Tutor -- As everybody has said, this amp will in fact require very stout EL34 tubes. Based on circuit values, the design of the comparitor circuits cause the output tubes to idle with a quiescent current of about 60 ma. With 10% of this current flowing through the screen circuit, that establishes the plate dissipation of each tube at about 27 watts, which is over the Design Center rating of 25 watts for these tubes -- although still within the Design Maximum rating. It also operates the tubes right at the Design Maximum rating with regards to screen grid voltage as well. Against these facts then, the comments about the use of today's quality tubes are well founded. Also however, it would be a very good idea to recenter the comparitor circuits to establish a quiescent current of 50 ma instead of the original 60 ma. This can be easily done by changing out he existing 20 Ohm cathode resistors for 24 Ohm pieces. This will reduce the plate dissipation of each tube to about 23 watts, which is much more reasonable for the output tubes to operate at relative to tube life, yet will still allow the unit to operate with low distortion.
But there is one more point that has not been mentioned yet, that may be the most important point of all. This design uses massive amounts capacitance in the power supply feeding the output tubes. With such designs (and particularly so with UL configurations), unless the screen grid circuit employs enough resistance in it, it becomes highly prone to producing output tube arcing -- and particularly so when the screen grid is run very close in voltage to the plate, at the limits of its voltage rating, and with high Gm tubes -- ALL ELEMENTS OF THIS DESIGN. As a result, it is almost a foregone conclusion that the 20 Ohm screen stopper resistance is not enough in this case -- a value set only to prevent garden variety parasitic oscillations due to the push-pull-parallel connection of the output tubes. However, it is not nearly enough to cause what is effectively a tank circuit formed at the screen grid element to remain stable under the conditions these tubes operate under, and again, particularly under dynamic conditions. As I result, I would suggest that it is absolutely necessary to increase the value of the screen stopper resistors to a value of 400 to 500 Ohms.
When you've got intermittent weird stuff happening, and you've addressed all the usual suspects that have been mentioned, then its time to look elsewhere. CJ made very good products, but that doesn't mean that they weren't infallible. Addressing the usual suspects, while also addressing the matters I've mentioned will very likely resolve the issue.
Good luck with it!
Dave
