Gain Enhancement for Amplifier Using a Replica Amplifier

This invention relates to amplifiers, and more particularly, to a gain enhancement technique for an amplifier. The amplifier preferably is electronic, but it may be mechanical or otherwise non-electrical.

Operational amplifiers typically require high open-loop gain and high bandwidth to minimize errors in output voltage. Obtaining high open-loop gain is difficult in modern MOS amplifiers, however, due to the small intrinsic gain of short-channel MOS transistors. Long-channel transistors can improve gain, but bandwidth is sacrificed.

Positive feedback or cascode amplifiers, such as regular cascode, double cascode, and active cascode, can improve gain, but the output swing and input common-mode range are concomitantly reduced, as is generally known in the art. The reduction in output swing (often termed "compliance" or increase of "headroom" relative to the power supply) is a problem, particularly in low voltage applications. In this context, another problem with using cascode circuits for improving gain is that the output resistance of the amplifier is increased, and this makes it more difficult for the amplifier to drive a resistive load.

According to the invention, these problems are addressed by an amplifier which may be modelled as having a first stage, a second stage, and a correction circuit. As used herein, the term "stage" is not meant to necessarily imply or indicate a discrete collection of components, such as transistors or resistors; rather, a single component may, for example, form all or part of the first and all or part of the second stage (as will be apparent from the Detailed Description below). Furthermore, a "stage" may be an implicit function of a component which is part of another stage. The first stage receives an input signal from an input source, and provides a first stage output signal. The correction circuit receives an input signal similar to the input signal received by the first stage, and provides a correction circuit output signal. The correction circuit output signal and the first stage output signal are then combined and provided to the second stage, which generates the amplifier's output signal.

In an embodiment of the present invention, the correction circuit has its own first and second stages which are substantially similar to the aforesaid first and second stages (hereafter, the "main amplifier first stage" and the "main amplifier second stage", respectively; the term "main amplifier" shall be used to refer in the aggregate to the main amplifier first stage and main amplifier second stage), and a coupling stage which is also similar to the main amplifier first stage. Each of the correction circuit and the main amplifier has a feedback circuit, and the two feedback circuits are substantially similar to each other. When a similar input is provided to the correction circuit as is provided to the main amplifier, the coupling stage of the correction circuit provides an output which is about the same as that which would have been provided by the first stage without the correction circuit. With the correction circuit, the output from the first stage is much smaller than the output from the correction circuit. The two outputs are combined and provided to the second stage.

Preferably, the first stage of the main amplifier includes a transconductance element and the correction circuit includes a similar transconductance element in its first stage and a like transconductance element in the coupling stage. The coupling stage receives the same signal as is input to the transconductance stage of the second amplifier and provides an output current to a summing node at the output of the main amplifier transconductance element. If the transconductance values for each transconductance stage are about the same, then the coupling stage provides a current which is about equal to what the current would have been from the transconductance element of the main amplifier were it not for the presence of the correction circuit (i.e., were the correction circuit absent).

The transconductance stage of the main amplifier need not necessarily match the coupling stage transconductance or the correction circuit second stage transconductance. The current provided by the coupling stage should, however, result in an output voltage which is the same as the output voltage from the replica amplifier. Accordingly, the product of the transconductance of the coupling stage and the output resistance of the main amplifier should be about the same as the open loop gain of the replica amplifier, where the open-loop gain is the product of the transconductance and the output resistance.

Because the coupling stage has an electrical effect (e.g., loading) on the transconductance element of the main amplifier, the correction circuit further includes mimicking circuitry which mimics these electrical effects on the transconductance element of the correction circuit.

The amplifier provided by the first and second stages of the correction circuit (hereafter the "replica amplifier") may be identical to the main amplifier, may be a scaled version, or may be different and even have different polarity transistors. Both the main and replica amplifiers have second stages comprising output resistances. In a scaled version, if the transconductance stage and the output resistance stage of the replica amplifier are substantially similar to the corresponding stages of the main amplifier, the transconductance provided by the transconductance element of the replica amplifier is decreased by a factor of N and the output resistance is increased by the same factor N. By decreasing the transconductance and increasing the output resistance, chip space can be saved while keeping the open loop gains equal.

