c Copyright 2008. W. Marshall Leach, Jr., Professor, Georgia Institute of Technology, School of Electrical and Computer Engineering.

Current Mirrors
Basic BJT Current Mirror
Current mirrors are basic building blocks of analog design. Fig. 1 shows the basic npn current mirror. For its analysis, we assume identical transistors and neglect the Early eect, i.e. we assume VA . This makes the saturation current IS and current gain independent of the collectorbase voltage VCE . The input current to the mirror is labeled IREF . This current might come from a resistor connected to the positive rail or a current source realized with a transistor or another current mirror. The emitters of the two transistors are shown connected to ground. These can be connected to a dc voltage, e.g. the negative supply rail.

Figure 1: Basic current mirror. The simplest way to solve for the output current is to sum the currents at the node where IREF enters the mirror. Because the two transistors have their base-emitter junctions in parallel, it follows that both must have the same currents. Thus, we can write the equation IREF = IO + Solution for IO yields 2IO

IREF 1 + 2/ Note that this equation predicts that IO < IREF unless . Because the Early eect has been neglected in solving for IO , the output resistance is innite. If we include the Early eect and assume that it has negligible eect in the solution for IO , the output resistance is given by IO = rout = r02 = VA + VCE2 IO

For a more accurate analysis, we can include the Early eect in calculating the output current. Consider the circuit if Fig. 2. If the transistors have the same parameters, we can write VBE IC1 IC1 = IS0 exp IB1 = VT 0 VBE IO VCB2 IO = IS0 1 + exp IB2 = VA VT 0 (1 + VCB2 /VA ) 1

By taking the ratio of IO to IC1 , we obtain VCB2 IO = 1 + IC1 VA Note that the voltage VCB is used for the Early eect, not the voltage VCE . This agrees with the Ebers-Moll model for the BJT that is used in SPICE. In deriving the small-signal and T models, the derivations are simplied by the use of VCE . Thus VCE is used in most contemporary texts for the Early eect. The correct value is VCB which is used here.

Figure 2: Circuit for including the Early eect. Summing currents at the node where IREF enters the circuit yields IREF = IC1 + Thus IC1 is given by IC1 = It follows that IO is given by VCB2 (1 + VCB2 /VA ) IREF IO = 1 + IC1 = VA 1 + 2/ 0 Note that this equation predicts that IO can be larger that IREF . The output resistance is given above. Note that the eect of a nite is to reduce IO but the eect of the Early eect is to increase it. Because of the Early eect, the output current can be greater than the input current. One way of obtaining a better match between the input and output currents is to use series emitter resistors on the transistors. If the current in one transistor increases, it causes the voltage across its emitter resistor to increase, which causes a decrease in its base-emitter voltage. This causes the current to decrease, thus causing the two transistors to have more equal currents. A typical value for the emitter resistors might be 100 . With these resistors, Rte2 is no longer zero so that the output resistance is increased. It is given by rout = ric2 which can be much greater than r02 . IC1 IO 2IC1 = IC1 + + 0 0 (1 + VCB2 /VA ) 0 IREF 1 + 2/ 0

BJT Mirror with Base Current Compensation

Figure 3 shows the basic current mirror with a third transistor added. The collector of Q3 must be connected to a positive reference voltage, e.g. the positive supply rail, which biases it in the active 2

mode. If we neglect the Early eect and assume all transistors are identical, we can write IREF = IO + Solution for IO yields IO = 2IO (1 + )

IREF 1 + 2/ [ (1 + )]

Figure 3: Mirror with base current compensation. For a non-innite Early voltage and VCB1 = VBE3 VA , it can be shown that the output current is given by IREF (1 + VCE2 /VA ) IO = 1 + 2/ [ 0 (1 + 3 )] where VCE3 3 = 0 1 + VA

BJT Wilson Mirror

A Wilson current mirror is shown in Fig. 4. We neglect the Early eect in the analysis and assume the transistors to have identical parameters. The emitter current in Q3 is IO /. This current is the input to a basic current mirror consisting of Q1 and Q2 . This current is mirrored into the collector of Q1 by dividing by (1 + 2/). At the node where IREF enters the mirror, we can write IREF = Solution for IO yields IO = IO 1 + IO IO / + = IO + 1 + 2/ 2+ IREF (1 + ) / (2 + ) + 1/

The advantage of the Wilson mirror over the current mirrors examined above is that it has a much higher output resistance. This is caused by two positive feedback eects. To see how this occurs, suppose a test current source is connected between the mirror output and ground. If the source delivers current to the output node, the voltage increases. This causes a current to ow through r03 , causing the emitter voltage of Q3 and the base voltage of Q1 to increase. The increase in voltage at the emitter of Q3 causes its collector voltage to increase because Q3 is a common-base stage for an emitter input. Because Q1 is a common-emitter stage for a base input, the increase 3

Figure 4: Wilson mirror. in voltage at its base causes the collector voltage of Q1 and the base voltage of Q3 to decrease. Because Q3 is a common-emitter stage for a base input, the decrease in voltage at its base causes its collector voltage to increase. Thus there are two positive feedback eects which cause the collector voltage of Q3 to increase to a larger value. Because rout is the ratio of the collector voltage of Q3 to the current in the test source, it follows that the output resistance is increased.

