0.

9 V rail-to-rail constant gm CMOS amplier input stage

C.-W. Lu and C.-M. Hsiao
Presented is a 0.9 V rail-to-rail constant gm CMOS amplier input stage consisting of complementary differential pairs and a gm control circuit. The gm control circuit eliminates the gm dead zone, which occurs in the conventional rail-to-rail amplier with ultra-low supply voltages. The proposed amplier input stage has a constant gm that varies by +2.3% for rail-to-rail input common-mode levels. To verify the proposed amplier design, an experimental prototype operational amplier is also implemented using 0.35 mm standard CMOS technology.

The transconductance is then described as gmn   VGn IDn I0n W enn Ut nn Ut 2nn Ut L nt 3

Similarly, the transconductance of the p-channel differential pair in region I (Vicom , VGp 2 npUt) is then described as:   VDD VGp IDp I0p W e np U t np Ut 2np Ut L pt

gmp

For VGp 2 npUt , Vicom , VGn nnUt , the drain currents of n- and p-channel pairs in region II can be approximately expressed as: IDn and   VDD Vicom Ip W np U t I0p e L p 2   Vicom In W I0n e nn U t L n 2 5

Introduction: The extensive usage of portable electronic products has created a large demand for low-supply voltage integrated circuits. This is particularly true for low-supply rail-to-rail CMOS operational ampliers. However, many low-supply voltage applications require a rail-torail input stage to increase the signal-to-noise ratio. Fig. 1a shows a well-known method for obtaining a rail-to-rail input stage based on connecting n-channel and p-channel differential pairs in parallel. At least one of the two differential pairs is active for any input common-mode voltage (vicom). However, Fig. 1b shows that the transconductance ( gm) of this input stage when both differential pairs are active (region II) is twice that of when only one pair is active (regions I and III). This large variation in gm prevents optimal frequency compensation and introduces severe signal distortions. Researchers have proposed several methods to solve this problem [1, 2]. Recently, we proposed a 1 V rail-to-rail constant gm scheme that keeps the sum of currents in the complementary differential pairs constant [3]. At ultra-low supply voltages, however, the sum of the common-mode voltage ranges of the n-channel and p-channel differential pairs may become larger than the available supply voltage. This creates a dead zone in region II [2]. Fig. 1c shows that the dead zone produces a smaller value of total transconductance ( gmT). Our simulation result shows that the dead zone occurs when the supply voltage goes down below 1 V. This makes the conventional schemes not applicable to a 0.9 V opamp. To avoid generating a dead zone in the input range under ultra-low supply voltages, this Letter employs a negative feedback loop to keep the current of the differential pair close to the value of a reference level, which is set at the bottom of the dead zone.
Mpt vGp Mp Mn vGn Mnt Mn In gmT II gmp vicm I gm II gmT III

IDp

The gmT of the complementary differential pairs can then be described as gmT   Vicom   VDD Vicom I0p W I0n W np U t e nn U t e nn U t L n np U t L p
25
VDD = 0.9 V

20

Vgs = 0.375 V

, A/V

15
Vgs = 0.36 V

mT

10
Vgs = 0.345 V

region I

region II

region III

simulation calculation

0
III

current summation

gm

Ip Mp

I gmn

0.2

0.4 0.6 vicom, V

0.8

VDD

Fig. 2 Simulated and calculated gmT against vicom for different bias voltages of tail current transistors

gmp

gmn

vicm VDD

Fig. 1 Rail-to-rail amplier input stage

Proposed circuit: The proposed amplier input stage is operated in the weak inversion. The tail current of the n-channel differential pair depicted in Fig. 1a, can be obtained as   W In 2I0n L n Vicom e nn U t    .   Vicom VGn W W 1 2 e nn U t L n L nt 1

Fig. 2 shows the simulated and calculated gmT against vicom for different bias voltages of the tail current transistors under a 0.9 V supply voltage. The gmT in regions I and III is strongly dependent on the bias voltages, but nearly independent of vicom. This is the opposite of gmT in region II. This study sets a reference gmT at the bottom of region II and depresses the gmT in regions I and III to the reference gmT by varying the bias voltages. The input common-mode voltage (vicom,min) which minimises the gmT can be obtained by setting the partial derivative of gmT with respect to vicom to zero. vicom,min is then expressed as   nn I0p W =Lp nn np nn VDD Ut ln 8 vicom;min nn np nn np np I0n W =Ln Substituting (8) into (5) and (6), the reference currents of the n- and p-channel pairs for achieving a constant gmT are:   W exp Irefn 2I0n L n 9    nn I0p W =Lp VDD np ln nn np Ut nn np np I0n W =Ln   W Irefp 2I0p exp L p 10    nn I0p W =Lp VDD nn ln nn np Ut nn np np I0n W =Ln

