Chapter 6

 

Differential and Multistage Amplifiers

 

 

Up to this point we have mainly been looking at transistor circuits driven by unbalanced inputs (i.e., one terminal of the signal source is grounded). A more important transistor arrangement is one that receives differential or balanced input signals.  The differential pair is one such example and is used extensively in present day monolithic IC operational amplifiers.  In this chapter we shall demonstrate how one generates the appropriate input signals for investigating the behavior of differential amplifiers.  Following this, we shall investigate several circuits involving differential and multistage amplifiers. This will also include an investigation of various types of current mirrors and current sources.

 

6.1 Input Excitation for the Differential Pair

 

The differential pair, shown in Fig. 6.1, is the most widely used circuit building block in analog integrated circuits.  Its operation is based on the fact that only the difference between the signals appearing at the two inputs is amplified. The signals appearing as common mode at the amplifier inputs are (ideally) not amplified.  In the following we would like to use Spice to investigate the effect of varying the input common-mode and differential-mode signal levels on the collector currents. The focus of this problem is not so much on the behavior of the differential pair, but rather on how we generate the input signals for differential amplifiers within Spice.

 

For example, to analyze the behavior of a differential pair subject to a differential input signal, first-time users of Spice are often tempted to apply a single ungrounded source to the amplifier input as illustrated in Fig. 6.1. Unfortunately, due to the lack of a defined common-mode input level, erroneous data is generated by Spice.  To see this, we perform a DC analysis of the differential pair shown in Fig. 6.1 assuming the BJTs had the following parameters: I_S=14 fA, B_F=100 and V_AF=100 V. The Spice input file is listed in Fig. 6.2.

 

The Spice results of the DC analysis are then found in the output file as follows:

 

 

 

As is evident from these results, we see that the DC voltage appearing at the two inputs to the differential pair (nodes 1 and 2), and at the emitters of the two transistors (node 3), are very large negative levels, outside the limits of the supply voltages.  Certainly, these levels would not be observed in any real circuit.  The reason for this, as just mentioned, is the lack of a common-mode input level and can be corrected by revising the input excitation to include a common-mode level amongst the differential component.

 

 

 

 

The circuit shown in Fig. 6.3 accomplishes this in a convenient way. It allows the input common-mode level to be adjusted by varying the value of V_CM, independent of the value of the differential-mode component being established by the two VCVSs connected across the input terminals of the differential pair. The level of each VCVS is one-half the voltage value set by the isolated voltage source V_d.  This voltage source is loaded arbitrarily in a 1-ohm resistor in order to satisfy the Spice requirement that every node in a circuit has at least two connections.

 

 

 

 

To demonstrate the versatility of this input-source arrangement, we shall investigate the effect of separately varying the input common-mode and differential signal components on the collector currents of the differential pair of Fig. 6.3.  The Spice input file describing this circuit is listed in Fig. 6.4.  The parameters of the npn BJTs are assumed to be the same as before (i.e., I_S=14 fA, B_F=100 and V_AF=100 V).  As our first analysis request, we are asking that Spice perform a DC sweep of the input common-mode voltage level V_CM beginning at -5 V and ending at +6 V in increments of 100 mV.  The differential input component V_d is set to zero during this analysis, thus also making V⁺ and V^- equal zero.  On completion of Spice we observe in Fig. 6.5 the behavior of the two collector currents as a function of V_CM.  Here we see that both Q₁ and Q₂ are conducting equal currents of 0.5 mA for most input common-mode levels; However, when V_CM exceeds 5.6 V the collector current of both Q₁ and Q₂ fall off very rapidly to 0 mA.  Any further increase in V_CM causes the collector currents to go negative. This observed behavior is easily accounted for by the fact that the base-collector junctions of both Q₁ and Q₂ become forward-biased when V_CM exceeds 5.6 V and conduct appreciable currents in the opposite direction to the normal flow of collector current.

 

To investigate the effect of a differential signal appearing at the input terminals of the differential pair, we simply revise the Spice input file given in Fig. 6.4 by replacing the DC sweep command given there by the following one:

 

.DC Vd -500mV +500mV 10mV

 

Here we are requesting that Spice vary the level of V_d between -500 mV and +500 mV in 10 mV increments. The common-mode input level V_CM is to remain at 0 V throughout this analysis.  Submitting the revised input deck to Spice, results in the display of the collector currents in Fig. 6.6. Here the results are as expected: Depending on the level of input differential signal, the biasing current of 1 mA is steered between the two transistors. 

 

 

6.2 Small-Signal Analysis of the Differential Amplifier: Symmetric Conditions

 

In this section we investigate the small-signal behavior of the BJT differential-pair amplifier configuration shown in Fig. 6.7 using Spice. We shall assume throughout this section that the circuit remains symmetric, i.e., the resistances in collectors are equal and transistors Q₁ and Q₂ are matched.  Our purpose here is twofold: to illustrate how one uses Spice to determine the small-signal behavior of a differential amplifier, such as that shown in Fig. 6.7, using the multiple source arrangement discussed in the previous section, and to determine the accuracy of the expressions that are derived by Sedra and Smith, 3^rd Edition, in Sections 6.2 and 6.3 of their text.

 

It is important to point out that when one compares results computed by Spice to those computed through the application of closed-form expressions derived by hand analysis from a small-signal circuit model of the transistor circuit, the accuracy of the results depend on two factors: (1) how precise the estimate of the values of the parameters of the small-signal model is, and (2) how accurate the small-signal expressions are; given that certain simplifying assumptions are made in their development. Within this section, we are simply addressing the second issue; that being, the accuracy of the expressions derived by Sedra and Smith for the BJT differential amplifier under various circuit conditions. This is accomplished by making use of the small-signal model parameters computed by Spice directly through a DC operating point analysis command rather than estimating them from the DC circuit conditions.

 

 

 

 

 

 

According to the development provided by Sedra and Smith, the small-signal analysis of the differential-pair shown in Fig. 6.7 is relatively straight forward. In Table 6.1 we summaries the results of their analysis. This table includes expressions for both the differential-mode and common-mode voltage gain (A_d and A_CM) and the input resistances (R_id and R_iCM). As well, it includes expressions for both the input-referred offset voltage, input bias current and input offset current (V_OS, I_B and I_OS) under asymmetric circuit conditions.

 

Let us consider using Spice to calculate the differential-mode voltage gain and input resistance of the differential amplifier shown in Fig. 6.7.  We shall assume that each transistor has device parameters: I_S=14 fA and B=100. For the moment we shall neglect the effect of the transistors Early Voltage (i.e., V_A=¥).  This is equivalent to stating on the transistor model statement that VAf=0.  The Spice input file describing the differential amplifier of Fig. 6.7 is seen listed in Fig. 6.8. The input to the differential amplifier is arranged using the multiple-source set-up described in the last section. It consists of two voltage-controlled sources (EV+ and EV-) and two independent input voltage signals (Vd and Vcm).  The DC level of both of these voltage sources are set equal to 0 V since we already know that the two transistors of the differential pair will remain in the active region under these bias conditions.  That is, with a transistor quiescent current of 0.5 mA, the voltage at the collector and emitter of each transistor will be approximately 10 V and --0.7 V, respectively. The differential input voltage source Vd also includes an AC component of 1 V.  The reason for this component will be made clear in a moment.

 

Our first analysis request is a .TF command written as

 

.TF V(5,4) Vd.

 

This analysis command will provide us with an equivalent circuit representation of the differential amplifier evaluated around 0 V DC as seen looking into the port made by the input differential signal source V_d and the output port denoted between nodes 5 and 4 in Fig. 6.7. Unfortunately, with this multiple source arrangement, the resistance seen by the input source V_d is just the 1 ohm resistor connected across V_d.  Recall that this resistor was added just to increase the number of connections made at node 101 and serves no circuit function. We therefore require another means of obtaining the resistance seen looking into this differential amplifier.

 

If we consider applying a 1 V AC signal across R_d, then this signal will also appear across the input terminals of the differential pair.  By calculating the current that flows into the terminals of the differential amplifier due to this AC signal, we can obtain the input resistance of this amplifier as the reciprocal of this current. In the Spice deck listed in Fig. 6.8, we have already indicated this 1 V AC component on the statement describing the differential input voltage source.  The AC current that flows through the input terminals of the differential amplifier can then be monitored with the two dependent voltage sources in series with the input terminals (i.e., EV+ and EV-).  To obtain this information, we have added the following two command statements to the Spice deck:

 

 

The first command describes to Spice that an AC analysis is to be performed at only one frequency point of 1 Hz. The second statement instructs Spice to send to the output file the magnitude of the AC current that flows into the input terminals of the differential amplifier. As a means of insuring that the correct AC voltage appears at the input terminals of the amplifier, we are also requesting that Spice print out the voltage that appears there.

 

Finally, as our last analysis request we have included an operating point command (.OP) in the Spice deck. This command will provide us with the parameters of the small-signal model of each transistor.

