Gain boosting in analog

In short: Gain boosting places a small auxiliary amplifier around the cascode device so it actively pins the drain of the input transistor, multiplying output impedance (and hence DC gain) by the auxiliary amp's own gain without adding stacked transistors or burning extra voltage headroom.

Why a cascode alone runs out of gain

In deep-submicron CMOS the intrinsic gain of a single transistor (g_mr_o) has collapsed. A 28 nm device might offer g_mr_o of only 15–25, so a simple common-source stage gives roughly 24–28 dB. The classic fix is the cascode, which stacks a second transistor on top of the input device and raises output impedance to about g_m2r_o2·r_o1. That squares the gain term and buys you another 25 dB — but it costs a V_DSsat of headroom and, crucially, you cannot stack a third or fourth device at a 0.9 V supply. You simply run out of rail.

Gain boosting (also called the regulated cascode, or RGC) breaks this trade-off. Instead of stacking more transistors to raise output impedance, it uses feedback to make the existing cascode transistor work harder.

Key fact: Gain boosting multiplies output resistance by the auxiliary amplifier gain A_aux, giving R_out ≈ A_aux·g_m2r_o2·r_o1. A 30 dB booster adds ~30 dB of overall gain for the price of one tiny amplifier.

The regulated cascode idea

Consider the cascode transistor M2 sitting on top of input device M1. Normally M2's gate is held at a fixed bias. The whole benefit of the cascode comes from M2 keeping the drain of M1 (node X) roughly constant as the output swings — but it only does this to the extent of its own 1/g_m2r_o2 feedback factor.

Gain boosting inserts an auxiliary amplifier that senses node X and drives M2's gate. If node X tries to move, the auxiliary amp reacts with gain A_aux and pushes M2's gate the opposite way, forcing X back. Node X is now regulated far more stiffly, so the current through M1 barely changes with output voltage — which is exactly the definition of high output impedance.

Where the extra gain actually comes from

Work the small-signal loop. Without boosting, a change ΔV_out feeds back to node X through M2 with factor ≈1/(g_m2r_o2). With the booster, that same disturbance is additionally suppressed by (1+A_aux). The output resistance becomes:

R_out ≈ (1 + A_aux) · g_m2 r_o2 · r_o1

Because DC gain of the stage is g_m1·R_out, you have multiplied the whole gain by (1+A_aux). In practice the auxiliary amplifier is itself just a small single-stage differential pair — it needs no large output swing, only to servo the gate of M2 — so it is cheap in area and current. Stacking two boosted cascodes (NMOS and PMOS sides of a telescopic OTA) can push a single-stage amplifier past 100 dB, territory that would otherwise demand a two-stage or three-stage design with its attendant compensation headaches.

Key fact: Gain boosting raises DC gain without adding stacked devices, so the output swing of a boosted telescopic cascode is identical to an ordinary cascode — a decisive advantage at sub-1 V supplies.

The catch: the pole-zero doublet and settling

The auxiliary loop introduces an extra pole, and the interaction between the main amplifier bandwidth and the auxiliary bandwidth creates a pole-zero doublet. Bult and Geelen's original 1990 analysis showed the safe design window: the auxiliary amp's unity-gain frequency ω_aux must sit above the main amplifier's 3 dB bandwidth but below the main amplifier's non-dominant pole (often the cascode node pole). Violate the lower bound and the doublet shows up as a slow-settling tail — deadly for switched-capacitor circuits that must settle to 0.01% in half a clock period. Violate the upper bound and the auxiliary loop goes unstable.

In a modern flow you verify this with real tools: an AC analysis in Cadence Spectre to confirm phase margin of both loops, a stb (stability) analysis to break the auxiliary loop explicitly, and a transient settling test to catch the doublet tail that an AC plot can hide. Monte Carlo over mismatch of the auxiliary pair matters because its offset appears directly at node X.

Watching gain boosting in motion

Practical design checklist

• Size the auxiliary amp for gain, not swing. It only servos a gate; a simple 5-transistor OTA with 20–40 dB is usually enough.
• Bias node X correctly. V_ref sets the drain of M1 just above its V_DSsat — too low kills M1's r_o, defeating the purpose.
• Mind the polarity. An NMOS cascode uses a PMOS-input auxiliary amp and vice versa, so the reference can be placed at the correct common-mode.
• Place the doublet deliberately. Set ω_aux between the main bandwidth and the main non-dominant pole; verify with a stability (stb) run.
• Check both PVT and mismatch. Auxiliary offset lands on node X; budget it in your Monte Carlo.

Teams building data converters, bandgap references, and high-resolution ADC frontends lean on this technique constantly. If you want hands-on layout and PVT closure for these stages, our partner Avecas analog design services runs boosted-cascode OTAs through full sign-off flows.

Ready to design your own regulated-cascode OTA from schematic to sign-off? Join the CourseTron Analog IC Design track and build a 100 dB single-stage amplifier with full Spectre verification.

Frequently asked questions

Does gain boosting change the amplifier's output swing?

No. This is its headline benefit. Because you are not adding stacked devices, the output swing of a gain-boosted cascode equals that of a plain cascode. You gain DC gain for free in the swing budget, which is why it thrives at low supply voltages.

How much gain can the auxiliary amplifier add?

The stage gain is multiplied by roughly (1+A_aux). A modest 30 dB booster adds about 30 dB of overall gain, taking a 52 dB cascode to around 82 dB. Boosting both the NMOS and PMOS cascodes of a telescopic OTA can exceed 100 dB in a single stage.

What is the pole-zero doublet and why does it matter?

The auxiliary loop creates a closely spaced pole and zero. If mispositioned it produces a slow-settling tail that degrades accuracy in switched-capacitor and sampled systems. The auxiliary unity-gain frequency must lie above the main amplifier's bandwidth but below its non-dominant pole.

When should I prefer gain boosting over a second gain stage?

Choose gain boosting when you need very high gain but must keep a single-stage architecture for fast settling and simple (single-pole) compensation — common in high-speed pipelined ADC residue amplifiers. A second stage buys swing but complicates Miller compensation and stability.