Tail current node equalization for a variable offset amplifier

Abstract

First and second differential transistor pairs, where each may be intentionally unbalanced or balanced, are provided. First and second variable current generators are coupled to control respective tail currents of the first and second differential pairs. A switch circuit is coupled to equalize the voltages of the respective tail current nodes. Applications of the amplifier circuit include sense amplifiers and comparators.

Description

BACKGROUND

[0002] The invention is in general related to amplifier circuits and in particular to those having variable offset capability.

[0003] Amplifier circuits are used to amplify an input electrical signal to provide current and/or voltage gains or reductions. They may be used to amplify a single ended or a differential signal. A basic component of many amplifier circuits is the differential transistor pair used as the input stage of the amplifier. When the differential pair is used in conjunction with an output regenerative latch stage, a digital output signal (having one of two stable states) can be obtained that is an indication of a comparison between two single ended input signals or a determination of the magnitude of a differential signal. Due to the high gain and bistable nature of the latch stage, the output nodes of the latch stage are kept equalized until the correct information represented by the input signal can be expected to be present.

[0004] Most practical implementations of amplifier circuits suffer from manufacturing process-induced variations in the structure of the circuit elements, which cause an offset in the amplifier's operation. This offset may be explained by, for instance, considering an amplifier that is designed to amplify a differential input signal. In the ideal case, the output of the amplifier would be zero volts if the input differential signal was zero volts. However, in practice, a zero voltage differential input signal often yields a small but nevertheless non-negligible output offset voltage.

[0005] Output offset may be corrected using a wide range of techniques known as offset cancellation techniques. In one such technique, the value of the input differential signal that actually yields a zero output voltage is measured and stored, and then is subsequently subtracted from each new input signal to thus cancel the offset of the amplifier.

[0006] To help reduce the inherent offset of the amplifier, the input differential transistor pair that receives the input signal is constructed using transistor matching techniques such that the pair is said to be balanced. The transistors of a balanced differential pair are typically replicates of each other in terms of size and as such are designed to mimic each other's electrical characteristics. This also allows the amplifier to respond in the same manner to variations in the input signal that have opposite polarity but the same magnitude.

BRIEF DESCRIPTION OF THE DRAWINGS

[0007] The invention is illustrated by way of example and not by way of limitation in the figures of the accompanying drawings in which like references indicate similar elements. It should be noted that references to “an” embodiment in this disclosure are not necessarily to the same embodiment, and they mean at least one.

[0008] FIG. 1 shows a circuit schematic of an embodiment of a variable offset amplifier circuit.

[0009] FIG. 2 depicts a circuit schematic of another embodiment of the variable offset amplifier circuit.

[0010] FIG. 2B illustrates a circuit schematic of a variable offset amplifier circuit that uses matched differential pairs.

[0011] FIG. 3 illustrates a circuit schematic of an embodiment of a digitally controllable current source.

[0012] FIG. 4 shows a circuit schematic of a digitally controllable current source cell.

[0013] FIG. 5 depicts a circuit schematic of an embodiment of the variable offset amplifier that also includes an embodiment of a common mode feedback circuit.

[0014] FIG. 6 illustrates a circuit schematic of an embodiment of a digitally controllable variable offset comparator circuit.

[0015] FIG. 6B shows a flow diagram of an embodiment of an offset trimming procedure.

[0016] FIG. 7 illustrates a waveform timing diagram of an embodiment of a process for operating the variable offset amplifier circuit with tail current node equalization.

[0017] FIG. 8 shows a flow diagram of an embodiment of a process for designing the variable offset amplifier circuit.

[0018] FIG. 9 depicts a high speed data link in which a variable offset comparator is used in the receiver portion.

