The implementation of the current mirror circuit may seem simple but there is a lot going on. The simple two transistor implementation of the current mirror is based on the fundamental relationship that two equal size transistors at the same temperature with the same V_GS for a MOS or V_BE for a BJT have the same drain or collector current. To best understand this important circuit building block and how it makes use of this relationship we need to deconstruct the circuit into input and output sections and examine each in turn.

A current mirror is a circuit block which functions to produce a copy of the current flowing into or out of an input terminal by replicating the current in an output terminal. An important feature of the current mirror is a relatively high output resistance which helps to keep the output current constant regardless of load conditions. Another feature of the current mirror is a relatively low input resistance which helps to keep the input current constant regardless of drive conditions. The current being 'copied' can be, and often is, a varying signal current. The current mirror is often used to provide bias currents and active loads in amplifier stages.

The ideal block level concept of the current mirror is shown in figure 11.1. Given a current source as the input, the input section of the current mirror looks like a virtual short circuit and reflects (swaps the direction of flow) this current to produce a current sink (the current exiting the mirror); as a result, we obtain a current sink (figure 11.1a). Conversely, given a current sink as the input, the current mirror reflects this current to control current source (figure 11.1b); as a result, now we obtain a current source. We can generalize this basic current mirror structure with this first observation: A current mirror consists of a low impedance input stage connected to a high impedance output current stage.

Conceptually, an ideal current mirror is simply an ideal current amplifier with a gain of -1. In Chapter 8 we explored the transistor and you should recall that the BJT device is a current amplifier of sorts (current controlled current source) in that the collector current is β times the base current. The problem with using this feature directly is that β is not a well controlled value from device to device and can vary with changes in temperature. Accurate current amplifiers are difficult to directly implement using conventional transistor amplifier configurations which are typically voltage amplifiers. For example the MOS transistor is generally modeled as a voltage controlled current source and can not be used directly as a current amplifier. The use of feedback and the current to voltage relationship of two terminal elements such as a resistor are most often used when manipulating currents as signals. Because in a current mirror the input and output are currents, it is easier to convert the input to a voltage first and then convert a voltage back to a current at the output.

It should be noted that these two stages of the current mirror may have a linear relationship (for example where V_OUT = I_INR and I_OUT = V_IN/R) like a resistor. In figure 11.1.1 we see the classic operational amplifier implementation of the current to voltage converter explored back in Chapter 4 section 2. The virtual ground at the negative input of the op-amp provides a very low input resistance. These circuits use the linear relationship between the current in resistor R1 and the voltage across the resistor. However, this linear relationship is not necessarily required. Any element or combination of elements could be used such as the V_BE or V_GS of a transistor as in (b) if the output voltage was taken at the gate of M1 (output of the op-amp).

Similarly, as an output stage we have the operational amplifier implementation of the voltage to current converter from section 1 of Chapter 4 in figure 1.1.2. Here the input voltage is forced across resistor R1 such that the resulting current in R1 flows through transistor M1.

Implementing the block diagram of the current mirror shown in figure 11.1 follows directly from these voltage / current converter stages from Chapter 4, if we connect the output of the I to V converter in figure 11.1.1(b) to the input of the V to I converter in figure 11.1.2. With the two resistors being equal, I_OUT would be the mirror image of I_IN. Note that the second op-amp is not actually necessary because the gates of the two NMOS transistors can be tied directly to each other with the same result. Remember that two equal size transistors at the same temperature with the same V_GS (or V_BE for a BJT) have the same drain current. This is an important simplification of the current mirror concept.

The converters might consist of non-linear devices having whatever I to V characteristics that may include another physical quantity (such as temperature); the only requirement is that the characteristics be the inverse of each other. For example, if the input I to V stage implements a function v = f(i) and the output stage implements the inverse function i = f -1(v) the total input to output transfer function is v = f(i) = f(f -1(v)) . We can make a second observation: A current mirror consists of two connected stages with inverse transfer functions of each other.