 

On completion of Spice, we obtain the following parameters of the 2-port equivalent circuit of the differential amplifier shown in Fig. 6.7:

 

                

We see here that this particular differential amplifier has a small-signal voltage gain of 191.4 V/V, and an output resistance of 20 kohm. The input resistance indicated here is not the input resistance of the amplifier but, rather, the 1 ohm resistance shunting the input signal generator V_d. To determine the input resistance of the amplifier, we use the results of the AC analysis found in the output file,

 

                 

The input differential resistance to the amplifier is then found to be 10.44 kohm.

 

At this point it would be interesting to check if the results obtained by hand analysis agree with those computed by Spice. To determine whether this is the case, consider that the parameters of the small-signal model of each transistor, as computed by Spice, are as follows:

 

Using the expressions given in Table 6.1, we can expect that this amplifier will have a differential gain of 191.0 V/V and an input resistance of 10.44 kohm, values that are very close to those computed by Spice.

 

To further analyze the differential amplifier shown in Fig. 6.7, let us compute the common-mode voltage gain and common-mode input resistance using Spice.  One proceeds in exactly the same way that the differential-mode analysis was performed above using the .TF command, except that we change the source reference from Vd to Vcm as shown below:

 

.TF V(5,4) Vcm

 

The results of this analysis are:

 

Here we see that the common-mode voltage gain of this amplifier is exactly zero, which agrees with what one would expect given that the resistance in the two collectors are equal. Note that the input resistance is 492.7 Gohm rather than the theoretically expected value of infinity; a numerical artifact of Spice. We also note that the output resistance computed during this analysis is identical to that found during the previous analysis for differential-mode behavior.

 

 

Fig.6.9: Including a 200 k-ohm current source resistance in the differential amplifier of Fig. 6.7.

 

In the above analysis, we assumed that the differential pair was biased by an infinite-output-resistance current source I. In practise, an infinite-output-resistance current source cannot be achieved.  It is therefore imperative to repeat the above analysis including this current-source resistance, such as that shown in Fig. 6.9, and thus verify whether the expressions seen listed in Table 6.1 remain accurate. After all, in the development of most of these formulae, the current-source resistance was assumed infinite.  Here we shall assume a current-source resistance of 200 kohm.

 

We begin by modifying the Spice input file of the differential amplifier circuit in Fig. 6.8 by including a Spice statement for the current-source resistance as:

 

R 3 0 200k.

 

Submitting the revised file to Spice, results in the following pertinent small-signal differential-mode information:

 

From the above information we find that the differential-mode voltage gain is 190.8 V/V and the input differential resistance is 10.48 kohm. These results are very close to the previous case when the resistance of the current source was assumed infinite. The small differences in these small-signal results are due not directly to the presence of the source resistance R, but rather, because the addition of the source resistance alters the DC bias conditions of the two transistors.

 

We can proceed and modify the Spice deck to compute the common-mode input equivalent circuit, as was done before, and arrive at the following Spice results:

 

                 

Using the formulas given in Table 6.1, together with the small-signal model parameters for Q₁ and Q₂, we would obtain exactly the same results as those computed by Spice.

 

In each of the analyses performed above in this section, the Early effect was neglected. Let us now consider the situation where the Early voltage of each transistor of the differential pair is 100 V.  We shall maintain the current-source resistance at 200 kohm. The Spice input file for this situation is very similar to that shown in Fig. 6.8, with the model statement for each transistor modified according to:

 

.model npn_transistor npn ( Is=14fA Bf=100 VAf=100V ).

 

The results of both the differential-mode and common-mode analysis computed by Spice are:

 

                  

As is evident from these results, the differential-mode voltage gain is 182.7 V/V. The input differential resistance is calculated from the input current of 86.76 uA to be 11.52 kohm.  The common-mode voltage gain is, for all practical purposes, zero and the input common-mode resistance is 7.825 Mohm.

 

As a means of verifying the formulae given in Table 6.1 under the conditions of finite Early voltage, we also list the parameters of small-signal model computed by Spice below:

 

 

Substituting the above parameter values into the expressions for A_d, R_id, A_CM and R_iCM seen listed in Table 6.1, and approximating r_u by 10 r_o B, we obtain the following estimates of the amplifiers small-signal behavior:  A_d=182.8 V/V, R_id=11.52 kohm, A_CM=0 and R_iCM=6.75 Mohm.  Comparing these with those computed directly by Spice we see that we are in good agreement with all of them except R_iCM. The formula for R_iCM seen listed in Table 6.1 seems to underestimate the actual input common-mode resistance by about 14%. One possible source of error may be due to our approximation of r_u by 10 r_oB.

 

To summarize the results of this section we have compiled a list in Table 6.2 that compares the results computed directly by Spice to those computed using the formulae presented in Table 6.1. Furthermore, the rightmost column of this table lists the relative error in percent. As is evident, the results predicted by the formulae of Table 6.1 agree quite well with the results computed by Spice.  We can therefore conclude that when given good estimates of the small-signal model parameters, the formulae of Table 6.1 will predict quite accurately the small-signal differential and common-mode voltage gain of a differential amplifier and its corresponding input resistances.

 

 

 

 

6.3 Small-Signal Analysis of the Differential Amplifier: Asymmetric Conditions

 

The previous section assumed that many of the components in the differential amplifier of Figs. 6.7 or 6.9 were matched. In practise this is rarely the case. In this section we shall investigate the effect of asymmetric circuit conditions on amplifier behavior. Specifically, we are interested in observing the effect of variations, or mismatches, in collector resistances, transistor saturation (scale) currents, and transistor beta's on circuit behavior.

 

 

 

 

 

 

Fig. 6.11: Highlighting the input and output offset voltage of the differential amplifier shown in Fig. 6.9 subject to two different mismatches: (i) DR_C / R_C = 10% and (ii) DI_S / I_S = 5%.

 

Input Offset Voltage

 

To begin with, let us consider using Spice to determine the input offset voltage V_OS for the differential amplifier shown in Fig.6.9 assuming that each collector resistor undergoes a change of 5%, one positive, the other negative - resulting in a net change of 10%.  The Spice deck for this particular example is shown listed in Fig. 6.10. The analysis that we are requesting here is a DC sweep of the input differential voltage V_d. The range of our sweep is limited to be between -10 mV and +10 mV. The range of this sweep was determined by considering the formula for input offset voltage V_OS in Table 6.1. Given that DR_C/R_C = 10%, we can expect an input offset voltage having a magnitude of 2.6 mV.  Thus, we decided that our sweep should not exceed this by very much so that details, such as zero crossings, can be easily seen. Also included in the Spice deck is a .TF command to compute the common-mode voltage gain. As we can see from Table 6.1, the magnitude of the common-mode voltage gain A_CM will no longer be zero but approximately 2.5 mV/V.

 

The results of the DC sweep of the input differential voltage are shown in Fig. 6.11 for a 10% variation in the collector resistance.  Here we see that the transfer function curve v_o vs. V_d for this particular case (as there are other cases also shown in this figure), no longer passes through the origin. Instead, careful probing using the cursor feature of the PROBE facility of PSpice indicates that the output offset voltage for zero input is -472.7 mV. Conversely, the input voltage that corresponds to zero output voltage is +2.6 mV. This, then, is the negative of the input offset voltage (i.e., V_OS=-2.6 mV).  Interestingly enough, this corresponds exactly with the magnitude of the value predicted by the formula given in Table 6.1, as demonstrated above.

 

The common-mode voltage gain of the differential amplifier shown in Fig. 6.9 when the resistances in the two collectors differ by 10% is found by Spice to be:

 

Spice reports that the common-mode voltage gain is -2.042 mV/V. The magnitude of this value is reasonably close to the value that was predicted by the formula given in Table 6.1 at 2.5 mV/V.

 

 

 

 

 

We can repeat the above analysis and determine the effect that a 5% difference between the saturation currents I_S of the two transistors on the amplifiers input offset voltage. According to the formula given in Table 6.1, we can expect an input offset voltage of 1.3 mV.  The Spice input file for this particular case is provided in Fig. 6.12. The results of this analysis are shown superimposed in the previous graph of output voltage versus input differential voltage shown in Fig. 6.11. Using PROBE, we can determine that the output offset voltage is -236.4 mV and the input offset voltage is -1.3 mV.  Clearly our estimate of the input offset voltage agrees with that obtained with Spice.

 

Input Bias and Offset Currents

 

Mismatches in transistor B's result in different base currents, which, in turn, give rise to differences in the input bias currents to the amplifier.  To demonstrate this, consider altering the beta of Q₁ in the differential amplifier shown in Fig. 6.9 by -5%, and the beta of Q₂ by +5%. Specifically, for the example used above, the following model statements would be used to describe this situation to Spice:

 

.model npn_transistor1 npn ( Is=14fA Bf=95  VAf=100V )

.model npn_transistor2 npn ( Is=14fA Bf=105 VAf=100V )

 

These two statements can be used to replace the transistor model statements shown in the Spice file listed above in Fig. 6.12. Only the .OP command is necessary to see the effect of beta variation on input bias currents. The input bias current to the amplifier can be seen directly from the bias information generated by Spice for the two transistors.