DETAILED DESCRIPTION

[0019] A variable offset amplifier circuit is described that may permit faster sensing of an input signal, by equalizing the tail current nodes of the amplifier until it is time to evaluate the input signal. According to an embodiment, the variable offset amplifier has first and second input differential transistor pairs, each of which has a matched pair of transistors (at least being the same size). According to another embodiment, first and second differential transistor pairs are provided that are intentionally unbalanced. In that case, each pair has first and second output nodes that are “inter-coupled”, i.e. the first output node of the first pair is coupled to the second output node of the second pair, and the second output node of the first pair is coupled to the first output node of the second pair. In both cases, variable offset can be achieved using variable current generators that are coupled to control the respective tail currents of the differential pairs.

[0020] Such an amplifier circuit also allows a variable non-zero offset to be introduced, by varying the respective tail currents of the differential pairs. This variable offset capability also allows the inherent offset of the amplifier to be cancelled so that zero offset is achieved. In addition, the non-zero offset may be beneficial in certain applications such as adders; rather than requiring a separate adder circuit, the amplifier can be used to inherently add a value to an input signal by setting a non-zero offset to include the value to be added. The amplifier circuit also permits the polarity of the offset to be changed as a function of changing tail current. This means that there may be a positive or a negative offset that is permissible about a center or nominal offset value. Yet another advantage of the amplifier circuit is that relatively large offset changes may be introduced using relatively small changes in the tail current. This feature helps keep the speed of operation of the amplifier circuit as a whole from varying too much across a wide offset range.

[0021] Referring now to FIG. 1, an embodiment of the amplifier circuit with tail current node equalization is shown as implemented using p-channel metal oxide semiconductor field effect transistors (MOS FETs).

[0022] The equalization capability is provided by a switch circuit 101 operated by a control signal. The switch circuit 101 may be, for instance, a single, switching p-channel FET that selectively provides a low impedance path that connects the tail current nodes 103 and 107 directly to each other, as dictated by the control signal. The equalization of the tail current nodes may alternatively be performed by a multiple transistor circuit which creates low impedance paths from the respective tail current nodes 103, 107 to another node which is kept at a fixed voltage. Operation of the circuit with tail current node equalization is described below in connection with a comparator application depicted in FIG. 6.

[0023] Referring to FIG. 1, this embodiment includes a first differential transistor pair 102, 104, and a second differential transistor pair 106, 108. Each pair is intentionally unbalanced. For instance, in this embodiment, the unbalanced characteristic is obtained by the transistor 102 being larger than the transistor 104 by a factor N>1 in the width of the transistor channel. Also, transistor 108 has a greater channel width than that of transistor 106, by, in this embodiment, the same factor N. To help keep manufacturing costs low, the length of the transistor channels may be the same across all transistors in the amplifier circuit.

[0024] The first and second transistor pairs may be referred to as being inter-coupled to each other because output node 120 of the first pair is coupled to output node 124 of the second pair, while output node 104 of the first pair is coupled to output node 126 of the second pair. Output nodes 120 and 126 are respectively coupled to load devices 110 and 112. These load devices may include passive and/or active circuits, depending on the application of the amplifier circuit. The load devices may alternatively represent a separate output amplification state that provides an output signal at further output nodes of that stage.

[0025] The respective tail currents of the differential pairs are controlled by variable current generators 114 and 116 as shown. In this embodiment, variable current generators 114 and 116 are variable current sources that also pass the tail currents of the respective differential pairs. Other types of variable current generators to control the tail currents may alternatively be used. See, e.g., the current sinks in FIG. 2 described below.

[0026] Continuing to refer to FIG. 1, the amplifier circuit provides output voltages Voutb and Vouta in response to the input voltages Vina, Vinb, Vinc, and Vind. In an embodiment of the amplifier circuit, a difference output voltage Voutb−Vina is generated in response to the difference Vina−Vinb and Vinc−Vind. The gain of the amplifier circuit may be determined by a variety of factors, including the impedance Rload of the load devices 110 and 112, and the transconductance gm of each transistor in the first and second differential pairs. The provision of variable non-zero offset by an embodiment of the amplifier circuit in FIG. 1 may be explained using the following example.