The converter circuits in figures 11.1.1 and 11.1.2 are rather complicated and require as many as two operational amplifiers. A much simpler implementation would be better.

We would like a simple configuration where the active element, a single transistor, serves as the desired current-to-voltage converter. However, the transistor is a unidirectional device, where for the BJT the base emitter voltage controls the collector current or for the FET the gate source voltage controls the drain current. Producing the opposite where the collector current controls the V_BE is not possible in the conventional use of the device as a common emitter amplifier. Referring back to Figure 11.1.1, the solution is to incorporate negative feedback. In this case that means making the transistor adjust its base emitter or gate source voltage, V_BE or V_GS, so that the collector or drain current is I_IN = (V₁-V_BE)/R. For this purpose, we simply connect the collector to the base or gate to drain or “diode connect” the transistor. This classic “diode” connection results in 100% parallel negative feedback (figure 11.2). As a result, with this diode connected transistor, the collector current serves as the input quantity while the base-emitter voltage V_BE serves as the output quantity with the logarithmic transfer function of the base emitter junction. Similarly, a diode connected enhancement mode MOS FET (gate tied to drain) will serve as a similar current to voltage converter with V_GS as the output quantity rather than V_BE.

ADALM1000 BJT as a diode lab activity
ADALM1000 MOS as a diode lab activity
ADALM2000 BJT as a diode lab activity
ADALM2000 MOS as a diode lab activity

A bipolar transistor can be driven by a voltage or by a current. If we consider the base emitter voltage, V_BE, as the input and the collector current, I_C, as the output (figure 11.3), we can think of a transistor as a non-linear voltage-to-current converter having an exponential characteristic. The base can be directly driven by the voltage output of the I-to-V converter we just discussed. The collector provides the output terminal of our simple current mirror: The output V to I converter stage of the simple current mirror is just a transistor acting as a non-linear (exponential for BJT) voltage-to-current converter. Again if a MOS transistor were used for the input stage the output stage would be a MOS transistor with the gate serving as the voltage input and the drain as the current output.

The final step is to connect the output of the input stage (the base emitter junction of Q₁) to the input of the output stage (the base emitter junction of Q₂) to build the basic BJT current mirror circuit (figure 11.4). At this point we will concentrate on the issues involved with the BJT current mirror and pick back up with the MOS current mirror in section 11.6.

If transistors Q₁ and Q₂ in figure 11.4 are identical (that is have the same size emitter and thus equal I_S) the input current to output current ratio or gain is ideally 1. There are often occasions when a gain other than one is required. When building circuits from discrete devices only simple integer ratios are possible while in microelectronic integrated circuits it is possible to make transistors with arbitrary emitter areas, A. However, even in integrated circuits the best design practice is to use identical unit size transistors when making current mirrors.

If on the input side of the mirror we connect N identical devices in parallel and connect M devices in parallel on the output side, the gain of the mirror will be M/N. In figure 11.4.1 we see an example where 2 (N=2) devices are connected together on the input and 3 (M=3) devices are connected together on the output. The mirror gain will thus be 3/2 or 1.5. Since all five transistors share the same V_BEvoltage, their collector currents, I_C, will all be equal. The input current I_IN splits equally in Q₁ and Q₂ such that:

, or

There are three primary error sources that make the simple 2 transistor mirror less than ideal. The first is the mirror gain. Ideally I_OUT should exactly equal I_IN. There are systematic and random factors that make this not the case. The second is the incremental output resistance, which determines how much the output current varies with the voltage seen at the mirror output. The third limitation is the minimum voltage drop across the output leg of the mirror necessary to maintain the desired output current. This minimum voltage, called output compliance, is determined by the need to keep the output transistor of the mirror in the active region. There are also a number of secondary performance issues with mirrors, for example, temperature stability and frequency response.

