The last of the narrow-panel tweed Tremolux circuits, and the one that stops
looking like a tweed amp. Where the 5E9 and 5E9-A cathode-bias their 6V6GT pair,
the 5G9 runs a −28 V fixed bias off its own rectifier; where the tweed
Deluxe splits phase with a cathodyne, this uses a long-tailed pair; and it
carries a tremolo that modulates the output tubes' bias directly. It is a tweed
amplifier with most of a brownface amplifier inside it.
The drawing's own title block says 5G9, and that is how this entry is filed.
A note on the same sheet reads "Late 5E9-A models are similar to this model."
Similar is not the same, and a note about a neighbouring designation is not a
licence to merge two of them: the 5E9-A has its own published drawing, and if it
is ever documented here it will be documented from that drawing. This archive
files a circuit under the designation its own title block prints — the rule the
5E4-A entry already carries.
Signal path
Two channels into one bottle. The chassis takes two Inst. and two Mic.
jacks; each channel's pair sits on its own two 68 kΩ stoppers over a single
1 MΩ grid leak, and each drives one half of a 12AY7 — 100 kΩ plate load, and
one 820 Ω cathode resistor with a 25 µF · 25 V bypass serving both halves. The
two channels are electrically identical. Nothing voices one against the other:
the difference is the panel.
The volume controls are driven at their wipers. This is the circuit's one
genuine oddity, and it is on both pages of the drawing, so it is not a drafting
slip. Each 12AY7 plate goes through a 0.1 µF · 400 V capacitor straight to its
own 1 MΩ pot's wiper. One end of each pot is grounded; the other end joins a
shared mixing node, and that node — not a wiper — is what feeds the phase
inverter through a 0.02 µF coupler. The layout sheet says the same thing in
hardware: the coupling capacitor lands on the centre lug, the outer lug goes to
the pot case, and the two volume pots' remaining lugs are strapped together with
the tone control's centre lug.
It works because the plate is not an ideal source. Turned toward its grounded
end, a pot shorts its own channel's coupling capacitor to ground through the
plate's ~20 kΩ of source impedance, and the channel goes silent; turned the
other way it passes nearly all of the signal into the mixing node. The
arrangement mixes the two channels the same way, and — as on the tweed Deluxe —
each control loads the other, which is why the two interact.
Tone. One 1 MΩ control serves both channels, hung on the Inst. channel's
coupling node: a 500 pF mica in at its hot end, a 0.005 µF · 400 V out of its
cold end to ground, and its wiper on the mixing node.
Phase inverter. A 12AX7 long-tailed pair with 82 kΩ and 100 kΩ · 5 %
plate loads fed at their junction from the +310 V rail, a 470 Ω resistor from
the joined cathodes to a tail junction, 10 kΩ from that junction to ground, and
both 1 MΩ grid leaks returned to it. The cold grid's signal reference is a
0.1 µF · 200 V capacitor to ground — this circuit has no negative-feedback
loop, which is what separates it from the brownface amps that inherit the rest
of its topology.
Output. Two 0.02 µF · 400 V couplers into a 6V6GT pair whose cathodes are
grounded, whose screens each sit behind their own 470 Ω · 1 W resistor on the
post-choke +368 V node, and whose grids hang on the −28 V bias line through
220 kΩ leaks.
The tremolo modulates the bias
The second 12AX7 is a phase-shift oscillator direct-coupled to a cathode
follower, which is more machinery than the brown-era tremolos use for the same
job. The oscillator half takes a 100 kΩ plate load from the +368 V node and a
1.5 kΩ cathode resistor with a 25 µF bypass, and swings through a three-section
0.03 µF / 0.01 µF / 0.01 µF ladder — its two 1 MΩ returns land one on ground and
one on that +1.7 V cathode node, not both on ground — with a 2 MΩ Speed
control (and its own 100 kΩ end resistor) setting the rate. Its plate is wired straight
to the second half's grid; that half's plate sits on the same +368 V node as the
6V6 screens and its 220 kΩ cathode resistor is the load the 0.1 µF output
coupler is taken from.
From there a 1 MΩ resistor feeds a 250 kΩ-L Depth control that sits in the
−28 V bias line itself, with its wiper on the junction of the two output grid
leaks. The output grids draw no grid current, so no DC flows through the pot and
the bias voltage arrives unchanged wherever the knob is set; what varies is how
much of the oscillator's swing rides on top of it. The footswitch jack shorts
the ladder's middle node to ground, which stops the oscillator.
What the DC check does and does not say
The whole tremolo bottle is excluded from the simulated operating point, and
the exclusion is a statement about both halves rather than a convenience. The
oscillator's printed pins — +270 V at the plate, +1.7 V at the cathode — are the
average a meter reads while it swings, set by grid-leak detection rather than by
a static bias point, and the follower is direct-coupled to that swinging plate,
so the +260 V printed at its cathode is not a static point either.