[0027] Consider the situation where the input voltages are equal, i.e. Vina=Vinb=Vinc=Vind. Also assume that the tail currents I1 and I2 are equal and the load impedances are equal. In such a configuration, the amplifier circuit provides a nominal offset that will appear at the output as Voutb−Vouta=Vnominal

={[N/(N+1)]I1+[1/(N+1)]I2}Rload{[N/(N+1)]I2+[1/(N+1)]I1}Rload=0.  (1)

[0028] Next, keeping the input voltages the same, if I1 is increased and I2 is decreased both by the same amount, namely I1+ΔI and I2−ΔI, then Vouta changes to the following:

V out a ={[N /(N+1)][I1+Δ1]+[1/(N+1)][I2−ΔI]}Rload={I+[(N−1)/(N+1)]ΔI}Rload  (2)

[0029] Similarly, the new value of Voutb is given by:

V out b ={[N /(N+1)][I2−ΔI]+[1/(N+1)][I1+ΔI]}Rload={I−[(N−1)/(N+1)]ΔI}Rload  (3)

[0030] Finally, the difference output voltage Voutb−Vouta is given by:

V out b −V out a =−[2(N−1)/(N+1)]ΔIRload  (4)

[0031] Thus, increasing I1 and decreasing I2 resulted in a decrease in the difference output voltage as given in the expression above. This decrease is the offset forced by the change in the tail currents. Now, if the tail currents are changed in the reverse direction, that is if I1 is decreased and I2 is increased by the same amount, then following an analysis similar to that above gives the following expression:

V out b −V out a =[2(N−1)/(N+1)]IRload  (5)

[0032] which is an offset in the output voltage that is opposite in polarity to that given by equation (4). Thus, this example illustrates how opposite polarity offsets may be obtained in proportion to a differential change in the tail currents.

[0033] It should be noted that although in the example above the change in tail current was exactly differential in that one tail current was increased by exactly the same amount as the other tail current was decreased, the amplifier circuit may still be useful if the tail currents are changed in unbalanced ways. Also, unequal load impedances may alternatively be provided to force a non-zero nominal offset.

[0034] Continuing to refer to FIG. 1, to improve the common mode rejection of the amplifier circuit, a common mode feedback circuit (CMFB) 130 may be coupled to adjust the variable current generators 114, 116 in response to voltages of the first and second differential pairs. In the embodiment shown, the CMFB 130 detects the voltage at the output of the variable current generators 114 and 116 and in response adjusts the tail currents of the differential pairs by adjusting the variable current generators 114, 116. The CMFB 130 does not add any significant capacitance to the inputs of the differential pairs, making the CMFB 130 compatible with applications that require low input capacitance.

[0035] Referring now to FIG. 2, what's shown is a circuit schematic of another embodiment of the variable offset amplifier, this time using n-channel MOS FETs 202, 204 for the first differential pair and n-channel MOS FETs 206, 208 for the second differential pair. The variable current generators in this embodiment are current sinks 214 and 216 which control the tail currents of the differential amplifiers, respectively. The output nodes 220 and 224 are coupled to a load 210 which in turn is coupled to a power supply node, whereas output nodes 222 and 226 are coupled to a load 212 which is also coupled to the power supply node. Thus, in contrast to the embodiment of FIG. 1, the amplifier circuit in FIG. 2 has its load devices 210, 212 referenced to a power supply node rather than to a power return (e.g. ground) node. Despite this difference, the same ability to adjust the offset is present and the example used to explain the operation of the embodiment of FIG. 1 also applies albeit with a slight difference in nomenclature to the n-channel embodiment in FIG. 2. In addition, a CFMB 230 may also be provided for the n-channel embodiment in FIG. 2, to further improve the common mode rejection of this amplifier circuit. Tail current node equalization may be obtained in this embodiment via a single, switching n-channel FET (not shown) coupled to provide a low impedance path directly between the tail current nodes under a gate control signal.