An error source in this simple BJT based current mirror is that the transistors Q₁ and Q₂ (figure 11.4) each remove a base current I_B from the input current I_IN. As a result, the output current is smaller than the input current:

As was already discussed, current mirrors can just as easily be made from MOS FET transistors. The I-to-V and V-to-I functions are different but of course are still the inverse of each other. A significant advantage of the MOS current mirror is the lack of base current induced error that BJT current mirrors suffer. There are methods to correct or compensate for the base current in BJT current mirrors which will be discussed in detail in later sections of this chapter.

It is necessary to keep the output (BJT) transistor out of saturation, V_CB = 0 V. Or from another perspective, not allow the collector base junction to forward bias. That means the lowest output voltage that results in the correct output current, the compliance voltage, is V_OUT = V_CV = V_BE under bias conditions with the output transistor at the output current level I_C and with V_CB = 0 V or, restating the V_BErelation from earlier:

Where V_T is the thermal voltage and I_S is the reverse saturation current.

The V_CB of Q₁ in the mirror is zero. If V_CB is greater than zero in the output transistor Q₂, the collector current in Q₂ will be somewhat larger than Q₁ due to the Early effect. In other words, the mirror has a finite output resistance given by the r_O of the output transistor, namely:

Where:
V_A is the Early voltage
V_CB is the collector-to-base voltage

As we learned in an earlier chapter, the inclusion of emitter degeneration resistors (R_E1 and R_E2 in figure 11.5) can increase the effective collector impedance seen at the mirror output. In order for the mirror gain to remain equal to 1, R_E1 must of course equal R_E2. The added voltage drop across the emitter resistor R_E2 (I_OUT x R_E2) adds to the minimum allowable output voltage (see section 11.5.2).

While resistors could also be added to the sources in an MOS based mirror, it is often more effective to simple increase the channel length, L, of the transistors. The longer the channel the less it is affected by the channel length modulation due to the increasing drain voltage.

It is important to note that the inclusion of emitter resistors does not reduce the reduction in the output current I_OUT caused by the finite beta of Q₁ and Q₂. A compensating voltage can be inserted by including resistor R_B, of the correct value, as shown in figure 11.6.

For V_CB close to zero (that is, neglecting base-width modulation errors) the necessary value for R_B is:

For example, using A = 2 ( a mirror gain of 2 ), I_IN = 1mA, so r_E = 26? and R_E1 = 500? (introducing about 500mV of degeneration) R_B should be 1578? if β is enough larger than A so that factor can be ignored, or 1626? if it is included and beta is 100. Clearly, this compensation method becomes unpredictable when β is small and is comparable to A and will never be precise because β₁ and β₂ are not in general equal. However, this technique can be of practical utility in many cases.

The simple current mirror can, obviously, also be implemented using MOSFET transistors, as shown in figure 11.7. We know that transistor M₁ is operating in the saturation region because V_DS is greater than or equal to V_GS. Transistor M₂will also be in saturation so long as the output voltage is larger than its saturation voltage. In this simple configuration, the output current I_OUT is directly related to I_IN.

The drain current of a MOSFET I_D is a function of both the gate to source voltage and the drain to gate voltage of the MOSFET given by I_D = f (V_GS, V_DG), a relationship derived from the functionality of the MOSFET device. In the case of transistor M₁ of the mirror, I_D = I_IN. Input current I_IN is a known current, and can be provided by a resistor as shown in the figure, or by a threshold-referenced or self-biased current source to ensure that it is constant, independent of voltage supply variations.

Using V_DG=0 for transistor M₁, the drain current in M₁ is I_D = f (V_GS,V_DG=0), so we find: f (V_GS, 0) = I_IN, implicitly determining the value of V_GS. Thus I_IN sets the value of V_GS. The circuit in the diagram forces the same V_GS to apply to transistor M₂. If M₂ also is biased with zero V_DG and provided transistors M₁ and M₂ have good matching of their properties, such as channel length, width, threshold voltage etc., the relationship I_OUT = f (V_GS,V_DG=0 ) applies, thus setting I_OUT = I_IN; that is, the output current is the same as the input current when V_DG=0 for the output transistor, and both transistors are matched.