 

Submitting this input file to Spice, results in the following DC operating point information for the two transistors:

 

               

As we can see, the base currents of Q₁ and Q₂ are not equal but differ by 450 nA. The input bias current to the differential amplifier is the average of these two currents, being 4.50 uA. Correspondingly, the offset current for the amplifier is 450 nA.  It is interesting to note that the formula provided in Table 6.1 generates the same value observed from the Spice simulation.

 

It is also interesting to note that according to the above list of small-signal model parameters, the BETADC of Q₁ is 105 and for Q₂ it is 116. This is quite different from the beta that was assigned on the transistor model statement. This difference arises from the definition Spice uses to compute BETADC. This parameter is computed by Spice as the ratio of the collector current to base current, which includes the effect of transistor output resistance. This is contrary to what beta represents in the model statement, i.e., short-circuit current gain.

 

 

Fig. 6.13: Various current mirror circuits: (a) A simple two-transistor current-mirror circuit. (b) A current mirror with base-current compensation.  (c) The Wilson current mirror circuit.  Various voltage sources are used to monitor different branch currents.

 

 

6.4 Current-Mirror Circuits

 

Current mirror circuits play a very important role in the design of IC current sources and current-steering circuits. A current mirror circuit consists of two or more transistors arranged in such a way that at least two transistors in the circuit have their bases and emitters connected together causing them to have equal v_BE's. There are many different current mirror circuits, each suitable for a different application. In the following we shall investigate several widely used current mirror circuits:  A two-transistor current mirror, herein referred to as the simple current mirror, the simple current mirror with base-current compensation, and the Wilson current mirror. Each of these is depicted in Fig. 6.13. In our investigation, we shall judge the behavior of these current mirrors based on their: (a) current gain accuracy, (b) output resistance, (c) minimum output voltage, and (d) input current range. Another important aspect of a current mirror circuit is their frequency response behavior. However, we defer discussion of this topic until the next chapter.

 

Current-Gain Accuracy:

 

The current-gain accuracy is an indication of how significantly the transistor base currents affect the operation of the current mirror circuit. The current gain is the ratio of the output current to the input current (I_out/I_in). In hand analysis, the current gain is usually evaluated with the Early effect neglected. Obviously, the more ideal a current mirror is, the closer this ratio is to unity.  For the simple current mirror, it has been shown by Sedra and Smith in Section 6.4 of their text that the current transfer ratio is 1 / {1 + 2/B}.  Conversely, the current transfer ratio for the remaining two current-mirror circuits shown in Fig. 6.13 can be approximated by 1 / {1 + 2/B²}.  For B=100, the simple current mirror has a current transfer ratio of 0.9803 A/A, whereas the other two mirror circuits have a ratio much closer to unity, of value 0.9998 A/A. Clearly, the simple current mirror is not very accurate.

 

With the aid of Spice, let us observe the current transfer ratio of the current-mirror circuits shown in Fig. 6.13 for various output voltage levels. We shall consider that the transistor is ideal except that it has a finite beta equal to 100.  Modeling the transistor in this way will enable us to compare the result generated by Spice with that calculated by hand.  Furthermore, we shall assume that each transistor of the mirror circuit is integrated on a common p-type substrate[1]. This is usually necessary in order to obtain matched devices without trimming. To ensure proper isolation, the pn junction formed by the substrate is reverse biased by connecting the substrate to the lowest potential in the circuit. In this particular case it would be the ground node (node 0). 

 

Let us consider creating Spice decks for each of the three current mirror circuits shown in Fig. 6.13. Since they are all quite similar, we shall only show the Spice deck for the Wilson current mirror.  The Spice input file describing the Wilson current mirror shown in Fig. 6.13(c) is provided in Fig. 6.14. The current mirror is supplied with a 1 mA input current source connected in series with a 10 V DC voltage source. This voltage source will enable us to monitor the input current with a Spice print or plot command. The value of this voltage source is not important, but its level is set typically at V_CC. The output terminal of the current mirror is connected to a 5 V DC voltage source. This voltage source acts to simulate various load conditions, as well as allowing us to monitor the output current of the mirror.  A DC sweep of the output voltage source is to be performed over a voltage range varying between 0 and 5 V in increments of 10 mV. A DC operating point analysis is also included in this Spice deck.

 

 

 

 

 

Fig. 6.17: Demonstrating the variations in the current transfer ratio of the Wilson current mirror over an input current range of 1 uA to 1 mA.

 

 

The results of the Spice analysis are shown in Fig. 6.15. Here we have plotted the output current normalized to the input current level of 1 mA. This operation was performed directly with the Probe facility of PSpice; unfortunately, Spice cannot perform this operation directly. Furthermore, with the cursor feature in Probe, we were able to determine from the graph in Fig. 6.15 that the base-current compensated mirror circuit and the Wilson current mirror circuit have nearly ideal current gain at 0.9998 A/A, whereas, the current transfer ratio for the simple current mirror was found to be 0.9804 A/A. All three sets of results agree with those found with the above hand calculations. We also notice from the current transfer characteristics shown in Fig. 6.15 that when the output voltage goes too low, the current transfer ratio drops significantly towards zero. This suggests that the current mirror has a limited range of operation. We shall have more to say about this in a moment when we consider a more sophisticated model for the transistor.

 

 

Output Resistance:

 

Based on the above analysis, one might be tempted to conclude that both the base-current compensated current mirror and the Wilson current mirror could be used interchangeably, given that their current gain accuracies are identical. Unfortunately, this is not the complete picture because the analysis above ignored the Early effect. As we shall see, the current gain ratio of a current mirror can be adversely affected by the Early voltage of a transistor, thus altering our conclusion about which circuit is the better current mirror.

 

To see this, let us modify the model statement for the npn transistor given in the Spice deck for each of the current mirror circuits by the following transistor model statement:

 

 

This Spice model was chosen here because it is representative of a typical small-sized integrated npn transistor found on a bipolar semi-custom analog transistor array manufactured by the Gennum Corporation [Gennum Data Book, 1991].  We shall maintain the input current to each mirror at 1 mA.  Typically, these transistors have a forward B_ac of approximately 90 at a bias current of 1 mA and an Early voltage in the neighborhood of 80 V.  One should bear in mind that current mirror circuits are generally called on to mirror a wide range of currents and not just a single current level. We shall address this issue below in the subsection entitled:  Input Current Range.

 

Submitting the revised decks for the three current mirrors to Spice, we see on completion of Spice in Fig. 6.16 a plot of the output current as a function of the output voltage.  As is evident, the behavior of the base-current compensated mirror circuit and the Wilson current mirror circuit now differ significantly. The Wilson current mirror has an i - v behavior that is very much independent of the output voltage, i.e., near zero slope.  On the other hand, both the base-current compensated mirror and the simple current mirror circuit have behavior that depends strongly on the output voltage.  To quantify the output resistance of each of these current mirrors, we can simply estimate this from the reciprocal of the slopes of the i - v characteristics of each mirror in their linear regions.  The Probe facility of PSpice will be used to obtain these slopes directly from the graph shown in Fig. 6.16. In the case of the simple current mirror and the base-current compensated mirror circuit, the output resistance is approximately 120 kohm.  The Wilson current mirror has a much higher output resistance of about 3.38 Mohm.

 

According to small-signal analysis, the output resistance of the simple current mirror, and the base-compensated current mirror, is simply the incremental resistance r_o of the output transistor (i.e., Q₂ in the circuits of Fig. 6.13(a) and (b)). However, according to the Early voltage and the transistor bias level, we estimate the incremental output resistance of this transistor to be V_A / I = 80 V / 1 mA = 80 kohm. (The results of a .OP command also confirm something very close to this value).  However, the Spice results indicate mirror output resistance of 120 kohm. Thus, some other effect must be playing a major role in increasing the output resistance of these current mirrors. A detailed investigation reveals that the integrated npn transistor used in our simulations has a 12 ohm resistor in series with the emitter (See the model statement for this transistor, given above). Although, the value of this resistor seems small, its effect on the output resistance of the current mirror is significant.

 

According to small-signal analysis, the resistance seen looking into the collector terminal of a transistor, denoted as r_o, with emitter resistance R_E is given by

(6.1)

 

In the situation described above for the two simple current mirror circuits, R_E of 12 ohm dominates the parallel combination of R_E||r_p because r_p is in the k-ohm range (i.e., r_p = B/g_m where g_m=I_C/V_T =1 mA /25 mV = 40 mA/V). Thus, according to Eqn. (6.1), the output resistance of this mirror is calculated as follows:

 

Clearly, our estimate of the output resistance is now in-line with that observed through computer simulation of 120 k-ohm.

