase the overall gain, but strated that sufficient gain can be obtained with CMOS
become correspondingly more complex [9]. composite cascode stages to implement a high gain op amp
This work increases the CMOS op amp gain to BJT-like in the classical Widlar architecture. It also demonstrated
levels by using the composite cascode stage operating in the that the compensation capacitance can be minimized with
subthreshold region. In addition, the parasitic capacitance this approach, requiring a 3.5 pF value for the op amp.
at the output of the first stage is used to compensate the IV. An Op Amp with no Compensation Capacitance
op amp. Not only are the effects of feedthrough and pole Rather than designing the op amp with a high-gain, mod-
splitting minimized in this way, a great deal of chip area is erate bandwidth first stage and then degrading bandwidth
conserved by eliminating the compensation capacitor. Be- with a compensating capacitor, the premise of this work is
cause of the low dc current required by the differential input to create a very high-gain, very low bandwidth first stage.
stage, dc power consumption can also be minimized. Op- This is done by using the composite cascode differential
eration in the weak inversion region can also lead to lower stage shown in Fig. 3. P-channel input devices, M1 - M4,
harmonic distortion than strong inversion operation [4]. are used for the differential pair while the current mirror
III. Operation of MOS Devices in Weak Inversion load is implemented with devices M5-M8. Source followers
An earlier work [4] demonstrated the high gain that can are inserted between the input signal and the differential
be obtained with the composite cascode stage operating stage to act as dc voltage level shifters. The scaling of the
in the weak inversion region and suggested the use of this devices and the tail current are selected to place devices M3
stage for the first differential stage of a CMOS op amp. and M4 along with M5 and M6 in the subthreshold region.
This work also suggested that the resulting lower corner This significantly increases the drain to source resistances
frequency of the first stage would result in a smaller com- of these devices. In addition, the load impedances of de-
pensation capacitor. In another work [10], it was suggested vices M3 and M4 are further increased due to the source
that voltage gains exceeding 10,000 V/V per stage could be degeneration provided by devices M1 and M2. Devices M7
achieved by combining operation in the weak or moderate and M8 provide source degeneration for devices M5 and M6
inversion region with the composite cascode configuration as well.
of Fig. 1. Aspect ratios are chosen to operate the device With the high impedance loads of M1 and M3, the gain
M2 in the subthreshold region while M1 operates in the increases and the bandwidth decreases due to the parasitic
moderate or strong inversion region. Both devices are in capacitance at the output of this first gain stage. This
the active region. The subthreshold drain current of device differential stage is similar to that of Fig. 2 [3], however,
M2 leads to a large output resistance, resulting in a very the drain current has been decreased and scaling has been
high voltage gain. Devices M3 and M4 provide the compos- changed from that circuit to achieve the proper quiescent
ite cascode load for this device which also exhibits a large current to lead to the increased voltage gain and decreased
output resistance. bandwidth.
This concept was implemented [3] in a high gain (ap- This high voltage gain stage is followed by an output stage
proximately 120 dB) CMOS op amp that used the Widlar consisting of a source follower (M19) that also acts as a
architecture. In that work, the overall voltage gain was level shifter and the output amplifier consisting of M20 and

601
VDD = 1.5V

M9 M18
M10
m=2 m=2
10/10 M13 10/10
10/10
m=8 1/8 M20

M1 M2 10/4
V- 2/20 V+

20/1 20/1 M14


m=8 m=8 Vout
V DD
20/2 1/2 0.5pF
M3 m=10 M4
M11 M19
1/1 100K
M5 M6 25/0.6
40/1 m=2
m=36 RLoad CLoad
M21
M 12
m=4
2/10 6/2
1/16 M15 10/30 M17
M7 M8 m=4
2/10 2/10
m=4 M16

VSS = -1.5V

Figure 3: Op amp without added compensation capac-


itor.

M21. Figure 4: Open loop magnitude response.


