ISSCC94 / SESSION 14 / AMPLIFIERS / PAPER FA 14.

FA 14J: A Compact Power-Efficient 3V CMOS value of the floating current source is dptermincbd 11y thih
Rail-to-Rail Inputloutput Operational Amplifier translinear loops M,,,M ,, M,,, MI,and M,,, M,, M,,, MI,. Thc
mirrors M,,-M,, and M,,-M,, are loaded by the input pairs. This
for VLSI Cell Libraries means that the gate-source voltage of M,, and MI, vary a s a
function of the common-mode input voltage. However, the
Ron Hogervorst, John P. Tero’, Ruud G.H. Eschauzier, Johan H. current ofthe floating current source changes only 3%,because
Huijsing an increase of the gate-source voltage of one mirror compen-
sates for a decrease in the other mirror.
Delft University of Technology, Deifl ,The Netherlands
’Philips USA , Sunnyvale, CA The floating current source has the same structure a s the class-
AB driver circuit, M19-q,,. Therefore, the supply voltage de-
The design of low-cost mixedmode VLSI systems requires pendence of the floating current source and class-AB driver
compact power-eflicient library cells with a good performance. largely compensate each other. The result is a quiescent cur-
Digital library cells fully benefit from the continuing down- rent independent of the supply voltage.
scaling of CMOS processes, since these cells contain mini-
mum-sizecomponents.Analog library cells, such as theopamp, The opamp is frequency compensated using the Miller tech-
cannot be designed using minimum-size transistors, for rea- nique [41. The capacitors, C,, and C,,, split the poles around
sons of gain, offset, etc. Moreover, low-voltage rail-to-rail the output transistors M,, and M,,, ensuringa 20dB per decade
requirements complicate the design. To obtain compact low- roll off of the amplitude characteristic.
voltage analog cells with a good performance, simple power-
ellicient designs need to be developed. In Figure 3, a second realization of the compact rail-to-rail
opamp is shown. This opamp is basically the same as the opamp
In principle, the two-stage opamp is suitable for a compact shown in Figure 2, except for the frequency compensation. This
design. Published two-stage rail-to-rail opamps, however, con- opampiscompensated byincludingthecascodestages,M,, and
tain complex class-AB output stages that useexcessivedie area MI,, in the Miller Imps, increasing the unity-gain frequency
111.Moreover, the class-AB control contributes significantly to from 2.6MHz to 6.4MHz.
the offset of the amplifier. This can be avoided, a t the expense
of die area, by inserting an intermediate stage between the The opamps are realized in a 0.8pm BiCMOS process. The n-
input and the output stage L21. channel transistors and p-channel transistors have threshold
voltages of 0.64V and -0.75V, respectively. Micrographs of the
this paperpresent sacompacttwo-stage3VCMOSopamp that compact opamp with Miller compensation and the op-amp with
is suitable as a VLSI library cell because of its small die area. cascoded Miller compensation are shown in Figure 4 and
Tho opamp, shown in Figure 1, contains a rail-to-rail input Figure 5, respectively. A list of specifications is given in Table
stage, MI-M, [31.Thecurrent mirrorsM,,-M,,andM,,-M,,add 1. Opampl contains the Miller compensation. Opamp2 con-
the signals ofthe rail-to-rail input stage. Both current mirrors tains the cascoded Miller compensation. At supply voltages
are biased by one side of the floating current source, Ibr This from 2.5V to 2.9V the common-mode input range is from Vss-
provides a constant current, I,,, in the cascode transistors, MI, 0.4VtoVDo-1.4V.At supply voltages above 2.9V, the input stage
and MI,, independent of the bias currents of the n-channel and has a common-mode input range from Vs,-0.4V to VD,+0.5V.
p-channel input pair. The CMMR of the opamps is 43dB in the transition regions of
the current switches. It increases up to 70dB, when the com-
The constant current in the cascode transistors ofthe summing mon-mode input voltage is near one of the supply rails. The
circuit biases the class-AB driver circuit, M,,-M,, [41. This Bode plots ofthe compact opamp with Miller compensation and
avoids a contribution to the noise and offset oftheopamp oftwo the compact opamp with eascoded Miller compensation are
independent current sources, that would otherwise be neces- shown in Figures 6 and 7, respectively.
sary to bias the floating class-AB driver.
References
The class-AB driver circuit biases the output transistors MZ5-
M,,;. The floating architecture of the class-AB control has the [l] Botma, J. H., et al.,‘A Low-Voltage CMOS Operational Amplifier
same advantages as the floating current source: it does not with a Rail-to-RailConstant-gm Input Stage and a Class AB Rail-to-
contribute to the noise andoffset ofthe amplifier. The result is Rail Output Stage”, Proceedings ISCAS 93, pp. 1314-1317.
that the offset and the noise of the opamp are of the same order [2] Pardoen, M.D., M. G. Degrauwe, “A Rail-to-Rail Inputloutput
as the offset and noise of a three stage opamp. CMOS Power Amplifier”,IEEE J. Solid-stateCircuits, vol. SC-25,pp.
501-504, Dec., 1990.
Using the topology described above, two opamps are realized.
The first is shown in Figure 2. The input pairs, MI-M,, operate 631 Huijsing,J. H., J. P.Tero,“Combinationdriver-summingcircuitfor
in strong inversion, which combines a relatively large gm and rail-to-rail differential amplifier”,US Pat. Appl. Ser. No. 36774, filed
relatively small input transistors. To obtain a constant transcon- March 25, 1993.
ductance over the whole common-mode input range, the cur- [4] Montecelli. D. M., “AQuadCMOSSingle-SupplyOp Amp withRail-
rent switches, M, and M,, and the 3x current mirrors, M,-M, to-Rail Output Swing”, IEEE J. Solid-state Circuits, vol. SC-21, pp
and M,,-MI,, are added to the input stage 151. 1026-1034, Dec., 1986.