 

For the case of the Wilson current mirror, its output resistance is given by

(6.2)

 

Attaching a numerical value of 90 to beta and 80 k-ohm to r_o, suggests that the output resistance is approximately 3.6 M-ohm. This, then, agrees quite closely with the value of the output resistance obtain through Spice of 3.38 M-ohm. The effect of the emitter resistance of each transistor is much greatly reduced due to the feedback action of the three-transistor loop which forms the Wilson current mirror circuit.

 

Based on the above observations of current mirror accuracy and output resistance, one would probably prefer the Wilson current mirror over the other two current mirror circuits shown in Fig. 6.13.

 

Minimum Output Voltage:

 

A drawback to the Wilson current mirror is that it requires a relatively high output voltage to operate in the linear region (1.14 V as opposed to 0.36 V for the other two circuits; see Fig. 6.16). When used as an active load in an amplifier configuration, the required voltage reduces the range of the output voltage swing and is therefore not desirable.

 

Input Current Range:

 

Current mirrors are expected to operate over a wide range of input current levels. In the above analysis, the various attributes of the different current mirror circuits were all evaluated at a single input current level of 1 mA. We shall now explore the performance of the Wilson mirror at various input current levels. Specifically, we shall evaluate its i - v characteristic at input current levels of 1 uA, 10 uA, 100 uA and 1 mA.

 

The Spice input file is a concatenation of four separate Spice files of the type seen listed in Fig. 6.12 with only the input current level altered according to the list described above. Once again, this allows us to plot all the results together on a single graph using the Probe facility of PSpice[2].

 

The results of these analyses are shown plotted in Fig. 6.17. Rather than plot the wide range of output currents on a single graph, where much detail would be lost with the scale used there, we instead plot the ratio of the output current to the input current, I_out / I_in, as a function of the output voltage. In this way, the same scale can be used for all input current levels and comparisons can easily be made.   The output resistance and the minimum output voltage can be derived from the data contained in this graph.

 

Over an input current range of 1 uA to 1 mA, the minimum output voltage can be seen to vary between 0.97 V and 1.2 V. In addition, over this same current range, we see that the slopes of I_out/I_in versus V_out curves are quite similar ranging between 210 to 292 uA/A per volt. Multiplying each one of these slopes by the corresponding input current level, we can convert these slopes to output conductances, and then to output resistances. On doing so, we find that for input current levels of 1 uA, 10 uA, 100 uA and 1 mA, the output resistance is 4.5 Gohm, 474 Mohm, 47 Mohm, and 3.4 Mohm, respectively. According to the output resistance formula given in Eqn. (6.2), with r_o=V_A/I, these values are in close agreement. 

 

 

 

Fig. 6.20: Comparing the I_out -- V_out behavior of the high-swing cascode current mirror with that of the Wilson current mirror. Each transistor is modeled after a small npn integrated transistor manufactured by Gennum Corp.

6.5 A High-Performance Current Mirror

One means of decreasing the minimum output voltage of a current mirror circuit whose output is derived from the transistor action of two transistors stacked one above the other is to reduce the voltage that appears at the base of the top transistor. In this way, the voltage that can appear at the output of the current mirror can be reduced without the top transistor saturating. The circuit shown in Fig. 6.18 accomplishes this task while maintaining the high output resistance through the cascode output. Here the voltage at the base of Q₃ and Q₄ is set by the external voltage source V_B. Correspondingly, the voltage at their respective emitters is one diode drop lower at about (V_B - 0.7) V. Clearly, then, if V_B is set between 1.0 V and 1.4 V, then the voltage at the emitter of Q₃ will fall somewhere between 0.3 V and 0.7 V. Thus, the voltage at the output terminal of the current mirror can be reduced to about 0.6 -- 1.0 V before Q₃ saturates (assuming Q₃ saturates at 300 mV). The other nodes in the circuit are set at levels that ensure that the other transistors are operating in their active regions. Transistor Q₅ is used to compensate for the base current of Q₁ and Q₂.

 

We shall now compare the behavior of this high-swing cascode current mirror circuit shown in Fig. 6.18 with the Wilson current mirror circuit shown in Fig. 6.13(c). The input to the current mirror will be set to 1 mA. The Spice input file for the high-swing cascode circuit is shown in Fig. 6.19. The Spice input file for the Wilson current mirror was already shown in Fig. 6.12. These two files will be concatenated together and submitted to Spice. A DC sweep of the output voltage beginning at 0 V and increased +5 V in increments of 10 mV is requested. The output current will then be plotted as a function of the output voltage.

 

The results of the Spice analysis are shown plotted in Fig. 6.20.  As can be seen, the output current from the high-swing cascode current mirror is much closer to the input current of 1 mA than in the case of the Wilson current mirror, i.e., the new circuit behaves in a more ideal fashion.  Although it is not readily apparent, the high-swing cascode current-mirror has an output resistance about twice that of the Wilson current mirror of about 6.5 Mohm. This output resistance value was determined with the aid of the cursor facility of Probe.  With regards to the minimum output voltage, we see that the high-swing cascode mirror circuit does indeed succeed at its primary objective of reducing the minimum output voltage below that of the Wilson current mirror, specifically it can operate with a voltage as low as 0.7 V.

 

 

 

6.6 Current-Source Biasing in Integrated Circuits

 

A typical approach for implementing a current source in IC technology is with current mirror circuits.  Figure 6.21 illustrates a simple current mirror arranged as a current source.  The input terminal of the current mirror is fed with a 942 kohm resistor connected to the positive power supply. This results in an input bias current of about 10 uA.  The output terminal is connected to a variable DC voltage source V_out to simulate the action of a load.  With the aid of Spice, we would like to determine the range of output voltages that can appear across the output port of the current mirror before the circuit ceases to operate as a current source.  In addition, we would also like to determine the Norton equivalent circuit representation of the output port of this current mirror.  We shall assume that the transistors are matched and are typical of the small npn integrated transistor variety manufactured by the Gennum Corporation.  As mentioned before, these transistors have a forward B_{ac} of approximately 90 at a bias current of 1 mA and an Early voltage in the neighborhood of 80 V.

 

The Spice input file describing this circuit is given in Fig.  6.22 where we have requested a DC sweep of the output voltage V_out beginning at 0 V and ending at 5 V.  The output collector current of Q₂ will then be plotted as a function of V_out.  The results of the Spice analysis are shown in Fig.  6.23.  For output voltages larger than 220 mV, we see that the output current remains relatively constant around 10 uA, increasing slightly at a rate of 0.13 uA per volt.  When the output voltage drops below 220 mV the output current drops quickly to zero amps.  Thus, the circuit operates as an effective current source provided the voltage appearing across the output remains above 220 mV. 

 

We can go further and characterize the output port of this circuit (assuming V_out > 220 mV) using the Norton equivalent circuit as shown in Fig. 6.24.  The level of the current source of 9.64 uA was found by finding the y-axis intersection of the extrapolated line joining the points on the i-v curve in Fig. 6.21 for V_out > 220 mV. The output resistance R_o=7.8 Mohm is simply the inverse of the slope of this line found using the Probe facility of PSpice. It is interesting to note that a similar resistance value is found through a .TF command with the output voltage biased somewhere inside the linear region of the current source.

 

It is also reassuring that hand analysis also generates a value of the output resistance for the current source that is in the neighborhood of that predicted by Spice, i.e., V_A / I = 80 V / 10 uA = 8 Mohm.  The effect of the emitter resistance of 12 ohm is not very significant at a current level of 10 uA because the transistor transconductance is small, i.e., 0.4 mA/V. Thus, from Eqn. (6.1}, we see that R_o does not change by much, e.g., R_o » [ 1 + (0.4 x 10^-3)(12)](8 x 10⁶) = 8.04 Mohm.

 

 

 

6.7 A CMOS Differential Amplifier with Active Load

 

Differential pairs, current mirrors and current sources are usually combined in MOS technology to form differential amplifiers. The current source is used to bias the differential pair and the current mirror acts as a large resistive load, thus providing large voltage gain.  An example of a CMOS differential amplifier is shown in Fig. 6.25. It is easily shown by hand analysis that the small-signal differential-mode voltage gain A_d of this stage is given approximately by A_d=- g_m1 (r_o2 || r_o4).  The corresponding common-mode voltage gain A_CM is, to a first order approximation, zero.  An expression for A_CM can be derived through a small-signal analysis, however, the result is usually too complex to be insightful.

 

Using Spice let us compute the differential-mode and common-mode voltage gains of the differential amplifier shown in Fig. 6.25, and thus, its common-mode rejection ratio (CMRR). We shall assume that the NMOS and PMOS transistors are fabricated with a CMOS process which can be characterized by the following Spice model parameters:  u_nC_OX=20 uA/V², u_pC_OX=10 uA/V², |V_t|=1 V, and lambda=0.04 V^-1.  For the time being we shall neglect the body effect of the transistor (i.e., gamma =0).  The length and width dimension of each transistor are listed in Table 6.3.   