[0036] In FIG. 2B, a circuit schematic of a variable offset amplifier circuit that uses matched differential pairs I and II is shown. This embodiment is designed to accept a single, differential input signal Vn, and in response provide a differential output signal Vo at the output nodes indicated. Note that each differential amplifier I, II in this case is configured to have a single-ended output such that together they provide the differential output signal Vo. Variable offset is obtained in this embodiment by setting the tail currents I1 and I2 to different values, using variable current generators 114 and 116. Tail current node equalization is provided by a single p-channel transistor 210 whose gate receives the control signal.

[0037] Referring now to FIG. 3, this figure depicts a circuit schematic of an embodiment of a digitally controllable current source that can be used in place of the variable current generator 114 or 116. This current source has a number of digitally variable current cells 302_1, 302_2 . . . 302_M that are coupled in parallel to provide their individual currents which are summed to yield Iout. This output current Iout may be the tail current of a differential transistor pair. Each individual cell current may be adjusted by varying a bias level in each cell in response to an input signal. The individual cell current may be turned on or off in response to a signal at a control input CTL as shown. This control signal may be digital in nature, that is have one of two stable states corresponding to two levels of current at each individual cell's output. The total current may thus be adjusted by setting a digital value at the control inputs of the individual cells 302.

[0038] It should be noted that the output currents provided by the individual cells 302 may be unbalanced. For instance, some of the cells may provide larger currents (for coarse granularity control of the output current) while others may provide smaller currents (for fine granularity control of the output current Iout) Use of such digitally controllable current sources allows the offset of the amplifier circuit to be trimmed digitally, by selecting the desired offset according to a multi bit digital value.

[0039] FIG. 4 shows a circuit schematic of an embodiment of a digitally controllable current source cell 302. When the input signal at CTL is at a relatively high voltage, such that the p-channel transistor 402 is cut off, the output drive p-channel transistor 404 is biased according to an input bias signal at BIAS via p-channel transistor 408. In the other state, i.e., when the input signal at CTL drops to a relatively low voltage such that p-channel transistor 402 has its channel inverted, the gate of the p-channel transistor 404 is pulled to a relatively high voltage that is sufficient to, in this embodiment, place the transistor 404 in cutoff. Thus, two different levels of cell current Icell, e.g. “on” and “off”, are obtained in response to the input control signal at CTL. As mentioned in the previous paragraph, the various cells may be designed to provide different levels of “on” currents so that some may be used for fine granularity control of the total current Iout (see FIG. 3) while others may be used for relatively coarse granularity control of this output current. Referring back to FIG. 4, the varying levels of “on” currents may be achieved by sizing the transistors 408 and 404 as known to those of ordinary skill in the art.

[0040] Turning now to FIG. 5, this figure depicts a circuit schematic of an embodiment of the CMFB 130 used to improve common mode rejection, in an embodiment of the amplifier circuit. This embodiment of the amplifier circuit is similar to the p-channel embodiment shown in FIG. 1 that includes mismatched p-channel transistors 102, 104 for the first differential pair and 106, 108 for the second differential pair. Variable current generators 114 and 116 are coupled to source tail currents into a common source node for each pair as shown.

[0041] The voltages at two common source nodes are provided to separate n-channel transistors 504 and 508. If the two common source node voltages are equal, then the transistors 504 and 508, which in this embodiment are assumed to be matched, will present the same impedance to a bias circuit that includes the n-channel transistor 512 coupled in series with a diode connected p-channel transistor 516. A predetermined bias voltage appears at the node 520 which controls both bias levels of the two variable current generators 114, 116. The CMFB 130 operates as follows. Any differential variation in the common source nodes of the differential pairs should not cause a significant change in the total impedance between the node 520 and power supply return (ground), because of the way transistors 504 and 508 are connected to the same output node 520. However, if there is a common mode voltage change such that the voltages at both common source nodes of the differential pairs either increase or decrease simultaneously, then the impedance presented to node 520 will change, thereby causing the voltage at that node 520 to also change in a predetermined manner. In the embodiment shown in FIG. 5, if the voltages at the common source nodes of the differential pairs were to increase simultaneously, then the voltage at node 520 decreases in response, which in turn decreases the bias level and therefore the output current of each variable current generator 114, 116. On the other hand, if the common source node voltages of the differential pairs were to decrease simultaneously, then the voltage at the node 520 would increase in response, thereby increasing the tail currents I1 and I2. Such a feedback mechanism helps improve the common mode rejection of the amplifier circuit when it is used as a difference amplifier.

