I recently had an experience using the Electro-Harmonix 7591 tubes in a
McIntosh 1500 that I feel warrants a comment. I noticed on a couple of
occasions that one of the 4 outputs tubes would suddenly overheat to the
point of glowing bright red. I rushed to the receiver and turned it off,
and of course immediately suspected bias. A complete examination of the
bias pots and power supply shows no problem.

Later, the same thing happend, except on a different tube. Now, I think "RF
oscillation", and so I hook up instrumentation to the amp and await the next
occurence.

Result - no RF or oscillation, just tube current running away (normal 0.7
volts across the 15 ohm bias measurment resistor became 4.5 volts!,
indicating a current of 300ma, where the normal bias level is 45ma). This
time, I was fortunate enough to observe that the tube began getting red hot
inside the plate first, then the plate followed. After closely examining
the tube, I indentified the element overheating as the screen grid. So, I
had a classic case of screen grid runaway, which means all of the tubes were
sitting there like little volcanoes ready to erupt, and the choice of which
one goes first was random.

Examination of the tube specs and the schematic of the Mac 1500 shows that
the screen is being operated off of a 450V supply, with just a 900 ohm
dropping resistor and additional filter cap between it and the plate supply
(at 460 volts). This is a very high screen voltage, and the 7591 tube data
I have shows that the screen voltage should never exceed 400V.

The receiver never exhibits this behavior on the old NOS 7591's, which just
goes to show that a new tube can be designed as a match, but unless the
exact same materials are used in the exact same geometry, Murphy can come
and bite you.

The solution? Screen stopper resistors don't "stop" it (no pun intended)
until you reach a value of over 2000 ohms, which reduces the output power of
the amp to 25W and increases distortion, measured on my own unit, from 0.1%
to over 3.0%. So that's a no go.

Another solution would be to downpower the screen voltage to below 400
volts, but I suspect that would have the same result of increased distortion
and less power (the McIntosh unity-coupled circuit requires a lot of voltage
swing to drive the output tubes, and I suspect that the high screen voltage
was used to make the tubes easier to drive, and the hand selected tubes of
the era could take it. Other McIntoshs that use 7591's, like the MC-225,
have swapped screen drive (driven from the opposite plate), bootstrapping to
increase the drive and cathode bias instead of the fixed bias in the 1500.
Cathode bias would stop this dead, since it automatically limits the maximum
current thru the tube).

So, I've had no problem with these in MC 225's, but in a 1500, watch out!!