The drain-to-source voltage can be expressed as V_DS=V_DG + V_GS. With this substitution, the Shichman-Hodges model provides an approximate form for function f(V_GS,V_DG):

Where:
Kp is a technology related constant associated with the transistor,
W/L is the width to length ratio of the transistor,
V_GS is the gate-source voltage, V_th is the threshold voltage, V_DS is the drain-source voltage
λ is the channel length modulation constant

Output resistance

Because of channel-length modulation, the mirror has a finite output resistance given by the ro of the output transistor, namely:

Where:
λ = channel-length modulation parameter
V_DS = drain-to-source bias.

Compliance voltage

To keep the output transistor resistance high, V_DG = 0 V. That means the lowest output voltage that results in correct mirror behavior, the compliance voltage, is V_OUT = V_CV = V_GS for the output transistor at the output current level with V_DG = 0 V, or using the inverse of the f- function, f^-1:

For the Shichman-Hodges model, f^-1 is approximately a square-root function.

Figure 11.8 shows a mirror where the simple wire connecting the collector of Q₁ to its base is replaced by an emitter follower buffer. This improvement to the simple current mirror is referred to as an emitter follower augmented mirror. The current gain (?_Q3) of the emitter follower buffer stage (Q₃) greatly reduces the gain error caused by the finite base currents of Q₁ and Q₂.

One thing to note that is different in this mirror configuration vs. the simple two transistor mirror is that the Collector-Base voltage, V_CB, of Q₁ is no longer zero. It is equal to the V_BE of Q₃. Given the effect of the finite output resistance (Early effect ) the output current I_OUT in Q₂ will most closely match I_IN when the collector voltage of Q₂ is the same as that of Q₁ which is 2XV_BE above the common voltage. Also note that when driven by a resistor, like R₁, I_IN will now be (V₁-V_BE1-V_BE2)/R₁.

Another consequence of adding the emitter follower buffer is, in general, a loss in the frequency response of the mirror. Transistor Q₃ is potentially operating at a very small current of 2I_B. If there were to be a significant capacitance to ground at the base connection common to Q₁ and Q₂ the current available to discharge this current will also be small equal to 2I_B. But the current available to charge this node is potentially equal to ?_Q3I_IN which is very much larger than 2I_B. This asymmetry in the charging vs. discharging current available for this node in the current mirror can lead to very undesirable response to fast changes to I_IN.

A Wilson current mirror or Wilson current source, named after George Wilson, is an improved mirror circuit configuration designed to provide a more constant current source or sink. It provides a much more accurate input to output current gain. The structure is shown in figure 11.9.

We will be making the following two assumptions. First, all transistors have the same current gain ?. Second, Q₁ and Q₂are matched, so their collector currents are equal. Therefore, I_C1 = I_C2 (= I_C) and I_B1 = I_B2 (= I_B) .

The base current of Q₃ is given by,

The emitter of Q₃ current by,

Looking at figure 11.9, it can be seen that _IE3 = I_C2 + I_B1 + I_B2 Substituting for I_C2, I_B1 and I_B2, I_E3 = I_C + 2I_B

so,

Substituting for I_E3

rearranging,

The current through R₁ is given by, I_R1 = I_C1 + I_B3

But, I_C1 = I_C2 = I_C

Substituting for I_C and since we get,

Therefore,

And finally,

From the above equation we can see that if

And the output current (assuming the base-emitter voltage of all transistors to be 0.7 V) is calculated as,

Note that the output current is equal to the input current I_R1 which in turn is dependent on V₁ and R₁. If V₁ is not stable, the output current will not be stable. Thus the circuit does not act as a regulated constant current source.

In order for it to work as a constant current source, R₁ must be replaced with a constant current source.

Advantages over other configurations

This circuit has the advantage of virtually eliminating the base current mismatch of the conventional BJT current mirror thereby ensuring that the output current I_C3 is almost equal to the reference or input current I_R1. It also has a very high output impedance due to the negative feedback through Q₁ back to the base of Q₃.