[0042] Turning now to FIG. 6, this figure illustrates a circuit schematic of an embodiment of a digitally controllable variable offset comparator circuit. The comparator circuit includes an amplifier circuit substantially as shown in the schematic of FIG. 1, including the first and second differential pairs which are defined by transistors 102, 104 and 106, 108, respectively. An alternative here would be to use the amplifier circuit of FIG. 2B. The variable current generators 114 and 116 are also coupled to control the tail currents I1 and I2 to the respective differential pairs. These current generators 114, 116 are controlled by a digital value that is received on multiple, offset select lines as shown. In this embodiment, each digital value of the offset corresponds to two oppositely varying tail currents I1 and I2 that are equidistant from a nominal tail current. This configuration is similar to the example given above in connection with FIG. 1 which helped explain the availability of opposite polarity offset using the amplifier circuit.

[0043] A single ended output voltage for this comparator may be available as either Vout or Vout#. To drive these output signals into one of two possible stable states, a regenerative load circuit 610 is provided as shown. After being reset by an input signal, this regenerative load circuit 610 will quickly amplify any difference between Vout and Vout#, where such amplification occurs at a relatively high gain due to the cross coupled n-channel pair 620 and p-channel pair 624, thereby ensuring that the output signals Vout and Vout# only assume one of two possible stable states. Thus, if Vin+ is greater than Vin− by at least the amount of offset that has been selected (as referred back to the input of the differential pairs), then the regenerative latch circuit 610 forcefully drives Vout to a low voltage level and simultaneously drives Vout# to a high voltage level. Other types of regenerative latch circuits may be used to provide the digital type output signal typically associated with a sense amplifier or a comparator application.

[0044] The variable offset comparator shown in FIG. 6 may behave as a sense amplifier which detects small differences between two analog signals Vin+ and Vin−. In this embodiment, a pair of differential signals are sensed, where the first differential signal is applied to the first differential pair 102, 104, while an inverted version of the differential signal is applied to the second differential pair 108, 106. This signal definition assumes that Vin+ is fed to the gate of the larger transistor 102 as well as the gate of the smaller transistor 106. Similarly, Vin− is fed to the gate of the smaller transistor 104 and to the gate of the larger transistor 108.

[0045] Turning now to FIG. 6B, a flow diagram for an offset trimming process is depicted. In operation 660, the comparator inputs are shorted to each other and, if necessary, to some reasonable voltage. The offset adjustment is then swept from the maximum negative to positive offset values (operation 664). Assuming the gain of the comparator is very large, the output of the comparator will change from low to high when the offset adjustment value is at the lowest offset setting (operation 668). At this setting, the offset of the comparator may be less than 1 least significant bit (LSB) of offset resolution. Note that changing the offset of the comparator may be viewed as changing the reference level to which a differential input signal Vin+−Vin−, as connected in the embodiment of FIG. 6, will be compared by the comparator. Thus, the reference level of the comparator may be set by the offset adjustment.

[0046] In FIG. 7, a waveform timing diagram of a process for operating the variable offset amplifier circuit with tail current node equalization is shown. In particular, this process refers to a comparator application such as depicted in FIG. 6 in which the output of the amplifier is a bistable output of the regenerative load circuit 610. In this embodiment, the control signal that is used to control the switch circuit 101 (in this case a single p-channel FET) is a digital signal applied to the gate of the FET and that has at least two stable states. The control signal also has a precise timing relationship with respect to the reset signal, another digital signal that is applied to equalize the output nodes of the regenerative load circuit 610. This control signal, labeled equalize#, is depicted by a separate waveform shown in FIG. 7. FIG. 7 also shows a waveform for the reset signal as well as a waveform for a latch signal. The latch signal is used to evaluate the first and second output nodes, by latching on, in this example, the rising edge, a discrete voltage state of the first and second output nodes. The structure for such a function is not shown in FIG. 6 as it is easily understood by one of ordinary skill in the art to include, for example, a flip flop.