Each amplifier may have a feedback network with a source impedance and a feedback impedance. The ratios of the two feedback impedances to the respective source impedances should be matched as closely as possible. The replica amplifier impedances can be identical to the main amplifier impedances, or they can be scaled, so long as the same ratio is maintained.

A circuit according to the present invention provides an increase in gain, without a loss in the output swing. Since the output swing is important in circuit applications with a low voltage power supply, this circuit is suitable for applications with a low supply voltage range, e.g, two volts. The circuit can be used with other elements, such as a cascode transistor, and with additional stages in the amplifiers.

Other objects, features, and advantages will become apparent from the following description of embodiments and from the claims when read in conjunction with the appended drawings in which:

FIG. 1 is a partially schematic, partially block diagram model of a prior art amplifier with feedback;

FIG. 2 is partially schematic, partially block diagram of an embodiment of the present invention;

FIG. 3 is a similar diagram of the embodiment of FIG. 2, but labeled to indicate mismatch;

FIG. 4 is a similar diagram of the embodiment of FIG. 2 with a scaled replica amplifier;

FIG. 5(a) is a schematic circuit diagram for an implementation of another embodiment of the present invention;

FIG. 5(b) is a partially schematic, partially block diagram of a portion of the circuitry in FIG. 5;

FIG. 6 is a partially schematic, partially block diagram of an embodiment of the invention used for testing;

FIG. 7 is a schematic circuit diagram for an alternative implementation of the invention;

FIGS. 8(a) and 8(b) are schematic circuit diagrams of another embodiment of the present invention, employing a fully differential amplifier; and

FIG. 8(c) is a partial schematic, partial block diagram representation of the circuitry of FIG. 8(b); and

FIG. 9 is a partially schematic, partially block diagram of an alternative embodiment of the present invention.

Referring to FIG. 1, an amplifier 2 with a single amplifying stage topology can be generally modelled as a transconductance element 4, constituting a first stage, and an output resistance 6, constituting a second stage. A feedback network couples an input terminal 9 to the first stage and an output terminal 11 from the second stage. The feedback network includes a source impedance 7, of value Z.sub.s, and a feedback impedance 8, of value Z.sub.f.

The ideal closed-loop gain of the amplifier is expressed as V.sub.o /V.sub.i =-Z.sub.f /Z.sub.s =-(beta) (for frequencies below a certain break point frequency). This ideal gain relationship assumes that the open-loop gain of the amplifier, A.sub.o =(g.sub.m)(r.sub.o), is infinite, where g.sub.m is the amplifier transconductance and r.sub.o is an intrinsic output resistance. In practice, however, open-loop gain is finite, producing an error factor of (1+beta)/A.sub.o, so that the actual gain is (beta)/(1+error), where "error" is the error factor. As this equation indicates, the greater the error, the lower the actual gain.

Referring to FIG. 2, an amplifier assembly 10 according to the invention has a main amplifier 12 and a correction circuit 20. The correction circuit includes a replica amplifier 14 and a transconductance coupling stage 16. Replica amplifier 14 has a feedback network 7R, 8R which is substantially similar to the feedback network 7, 8 in main amplifier 12. In this embodiment, coupling stage 16 includes a transconductance element which has the same transconductance g.sub.m as that of transconductance stage 17 of the main amplifier and transconductance stage 19 of the replica amplifier. As a result, coupling stage 16 provides an output current i.sub.x which has the same value as the current from transconductance stage 19 of the replica amplifier. This current would also have been the output of main amplifier transconductance stage 17 in the absence of correction circuit 20. As a result, current i.sub.x, which is injected from the replica amplifier to summing node 21 of the main amplifier (at the input to the output resistance stage), is close to the total current needed to bring output Main V.sub.o of the main amplifier 12 to an ideal output voltage. Consequently, the current needed from the transconductance stage 17 of the main amplifier, .DELTA.i.sub.x, is much smaller than the output of transconductance stage 4 in FIG. 1, and represents a portion of the additional current needed to bring the current i.sub.x of the replica amplifier 14 up to the ideal output current.