The biasingcircuit, M,,-M,,, switches off M, a t supply voltages [5] Hogervorst,R., et al., “CMOSLow-Voltage OperationalAmplifiers
lower than 2.9V. This prevents M,-MI, from forming a positive with Constant-gm Rail-to-Rail Input Stage”, Proceedings ISCAS 92,
pp. 2876-2879
fcedback loop a t supply voltages larger than 2.9V.
[6] Huijsing, J. H., D. Linebarger, “Low-voltageoperational amplifier
The summing circuit, M,,-M,,, and the class-AB output stage, withrail-to-railinputandoutput ranges”,IEEEJ.Solid-StateCircuits,
M,,,-M,,,,are biased by a floating current source, M27-MzgThe vol. SC-20, pp, 1144-1150,Dec.. 1985.

Authorized licensed use limited to: National University Fast. Downloaded on May 28,2021 at 04:31:06 UTC from IEEE Xplore. Restrictions apply.
 I
ISSCC94/ FRIDAY, FEBRUARY 18,1994/ PRESIDIO / 9:00 AM

DIV DIV START 100 0 0 0 . 0 0 0 Hz

10.00 50.00 STOP 20 0 0 0 0 0 0 , 0 0 0 H Z
RBW: 100 Hz ST: 43.1 s e c RANGE: R = - 1 0 , T = - l O d 0 r r

Figure 1: Basic topology of rail-to-rail opamp. Figure 6: Frequency response of o p a m p w i t h Miller

compensation.

DIV OIV START 100 0 0 0 000 H Z

10.00 5 0 00 S70P 20 000 000.000 H Z
ROW 3 0 0 I-lz ST: 1 4 3 s e c RANGE: R=-10. T = OdBm

Figure 2: O p a m p w i t h Miller compensation. Figure 7: Frequency response of o p a m p

w i t h cascoded Miller compensation.

Parameter oDampl oDamD2 &&

Die area 0.04 0.04 mm2
Supply:
voltage range 2.5-6.0 2.5-6.0 V
quiescent current 180 180 PA
Peak output +3 *3 mA
Offset voltage 4.0 5.0 mV
Input noise a t lOkHz 22 31 nVldHz
Open-loop gain 87 85 dB
Unity-gain:
frequency 2.6 6.4 MHz
margin 66 53

Figure 3: O p a m p w i t h cascoded Miller compensation. Table 1: Measurement results. (Vsupp,,,=3.3V,

Figures 4 and 5: See page 345. R,a,=10kS2, C,-,,=lOpF, TA=27"C)

Authorized licensed use limited to: National University Fast. Downloaded on May 28,2021 at 04:31:06 UTC from IEEE Xplore. Restrictions apply.