[0047] The waveforms in FIG. 7 are interpreted to show that equalize# is used to equalize voltages of the tail current nodes, while at the same time the reset signal is asserted, the latter causing the output nodes to be equalized. The reset signal is deasserted, just after equalize# is deasserted. This causes the comparator offset to be more predictable when the input differential pairs I, II are to detect the information that is present at their inputs. This detection is exhibited at the output nodes whose voltages will be in a state that is then evaluated in the time interval indicated in FIG. 7, by the rising edge of the latch signal. This process of resetting, equalization, and latching can be repeated, based on the timing of a single clock signal (not shown) to repeatedly and accurately perform comparisons or sensing of an input signal.

[0048] The waveforms of FIG. 7 also illustrate some of the advantages of tail current node equalization, as applied to for instance a variable offset comparator application. Referring to FIGS. 1, 2, and 2B, consider that if a large transient were to appear on the inputs of the differential pairs I, II just before evaluation of the output nodes, such transient might inject noise (and in particular, differential noise) into the tail current nodes. An example of such a transient would be one caused the output of a switched capacitor, discrete arithmetic circuit. In such a case, the tail current values I1, I2 may change, due to the finite output impedance of the variable current generators 114, 116. This differential current error effectively changes the input offset of the amplifier and comparator, as the offset is controlled by the tail currents I1, I2. Although the tail current nodes will settle, unassisted, to their predetermined, quiescent value in time, the nodes may move towards their quiescent voltage values at substantially different rates. This may also require a significant period of time, (e.g. 20-30 pico seconds, where the period of the waveforms in FIG. 7 is about 200 pico seconds) thereby lowering the frequency at which the comparator can be used to repeatedly make comparisons. Accordingly, a solution to this problem may be to equalize the tail current nodes, as described above, without waiting the length of time needed for the tail current nodes to settle unassisted, only until the earliest point in time at which the correct information represented by the input signal can be expected to be evaluated at the output nodes.

[0049] Another advantage of tail current node equalization is particularly apparent in comparator applications with a complementary MOS type latch in the output stage. Such an output stage can yield a full swing (nearly rail to rail) output voltage for the comparator. However, this full swing can change the tail current nodes from their quiescent voltage values, thereby affecting the offset (and hence decreasing the sensitivity) of the comparator for the next comparison cycle (see FIG. 7). Tail current equalization “dissipates” this affect, without having to wait a relatively longer period of time for the tail current nodes to settle unassisted.

[0050] The various embodiments of the amplifier circuit described above may be designed using conventional computer aided design/electronic design automation tools. FIG. 8 depicts a flow diagram of an embodiment of a computer-implemented method for designing the amplifier circuit. The method involves the creation of a representation of the differential transistor pairs where each pair is either intentionally unbalanced as in FIGS. 1 and 2, or balanced as in FIG. 2B (Operation 704). The size of the transistors that constitute an unbalanced pair may be selected according to the width of the channel of each transistor, while keeping the length of the channels the same. Other techniques for sizing the transistors may alternatively be used.

[0051] The method also includes creating a representation of first and second variable current generators that are coupled to control respective tail currents of the first and second differential pairs (Operation 708). This representation of the variable current generators may include representing each generator as being digitally controllable as, for instance, shown in FIG. 3 (operation 712).

[0052] As mentioned above, there are various different scenarios for applying input signals to the amplifier circuit. For instance, a differential input signal may be applied to the larger and smaller transistors of the first pair, while an inverted version of the input signal may be applied to the larger and smaller transistors of the second pair. In a comparator embodiment, a magnitude of this differential input signal may be compared to a reference value that has been defined by a digital offset selection value. In a sense amplifier embodiment, this reference value may be set to a very small value. A representation of these input signals may be created using a computer-implemented method according to various well known techniques.

