Analog IC · Lesson 2
Interactive current mirror
A basic two-transistor NMOS mirror. Drag components to rearrange, change the transistor sizes (W/L) and supply, and watch I_REF, I_OUT, the mirror ratio, operating region, and the output curve respond live.
📘 Chapter introduction(click to show)
In this chapter you will learn to
- ✓Explain why a diode-connected transistor is always in saturation.
- ✓Calculate the output current of a mirror from the W/L ratio.
- ✓Find the output compliance limit and explain the effect of finite r_o.
1. Why this circuit matters
Analog chips need many accurate, steady currents: to bias amplifiers, to act as loads, and to charge capacitors at a known rate. Making each one with its own resistor would be inaccurate and would waste silicon area. Instead, a chip generates one good reference current and copies it wherever it is needed. The circuit that copies it is the current mirror.
💡 Think of it like this
A current mirror is a photocopier for current. You feed in one original, I_REF, and each output transistor prints a copy. Make the output transistor twice as big and the copy is enlarged 2×; make it half the size and the copy is reduced to ½.
2. Definition
A current mirror is a circuit that forces the current in one or more output branches to equal, or be a fixed multiple of, a reference current, almost independent of the output voltage.
3. Meet the components
R_REF, reference resistor
Sets the reference current I_REF drawn from the supply.
M1, diode-connected NMOS
Gate tied to drain. Converts I_REF into a gate voltage V_GS. Always in saturation.
M2, output NMOS
Converts that same V_GS back into a current: the copy, scaled by its size.
Supply V_DD
Provides the energy for both branches.
Ground
The shared source connection for M1 and M2, so both see the same V_GS.
4. How it works, step by step
- 1
Inject a reference. R_REF pushes a current I_REF into transistor M1.
- 2
M1 finds its own V_GS. Because M1's gate is tied to its drain, V_DS = V_GS, which is always above V_OV, so M1 is always saturated. Its V_GS settles at exactly the value that carries I_REF.
- 3
Share the voltage. M1's gate is wired to M2's gate, and both sources are at ground. So M2 has the same V_GS. No current flows into the gates; only the voltage is shared.
- 4
M2 copies the current. Same V_GS and same process means the same current per unit of W/L. So I_OUT = I_REF × (W/L)₂ / (W/L)₁.
- 5
Stay in saturation. The copy holds only while V_OUT ≥ V_OV. Below that, M2 enters triode and the output current collapses. That limit is the output compliance.
🔑 The one thing to remember
Current → voltage (in M1) → current (in M2). The shared gate voltage carries the information, and the size ratio sets the copy ratio.
Transistor sizing
Ideal mirror ratio = (W/L)₂ / (W/L)₁ = 1.00×
Bias & process
Key formulas
I_REF = (V_DD − V_GS) / R_REF
I_D = ½·k'n·(W/L)·(V_GS − V_th)²·(1 + λ·V_DS)
I_OUT / I_REF = (W/L)₂ / (W/L)₁
r_out = 1 / (λ·I_OUT)
saturation while V_OUT ≥ V_OV = V_GS − V_th
Schematic
Operating point
- V_DS of M1 and M2 differ, so channel-length modulation makes the actual ratio deviate slightly from the ideal (W/L) ratio.
Output characteristic (I_OUT vs V_OUT)
Beyond the basic mirror: cascode, Wilson and BJT mirrors
M4 holds M2's drain almost fixed, so the output current barely changes with V_OUT. That is why r_out rises by about g_m·r_o.
| Mirror | I_OUT/I_REF | r_out | × basic | Min V_OUT |
|---|---|---|---|---|
| Basic MOS mirrorr_out = r_o. Needs only V_OV of headroom. | 1.0000 | 200 kΩ | 1.0× | 0.32 V |
| Cascode MOS mirrorr_out ≈ g_m·r_o². Costs an extra V_th + V_OV of headroom. | 1.0000 | 25.7 MΩ | 128.5× | 1.13 V |
| Wilson MOS mirrorNegative feedback raises r_out to ≈ g_m·r_o²/2. | 1.0000 | 12.6 MΩ | 63.2× | 1.13 V |
| Basic BJT mirrorThe two base currents are taken from I_REF: ratio = 1 / (1 + 2/β). | 0.9804 | 510 kΩ | 2.6× | 0.20 V |
| BJT mirror with β-helperAn emitter-follower supplies the base current: error drops to ≈ 2/β². | 0.9998 | 500 kΩ | 2.5× | 0.20 V |
Trade-off: the cascode and Wilson mirrors gain tens of times more output resistance but need about V_th + V_OV more headroom. BJT mirrors lose accuracy to base current; a β-helper fixes most of it.
Symbol & pin reference
NMOS transistor — D (drain), G (gate), S (source). Channel colour shows the operating region.
PMOS transistor — note the gate bubble; source (S) sits at the top toward VDD.
Resistor — two leads. Here R_REF sets the reference current.
Ground (0 V reference). The small cyan dots are connection pins.
How it works, step by step
- Set the reference. R_REF pulls a current from VDD into M1. M1 is diode-connected (gate tied to drain), so it is always in saturation.
- M1 finds its V_GS. M1's gate-source voltage self-adjusts until its drain current equals I_REF = (V_DD − V_GS)/R_REF.
- Share the voltage. That same V_GS is wired to M2's gate (the grey mirror node). No gate current flows — it just copies the voltage.
- M2 copies the current. With the same V_GS and overdrive, M2 produces I_OUT = I_REF × (W/L)₂/(W/L)₁ — scaled purely by the size ratio.
- Stay in saturation. As long as V_OUT ≥ V_OV, M2 acts as a stiff current source; below V_OV it drops into triode and the current collapses.
Where it lives: a differential amplifier
Current mirrors are everywhere in analog ICs. This single-stage differential amplifier (the front end of an op-amp / OTA) uses two of them.
Blue = left leg, green = right leg, amber = tail. Wire thickness follows the current. M4 copies I_1; the difference I_1 − I_2 goes to the output.
1. PMOS mirror as active load (M3 & M4)
M3 is diode-connected and mirrors into M4. Instead of plain resistors, the mirror gives each leg a very high load resistance (r_out), which is what makes the voltage gain large. It also combines both input legs into a single-ended output at Vout.
2. Current-mirror tail bias (I_TAIL)
The tail is itself a current mirror copying a reference current. It sets the total bias current, which splits as I_TAIL/2 in each leg of the input pair M1/M2. A stiff tail keeps the bias steady as the inputs move.
Why mirrors here?
They provide accurate, temperature-tracking bias currents and high-impedance loads using only transistors — no big resistors — which is exactly what integrated circuits need.
Things to try
- Set (W/L)₂ = 2 × (W/L)₁ and confirm I_OUT doubles (a 2:1 mirror).
- Drag V_OUT below the overdrive and watch M2 turn amber (triode) as the current collapses.
- Increase λ and see the output resistance r_out fall — a less ideal current source.
- Set λ = 0 and the actual ratio snaps exactly to the ideal size ratio.
- Lower R_REF and watch I_REF (and therefore I_OUT) rise.
- Drag the components around — the wiring and current flow stay connected.