Adding a fourth transistor to the simple Wilson current mirror in figure 11.10, we have the modified or improved Wilson mirror. The improved input to output current accuracy is accomplished by equalizing the collector voltages of Q₁ and Q₂ at 1 V_BE. This leaves the finite ? and voltage differences of each of Q₁ and Q₂ as the remaining unbalancing influences in the mirror.

A Widlar current source is a modification of the basic two-transistor current mirror that incorporates an emitter degeneration resistor for only the output transistor, enabling the current source to generate low currents using only moderate resistor values. This circuit is named for its inventor, Robert Widlar, and was patented in 1967.

The Widlar circuit may be used with bipolar transistors or MOS transistors. An example application is in the now famous uA741 operational amplifier, and Widlar used the circuit in many of his designs.

Figure 11.11 is an example Widlar current source using bipolar transistors, where the emitter resistor R₂ is connected in series with the emitter of output transistor Q₂, and has the effect of reducing the current in Q₂ relative to Q₁. The key to this circuit is that the voltage drop across the resistor R₂ subtracts from the base-emitter voltage of transistor Q₂, thereby reducing the collector current compared to transistor Q₁. A simulation plot showing this reduction in I_C2 is presented in figure 11.12.

This observation is expressed by using KVL around the base emitter loop of the circuit in Figure 11.11 as:

Where ?₂ is the beta of the output transistor, which may not be the same as that of the input transistor, in part because the currents in the two transistors are very different. The variable I_B2 is the base current of the output transistor, V_BE refers to base-emitter voltage. If we neglect the effect of finite ? and use the V_BE equation we can obtain a useful formula for the output current:

where V_T is the thermal voltage, I_IN = I_C1 and I_OUT = I_C2.

Suppose we want to create a 100uA output current from a 300uA input current as in the simulation plot of figure 11.12. V_T is 26mV times ln(3) is 28.5mV. 28.5mV divided by 100uA is 285 ohms. This equation makes the approximation that the currents are both much larger than the saturation currents I_S1, I_S2, an approximation valid except for very low current levels. In the following the distinction between the two scale currents is dropped, although the difference can be important, for example, if the two transistors are designed with different emitter areas.

A test current I/_x is applied at the output, and the output resistance is then R_O = V_x / I_x.

An important property of a current source is its small signal incremental output impedance, which should ideally be infinite. The emitter degeneration resistance introduces local current feedback for transistor Q₂. Any increase in the current in Q₂ increases the voltage drop across R₂, reducing the V_BE for Q₂, thereby countering the increase in current. This feedback means the output impedance of the circuit is increased, because the feedback involving R₂ forces use of a larger voltage to drive a given current.

Output resistance is found using a small-signal model for the circuit, shown in Figure 11.13. The transistor Q₁ is replaced by its small-signal emitter resistance r_E because it is diode connected. In a diode-connected transistor the collector is short-circuited to the base, so the transistor collector-base junction has no time-varying voltage across it. As a result, the transistor behaves like the base-emitter diode, which at low frequencies has a small-signal circuit that is simply the resistor r_E = V_T / I_E, with I_E the DC Q-point emitter current. The transistor Q₂ is replaced with its hybrid-pi model. A test current I_x is attached at the output.

Using the figure, the output resistance is determined using Kirchhoff's laws. Using Kirchhoff's voltage law from the ground on the left to the ground connection of R₂:

Rearranging:

Using Kirchhoff's voltage law from the ground connection of R₂ to the ground of the test current:

or, substituting for I_b:

Eq. 4

According to Eq. 4, the output resistance of the Widlar current source is increased over that of the output transistor itself (which is r_O) so long as R₂ is large enough compared to the r_p of the output transistor. (Large resistances R₂ make the factor multiplying r_O approach the value (? +1).) The output transistor carries a low current, making r_p large, and increase in R₂ tends to reduce this current further, causing a correlated increase in r_p. Therefore, a goal of R₂ ? r_p can be unrealistic, and further discussion is provided below. The resistance R₁ / r_E usually is small because the emitter resistance r_E usually is only a few ohms.