[0053] In addition, an output of the amplifier circuit may be defined in different ways as well. For instance, a differential output voltage may be defined as the difference between Voutb and Vouta (see FIG. 1). Other types of outputs, such as the single ended output as defined for the sense amplifier/comparator application of FIG. 6, may also be represented in the computer-implemented method. In general, the representations of all the embodiments described above may be created using computer aided design/electronic design automation tools that are well know to those of ordinary skill in the art.

[0054] A particular system application of the variable offset comparator with tail current node equalization is illustrated in FIG. 9. That figure shows a digital high speed data link involving two integrated circuit dies 904, 906 that are connected by a bus 905. The bus 905 is designed to carry a differential signal (data, address, or control) between the dies 904, 906. In the receiver portion 910 of the bus interface 914 (either in die 904 or in die 906), this differential signal is forwarded to the two input nodes of an on-chip variable offset comparator (such as the one depicted in FIG. 6). The comparator is controlled (via the equalize# and reset signals) so that data symbols in the differential signal are detected as latched voltage states of the comparator output nodes.

[0055] To summarize, various embodiments of an amplifier circuit with tail current node equalization, and a process for designing such a circuit have been described. In the foregoing specification, the invention has been described with reference to specific exemplary embodiments thereof. It will, however, be evident that various modifications and changes may be made thereto without departing from the broader spirit and scope of the invention as set forth in the appended claims. For instance, the particular circuit schematic shown in FIG. 4 for the digitally controllable current cell is just an example of a wide range of different designs that may be used to provide digital control of cell current. The specification and drawings are, accordingly, to be regarded in an illustrative rather than a restrictive sense.

Claims

1. An amplifier circuit comprising: first and second differential pairs; first and second variable current generators coupled to respective tail current nodes of the first and second differential pairs, to control respective tail currents of the first and second differential pairs; and a switch circuit coupled to equalize the voltages of the respective tail current nodes.

2. The amplifier circuit of claim 1 further comprising: a regenerative load circuit whose input is coupled to first and second output nodes of the first and second differential pairs.

3. The amplifier circuit of claim 2 further comprising: a further switch circuit coupled to equalize the voltages of the first and second output nodes of the first and second differential pairs.

4. The amplifier circuit of claim 3 wherein a plurality of input nodes of the first and second differential pairs are connected to each other so that when a differential input signal is received at a pair of input nodes of the first differential pair, an inverse of that input signal is at a pair of input nodes of the second differential pair.

5. The amplifier circuit of claim 4 wherein the first and second variable current generators are digitally controllable.

6. The amplifier circuit of claim 1 wherein the transistors of each differential pair are intentionally mismatched, and wherein each differential pair has first and second output nodes, the first output node of the first pair being coupled to the second output node of the second pair, the second output node of the first pair being coupled to the first output node of the second pair.

7. The amplifier circuit of claim 1 wherein the transistors of each differential pair are matched, and each pair having first and second transistors, the first transistors of each pair being coupled to a common input node, and the second transistors of each pair being coupled to further common input node, an output node of the amplifier being coupled to be driven by one but not both of the first and second transistors of one of the pairs, and not driven by the other pair.

8. A circuit comprising: means for driving the voltages of first and second output nodes in opposite directions, in accordance with an input voltage difference, the driving means having an offset that is a function of first and second currents through respective first and second bias nodes of the driving means; means for changing the first and second currents in accordance with a control input; and means for reducing changes in the first and second currents, said changes being caused by noise injected into the first and second bias nodes.

9. The circuit of claim 8 further comprising: means for regeneratively driving the first and second output nodes.

10. The circuit of claim 9 further comprising: means for equalizing the voltages of the first and second output nodes.

11. The circuit of claim 8 further comprising: means for improving the common mode rejection of the driving means as a function of the voltages of the first and second bias nodes.