 

 

 

 

 

The Spice description of this circuit is seen in Fig. 6.26.  The input is excited using the multiple-source arrangement depicted in Fig. 6.3. The first analysis that is requested is a DC sweep of the input differential voltage. The input differential voltage is swept between the two voltage supply limits (i.e., -5 V and +5 V) with a voltage increment of 50 mV.  This Spice analysis will then be followed by a DC sweep of the input common-mode voltage.  These two DC sweeps are necessary to locate the high-gain linear region of the amplifier.

 

The large-signal differential-input transfer characteristic of the CMOS amplifier as calculated by Spice is shown in Fig.  6.27. Here we see that the high-gain region is in the vicinity of 0 V.  However, pertinent details of this region are not clearly evident. We shall therefore re-run the Spice job using a more refined step size for V_d.  Specifically, we shall replace the DC sweep command given earlier in Fig. 6.26 by the following one:

 

.DC Vd -100mV +100mV 1mV

 

in Fig. 6.28. As is evident, this particular amplifier has an output DC offset of -3.5 V, or equivalently an input offset voltage of -50 mV. The linear region of this amplifier is between V_d=-10 mV and +65 mV. To obtain the small-signal differential gain in this region we can either estimate it from the slope of the line forming the amplifier linear region, or calculate it directly using a .TF command.  We shall choose the latter and enter the following .TF command in the Spice deck listed in Fig. 6.26:

 

.TF V(8) Vd

 

Most applications of differential amplifiers involve using them in a negative feedback configuration where the negative feedback forces the input offset voltage towards zero. It seems reasonable, therefore, to evaluate the differential gain and other characteristics with the transfer characteristics shifted so that the input referred offset voltage is zero. This can be achieved by applying a differential offset voltage of +50 mV to the input of the amplifier, i.e., modify the Vd source statement to:

 

Vd 101 0 DC 50mV.

 

On completion of Spice, we observe the expanded view of the CMOS amplifier high-gain region

 

An operating point command (.OP) is included to obtain information about the small-signal transistor model parameters.

 

The results of this analysis are then found in the Spice output file as follows:

 

 

The small-signal differential gain A_d is therefore 67.84 V/V.  To compare this quantity with that estimated by hand, we recall that A_d=- g_m1 (r_o2||r_o4) and

 

 

Thus, we can estimate g_m1 to be 86.6 mA/V by assuming I_D1=12.5 uA. Similarly, the output resistance of M₂ and M₄ is given by r_o= 1/lambdaI_D = 1 /(0.04 * 12.5 x 10^-6) which gives r_o=2 Mohm. Substituting these values into the expression for A_d results in A_d=86.6 V/V. When compared to the gain computed by Spice, our estimate here has a relative error of about 28%.  The reason for this error is largely due to the inaccuracy in estimating the drain bias current of each transistor, which in turn is the result of neglecting the Early effect.

 

A better estimate of the differential voltage gain can be obtained by using the bias point and thus the small-signal model parameters generated by Spice. These are listed below:

 

Using these, we compute the differential voltage gain A_d to be 66.4 V/V.  This is obviously much closer to the value computed by Spice at 67.84 V/V. 

 

Once again, we see that the small-signal gain expression derived by Sedra and Smith results in accurate gain prediction provided that good estimates of the bias points and hence of the small-signal model parameters of the transistors are used. 

 

In a similar fashion, the large-signal common-mode transfer characteristic of the amplifier is computed by replacing the DC sweep command in the Spice deck listed in Fig. 6.25 by one that sweeps the input common-mode voltage (V_CM) between -5 V and +5 V in 50 mV increments. The syntax of such a Spice statement would appear as follows:

 

.DC Vcm -5V +5V 50mV.

 

The revised Spice input file is then re-run and the effect of the input common-mode signal level on the output is then graphically displayed as in Fig. 6.29. Here we see that an input common-mode level ranging between -1.2 V and +3 V has little effect on the output signal.  For instance, a 1 V change in the input common-mode level causes a +180 mV change in the output voltage level and this is fairly consistent over the -1.2 V to +3 V range.  In other words, the common-mode gain A_CM is approximately 180 mV/V.  This also can be confirmed by using a .TF command. However, the estimate we obtained directly from the transfer characteristic shown in Fig. 6.29 is sufficiently accurate for our purposes here.

 

We should note here that once the common-mode range is known we must check to see whether the V_CM used to determine the large-signal differential characteristic is valid. Specifically, the large-signal differential characteristic computed earlier was obtained with V_CM=0 V. Fortunately, this lies within the common-mode range of the amplifier, and the differential characteristics obtained are therefore valid.

 

Combining the above estimate of the amplifier common-mode gain with the small-signal differential gain calculated earlier, we can compute the CMRR of the amplifier to be 67.84/180 x 10^-3=376.9 or 51.5 dB.

 

 

Fig. 6.29: The large-signal common-mode DC transfer characteristics of the CMOS differential amplifier shown in Fig. 6.25. An input differential offset voltage of +50 mV is applied to the amplifier input to ensure that the amplifier is biased in its linear region.

 

 

In the above analysis we ignored the presence of transistor body effect (i.e., gamma=0). In the following we shall repeat the above analysis with gamma=0.9 V^1/2. This requires that we alter the two MOS model statement provided in the Spice deck seen listed in Fig. 6.26 according to:

 

.model pmos_transistor pmos (kp=10u Vto=-1V lambda=0.04 gamma=0.9 )

.model nmos_transistor nmos (kp=20u Vto=+1V lambda=0.04 gamma=0.9 )

 

 

Repeating each Spice analysis suggested above, we then find the following differential-mode and common-mode voltage gains:

 

 

Comparing these results with those computed without the body effect taken into account, we see that both A_d and A_CM have changed little. The new CMRR then becomes 54.0 dB; an increase of 3.5 dB. 

 

 

 

6.8 GaAs Differential Amplifiers

 

In Fig. 6.30 we show a simple GaAs MESFET differential amplifier. Each MESFET is assumed to have minimum length (1 um in this case), but their widths vary and are specified alongside each transistor. Using PSpice (recall that Spice does not have a built-in model for MESFETs) we would like to compute the large-signal differential and common-mode DC transfer characteristics of this amplifier.  From this, we would like to determine its common-mode rejection ratio (CMRR).  We shall assume that the MESFETs are characterized by the following parameters:  B=0.1 mA/V² (for a 1-um wide device), V_t=-1.0 V, and lambda=0.05 V^-1.

 

The PSpice input file describing the circuit of Fig. 6.30 is listed in Fig. 6.31.  The inputs to this amplifier are assumed to be driven by the same multiple voltage source combination shown connected to the input terminals of the circuit in Fig. 6.3. To obtain the large-signal DC differential transfer characteristic we shall perform a DC sweep of the input differential voltage between the voltage limits of the two supplies (i.e., V_DD=+5 V and V_SS=-5 V) using a 50-mV step.  The input common-mode level V_CM shall be set equal to zero, as it is usually assumed that V_CM=0 V is an input that will be within the linear range of the amplifier.

 

On completion of the PSpice run the large-signal differential characteristic shown in Fig. 6.32 is found. Here we see a rather strange large-signal differential transfer characteristic. Instead of the output voltage swinging between the limits of the two power supplies, the output only swings between V_DD and ground - about one-half the normal voltage swing. We also notice that below the high-gain region of the amplifier, for input levels less than -1 V, the output level does not saturate, but instead begins to rise in a very linear manner. This suggests that this amplifier has a limited range of operation as a differential amplifier and therefore the input differential voltages must be restricted to be greater than -1 V. 

 

To obtain a better view of the linear region of this amplifier, we shall re-sweep the input differential voltage between -200 mV and +200 mV using a step size of 10 mV.  This requires that we alter the DC sweep command seen listed in the Spice deck of Fig. 6.31 to the following one:

 

.DC Vd -200mV +200mV 10mV.

 

Re-running the Spice analysis, we obtain the large-signal characteristic of the amplifier shown in Fig. 6.33. We see that the slope of the linear region is rather low, approximately 40 V/V, extending between 0 and +50 mV along the Vd axis. Such low voltage gain can be attributed to the rather low channel-length modulation factor of the MESFET (i.e., lambda=0.05 V^-1). As an estimate of the small-signal voltage gain in the high gain region of this amplifier, consider evaluating the small-signal transfer characteristic of this amplifier using the .TF command of PSpice around an input differential voltage of 25 mV. This point was chosen because it lies approximately midway between the two extremes of the linear region. To accomplish this, the following .TF command is placed in the PSpice deck shown in Fig. 6.31,

 

.TF V(3) Vd.

