For
Tube-amp builders who want a real Studio: design any tube regulator from spec, simulate, diagnose and export.
You will learn
  • Compare every tube-only regulator topology on a shared bench
  • Master cold-cathode VR physics, stacking and ballast sizing
  • Design series + error-amp regulators with predictable Zout, ripple and stability
  • Diagnose live faults (oscillation, sag, hum, no-strike) with the interactive assistant
  • Export Markdown reports + SPICE netlists wired to Ampera's Koren tube models
Before you start
Power supply & rectification
Time & level
45 minAdvanced

Cascade and feed-forward

4 min7 min leftPrevNext
Chapter 6 / 84 min

Cascade and feed-forward

Multi-stage error amps, feed-forward null tricks for sub-mV ripple.

When the loop in chapter 5 runs out of headroom — Zout stuck at 1 Ω, ripple stuck at 1 mV — two patterns add gain without compromising stability: cascade (more loop gain) and feed-forward (parallel cancellation path).

ConceptCascade — multiplying µ

Two error-amp stages in series. The first compares Vout · β to Vref; the second adds gain before driving the pass tube. Effective µ ≈ µ1 × µ2. A 12AX7 cascode followed by another 12AX7 reaches µeff > 5000, bench Zout < 0.5 Ω, ripple atten 65 dB.

Cascade error amplifier regulator (Module 06)Two-stage cascaded error amp. V2 (common-cathode) compares V_out · β to V_ref; V3 (common-cathode) adds a second stage of gain before driving the pass tube V1. Effective loop gain ≈ µ₁ × µ₂.Two stacked 12AX7 → 12AX7 → 6080 pass6080 · V16080V112AX7 · V212AX7V2Ra2 220 kΩRa2220 kΩ12AX7 · V312AX7V3Ra3 220 kΩRa3220 kΩRvr 22 kΩRvr22 kΩ0A2 (VR tube)0A2VR1R1 68 kΩR168 kΩR2 100 kΩR2100 kΩClick to copy "V_raw"V_rawClick to copy "V_out"V_outClick to copy "GND"GNDClick to copy "V_ref"V_ref
ConceptFeed-forward — cancelling before the loop

The error amp can only correct what reaches it through R1. A spare tube samples Vraw ripple directly via a Cffcoupling cap, scales and inverts it, and injects the correction at Vout through Riff. Done right, the cancellation pushes ripple atten to 95 dB on the bench.

Feed-forward correction (Module 06)Series + error amp loop with a spare triode (V3) injecting a feed-forward correction sampled from V_raw ripple via Cff. The correction sums at V_out through Riff.Feed-forward path — C_ff samples V_raw, R_iff injects into V_out6080 · V16080V1Rgs 47 kΩRgs47 kΩRa 220 kΩRa220 kΩ6BM8 · V26BM8V2Rvr 22 kΩRvr22 kΩ0A2 (VR tube)0A2VR1R1 68 kΩR168 kΩR2 100 kΩR2100 kΩCff 100 nFCff100 nFRg3 1 MΩRg31 MΩRa3 100 kΩRa3100 kΩ6BM8 · V36BM8V3Riff 33 kΩRiff33 kΩClick to copy "V_raw"V_rawClick to copy "V_out"V_outClick to copy "GND"GND

The hidden cost: feed-forward has two sweet spots — one for AC ripple (trim ≈ 35 %), one for DC line drift (trim ≈ 65 %). With a single physical pot you can only nail one. Most builds optimise for the AC null because that's where the audible noise lives.

Calc · feed-forward
Open →
Feed-Forward Lab
Two trim sliders + a sweep plot showing the two optima and the 10 % cross-coupling between them. Watch the AC and DC residuals fight each other.
Lab · multi-design
Run →
Multi-design comparator
Side-by-side step response of self-bias / series + err / feed-forward — see the cancellation kick in faster than the loop can react.
Check yourself
Why does cascading two 12AX7 stages NOT give you µ_eff = 100 × 100 = 10000 in practice?
In the same category