12. A circuit comprising: first and second pairs of source coupled transistors, the transistors of each pair being intentionally mismatched and connected to drive first and second output nodes; a first variable current generator coupled to provide a current being one of (1) directed into, and (2) directed out of, a common source node of the first pair of transistors; a second variable current generator coupled to provide a current being one of (1) directed into, and (2) directed out of, a common source node of the second pair of transistors; and a switch circuit coupled to equalize the voltages of the common source nodes.

13. The circuit of claim 12 wherein an output node of a larger one of the first pair of transistors is shorted to an output node of a smaller one of the second pair of transistors.

14. The circuit of claim 13 further comprising: a latch circuit whose input is coupled to the first and second output nodes.

15. The circuit of claim 14 further comprising: a further switch circuit coupled to equalize the voltages of the first and second output nodes.

16. The circuit of claim 15 wherein an input node of the larger one of the first pair of transistors is shorted to an input node of the smaller one of the second pair of transistors, and an input node of a smaller one of the first pair of transistors is shorted to an input node of a larger one of the second pair of transistors.

17. The circuit of claim 16 wherein the first and second variable current generators have digital control inputs.

18. The circuit of claim 16 wherein the transistors of each pair have the same size ratio 1:N where N is a positive integer greater than one.

19. A method for sensing a differential signal, comprising: equalizing voltages of tail current nodes of first and second differential pairs, that are coupled to drive first and second output nodes with a regenerative circuit; and then releasing the tail current nodes of the first and second differential pairs, while a differential signal is being applied to first and second input nodes of the first and second differential pairs; and then evaluating the first and second output nodes.

20. The method of claim 19 further comprising equalizing voltages of the first and second output nodes, and then releasing the first and second output nodes while the differential signal is being applied to the first and second input nodes.

21. The method of claim 20 wherein the evaluation comprises latching a discrete voltage state of the first and second output nodes.

22. The method of claim 19 further comprising: prior to equalizing the voltages of the tail current nodes, trimming an offset of the first and second differential pairs, by changing respective tail currents of the first and second differential pairs, while the first and second input nodes are equalized, until the first and second output nodes change to another discrete state.

23. The method of claim 22 wherein the respective tail currents are changed by changing a binary variable which causes the tail currents to change in discrete, predetermined amounts.

24. An article of manufacture comprising: a machine-readable medium having data that when accessed by a processor supports a representation of a circuit that includes a pair of differential amplifiers, a pair of variable current generators whose outputs are respective tail current nodes of the pair of differential amplifiers, and a switch circuit coupled to equalize the voltages of the tail current nodes.

25. The article of manufacture of claim 23 wherein the machine-readable medium has further data that supports a representation of the circuit that further includes a latch circuit whose input nodes are output nodes of the pair of differential amplifiers, each differential amplifier is configured with a single-end output and coupled to the other to receive a differential input signal for the circuit.

26. The article of manufacture of claim 25 wherein the machine-readable medium has further data that supports a representation of the variable current generators as being digitally controllable so that they supply discrete, predetermined amounts of respective tail currents of the pair of mismatched differential amplifiers.

27. The article of manufacture of claim 26 wherein the machine read medium has further data that supports a representation of the circuit as part of a data communications receiver portion of a bus interface in an integrated circuit die.

28. A system comprising: first and second integrated circuit dies coupled to communicate with each other via a data communications bus, the bus being designed to carry a differential signal between the first and second dies, the first die having a bus interface in which an on-chip variable offset comparator has a pair of input nodes coupled to receive the differential signal, the comparator having a pair of differential amplifiers that have the input nodes, a pair of variable current generators whose outputs are respective tail current nodes of the pair of mismatched differential amplifiers, and a switch circuit coupled to equalize the voltages of the tail current nodes.

29. The system of claim 28 wherein the comparator further includes a latch circuit whose input nodes are output nodes of the pair of differential amplifiers.

30. The system of claim 29 wherein the variable current generators are digitally controllable so that they supply discrete, predetermined amounts of respective tail currents of the pair of differential amplifiers.