One tube, five flavours

A triode wired as a cathode follower drops V_ak between anode (at V_raw) and cathode (at V_out), and a further I·Rk on the cathode resistor that auto-biases the grid. The output voltage is simply what is left after both drops.

The grid is grounded through Rg, so the grid-to-cathode bias is simply −I·Rk. To pick Rk you choose the desired bias voltage from the tube’s datasheet:

Open-loop output impedance

The small-signal Z_out of a CF triode (looking into the cathode, with grid grounded for AC) is the reflected plate resistance r_p/(µ+1). The familiar 1/g_m form is only the µ≫1 approximation of this — fine for high-µ tubes, but it overestimates badly for a low-µ power triode like the 6080 (µ=2).

The 1/g_m form is the µ≫1 approximation — it overestimates here since µ=2. Use the exact r_p/(µ+1) for the 6080:

Pass-tube dissipation

The triode sees V_ak across it and conducts I_load (cathode current ≈ load current at this scale). Multiply for the average anode dissipation.

Ripple attenuation: — a self-bias CF has no error-amplifier loop, so V_raw ripple passes through almost unchanged. The big input filter (LC / CRC) does most of the work.

Same CF topology as the self-bias variant, but the grid is no longer pulled to ground via Rg — instead it is sat at a chosen DC potential. The pass-tube equation is unchanged except R_k is gone.

V_grid can come from a negative bias supply (clean but adds a winding/rectifier), or from a stiff divider on V_raw. In divider form the upper resistor sets the loading and the lower one (R_g) sets the grid voltage:

Why Z_out is (a hair) better

In self-bias the cathode resistor R_k introduces local NFB that raises Z_out by R_k/(1+µ) at low frequencies. Fixed bias removes R_k entirely (or fully bypasses it with a large cap), so only the intrinsic CF impedance remains:

Practically identical to self-bias when Rk is well bypassed — the win is mostly that no Rk dissipates a few watts as heat.

Pass-tube dissipation

Note: the divider taps V_raw, so any V_raw ripple shows up on the grid (and thus on V_out) with full µ. A bypass cap on R_g shunts that ripple to ground above the corner frequency — this is what the Cg in the schematic does.

A series diode + reservoir cap on the grid drive samples the peak of V_raw instead of its DC average. The grid then sits at (V_peak − V_diode), and the cathode follower lands |V_bias| volts below that. Net effect: a few tens of volts of extra V_out headroom when V_raw is barely above what you need.

How long must the cap hold?

Between successive peaks the cap discharges through the bleed resistor R_bleed (plus any grid current). The fractional droop over one ripple period is approximately:

Rule of thumb: keep R·C at least 100× T_ripple to stay under 1 % droop and avoid imposing fresh ripple on V_out.

Same Z_out, same P_diss — only the bias rail moves

Z_out and P_diss are unchanged from the self-bias / fixed-bias formulae — peak detection only redefines V_grid. Effectively, you trade a few volts of average headroom for the peak-minus-diode value.

Ripple atten: — peak detection lifts the operating point but, like the other bias schemes, has no error loop. The grid-cap RC also adds a (slow) HP corner if R_bleed is too small.

A pentode in CF duty keeps the same V_out = V_raw − V_ak relation as the triode case, but the screen grid (G2) draws its own current from a separate clean reference. The dissipation budget therefore has two terms, and the small-signal Z_out is much higher because pentodes have essentially no µ-feedback in CF.

Why Z_out is higher than for a triode CF

A pentode acts as a near-ideal current source from anode to cathode — r_p is in the tens of kΩ, and the screen decouples plate voltage from cathode current. In CF, the small-signal output impedance reduces to:

In practice r_p is so high it drops out, leaving 1/g_m in parallel with the actual load impedance. With a high g_m pentode like the EL509 you can get tens of ohms — but the lack of a µ-feedback term means line/load reg is poor without an outer loop:

Total pass-tube dissipation

Two terms — anode and screen. Both must be checked against the tube datasheet maxima (typically P_a,max ≫ P_g2,max).

Note: the screen *must* be referenced to a stable, well-bypassed node (often its own VR tube — 0A2 / 0B2). Any ripple on V_g2 transfers directly to I_a, then to V_out through the cathode resistor of the load.

Removing the input choke gives a cap-input (a.k.a. Π / "pi" if a second cap is added). DC at the reservoir is now near the rectified peak rather than 0.9·V_rms, which is great for headroom — but ripple at the regulator input is several volts peak-to-peak instead of millivolts. The regulator equations are the same as the pentode CF case; the differences live on the PSU side and in how much ripple reaches V_out.

Ripple at the regulator input

Standard full-wave cap-input ripple, in the textbook approximation:

With a 50 Hz mains and a full-wave rectifier, f_ripple = 100 Hz; this puts the ripple in the volts, not millivolts.

Open-loop ripple atten of the pentode CF

The pentode CF, without an outer error loop, has only its g_m·R_load to fight V_raw movement. Ripple at the output is:

Dropout check

The cap-input PSU swings between V_peak and (V_peak − V_ripple,pp). Even the *bottom* of the ripple must keep V_ak above the pentode’s minimum operating value, or the regulator falls out of regulation once per half-cycle:

Note: removing the choke also removes its critical-inductance constraint (no minimum load needed), and is cheaper / lighter. Cost: visible 100 Hz hum on V_out unless you close an error loop or add a CRC stage downstream.