WO1999003197A2

WO1999003197A2 - A high speed and high gain operational amplifier

Info

Publication number: WO1999003197A2
Application number: PCT/SE1998/001347
Authority: WO
Inventors: Niaxiong Tan
Original assignee: Telefonaktiebolaget Lm Ericsson (Publ)
Priority date: 1997-07-08
Filing date: 1998-07-08
Publication date: 1999-01-21
Also published as: WO1999003197B1; EP0996996A2; KR20010014373A; WO1999003197A3; SE9702641D0; CN1111947C; SE519691C2; AU8366598A; CA2295840A1; EP0996996B1; SE9702641L; DE69836329D1; CN1262811A; JP2002511995A; DE69836329T2; TW393831B; US6018268A

Abstract

The present invention relates to the design of high speed and high gain operational amplifiers for use in for example high performance switched-capacitor analog circuits. Increasing the gain without suffering the speed may be achieved by designing the amplifier as a single-stage operational transconductance amplifier with single cascode for the N-type transistors (M8, M9) and double cascode for P-type transistors (M4, M5 and M10, M11). The invention may also comprise a continuous-time common-mode feedback. With this design of the invention, high speed and high gain can be maintained with a large phase margin to guarantee the stability.

Description

A high spead and high gain crjera ±xol aτpi i-Pipr Field of the invention

The present invention relates to the design of high speed and high gain operational amplifiers for the use in high performance switched-capacitor analog circuits, e.g., high performance analog-to-digital converters. Operational amplifiers are the most crucial building blocks of analog circuits. For high perfomance analog-to-digital converters in wideband radio systems the operational amplifiers set the limit of speed and accuracy.

Background of the invention

Operational amplifiers are the heart of most voltage-mode analog circuits. They usually dictate the operation speed and the accuracy of the switched-capacitor (SC) cicuits . They also consume most of the power in the SC circuits. High performance analog-to-digital (A/D) coverters usually use the SC circuit technique. Therefore, the performance of the operational amplifiers determines the perfomance of the A/D converters .

For the SC citcuits, the load is purely capacitive. Usually single stage operational transconductance amplifiers (OTAs) are preferred over multi-stage operational amplifiers. In OTAs, the capacitive load is used to create the single dominant pole , which usually yields high unity-gain bandwidth. The DC gain is usually moderate but can be improved by cascoding. For multi-stage operational amplifiers, internal miller capacitors and sometimes resistors are used to split poles and introduce zeros to compensate for the phase lag and the frequency response can be independent of the load. However, the unity-gain bandwith is usually lower than the single-stage OTAs, though the DC gain is higher due to the cascading of more stages. For high speed A/D converters, usually single-stage architectures are preferred in that it is possible to achieve a single-pole settling and to have a very wide bandwidth. However, the gain is usually not enough for high accuracy A/D converters.

In for example the document US-A-4 749 956 is shown a fully- differential operational amplifier for MOS integrator circuits, where the operational amplifier has one cascode transistor pair in the P-branch and one cascode transistor pair in the N-branch, see figure 5 in the said document.

Summary of the invention

The aim of the invetion is to increase the gain without suffering the speed and that may be achieved by designing a high speed and high gain operational amplifier for the use in high performance switched-capacitor analog circuits, e.g., high performance analog-to-digital coverters. The invented operational amplifier is a single-stage operational transconductance amplifier type with single cascode for the N-type transistors and double cascode for the P-type transistors. With reference to the cited document there should be a single-cascode in the N- and P-branches. The invention may also comprise a cotinuous-time common-mode feedback. With this design of the invention , high speed and high gain can be maintained with a large phase margin to guarantee the stability.

Brief description of the drawings

Figure 1 is schematic view of the invented operational tranconductance amplifier OTA. Figure 2 is a schematic view of the common-mode feedback circuit according to the embodiment.

Figure 3 shows a simulated frequency response of the OTA according to the invention.

Detailed description of the preferred embodiments

The operational amplifier shown in figure 1 is a folded- cascode OTA. Unlike coventional OTAs a double cascode is used in the P-branch to increase the gain without much speed penalty.