The first stage is designed to achieve a voltage gain of tens
of thousands to hundreds of thousands along with a band-
width of a few Hertz. The output stage is designed to have
a gain between 1 and 10 V/V with a relatively high band-
width. With a dominant pole of a few Hertz, the phase
margin of the circuit should be acceptable even in the worst-
case, unity gain feedback configuration. No additional com-
pensation is required for this amplifier.
V. Simulation Results
Operation of this amplifier was simulated with Cadence
using the C5 AMI 0.5-µm technology and BSIM3v3 device
models. The load impedance consisted of a 100-kΩ resis-
tance in parallel with a 0.5 pF capacitance, a heavier load
than that presented by the IC lowpass filter that is intended
to follow the amplifier stage. For the circuit of Fig. 3, the
loaded voltage gain of the differential input stage is 43,550
V/V and the bandwidth is about 1 Hz. The second stage
has a loaded voltage gain of 7.46 V/V and a bandwidth of
about 2.23 MHz. The overall voltage gain is 325,000 V/V
(110 dB) while the open loop bandwidth is about 1 Hz.
The frequency response showing magnitude and phase
is plotted in Figs. 4 and 5. The phase margin is 95◦ .
For power supply voltages of ±1.5 V, the output of the
amplifier can swing from -1.25 V to +1.35 V. In the unity
gain configuration, the total harmonic distortion or THD is
0.24% with a 2.4 V peak-to-peak output voltage and 1.5%
for a 2.6 V peak-to-peak output.
Table 1 summarizes several important parameters for the
op amp when driving the capacitor/resistor load. Figure 5: Open loop phase response.

602
realization of amplifier arrays in low-frequency biomedical
measurements.

References

[1] L. Fay, V. Misra, and R. Sarpeshkar, “A micropower


electrocardiogram amplifier,” IEEE Transactions on Biomed-
ical Circuits and Systems, vol. 3, pp.312-320, October 2009.
[2] M. Mollazadeh, K. Murari, G. Cauwenberghs, and N.
Thakor, “Micropower CMOS integrated low-noise amplifi-
cation, filtering, and digitization of multimodal neuropoten-
tials, IEEE Transactions on Biomedical Circuits and Sys-
tems, vol. 3, pp. 1-10, February 2009.
[3] D. T. Comer, D. J. Comer, and L. Li, “A high-gain
CMOS op amp using composite cascode stages,” Interna-
tional Journal of Electronics, vol. 97, pp. 637-646, June
2010.
[4] D. J. Comer and D. T. Comer, “Using the weak in-
version region to optimize input stage design of CMOS op
Figure 6: Layout of op amp core. amps,” IEEE Transactions on Circuits and Systems-Part
2, vol. 51, pp. 8-14, January 2004.
[5] T. H. Lee, “Tales of the continuum: A subsampled his-
tory of analog circuits,” IEEE Solid-State Circuits Newslet-
ter, issue 4, 38-51, Fall 2007.
Table 1 Summary of results [6] D. J. Comer and D. T. Comer, Fundamentals of Elec-
tronic Circuit Design, New York: John Wiley and Sons,
Open loop gain, dB 110 2003.
Open loop bandwidth, Hz 1 [7] P. R. Gray, P. J. Hurst, S. H. Lewis, and R. G. Meyer,
Spectral noise density,
√ Analysis and Design of Analog Integrated Circuits, 4th Ed.,
@ 100 Hz RTI, nV / Hz 6.6 New York: John Wiley and Sons, 2001.
DC supply current, µA 9.2 [8] D. A. Johns and K. Martin, Analog Integrated Circuit
No signal dissipation, µW 27.6 Design. New York: John Wiley and Sons, 1997.
GBW, kHz 320 [9] P. E. Allen and D. R. Holberg, CMOS Analog Cir-
cuit Design, Second Edition. New York: Oxford University
As expected, the voltage gain for this op amp is high, the Press, 2002.
bandwidth is low, and the dc power supply current is low. [10] D. J. Comer, D. T. Comer, and C. S. Petrie, “The
A phase margin of 95◦ guarantees stability. Although the utility of the active cascode in analog CMOS design,” The
bandwidth is somewhat low, this high gain op amp would International Journal of Electronics, vol. 91, pp. 491-502,
perform sufficiently well for many high-gain, low-frequency, August 2004.
biomedical applications. The voltage gain varies from 106
dB to 112 dB over the process corners with a temperature
variation from -5◦ C to 80◦ C.
The chip area for this op amp is reduced considerably by
eliminating the need for a compensating capacitance. The
layout of the core op amp, not including the load or the
bond pads, is shown in Fig. 6. The chip area required is
approximately 0.034 mm2 (192 µm × 180µm) compared to
0.05mm2 for the compensated op amp of Fig. 2 that uses
a 3.5 pF capacitor for compensation.

VI. Conclusion

This work demonstrates the feasibility of the composite cas-


code connection, operating in the subthreshold region, for
the input differential stage of an op amp. The resulting am-
plifier has a high gain and low bandwidth while eliminating
the need for the addition of a compensation capacitor. It
also has reduced chip area, low nonlinear distortion, and
low noise. A natural application of such amplifiers is in the

603

You might also like