The effective open-loop gain of the main amplifier is significantly increased by this technique, so that the error can be shown to be ((1+beta)/A.sub.o).sup.2. Comparing this error with the error of an amplifier without a replica as discussed above, the error is effectively reduced by a factor of A.sub.o /(1+beta), thus increasing the effective open-loop gain by this factor. If beta is small relative to A.sub.o, as is typically true, there is a substantial increase in the open-loop gain.

Preferably, the main amplifier and the replica amplifier are matched as closely as possible, since mismatch reduces the gain. Referring to FIG. 3, for example, if the transconductances have equal value and if the output resistance of the replica amplifier is r.sub.o +.DELTA.r.sub.o, the effective gain is limited to A.sub.o (r.sub.o /.DELTA.r.sub.o). Consequently, if there is a 5% mismatch for example, the gain improvement is limited to a factor of 20.

Referring to FIG. 4, the replica amplifier does not necessarily have to be identical to the main amplifier, but rather, can be scaled down in size and power without detracting from the effectiveness of the gain enhancement technique. To scale down by a factor of N, transconductance stage 19a has a transconductance of g.sub.m /N and output resistance (r.sub.o)(N); the impedances 7a, 8a are scaled to (N)(Z.sub.s) and (N)(Z.sub.f), respectively. Beta and A.sub.o are thus the same as for the non-scaled embodiment of FIG. 2. Accordingly, the two feedback networks and the two amplifiers may be substantially similar to each other even though they may have different values. By scaling the output resistance and feedback impedances by the same factor, certain parasitic effects can be minimized. The scaled down embodiment is beneficial since it can take up less space on a chip when fabricated, and can reduce power consumption.

When a square pulse or a step voltage is presented at the input to an amplifier, generally it is desired that the output settle as quickly as possible. If the unity-gain bandwidth of the replica amplifier is made equal to that of the main amplifier, however, there is an increase of about 20% in the settling time due to a pole-zero pair introduced near the closed loop bandwidth of the circuit. This effect can be removed by making the unity-gain bandwidth of the replica amplifier greater than that of the main amplifier. Since the replica amplifier typically does not have to drive large capacitive loads, it is easy to make the unity-gain bandwidth of the replica amplifier higher than that of the main amplifier, which may be required to drive considerable capacitive loads.

Referring to FIG. 5(a), an exemplary CMOS circuit implementation of the invention (exclusive of feedback networks) includes a main amplifier 40, a replica amplifier 42, and a coupling transconductance stage 44 which is MOSFET MN45. Since transistors MNX35 and MN45 are matched, if no voltage is applied to the inputs of the main amplifier, Main V.sub./ , the output of the main amplifier 40, Main V.sub.o, is the same as the replica output, Replica V.sub.o, which is already close to the desired output voltage, as described above.

In this example, the first transconductance stage and the first output resistance stage provide a voltage at node 41. Transistors MN35 and MN45 convert input voltages from node 41 and from node 43, respectively, to an output current at the combined drain of the transistors. The output resistance is implicit looking into the drains of transistors MN35, MN45, and MP45 which are in parallel. The product of the output current and the output resistance is the main amplifier output voltage 48M.

Referring to FIG. 5(b), an amplifier, such as that shown in FIG. 5(a), may be referred to as having a two-stage topology, where each amplifying "stage" includes a transconductance stage and an output resistance stage. As discussed above, the term "stage" is used broadly to include each transconductance stage, g.sub.mlm and g.sub.mlr' and each output resistance stage, r.sub.ol and r.sub.o2, individually, or the combination of a transconductance stage and an output resistance stage. In this representation of the two-stage topology, a correction circuit 72 provides current i.sub.xc to the output of a second transconductance stage 74 of a main amplifier 70.

While transconductance stages 74, 75, and 76 can be matched so that g.sub.m2m =g.sub.m2c =g.sub.m2r, and r.sub.o2m =r.sub.02r', these equalities are not required. It is important, however, that the output voltage produced by the coupling current i.sub.xc be substantially similar to Replica V.sub.o. Accordingly, (g.sub.m2c)(r.sub.o2m)=(g.sub.m2r)(r.sub.o2r). As the mismatch between the main and replica amplifiers increases, the improvement in open-loop gain is correspondingly reduced.