Transistors MO and Ml are the input devices and a transistor M12 provides bias current for them. Input signals Vin+ and Vin- are applied to the gates of transistors MO and Ml, respectively. Transistors M2 and M3 are bias transistors for the P-branch. Transistors M4 and M5 and the first cascode transistor pair in the P-branch and transistors M10 and Mil are the second cascode transistor pair in the P-branch. Transistors M6 and M7 are bias transistors for the N-branch and at the same time they provide a means to control the common-mode component via a signal CMFB generated in a common-mode feedback circuit. Transistors M8 and M9 are the cascode transistor pair in the N-brach. Vout+ and Vout- are the fully differential outputs. VbiasO is the bias voltage for transistor M12, Vbiasl is the bias voltage for transistors M8 and M9 , Vbias2 is the bias voltage for transistors M10 and Mil, Vbias3 is the bias voltage for transistors M4 and M5 , and Vbias4 is the bias voltage for transistors M2 and M3. AVCC and AVSS are the supply voltages usually having values of 5 and 0V, respectively. The invented operational amplifier shown in figure 1 is a single-stage OTA-type oprational amplifier and the unity-

1 8 ^• gain bandwidth is given by: / = — _t

2π C_L where g_min is the transconductance of the input transistors MO and Ml, and C_r is the load capacitance of the OTA.

Suppose the frequence of the parasitic poles formed at the sources of the cascode transistors are considerably larger than the dominant pole frequence, a single-pole settling results. The settling error in unity-gain buffer configuretion is given by B. Kamth, R. Meyer and P. Gray, "Relationship between frequency response and settling time of oprational amplifiers", IEEE J. Solid-State Circuits, vol. SC-9. Dec. 1974, pp. 347-352 by: . _{r) f}

where A is the DC gain of the oprational amplifier .Suppose a 12 -bit accuracy is needed. The operational amplifier needs to settling within a half clock sampling period with a 12-

,-12 bit accuracy and the relation will be: exp (-2π •/ ■ t) < -— ,

" 2 and thus : f > — = -^- = — = 2.8 •/ ,

" t 0.57 T ^■'sample , where T is the sampling period and f is the sampling sample frequency. The unity gain bandwidth must be three times larger than the sample frequence to guarantee a 12 -bit settling accuracy.

Considering the parasitic poles and the different surroundings for the OTA during different clock phases, there will be required that the unity-gain bandwidth will be at least six times larger than the sample frequency. Suppose a sampling frequence of 50 Mhz, then the unity-gain bandwidth should be over 300 Mhz. The smaller the load capacitance is, the larger unity-gain bandwidth v/ill be. However, there are two advers effects of using small load capacitance. The thermal noise pov/er and other noise power is inversely proportional to the sampling capacitance. Also the non-dominant poles can decrease the phase margin, if the non dominant poles are not so far apart from the dominant pole that is inversely proportional to the load capacitance. Therefore the load capacitance will be chosen 2* >4 pF. With this large sampling capacitance, the thermal noise does not limit the dynamic range of 12 bits if the peak input signal is larger than 0.5 V.

As a general rule, the phase margin should be larger than 45 deg for SC applications. With this large load capacitance, the phase margin is easy to guarantee.

The accuracy is directly related with the DC gain of the OTA and its capacitive surroundings. Suppose a 12 -bit accuracy is needed, a rough estimation of DC gain is given by: A_{D C} > 2 - 2ⁿ * 7ZdB .

Considering the design margin, the DC gain is reqiured to be larger than: 78 + 3 = 81 dB. To achieve this high gain, it is necessary to use cascode technique. Since the gain and output resistance of a PMOS transistor is considerably smaller than the gain and output resistance of an NMOS transistor, double cascodes are used for the upper branch as shown in figure 1. The DC gain is given by:

^ΛDC ⁼ 8_{m in} (^r ₀6 ^{- Λ}Λ/3 'I ^ro2 ^{' A}M-i ^{' A}M\0> ' where r_D and r₀ ^are the output resistance of the transistors M6 and M2 , respectively, A _M • _M ^{and A}M_W ^{are t le} 9^a^^{n of} the transistors M8 , M4 and M10 , respectively . The drawback is the limited output voltage range. However, it is of benefit to reduce the voltage swing to reduce distortion due to the sampling. Since the mobility in NMOS transistors is more than 4 times larger than that in PMOS transistors in certain aviable CMOS process, it is a good choice to design the common-mode voltage as low as possible to reduce the switch-on resistance of NMOS switches . The common-mode voltage is set to be 2V. The output voltage can swing more than +/- 1.2V without degradation in performance.

