The voltage regulator on the welder pictured turned out to
be non-functional. There were no descriptions of how it worked
and very limited information on how to troubleshoot it.

In the hope it might help somebody else, here's what I learned.

Component references are from 
http://www.zefox.net/~bob/welder/miller_legend_tech_manual_1996.pdf
The voltage regulator schematic detail is on page 52.

At ignition on, battery current flows through D4, through the
center slip ring, through the exciter field and back to ground.

On startup, the exciter winding and SR2 exceed battery voltage,
taking over the excitation requirements and reverse biasing D4,
effectively taking it out of the circuit.

Once the exciter is self-sustaining R4 serves as a trim, allowing
maximum voltage to be set. Nominally one should get 80 VAC at the
input of SR2 at 1800 RPM and 140 VAC at 3000 RPM. Note that the
exciter circuit floats: The loop is grounded at the exciter brush,
not at either output of SR2.

Exciter current thus circulates in at the center brush and out
at the two field brushes. The generator field current flows through
the regulator and back to SR2.

The regulator is actually a phase controlled DC power supply connected
in series with the exciter. During welding the exciter voltage is sufficient 
to fully energize the generator field, with the power supply's rectifier  
allowing the current to bypass the regulator power winding.

During generator operation at 1800 RPM the exciter alone does not produce
enough output to provide regulated AC power. In power mode the regulator
is connected to the AC output via transformer T1. The voltage from T1
serves two purposes: It samples the AC voltage, and provides power to
activate a pair of SCRs which connect the regulator power winding in
series with the exciter current, boosting the voltage and current applied
to the generator field. With zero voltage from T1, the SCR circuit is
inactive, with high voltage the SCRs are active but untriggered and
within the regulating range the SCRs are triggered to boost the exciter
voltage to provide 120 VAC at the generator output. 

In this particular case the regulator board showed signs of overheating
at R70, a 150 ohm resistor which, together with a zener diode provided
a coarse pre-regulator to VR50, an LM7815 which furnished regulated 15 V
power to the error amplifiers and SCR control circuits. This by itself
wasn't significant; the resistor tested 200 ohms, close enough.

However, in-circuit tests of the zener diode revealed low back voltage in
both directions, less than .1 volt. It wasn't obvious whether the diode
or the regulator was at fault, but the diode was less hard to isolate.

Lifting the anode lead revealed the diode to be shorted. This wasn't a
trivial exercise: Lifting the lead was made difficult by a tough conformal
coating, lead wires bent to anchor them in the holes and thin, fragile
traces that de-laminated readily at soldering temps. In hindsight it 
would have been wiser to just clip the lead, make the test and then
solder to the stub.

With the zener diode and resistor set replaced further in-circuit 
multimeter tests were ambiguous. It wasn't clear if the 7815 regulator
was functional, but the difficulties of working on the board made
isolating it very daunting. Instead, a battery charger and current
limiting resistor (1157 taillight bulb) were used as a test source
to drive terminal B positive relative to terminal A. 15 volts appeared
at the 7815's output, to my immense relief.

When reinstalled in the welder output at 1800 RPM was 121 volts, 
sagging to 118 volts at about 5 kW.

20241201