Referring again to FIG. 5(a), mimicking circuitry 46 is provided to better maintain a match between main amplifier 40 and replica amplifier 42. Transconductance stage 44, which includes transistor MN45, has an effect on main amplifier 40 since it has a common drain with transistor MN35. Transistor MNX45 is added to share a common drain with transistor MNX35, and thus to match the effect of transistor MN45 on main amplifier transistor MN35. Transistors MPX55, MPX65, and MNX55 are also added provide a bias current to transistor MNX45 to match the current to transistor MN45. Circuitry 46 thus causes transistor MNX45 to mimic transistor MN45 and, hence, its loading effect on the main amplifier, and provides this effect to the replica amplifier.

In this example, the supply voltages are +1 V and -1 V, for a 2 volt total possible output swing. The capacitors C.sub.1 and C.sub.2 in the circuitry are provided for stabilization.

In a representative embodiment, a CMOS operational amplifier as shown in FIG. 5(a) is implemented, for example, in a 1.2 micron CMOS technology. These minimum length transistors are employed for high speed and good output swing, except for transistors MP45 and MPX45 which are implemented with 1.8 micron transistors. The present invention can be used in conjunction with a cascode if desired.

The circuit in FIG. 6 is used for testing the performance of the amplifier. For ease of understanding, the coupling stage is not shown in full and is indicated only by dashed line 62, because the internal stages of the main and replica amplifiers are not shown. For the main amplifier 60, load capacitance C.sub.LM is 22 pF; adding that to the feedback capacitance C.sub.FM, which is in series with the two parallel 10 pF capacitors 64 and 66 yields a total of 28.6 pF load capacitance to be driven by the main amplifier 12. For the replica amplifier, the load capacitance CLR is approximately 10 pF and the total load capacitance driven by replica amplifier 20 is 16.7 pF. The 1% settling time has been measured to be 55 ns. A summary of measured performance is provided in Table I as follows:

Table I indicates different parameters with the replica amplifier present (right column) and with the replica amplifier absent (middle column). As indicated by the table, the effective DC gain is increased, using the replica amplifier, by a factor of about 13 with no load resistor, and a factor of 50 with a one kilohm load, without degrading other performance parameters. While the power dissipation and the die area are greater with the replica amplifier and coupling circuit than without, the power and area increase can be reduced by scaling down the replica amplifier as discussed above in connection with FIG. 4.

The principles of the present invention can be used with other designs, and in conjunction with other elements, such as the addition of a cascode transistor. For example, referring to FIG. 7, a replica amplifier 52 is coupled to a main amplifier 50 through transconductance stage 54. This embodiment is similar to that described in conjunction with FIG. 5(a), except that the input stages of both the main amplifier 50 and replica amplifier 52 are connected to receive differential inputs, and there are no transistors in series with capacitors C.sub.1 and C.sub.2.

Referring to FIG. 8(a), the replica amplifier technique can be applied to other types of amplifiers, such as a folded-cascode topology. In main amplifier 80, the input differential pair formed by transistors MN1a and MN2a serve as the transconductance stage. These transistors convert the input voltage to a differential current which is conveyed to the output terminals 81 and 82 by transistors MP3a and MP4a. The current is converted to an output voltage by an implicit output resistance given by the parallel combination of the effective resistances looking into the drains of MP3a and MN4a for output terminal 81, and MP4a and MN5a for output terminal 82. Transistors MP1a, MP2a, MN4a, MN5a, MN6a, and MN7a are used as current sources. Since the amplifier has only one g.sub.m stage and one effective r.sub.o (output resistance), this folded-cascode topology is considered a single-stage topology.

Replica amplifier 84 has a corresponding input differential pair of transistors MPX5a and MPX6a which convert the voltage difference between Rep V.sub. and Rep V.sub.+ to a current, and output resistances inherent from looking into the drains of transistors MPX3a, MNX4a, MPX4a, and MNX5a. In this example, the main amplifier and the replica amplifier have transconductance stages which comprise transistors of opposite polarity-type in that the main amplifier has NMOS transistors in the transconductance stage, while the replica amplifier has PMOS transistors in the transconductance stage.