Shown in figure 2 is the common-mode feedback circuit. Transistors M35 and M36 are the input devices of the common- mode feedback circuit and their gates are connected with the input voltages Vin+ and Vin- , respectively, which are the fully-differential outputs Vout+ and Vout- of the operational amplifier of figure 1. Transistors M33 and M34 provide bias currents for the input devices M35 and M36. Resistors 137 and 138 are used to generate the common-mode voltage in the fully differential input voltages at the gate of transistor M66. Noticable is that the common-mode voltage is level-shifted due to the gate-source voltage of transistors M35 and M36. The common-mode input voltage Vcm is applied to the gate of transistor M67 via transistor M39 and level-shifted by the gate-source voltage of transistor M39. Transistor M40 provides the bias current for transistor M39. The difference between the voltages applied at the differential pair M66 and M67 i.e., the level-shifted common-mode voltage in the fully differential signals and the level-shifted common-mode input voltage is used to generate the common-mode control signal CMFB used in the operational amplifier of figure 1. Transistors M68 and M69 are the loads for the differential transistor pair M66 and M67 and the current in transistor M69 is used to control the common-mode voltage in the operational amplifier of figure 1 via the signal CMFB. Transistor M64 is the bias transistor for the differential pair M66 and M67 and transistor M65 is the cascode transistor for transistor M64. VbiasO is the bias voltage for transistors M33, M34, and M40, Vbias3 is the bias voltage for transistor M65, and Vbias4 is the bias voltage for transistor M64. AVCC and AVSS are the supply voltages having values of 5 and 0V, respectively.

To verify the performance a SPICE simulation is carried out within the CADENCE platform. An optimization of DC operation points is prioritized to make the circuit less sensitive to process variation. The optimization is carried out in such a way that there is enough source-drain voltage to guarantee all the transistors in saturation region even when there is a considerable change in threshold voltage and transistor dimension. The simulation result is shown in figure 3, where both the amplitude and the the phase response are shown.

To check the robustness of the circuit the bias current is varied by 20%, and both the input and the output common-mode voltage are varied from 1.8 to 2V. Under all these variations, the DC gain is larger than 83dB, the unity gain bandwidth is larger than 400Mhz, and the phase margin is around 60deg with a 4-pF capacitance, as seen in figure 3. The performance of the OTA is summarized in table 1.

While the foregoing description includes numerous details and specificities, it is to be understood that these are merely illustrative of the present invention, and are not to be construed as limitations. Many modifications will be readily apparent to those skilled in the art, which do not depart from the spirit and the scope of the invention, as defined by the appended claims andtheir legal equivalents .

Table 1: Summary of the performance of the OTA

Claims

1. A method for using unsymmetrical cascode for the N- and P-branch in folded-cascode operational amplifier architectures for use in high performance switched-capacitor analog circuits, e.g., high performance analog-to-digital converters, characterized in that more cascode devices are used for the branch in which the devices have lower gain.

2. A method according to claim 1, characterized in that one more cascode device is used for the P-branch in folded- cascode operational amplifier architectures.

3. A method according to claim 2 to generate the control signal for a NMOS current source transistor pair (M6 and M7) in a single cascode NMOS current source pair (M6, M8 and M7, M9) , characterized in that the control signal contains the information of the common-mode component in the fully differential output signals.

4. A device for using unsymmetrical cascode in the N- and P- branch in folded-cascode operational amplifier architectures, characterized in that fully differential input signals are provided to be applied to an NMOS pair (MO and Ml) , wherein the signals pass through a double cascode PMOS pairs (M4 and M5 , M10 and Mil) and fully differential output signals are provided by terminating the signals with a single cascode NMOS current source pair (M6, M8 and M7, M9) .

5. A device according to claim 4, characterized in that an NMOS transistor (M12) is used to provide the bias current for the said NMOS pair (MO and Ml) , in that a PMOS transistor pair (M2 and M3 ) are used to provide the bias current for the double cascode PMOS pairs (M4 and M5 , M10 and Mil) and in that a control signal is used to bias the NMOS current source transistor pair (M6 ans M7) in the said single cascode NMOS current source pair (M6, M8 and M7, M9) .

6. A device for generating a control signal for an NMOS current source transistor pair (M6 and M7) in a single cascode NMOS current source pair (M6, M8 and M7 , M9) , characterized in that fully differrential signals from a device for realizing a folded-cascode operational amplifier are provided to be applied to an NMOS pair (M35 and M36) , in that a resistor pair (137 and 138) are provided to be used to generate a signal proportional to the summation of the fully differential signals applied to the said NMOS pair (M35 and M36), in that the said signal and a common-mode reference signal (Vcm) that is level shifted by an NMOS transistor (M39) are provided to be applied to a PMOS pair

(M66 and M67) , in that two diode connected NMOS transistors

(M68 and M69) are provided as the load for the said PMOS pair (M66 and M67) and wherein the control signal is generated at the drain of one of the NMOS load transistors (M69) .

7. A device according to claim 6, characterized in that an NMOS transistor pair (M33 and M34) are provided for the bias current for the said NMOS pair (M35 and M36) , in that an NMOS transistor (M40) is provided for the bias current for the said NMOS transistor (M39) and in that PMOS transistors (M64 and M65) are provided to form a cascode current source and to provide the bias current for the said PMOS pair (M66 and M67) .