 

and the statement for the input differential voltage is modified to include the input +25 mV offset voltage according to

 

Vd 101 0 DC +25mV.

 

The results of this analysis are then found in the Spice output file as follows:

                   

Thus, we see that this particular amplifier has a somewhat low differential voltage gain of 47.2 V/V. This low gain can be attributed to the rather low output resistance of the amplifier (i.e., R_o=12.38 kohm).

 

According to a hand analysis the voltage gain and output resistance of the differential amplifier shown in Fig. 6.30 are given by A_d=g_m2R_o and R_o=(r_o2||r_o4). To check these expressions against the values computed by PSpice, we list below the parameters of the small-signal model of each MESFET as computed by Spice:

 

                 

                 

Substituting the appropriate parameter values into the expression for R_o and A_d above, we get R_o=9.45 kohm and A_d=66.25 V/V.  When comparing these with those generated by PSpice (i.e., R_o=12.38 kohm and A_d=47.2 V/V), we see that our small-signal hand calculations are not as accurate as we've seen previously for Bipolar and CMOS technologies.  The reason for this is that the large-signal model for the MESFET described by Sedra and Smith in section 6.9 is different than that used by PSpice.  As a result, the small-signal models are slightly different. This was not the case for the other technologies. Here it should be noted that the development of accurate GaAs MESFET models is still a subject of current research.

 

To determine the large-signal DC common-mode transfer characteristics of the MESFET amplifier in Fig. 6.30 we require that PSpice compute the output voltage as a function of the input common-mode voltage. This is easily achieved by modifying the PSpice input file given in Fig. 6.33 by replacing the DC sweep command given there by the following one:

 

.DC Vcm -5V +5V 100mV.

 

 

 

Fig. 6.34: The large-signal common-mode DC transfer characteristics of the MESFET differential amplifier shown in Fig. 6.30. The input differential offset voltage is set equal to 25 mV.

 

 

The input differential voltage V_d will remain offset by 25 mV in order to keep the amplifier in its linear region.

 

Re-submitting this revised input file to PSpice results in the large-signal common-mode transfer characteristic shown in Fig. 6.34.  Here we see that the common-mode characteristic consists of two almost linear regions with opposite signed slopes. For input common-mode voltages less than 0.5 V, the slope of the large-signal characteristic is negative with a magnitude estimated directly from the graph to be approximately 60 mV/V. Whereas, for input common-mode voltages larger than 0.5 V, the slope is estimated to be about +75 mV/V.  This rather unusual looking characteristic is a result of the low output resistance of MESFETs.

 

 

The input common-mode range of this amplifier is not obvious from the graph of the output voltage as a function of the input common-mode voltage shown in Fig. 6.34. But it can be determined by plotting the drain-source voltage V_DS of each MESFET as a function of the input common-mode voltage and compare it to the corresponding gate-source voltage minus the threshold voltage V_GS-V_t of each MESFET.  If V_DS ³ V_GS - V_t, then the MESFET is in its linear region, otherwise it is not.  The same PSpice input file can be used here without any modifications.  Recall that V_t is equal to -1 V.  The results, as further calculated and displayed by Probe, are shown in Fig. 6.35. From these results we see that the upper linear region of this amplifier is determined by MESFET B₂ entering the triode region for V_CM exceeding +3 V. The lower common-mode input range limit is determined solely by B₃: When the V_CM decreases below -4.7 V, this transistor enters triode. Therefore, the common-mode input range (CMR) for this amplifier is between -4.7 and +3 V. We note that a common-mode input voltage of 0 V is within the CMR of this amplifier and thus our previous calculation of differential-mode gain is valid.

 

To estimate the CMRR of this amplifier, we have to consider that when the input common-mode voltage is less than 0.5 V, the common-mode gain A_CM is about -60 mV/V and when the input common-mode voltage is greater than this amount, A_CM=+75 mV/V. Thus, using the worst-case situation, that is, when |A_CM| is largest, the CMRR is computed according to 20log(|A_d|/|A_CM|) to be 56 dB. 

 

 

6.9 A BJT Multistage Amplifier Circuit

 

Fig.  6.36 illustrates the circuit of a simple operational amplifier.  It consists of a cascade of several gain stages, two of which are made from differential pairs, and an output buffer.  The positive and negative input terminals to the amplifier are labeled as V⁺ and V^-, respectively.  The output terminal is denoted as V_o.  This particular operational amplifier circuit is presented in Examples 6.2 and 6.3 in Sedra and Smith, 3^rd Edition, and the results of their analysis will be summarized at the end of this discussion.  For the purpose of our analysis we shall assume that both the npn and pnp transistors have the following device parameters: I_S=18 fA, B_F=100, V_AF=100 V.  In addition, Q_6 is considered to have 4x the area of Q₉ and Q₃.  Using Spice, we would like to analyze this operational amplifier to determine its DC operating point which includes such information as output offset voltage, input bias currents and quiescent power dissipation.  In addition, we would also like to compute the small-signal differential voltage gain (in the high gain region of the amplifier) and the input common-mode range.

 

The Spice input file describing the make-up of the operational amplifier shown in Fig. 6.36, together with the multiple input voltage source arrangement discussed in section 6.1, is given in Fig. 6.37. Notice how Q₆ is specified to have 4x the area of Q₃ and Q₉. Two analysis request commands are listed here: a DC operating point command and a DC sweep command. Also seen listed in the analysis request section of the Spice deck is a .NODESET command. We shall explain its role in a moment.

 

The DC operating point command will, in addition to calculating the DC operating point of the circuit and the small-signal model parameters of each transistor, determine the power dissipated by the amplifier and the input bias currents.  The DC sweep command will be used to compute the large signal transfer characteristic of the amplifier by varying the DC level of the input differential source V_d between -15 V and +15 V in 100 mV increments.  The common-mode input V_CM will be held at zero volts throughout this DC sweep.  One may be tempted to add a small-signal transfer function analysis request here; however, this should be deferred until one sees the large signal differential transfer characteristic of the amplifier and can determine what DC input conditions are required so that Spice linearizes the amplifier around a known operating point.

 

To assist in the DC bias calculation, we include in the Spice input file a list of initial guesses of some node voltages in the circuit on the statement beginning with the keyword .NODESET.  If one provides good estimates of the circuit's DC bias point, then the time required by Spice to compute the DC bias point is reduced, but more importantly, including a .NODESET statement can make the difference between obtaining a DC solution and Spice terminating the analysis with no solution because of a non-convergence problem.  In older versions of Spice (version 2G6 and less), this step is usually required but newer versions, such as PSpice, have improved algorithms that have more robust convergence capabilities and rarely require guesses of the node voltages.

 

Submitting the Spice input file listed in Fig. 6.37 to Spice, results partially in the following DC analysis output:

 

 

From these results, we see that this amplifier has an output DC offset of +2.0677 V and input bias currents of 2.443 uA. The static power dissipated by the amplifier is 0.298 W. 

 

 

 

The collector currents of each device have also been computed by Spice; these can be seen listed in Table 6.4}. Also shown in this table are the currents computed by hand analysis (performed in Example 6.2 of Sedra and Smith, 3^rd Edition, assuming that beta >> 1 and ignoring the effect of transistor Early voltage). A third column has also been added showing the relative error (in percent) between these two currents.  As we can see, our hand estimates are reasonably close to the Spice results; the largest error in our estimates never exceeds 15.3%.  It is reassuring that reasonably good estimates of the DC collector currents of a complicated transistor circuit can be obtained by assuming ideal transistor behavior (i.e., beta  >> 1 and V_A=¥).

 

As a further check on this, we compiled a table of collector current values that were computed by Spice for various combinations of beta and V_A values (I_S remains at 18 fA) seen listed in Table 6.5. As one can see from this table, as beta and V_A approach infinity (i.e., the transistors become more ideal), the results approach those computed by the simplified hand analysis.

 

The large-signal differential transfer characteristic of this amplifier is displayed in Fig.  6.38.  This figure illustrates a view of the operational amplifier differential-input transfer characteristics between -15 V and +15 V. We recognize that the high-gain region of the amplifier is in the vicinity of 0 V; however, the resolution of the input voltage axis does not enable us to be certain of the boundaries of this high-gain region. Therefore, we shall re-run the Spice input file with the DC sweep command modified to include an expanded view of this high-gain region between -5 mV and +5 mV. This will require that we replace the DC sweep command in the previous Spice input file with the following one:

 

.DC Vd -5mV +5mV 10uV

 

The results of this analysis are on display in Fig. 6.39.  For inputs less than -2 mV, the output remains saturated at -15 V.  In the region between -2 mV and +1 mV, the output level change from -15 V to +10 V in a linear manner. Thus, the output voltage swing for this amplifier is bounded between -15 V and +10 V, a somewhat unsymmetrical voltage swing. The gain experienced by input signals in this linear region is approximately (10 - (-15) )/3 mV = 8.33 kV/V.  For inputs greater than +2 mV, the output levels' off at +8.5 V. We also see from these results that the input offset voltage V_OS for this amplifier is +230.0 uV.  This is of course the systematic offset of the amplifier and does not include the components due to various imbalances in the circuit (see Section 6.3).