The use of opposite polarity-type transistors demonstrates that although the main and replica amplifiers may be identical or at least substantially similar, only the critical parts of the amplifier need be substantially similar. As discussed in conjunction with FIG. 5(a), the product of the transconductance of the coupling amplifier and the output resistance of the main amplifier, and the product of the transconductance of the replica amplifier and the output resistance of the replica amplifier should match. Matching between the transconductance stages of the main amplifier and the replica amplifier is not necessarily required. Thus despite the opposite polarity transistors, the circuit of FIG. 8(a) still has an effective enhanced gain.

The input to differential transistor pair MPC1a and MPC2a of the coupling amplifier 86 is connected to the input to the differential transistor pair MPX5a and MPX6a of replica amplifier 84. In this case, since the coupling amplifier has a transconductance that is substantially similar to the transconductance of the replica amplifier, the coupling amplifier delivers a current to the output of the main amplifier that is substantially similar to the current which the replica transconductance stage delivers to the output of the replica amplifier. In this example, the output resistances of the main and replica amplifiers match.

Mimicking circuitry 88 includes transistors MNX1a and MNX2a, which form a differential pair biased by MNX3a. These transistors mimic the loading effect of the differential input stage of the main amplifier formed by transistors MN1a, MN2a, and current source transistor MN3a. Thus, the effective output resistance of the main and the replica amplifiers are matched. As in the circuit of FIG. 5(a), the output of the main amplifier is already close to the ideal voltage.- Hence, the input of the main amplifier only has to change by a small amount to bring the main output even closer to the ideal voltage, and to thus increase the effective open-loop gain of the main amplifier.

In the main amplifier, transistors MP3a, MP4a, MN4a, and MN5a are the cascode transistors. These transistors increase the output resistance and hence the open-loop gain. This increased gain is achieved at the expense of the output swing, which is reduced by the saturation voltages of the transistors. In applications in which the output swing can be reduced, or where a cascode topology is favored, the replica technique allows a further increase in gain without a further decrease in the output swing. The replica amplifier has similar corresponding cascode transistors.

Since this embodiment is a fully differential amplifier, a common mode feedback (CMFB) signal is added to keep the common mode range at a desired value, preferably close to zero. As is generally known, additional circuitry (not shown) is provided which monitors the output and feeds back a signal to the CMFB input terminal for each amplifier.

Referring to FIG. 8(b), the implementation in FIG. 8(a) can be simplified. Since the coupling amplifier input is connected to the replica amplifier input, the coupling amplifier and the replica amplifier can be merged into one correction circuit 92 which has a combined replica/coupling amplifier. The current from the replica/coupling transconductance stage is split between the main output and the replica output at nodes 94 and 96. Due to the relationship discussed above in conjunction with FIG. 5(a), if the output resistances are the same, then the coupling transconductance and the replica transconductance would be the same, and the current is split equally. This circuit achieves the same functionality as the circuit in FIG. 8(a). Since the circuit in FIG. 8(b) has a smaller number of transistors than the circuit in FIG. 8(a), it can require less area to implement.

Referring to FIG. 8(c), the embodiment of FIG. 8(b) can be represented generally as having a main amplifier and a replica amplifier. The current output of combined transconductance stage 99 is provided to a current splitter 98 which provides current to the output resistance of the replica amplifier, r.sub.orc, and to the output resistance of the main amplifier, r.sub.om. To maintain a relationship so that the product of the current provided from the replica amplifier and the output resistance of the main amplifier equals the output voltage of the replica amplifier, the current is split in a ratio of r.sub.om :r.sub.orc. In effect, this representation of the circuitry of FIG. 8(b) is similar to previously described representations, such as FIG. 3, in which a coupling transconductance stage is explicitly indicated.

Referring to FIG. 9, circuit 110 is an alternative embodiment that is generally similar to that of FIG. 5(b) except that it has two transconductance coupling stages 112, 114. The current from stages 112, 114 are combined with the output currents from respective transconductance stages 116, 118, at respective summing nodes 120, 122.

Having described various embodiments of the present invention, it will become apparent to those skilled in the art that still other variations are possible without departing from the scope of the invention as defined by the appended claims. For example, it should be understood that the principles here could be applied to non-electrical feedback systems, and that gain, ideal or actual, is not necessarily greater than one, but may be one or less than one. In addition, the circuit described above can be incorporated in amplifier circuits with additional stages and elements.