ADALM1000 Lab Activity 6, BJT Current Mirror
ADALM1000 Lab Activity 6M, MOS Current Mirror

ADALM2000 Lab Activity 6, BJT Current Mirror
ADALM2000 Lab Activity 6M, MOS Current Mirror

When designing a circuit it is important to take into account the wide variation in certain device values from one to another. A central objective of the designer is to desensitize the circuit to these variations to produce a circuit which meets the specifications across all possible conditions. One aspect of design which is common to nearly all circuits is the establishment of stable bias or operating point levels. This seemingly minor portion of a design can provide the most challenging and interesting circuit problems. Many bias generators are centered around the generation of currents to operate the core of the circuit. Current generated from simple resistors and diodes or diode connected transistors connected across the power supply will vary approximately proportional to the variation of the supply voltage. This variation in the resulting bias current is frequently undesirable.

This is to introduce another kind of current mirror, actually a stabilized current source, which has an output which had been desensitized to variation in input current. To understand this configuration, it is helpful to examine the behavior of a zero gain amplifier. A NMOS version is shown in figure 11.14 but PMOS, NPN or PNP transistors will just as well function in this configuration

Remembering back to the previous explanation of the common emitter/source amplifier (figure 11.14(a)), the gain is a function of the drain (or collector) current and the load resistor. For a given drain current, if the drain resistor R_L is set equal to r_s then the gain A will be minus 1.

Alternatively. if the gate is held fixed at the same DC bias level which produces an identical drain current as in (a) and an input signal is applied to the top of load resistor R_L (figure 11.14(b)), then the gain will be plus 1. That is if the drain/source output impedance of the transistor is neglected. If we now connect the gate to the top of resistor R_L as in figure 11.14?, the net gain superimposing both paths will be 1 - 1 = 0.

In figure 11.15 we have an NPN transistor biased into conduction with a collector voltage V_C which is less than the base voltage V_BE by the thermal voltage V_T= kT/q, (equal to Ic times R_L) and is essentially constant with input voltage changes applied from V_IN. The voltages seen at V_BE and V_C are plotted vs. the applied voltage at V_IN in figure 11.14. As we can see while V_BE continues to rise, V_C remains much more constant and actually decreases above a certain level of V_IN.

A zero gain amplifier made using an enhancement mode NMOS 2N7000 transistor was simulated where the small signal AC gain and phase was calculated as the drain current was swept. As can be seen in figure 11.17 there is a sharp null or dip in the gain curve at around 345uA. This also occurs at the point where the phase makes a sharp transition from 0 degrees to 180 degrees.

ADALM1000 Lab Activity 7, BJT Zero Gain Amplifier
ADALM1000 Lab Activity 7m, NMOS Zero Gain Amplifier

ADALM2000 Lab Activity 7, BJT Zero Gain Amplifier
ADALM2000 Lab Activity 7m, NMOS Zero Gain Amplifier

Now that we understand the concept of the zero gain amplifier, the objective is to investigate its use to produce an output current which is stabilized (less sensitive) to variations of the input current level. This current source configuration, figure 11.17, is also sometimes called a Peaking Current Source or g_m-compensated mirror. Because the collector voltage V_C of transistor Q₁ is now more constant with changes in the input supply voltage as represented by V_IN, it can be used as the base voltage of Q₂ to produce a much more constant collector current in transistor Q₂.

ADALM1000 Lab Activity 8, BJT Stabilized current source
ADALM1000 Lab Activity 8m, NMOS Stabilized current source

ADALM2000 Lab Activity 8, BJT Stabilized current source
ADALM2000 Lab Activity 8m, MOS Stabilized current source

Return to Previous Chapter

Go to Next Chapter

Return to Table of Contents