 

Now that we know the boundaries of the high-gain region of the amplifier, we can use the transfer function command of Spice to compute the small-signal equivalent circuit parameters of the amplifier in its linear region.  But first we must decide on which point inside the linear region of the amplifier we should linearize about. Consider that, for most applications, a high-gain amplifier such as that shown in Fig. 6.36 is usually used in conjunction with negative feedback, and, as a result, the output potential of the amplifier is held close to ground potential (when no input signal is applied). Thus, the small-signal parameters of the amplifier should be obtained around the bias point that has the amplifier output voltage close to 0 V.  This is easily obtained by applying the negative of the amplifier input-referred offset voltage (i.e., -V_OS) across the input terminals of the amplifier. For the multiple source arrangement suggested in Fig. 6.3, this is easily achieved by setting V_d equal to -V_OS.

 

Returning to the example at hand, we shall modify the Spice statement for V_d according to

 

Vd 101 0 DC -230.0uV AC 1V.

 

A 1 V AC voltage has also been appended to this input voltage source statement. This will enable us to compute the AC current that circulates around the input terminals of the op-amp when a 1 V AC voltage signal is placed across its input terminals. From this we can then calculate the input differential resistance. For more details on this approach, the reader should refer back to Section 6.2.  Two analysis requests and a .PRINT command will be added to the input file in order to compute the small-signal parameters of the amplifier. These appear as follows:

 

The revised Spice input deck would then be re-submitted to Spice.  Some of the small-signal circuit parameters of this amplifier are then found in the output file as follows:

 

Here we see that the actual voltage gain in the amplifier linear region is 8.834 kV/V, very close to the 8.83 kV/V we estimated from the amplifier large-signal differential transfer characteristic displayed in Fig. 6.39.  The input resistance listed here is not for the amplifier but rather the resistance seen by the voltage source V_d.  In this particular case, this is the 1-ohm resistor connected in series with V_d.  In contrast, the output resistance of 133.7 ohm listed above is the actual output resistance for this amplifier. As an estimate of the input differential resistance of the amplifier, we use the results of the AC analysis, given below:

            

We see that current IM(EV+) is not quite the same as the current IM(EV-), but very close.  The reason for this difference lies in the amplifiers systematic offset that arises because of transistor Early voltage.  To determine the input differential resistance of the amplifier under such asymmetric conditions, we shall work with the average of the two base currents. In this way, we can eliminate the presence of the offset current in the resistance calculation.  Thus, we compute the average input base current to be 47.235 uA and therefore obtain the input differential resistance at 21.17 kohm.

 

The final analysis that we would like to perform on the operational amplifier shown in Fig. 6.36 is to determine its input common-mode range.  Consider performing a DC sweep of the input common-mode voltage V_CM between the rails of the two power supplies.  This requires that we replace the DC sweep command stated in the Spice listing in Fig. 6.37 by the following one:

 

.DC Vcm -15V +15V 0.1V

 

We also maintain an input differential offset voltage of V_d=-230 uV.  This is necessary to ensure that the amplifier is biased inside its linear region.

 

Re-submitting the revised input file to Spice, and observing the results in the output file, we obtain the amplifier large-signal common-mode transfer characteristic given in Fig. 6.40. As we can see from these results, the transfer characteristic is linear behavior over the range of V_CM between -14.3 V and +9.6 V. Outside these limits the characteristic becomes nonlinear. Thus, the input common-mode range for this amplifier is between -14.3 V and +9.6 V. We should also note that our large-signal differential transfer characteristic computed earlier is valid since it was obtained with an input common-mode voltage that falls within the input common-mode range of the amplifier. 

 

It is interesting to correlate the limits of the amplifier common-mode range with the mode of operation of the transistors in the front-end stage. Specifically, Sedra and Smith, 3^rd Edition,  mention in their text that the upper limit to the common-mode range is determined by Q₁ or Q₂ saturating, and that the lower limit is determined by Q₃ saturating.  We can determine when these transistors saturate by observing the voltage across the base-collector junction of each transistor and recalling that a transistor enters its saturation region when the base-collector junction becomes forward biased.  For instance, in Fig. 6.41 we display the voltage across the base-collector junctions of Q₁ and Q₃ and observe that Q₁ saturates when V_CM exceeds +9.6 V. In contrast, transistor Q₃ saturates when V_CM goes below -14.3 V.

 

Finally, as a summary of what we have learned about this op-amp through the application of Spice, Table 6.6 presents a collection of the results. Moreover, these results are compared with the simplified hand analysis performed by Sedra and Smith, 3^rd Edition, in examples 6.2 and 6.3 of their text.

 

 

6.10 Spice Tips

 

·      Spice can be conveniently used to compute both the large and small-signal characteristics of amplifier circuits. 

 

·      Inputs to differential amplifiers should consist of both a differential and common-mode level. An interesting arrangement of several voltage sources was given in this chapter (see Fig. 6.3, for instance), illustrating how the differential and common-mode levels can be independently adjusted. 

 

·      The high-gain linear region of an amplifier is located by first sweeping the input differential voltage v_d between the limits of the power supplies with the common-mode voltage V_CM equal to 0 V. This analysis is repeated with a reduced sweep range centered more closely around the high-gain region of the amplifier until a smooth transition through the high-gain region is achieved. Following this, one must check to see whether the input common-mode voltage of 0 V keeps the amplifier in its linear region.

 

·      Care must be exercised when computing the small-signal characteristics of an amplifier using Spice.  One should first decide what DC input conditions are required so that the amplifier is linearized around an appropriate DC operating point.  Generally, selecting an input differential offset voltage that forces the output voltage to 0 V will bias the amplifier at an operating point that is quite close to the operating point that results when some external negative feedback connection is made around the amplifier.

 

·      Any time an IC transistor is used in a circuit and its model is obtained from a library; the substrate connection must be defined.

 

·      Spice version 2G6 does not have a built-in model for the MESFET but PSpice does.

 

·      At times, Spice will not be able to compute the DC node voltages in a nonlinear circuit due to a DC convergence problem.  Sometimes this convergence problem can be alleviated by providing Spice with a set of initial estimates of the DC node voltages of the circuit. These are entered into the Spice deck using the .NODESET command of Spice.

 

·      The small-signal input resistance to a differential amplifier is computed using Spice by applying a known AC voltage across the input terminals of the differential amplifier and computing the AC currents that flows into the amplifier terminals. In many practical amplifier situations, these currents will not be equal. So, instead, the average of these two currents is used in the input resistance calculation.

 

6.11Bibliography

 

Staff,1990-1991 IC Data Book, Gennum Corporation, Burlington, Ontario, Canada.

 

6.12 Problems

 

6.1.     A BJT differential amplifier is biased from a 2-mA constant-current source and includes a 100-ohm resistor in each emitter. The collectors are connected to +10 V via 5 kohm resistors.  A differential input signal of 0.1 V is applied between the two bases.  Assume that the transistors are matched and have B=100 and I_S=14 fA.

 

(a)        With the aid of Spice, determine the signal current in the emitters i_e and the base-emitter voltage v_be for each BJT.

 

(b)        What is the total emitter current in each BJT?

 

(c)        What is the signal voltage at each collector?

 

(d)        What is the voltage gain realized when the output is taken between the two collectors?

 

 

Fig. P6.2                                                                                               Fig. P6.3

 

 

6.2.     For the circuit in Fig. P6.2 in which the transistors have high B, with the aid of Spice, determine the value of v₂. If the resistor R₁ is reduced to 2.5 kohm, what does v₂ become?

 

6.3.     Find the voltage gain and the input resistance of the amplifier in Fig. P6.3 using Spice assuming that B=100.

 

 

Fig. P6.4                                                                Fig. P6.5

 

 

6.4.         Find the voltage gain and the input resistance of the amplifier shown in Fig. P6.4 using Spice assuming B=100.

 

6.5.         The differential amplifier circuit of Fig. P6.5 utilizes a resistor connected to the negative power supply to establish the bias current I.

 

(a)   For v_B1=v_d/2 and v_B2=-v_d/2, where v_d is a small signal with zero average, find the magnitude of the differential gain, |v_o/v_d| using Spice.

 

(b)   For v_B1=v_B2=V_CM, find the magnitude of the common-mode gain, |v_o/V_CM| using Spice.

 

(c)   Calculate the CMRR.

 

(d)   If v_B1=0.1 sin(2p x 60t) + 0.005sin(2p x 1000t) volts, v_B2=0.1sin(2p x 60t) - 0.005sin(2p x 1000t) volts, plot the output voltage v_O using the .PLOT command of Spice.

