<!– Generated by hand for SushiDSP phase D6a. –>

jcm_pp_ot_nfb

JCM800 2203 block 5 in full: the long-tailed pair, the EL34 push-pull pair, the output transformer into a 16 Ω resistive load, and the global negative-feedback loop with the presence pot.

Status: simulated. Commit 2bdbf91 ran this block and committed its CSVs.

LTspice version: LTspice 24, the only detail 2bdbf91 records; the exact build is not recorded by that commit.

How the schematic is written

jcm_pp_ot_nfb.asc carries the whole circuit as SPICE directive text on the sheet rather than as placed symbols with wires. Directive lines are netlisted verbatim by LTspice, so node names are explicit and the netlist cannot depend on symbol pin geometry that could not be checked without running the tool. Tube subcircuits are instantiated as X lines against ../models/*.sub with the pin order documented there:

• 12AX7_KOREN: P G K
• EL34_KOREN: P G2 G1 K

Analyses

Both analysis lines live on the sheet as comments and tools/spice/export.py enables exactly one of them per run:

• ;ac dec 40 10 20k
• ;tran 0 40m 20m 1u (40 ms run, the last 20 ms saved as steady state)

.options plotwinsize=0 keeps the raw file uncompressed.

Probe and stimulus

• Probed node: out – the 16 Ω load, i.e. the transformer secondary after its winding resistance.
• Input node: g1
• Source: V1 g1 0 SINE(0 {AMP} 1k) AC 1
• Input level(s): AMP = 1 V fixed (this block steps Presence instead).

Stepped control

.step param PRESENCE list 0.1 0.5 0.9

PRESENCE is the normalised 5 kΩ presence-pot fraction, used as a rheostat in series with the 0.1 uF shunt on the NFB node. BIAS defaults to -38 V.

Components

Output transformer

The D6a brief specifies a 20 H primary and a secondary sized for a 3.4 kΩ : 16 Ω ratio, so Ls = 20 H * 16 / 3400 = 94.12 mH. The primary is centre-tapped to B+, which a single 20 H inductor cannot express, so it is written as two series-aiding halves of 5 H each: for a perfectly coupled continuous winding, L_total = L1 + L2 + 2*sqrt(L1*L2) = 5 + 5 + 10 = 20 H. Lpa runs pa1 -> B and Lpb runs B -> pb1, which puts the dots at opposite ends of the centre tap exactly as one continuous winding does. The 50 Ω of primary copper is split 25 Ω per half. Leakage is carried by the K1 = 0.999 coupling coefficient rather than an explicit leakage inductance.

Judgment calls and open questions

• NFB polarity is unverified. Whether the loop is negative depends on the transformer’s dot orientation relative to which LTP plate drives which EL34. If a simulation shows the loop is positive (oscillation, or gain rising instead of falling as Rnfb is reduced), swap the secondary’s terminals – write Ls 0 sec instead of Ls sec 0 – and rerun. Nothing else needs to change.
• EL34 cathodes are tied directly to ground: this is the fixed-bias arrangement D6d describes, with the bias supply reaching the grids through the 220 kΩ leaks.
• Screens are fed from the same 360 V node as the plates through their 1 kΩ stoppers. D6d’s separate screen supply belongs to block 6 (power-supply sag), which is not part of this corpus.
• The reactive speaker load is deliberately out of scope (D6d: the cab IR carries the speaker’s response), so the secondary sees a plain 16 Ω resistor.