Current mirrors

Transistor current mirror with emitter degeneration resistors

Introduction

There are situations in which you need to make an exact copy of a current.
When you look at the internal circuit diagram of f.e. an OPAMP or comparator, you will see multiple current mirrors.
In the examples below, the transistors have to be matched so have the same current gain (beta) and they also have to be thermally coupled so their temperature is equal.
In practical situations where you don't need an exact copy of the current, you can get away with unmatched transistors that are not thermally coupled. In case of using transistors that are not thermally coupled (not on one chip), it is wise to add emitter degeneration resistors, that provide some negative feedback to decrease the effects of temperature difference. They have a stabilizing effect, but with the resistors in place, you will need a higher voltage to get the same current.

Transistor current mirror variants

In the figures above, you see some variants of current mirrors, and more specific : current sinking mirrors, sorted from good to excellent.

If you need current sourcing mirrors instead of current sinking mirrors, then exchange ground with the positive supply and replace the NPN transistors with PNP transistors.

Figure 1 shows the most basic type of current mirror. The current through R1, being Iref, is slightly higher than the current through R2, being Iout, because not only the collector current of Q1 flows through R1, but also the base current of both Q1 and Q2.
Figure 2 is an improved variant in which the current mismatch described above. is decreased by using transistor Q3 to provide the base current for both transistors (Q1 and Q2) via the power supply. So now the extra current flowing through R1 is not the sum of the base currents of Q1 and Q2, but the sum of these currents, divided by the current amplification factor (=beta) of Q3.
Figure 3 shows the Wilson current source with a similar smart trick to decrease the mismatch between Iref and Iout. In this case, Q3 will draw base current via R1 and deliver this current to Q2. So now the base currents of Q1 and Q2 are not only drawn through R2 but a part of this current is also drawn through R1. This way the mismatch between Iref and Iout is minimized.
Figure 4 shows the best variant in which the current mismatch is pretty much absent by combining 2 mirrored current mirrors.The current mismatch of the upper current mirror is compensated by the current mismatch of the bottom current mirror.

Widlar transistor current mirror

In the current mirror above, which is called "the Widlar current mirror", resistor R3 is added to a standard current mirror. This allows to generate a very low output current Iout without needing a very large value for R1 to generate a low reference current Iref. Additionally, the emitter resistor causes negative feedback, causing a higher output impedance, as we will see later.

The resistor R3 decreases the base-emitter voltage of Q2, so the transistor gets less base current, resulting in a lower collector current.

Transistor current mirror with emitter degeneration resistors

In the current mirrors above, emitter resistors are added. These so-called emitter degeneration resistors provide extra negative feedback in order to make the current mirror more stable. Q1 is less sensitive to changes because it is connected as a diode and therefore its Vce is fixed. But Q2 is more sensitive to changes that affect transistor parameters. Vbe of Q2 depends on the collector current (Ic) and the collector-emitter voltage (Vce). By adding the emitter resistance, the emitter voltage of the transistor will rise when the current through the resistor rises, meaning that the base to emitter voltage decreases, so the transistor will conduct less and the collector-, and thus als emitter-current, will fall. Essentially, Vbe of Q2 will vary less with Ic and Vce by adding the emitter resistors, because this increases the output resistance of Q2. The higher the output resistance, the less Ic will vary with variations in Vce. This is the main characteristic of a constant current source : very high output impedance. This kind of negative feedback, that increases output impedance, provides more stability. Furthermore the emitter resistors decrease the effect of Vbe mismatch between the 2 transistors, when the voltage over the emitter resistors is much higher than base-emitter voltage drop.

The emitter resistor for Q1 is only necessary when you want both currents in the mirror to be equal. Otherwise the emitter resistor of Q1 can be left out, so we get a Widlar current mirror.

OPAMP current mirror variants

The circuits above are current mirrors based on an OPAMP instead of bipolar transistors.

In both left circuits, the OPAMP is configured as a voltage buffer. The voltage at the output of the OPAMP will follow the voltage on the non-inverting input. So we can state that Iref * R1 = [current through R2] * R2. We also know that : Iout = Iref + [current through R2]. So when combining both formula's, we find that Iout = Iref + (Iref * R1) / R2 = Iref * (1 + R1/R2). So when R1 = R2, then Iout = 2 * Iref.
In both right circuits, the OPAMP is configured in a different way to create a current mirror. Because with an ideal OPAMP, the current flowing into the non-inverting and inverting inputs is 0, we can state that all of Iref flows through R1 and all of Iout flows through R2. The voltage at the output of the OPAMP will be Iref * R1 = Iout * R2. This results in Iout = (Iref * R1) / R2. So when R1 = R2, then Iout = Iref.