 

6.6.         A BJT differential amplifier is biased from a 300 uA constant-current source and the collectors are connected to +7.5 V via 50 kohm resistors.  If the scale currents I_S of the two transistors have a nominal value of 10 fA but differ by 10%, what is the resulting input offset voltage?

 

6.7.         A BJT differential amplifier is biased from a 1 mA constant-current source and the collectors are connected to +15 V via a 10 kohm resistors.  If the B's of the two transistors are 100 and 200, what is the resulting input offset voltage?

 

6.8.         A BJT differential amplifier is biased from a 1 mA constant-current source and the collectors are connected to +15 V via a 10 kohm resistors.  If V_A of the two transistors are 100 and 200, what is the resulting input offset voltage?

 

 

Fig. P6.9                                                                Fig. P6.10

 

 

6.9.      For the simple MOS current mirror shown in Fig. P6.9, the devices have V_t=1 V, u_nC_OX=200 uA/V² and V_A=20 V. Further, V_DD=+5 V and V_SS=-5 V.  Using Spice, determine the Norton equivalent of this circuit when the input current is 100 uA. Also, what is the minimum voltage that can appear at the output of this mirror circuit while maintaining linear operation?

 

6.10.  For the cascode current mirror shown in Fig. P6.10, with V_t=1 V, u_nC_OX=200 uA/V² and V_A=20 V. Further, I_in=100 uA, V_DD=-5 V, V_SS=-5 V, and V_O=+5 V. Using Spice, determine the Norton equivalent of this circuit when the input current is 100 uA. Also, what is the minimum voltage that can appear at the output of this mirror circuit while maintaining linear operation?  Note that, although, the output resistance of this current source is much larger than the simple current mirror of Problem 6.9, its linear region is somewhat reduced.

 

(a)                                                                    (b)

 

Fig. P6.11

 

6.11.  In Fig. P6.11(a) we present a NMOS version of the Wilson current mirror. In (b), an additional transistor has been added to the Wilson circuit to make it more symmetrical.  Assuming that the MOS devices have Spice model parameters V_t=1 V, u_nC_OX=200 uA/V² and V_A=10 V, compare the input - output current behavior of these two mirror circuits.

 

 

(a)                                                                                (b)

 

Fig. P6.12

 

 

6.12.  Compare the input -- output current behavior (i.e., accuracy, output resistance, minimum output voltage and input current range) of the modified Wilson current mirror circuit shown in Fig. P6.12(a) with the behavior of the cascode current mirror shown in Fig.  P6.12(b).  Model each bipolar transistor after the Gennum Corporation integrated npn transistor described in section 6.4.  Which is the better current mirror?  How do these two current mirrors compare with the high-swing cascode current mirror circuit shown in Fig. 6.18.

 

 

(a)                                                                                (b)

 

Fig. P6.13

 

6.13.  Compare the input-output current behavior (i.e., accuracy, output resistance, minimum output voltage and input current range) of the modified Wilson current mirror circuit shown in Fig. P6.13(a) with the behavior of the cascode current mirror shown in Fig.  P6.13(b).  Assume that the MOS devices have Spice parameters:  V_t=1 V, u_nC_OX=200 uA/V² and V_A=70 V.  Which is the better current mirror?  Why are your conclusion here different than that found above in Problem 6.12 for the bipolar case?

 

6.14.  For the high-swing cascode current mirror shown in Fig. 6.18}, determine the maximum and minimum voltage limits of V_B that maintain the current mirror in its linear region. Assume that the transistors are modeled after the integrated npn transistors first suggested in Section 6.4.

 

 

(a)                                                                                (b)

Fig. P6.15

 

 

6.15.  In Fig. P6.15 we present two different ways of realizing a two-output current mirror circuit. The circuit in part (a) is a simple extension of the two transistor cascode current mirror circuit. The circuit in part (b) is a generalization of the Wilson current mirror circuit. Assuming that the transistors are modeled after the integrated npn transistor first suggested in Section 6.4, determine using Spice which is the better current mirror.  Base your judgment on current transfer function accuracy, output resistance, minimum output voltage and input current range.

 

6.16.  Compare your results found in Problem 6.15, with a multiple-output current mirror circuit created by extension of the high-swing cascode current mirror circuit of Fig. 6.18.  Which makes the better multiple-output current mirror?

 

 

(a)                                                                                                                        (b)

 

(c)

 

Fig. P6.17

 

 

6.17.  To improve the quality of a current mirror, an op-amp of gain A is sometimes used. In Fig. P6.17 we present several possible designs of a MOS current mirror circuit that incorporate an op-amp. For each design, determine using Spice how the op-amp gain affects the quality of the current mirror, i.e., (i) current gain accuracy, (ii) output resistance, (iii) minimum output voltage and (iv) input current range. Of the three circuits shown in Fig. P6.17, which behaves closer to the ideal current mirror? Assume the following parameter values for each MOSFET:  W=100 um, L=10 um, u_nC_OX=20 uA/V², |V_t0|=1 V and lambda=0.04 V^-1, and gamma=0.9 V^1/2.

 

Fig. P6.18

 

6.18.  In Fig. P6.18 we present a high-performance current mirror that is referred to in the literature as a regulated cascode current mirror. It is assumed that the two input current signals track one another. For an input current level of 1 uA, 10 uA, and 100 uA, determine (a) the current gain accuracy, (b) the output resistance, and (c) the minimum output voltage.  Assume the following parameter values for each MOSFET:  W=100 um, L=10 um, u_nC_OX=20 uA/V², |V_t0|=1 V, lambda=0.04 V^-1, and gamma=0.9 V^1/2.

 

Fig. P6.19

 

6.19.  Compute the voltages at all nodes and the currents through all branches in the circuit of Fig. P6.19 assuming beta is very large using Spice.  Compare this to the case when B=100.

 

6.20.  Compute the voltages at all nodes and the currents through all branches in the circuit of Fig. P6.19 assuming V_A is very large using Spice. (Setting V_A=0 in the model statement for the BJT is equivalent to setting V_A=¥).  Compare this to the case when |V_A|=100 V.

 

6.21.  Consider the effect of power-supply variation on the dc bias of the op-amp circuit of Fig. 6.36: If +15 V is lowered to +14 V, what is the effect on the amplifier parameters listed in Table 6.6? If, separately, -15 V is raised to -14 V, what is effect on the amplifier parameters in Table 6.6?

 

Fig. P6.22

 

6.22.  The differential amplifier in Fig. P6.22 is operated with I=100  uA, with devices for which V_A=200 V and B=100. What is the differential input resistance, output resistance, equivalent transconductance, and open-circuit voltage gain of this amplifier?

 

6.23.  In the multistage amplifier of Fig 6.36, 100-ohm resistors are introduced into the emitter lead of each transistor in the differential pair of the first stage (Q₁ and Q₂) and 25 ohm for each of the second-stage transistors (Q₄ and Q₅). With the aid of Spice, find the effect that these additions have on the input resistance, the voltage gain of the first stage, and the overall voltage gain.

 

6.24.  If, in the multistage amplifier of Fig. 6.36, the resistor R₅ is replaced by a constant-current source of the same value that flows through R₅ (i.e., maintain the same bias situation), calculate the overall voltage gain of this amplifier using Spice.

 

6.25.  A differential amplifier, utilizing JFETs for which I_DSS=2 mA, |V_P|=2 V, and V_A=100 V, is biased at a constant current of 2 mA. For drain resistors of 10 kohm, what is the gain of the amplifier for differential output? If the drain resistors have ±1% tolerance, what is the worst-case common-mode gain and CMRR?

 

 

Fig. P6.26

 

6.26.  The JFET circuit shown in Fig. P6.26 can be used to implement the current-source bias in a differential amplifier. With the aid of Spice, determine the output resistance if I_DSS=2 mA, V_P=-2 V, and V_A=100 V.

 

6.27.  A JFET differential amplifier is loaded with the basic BJT current mirror. The JFETs have V_P=-2 V, I_DSS=4 mA, and V_A=100 V. The BJTs have |V_A|=100 V and beta is large. The bias current I=2 mA. Find R_i, G_m, R_o, and the open-circuit voltage gain.

 

6.28.  An NMOS differential pair is to be used in an amplifier whose drain resistors are 100 kohm ±1%. For the pair, u_nC_OX=200 uA/V² and V_t=1 V. If the output is taken differentially, contrast the voltage gain and input offset voltage of this amplifier for a bias current of 100 uA and 200 uA.

 

Fig. P6.29

 

6.29.  For the amplifier circuit in Fig. P6.29 let W₁=100 um, W₂=500 um, W₃=50 um, and W₄=250 um. With the aid of Spice, compute the voltage gain of this circuit. Assume that the MESFETs are modeled with parameters:  V_t=-1 V, B=0.1 mA/V², and lambda=0.1 V^-1. Assume that the length of each device is 1 um.