BJT Low-Level Mirror

The circuit shown in Fig. 5 is a low-level current mirror. It can be used when it is desired to have a much lower output current than input current. For the analysis, we neglect the Early eect, assume identical transistors, and assume that . We can write VBE1 VBE1 IO RE IO = IS exp IREF = IS exp VT VT By taking ratios, we obtain IREF = exp IO IO RE VT

As an example, suppose IREF = 1 mA, VT = 25 mV, and IO = 50 A. It follows from this equation that RE = 1498 . The eect of this large a value of RE on rout is to make it greater than r0 . To calculate rout , we must know the small-signal Thvenin resistance Rtb2 looking out of the base of Q2 . Note that Q1 is a bjt connected as a diode and exhibits a small-signal resistance r01 k [rx1 / (1 + 1 ) + VT /IE1 ] ' VT /IE1 = 25 . This is in parallel with the small-signal resistance looking up into the IREF source. Thus an upper bound on Rtb2 is 25 . Let us assume r02 = 40 k, rx2 = 0, 2 = 0.995, and 2 = 199. It follows that rie2 = Rtb2 / (1 + ) + VT /IC2 = 497.6 . Thus rout is given by rout = ric2 = 40000 + 497.6k1498 = 159.5 k 1 0.995 1498/ (497.6 + 1498)

This equation cannot be solved for IO . If IREF and IO are specied, it can be solved for RE to obtain IREF VT RE = ln IO IO

This is larger than r02 by a factor of almost 4. 4

Figure 5: Low-level mirror.

BJT Transconductance Op Amp

An example application of the current mirror is the transconductance op amp. The circuit is shown in Fig. 6. The circuit consists of an input di amp and four Wilson current mirrors. For the analysis, we assume and VA for each bjt so that the output current from each mirror is equal to the input current. We assume that IABC splits equally between the emitters of Q1 and Q2 . Thus the total currents in Q1 and Q2 , respectively, are given by IABC IABC IABC + ic1 + ic2 = ic1 iC2 = 2 2 2 The latter expression for iC2 follows because ic1 + ic2 = 0. It follows from the mirrored currents that the output current is given by iC1 = io = 2ic1 If we neglect base currents and the Early eect, ic1 = ie1 = (vi1 vi2 ) /2re , where re = 2VT /IABC . Thus io is given by IABC (vi1 vi2 ) io = 2VT It can be seen that the transconductance gain is set by the current IABC . The gain can be varied by varying IABC . Because IABC 0, the circuit operates as a two-quadrant multiplier. The circuit symbol for the transconductance op amp is shown in Fig. 7. An example application of the transconductance op amp is a circuit which generates an amplitude modulated signal. The circuit is shown in Fig. 8. Let vi and IABC be given by vi = V1 sin c t IABC = IQ (1 + m sin m t)

where c is the carrier frequency, m is the modulating frequency, and m is the modulation index which must satisfy 1 < m < 1. The current io is given by IABC vi R2 io = 2VT R1 + R2 If we assume that CF is an open circuit at the operating frequencies, the current io must ow through RF . Because the second op amp forces the voltage at its inverting input to be zero, the output voltage is given by vo = io RF = IQ RF V1 R2 IABC vi R2 RF = sin c t (1 + m sin m t) 2VT R1 + R2 2VT R1 + R2 5

Figure 6: Transconductance op amp.

Figure 7: Transconductance op-amp symbol.

Figure 8: AM modulator.

Figure 9: AM modulated waveform.

An example waveform for vo for the case m = 0.6 is shown in Fig. 9 The purpose of the voltage divider formed by R1 and R2 at the input to the transconductance op amp is to attenuate the input signal so that it does not overload the input dierential amplier. Ideally, the peak voltage should not exceed about 50 mV to prevent distortion in the output waveform. The resistor R1 kR2 is in series with the noninverting input so that both inputs to the dierential amplier see the same source resistance. Typically, R1 and R2 are chosen so that R1 kR2 is no larger than about 100 . The capacitor CF is necessary for proper high-frequency response. The capacitor must be chosen experimentally to prevent the gain of the circuit from peaking up at some high frequency where the circuit can oscillate. A method of determining the optimum value of CF is to drive the circuit with a square wave for vi and m = 0 so that IABC = IQ , i.e. a dc current. The capacitor can be experimentally adjusted to minimize any ringing on the output waveform without degrading the bandwidth of the circuit.