The exclusion is bounded, and the bound is worth stating rather than asserting:
the pair draws about 2 mA from the +368 V node, so leaving it out raises that
simulated node by roughly a quarter of a volt and changes nothing below the
10 kΩ dropper, whose current is set entirely by the preamp and the inverter. The
DC comparison assumes the footswitch jack empty; plugging in and closing the
switch stops oscillation and moves no node in the table either, because the
Depth control it feeds carries no DC.
A DC pass here is a statement about bias points and nothing else. It says
nothing about whether the oscillator runs, at what rate, or how deep the tremolo
goes. Those are questions for a simulation that runs in time or for a bench.
Power
An 8160 power transformer with a 300-0-300 V secondary feeds the 5U4GB,
which delivers +370 V at the reservoir — the node the output transformer's
centre tap sits on. A 14684 choke takes it to +368 V for the 6V6 screens
(each through its own 470 Ω · 1 W) and the tremolo bottle, and a 10 kΩ · 1 W
dropper to +310 V for the 12AY7 plate loads and the inverter's plate-load
junction. Four 20 µF · 500 V cans do the filtering: two on the reservoir, one on
each rail below it, all of them in the condenser box on the back of the chassis.
The bias supply is a rectifier straight off a dedicated tap on the transformer's
high-tension winding — a fourth terminal between one end and the centre tap, a
separate lead on the layout, not a rectifier plate — into an 8 µF · 150 V can
— no series feed resistor — making −69 V, which an 82 kΩ / 56 kΩ divider and
a second 8 µF can bring to the −28 V line. Fixed, with no adjustment
trimmer.
Reading against the printed chart
The drawing prints a voltage chart, every value set at ±20 %, read to ground
with an electronic voltmeter, and every one of them was read at the scan
resolution named below. Two nodes are driven in the simulation and both are
values the sheet prints — the +370 V reservoir and the −69 V bias rectifier
— and everything below either of them is solved.
The supply rails land where the sheet says: the screen node at +368.5 V against
a printed +368 V, and the preamp / inverter rail at +313 V against +310 V. The
bias divider solves to −28.0 V against a printed −28 V, and the output grids to
−28.0 V against a printed −27.5 V. In the stages, both inverter plates sit
within 3 % of their printed +210 V and +200 V, the joined cathodes and the tail
junction within 3 % of +27 V and +25.5 V. The worst gated node is the 12AY7
plate pair, simulated at +160 V against a printed +175 V — 8.6 % low, and the
largest deviation on the sheet.
Three things about that comparison are worth a reader's scepticism, and are
recorded rather than smoothed over:
- One rail rests on an estimate. The 14684 choke's winding resistance is not
printed anywhere, so the netlist models it at 130 Ω — the figure this archive
already carries for the same choke on the 5F4, 5F6-A and 5F8-A. The +368 V
node is therefore the one prediction here that leans on a number the drawing
does not give. It happens to be a small lean: at the ~14 mA this model draws,
130 Ω costs 1.8 V against the sheet's own +370 → +368 step.
- Every value here was read at 700 ppi, and it took a second copy to get
there. The drawing survives in at least two scans. The one this entry is
read from is a CCITT stencil whose pages are 6518 × 4128 and 6376 × 4450
pixels; the other, easier to reach, is 1064 × 775 and 1494 × 976. On the
coarse copy two figures could not be separated — the inverter's tail junction
(
25.5 or 255) and the tremolo oscillator's cathode resistor (1500 or
1800) — and arithmetic from the sheet's own printed voltages resolved the
first correctly and the second wrongly, arguing for 1.8 kΩ where the
drawing plainly letters 1500 Ω. Both now read unambiguously at full
resolution. The lesson is recorded rather than quietly fixed: a chart value
read from a blurred digit passes every gate this project has, because the
simulation is then calibrated to it.
- The output tubes are modelled well outside their ratings. With +370 V on
the plates, +367 V on the screens and −28 V on the grids, the clean-room
6V6GT model puts about 41 mA through each tube — over 15 W of plate
dissipation on a valve rated for 12 W. That is an extrapolation: the model is
fitted to datasheet anchors taken far below this operating point, and the
drawing prints neither a plate nor a screen voltage for the output tubes, so
no node in the chart is gated on it. What it does say plainly is that this
circuit works its 6V6GTs hard.
A note on the rectifier
The drawing letters the rectifier 5U4GB, and so do the parts list, the
schematic and this page. The tube model in models/ is an explicit
5U4G / 5U4GA / 5U4GB class approximation, and it is not instantiated in the
simulation at all: as with every rectifier in this corpus, the power-supply
front end is replaced by an ideal source at the first rail. The mapping is
stated in the netlist header so nobody has to work out why a bottle the drawing
calls a 5U4GB resolves to a model file named for its older sibling.
Lineage
No ancestry arrow is drawn from this entry. The Tremolux's own chronological
chain — 5E9, 5E9-A, 5G9, and the piggyback circuits that follow — is recorded in
the history tier, where a sequence of models is what the page is for. A drawn
lineage edge is a derivation claim, and the resemblance between this preamp and
the tweed Deluxe's, real as it is, is not evidence of one.