where Ut is the thermal voltage, I0n is the zero bias current, n is the subthreshold slope coefcient, and (W/L)n and (W/L)nt are the aspect ratios of the n-channel differential pair and the tail current device, respectively [4]. Neglecting the unity in the denominator of (1), the drain current of the differential pair in region III (Vicom . VGn nnUt) can be approximately expressed as:   VGn In I0n W enn Ut 2 2 L nt

IDn

ELECTRONICS LETTERS 5th November 2009 Vol. 45 No. 23

Authorized licensed use limited to: National Chi-Nan University Library. Downloaded on December 25, 2009 at 02:04 from IEEE Xplore. Restrictions apply.

Fig. 3 illustrates the proposed amplier input stage, which consists of a complementary differential pair (M1M6) and a gm control circuit (M7M12, Rref , single-stage differential ampliers A1A2, Irefn and Irefp). Irefn and Irefp are set according to (9) and (10). The amplier, M7 M9, is a replica of the differential amplier, M1M3. Hence, the current in the devices of these two ampliers is equal. A negative feedback loop consisting of A1, M7M9, Rref , and Irefn maintains the tail current of the n-channel differential pair close to the value of Irefn. A similar feedback loop keeps the currents in the p-channel pair at the value of Irefp.

Conclusion: This Letter presents a 0.9 V rail-to-rail constant gm CMOS amplier input stage. The complementary differential pairs have a constant transconductance that varies by +2.3% for rail-to-rail input common-mode levels. Experimental results show that the proposed amplier input stage is suitable for low-speed, low-power and ultralow supply portable electronic products. # The Institution of Engineering and Technology 2009 14 July 2009 doi: 10.1049/el.2009.9033 C.-W. Lu and C.-M. Hsiao (Department of Electrical Engineering, National Chi Nan University, 1 University Rd, Puli, Nantou Hsien, Taiwan ) E-mail: cwlu@ncnu.edu.tw

Rref

2Rref
in-

2Rref
in+ inM2

Irefn

A1 M7

Irefn
M3

summing circuit

Irefn

M8

M1

M9

References
1 Achigui, H.F., et al.: 1-V DTMOS-based class-AB operational amplier: implementation and experimental results, IEEE J. SolidState Circuits, 2006, 41, (11), pp. 2440 2448 2 Stockstad, T., et al.: A 0.9-V 0.5-mA rail-to-rail CMOS operational amplier, IEEE J. Solid-State Circuits, 2002, 37, (3), pp. 286 292 3 Lu, C.-W., et al.: 1 V rail-to-rail constant-gm CMOS opamp, Electron. Lett., 2009, 45, (11), pp. 529530 4 Vittoz, E.A., et al.: CMOS analog integrated circuits based on weak inversion operation, IEEE J. Solid-State Circuits, 1977, SC-12, (3), pp. 224 231

gm control circuit Irefp A2 inM10

M12

M6

Irefp

Irefp
M4 M5 in-

M11

in+

Rref

2Rref

2Rref

Fig. 3 Proposed ultra-low supply voltage rail-to-rail constant gm amplier input stage

Simulation and experimental results: The proposed amplier was simulated and fabricated using 0.35 mm CMOS technology. Fig. 4 shows the simulated normalised amplier transconductances against Vicom. Variations in gm lie within an error interval of +2.3%. The amplier was connected as a unity gain closed-loop conguration and measured under a 25 pF capacitive load. A square wave input of a 0.9 V swing is applied to the input of the amplier. The measured slew rates are 0.162 and 0.135 V/ms for the rising and falling edges, respectively. The total harmonic distortion (THD) with the input of a 0.9 V and 1 kHz sinusoidal input waveform is 259 dB. The DC power consumption is 8 mW.
2.5 gmT normalised transconductance 2.0

1.5

1.0

0.5

gmn

gmp

0 0 0.2 0.4 0.6 Vicom, V 0.8

Fig. 4 Simulated normalised transconductances against input commonmode voltage

ELECTRONICS LETTERS 5th November 2009 Vol. 45 No. 23

Authorized licensed use limited to: National Chi-Nan University Library. Downloaded on December 25, 2009 at 02:04 from IEEE Xplore. Restrictions apply.