What happened? 'Tis a mystery.

I usually do not have problems with my amplifiers.  They are a pair of Dynaco
Mark II, with beefed-up capacitors in the power supply and a homebuilt driver
board patterned after...OH NEVER MIND!

I do not roll tubes.  I run the outputs at low cathode idle current (what we mostly
call "bias" but it's not).  I check that idle current every six, eight months unless I
hear problems (which is almost never, but read on). The tubes have been in the
amplifer for about fifteen years and still test OK.  The amp still puts out full power
on the workbench.  All is good.  Almost always.

ANYWAY, today I was listening to a Rachmaninoff symphony (as if that mattered)
and heard a couple of buzzy "barks" from the left channel loudspeaker.  Then
the audio in that channel resumed but quickly faded a-w--a---y.  

The left channel amp fell silent.

I moved the left channel subwoofer out of the way, disconnected the speaker cable from the amp, pulled
the input cable, and r-e-a-c-h-e-d behind the cabinet to unplug the amp from the power strip which is
controlled by a switch on the preamp.

Opened up the amplifier.  Blown fuse.  No evidence of a swollen or burst capacitor.  No evidence of burned
plates on the output tubes.  No evidence of damage to the rectifier, a stock 5U4.

No bad smell.  No phone, no pool, no pets (oops, that's from King of the Road).

I removed all the tubes and replaced the fuse.   No fuse blow.  Normal AC at the inputs to
the rectifier, both the 5v and 6.3v filament connections, and to the bias circuitry.  Power transformer OK.  
Good.

I put in the rectifier and ran the voltage up slowly on a VariAC.  At 85 VAC I stopped because the B+ was
approaching normal-operation voltage (with the output tubes removed, it was going to go too high if I let
the amp see full line voltage).

Rectifier OK.

So...don't tell me...it's a problem with one or more of the EL34 output tubes.  They're old.  Original
Mullards.  Original British Mullards, not rebranded SinoRussian tubes.  If they're blown, those are
expensive tubes to replace.  I probably wouldn't and would go SinoRussian.  Sigh.

I put the Mullards back anyway and powered up.  The filaments lit, the tubes warmed up.  No fuse blow.

I let it warm up while I was ready to lunge for the off switch if anything bad happened.

Nothing bad happened.

I set the cathode current, by adjusting the grid bias, to 1.3 volts across the cathode
resistor ("normal" would be 1.56) for longer tube life, and put the amp's covers back on.  
Hooked it back up to the system. Had a problem reaching into behind the cabinet far
enough to plug the amplifier back into power strip.  Do not have arms like a gorilla, so
I couldn't.  So I taped the cord to a 3-foot long stick and reached in with
that to where I could reach in a different way, grab the end of the stick, remove the
tape holding the cord to the stick, and plug in the amp. Got it.  BARELY.

I HATE CABINETS!!!

Turned on.  Listened.  Okay.

Only thing I can figure is that vibration from the subwoofer shook something loose in a tube
(the subwoofer is about four inches from the amp).  Whatever it was, touched a grid or the
cathode or the plate when it wasn't supposed to.  Hopefully it then fell all the way to the
bottom.

I'm listening to Stravinsky now...thinking about moving the left channel subwoofer
farther from the amplifier.

It sounds like you followed a good procedure to track the problem down. Tis a mystery. But the subwoofer seems like a long shot. Your problem may come back and stay a problem. I have had a few failed Dynaco power transformers but it has been two in 50 years. They both went out once (not intermittent) and were easy to find (shorted red/red winding).

If the output tube has a failure I do not think it would be intermittent either. It may sound weird but a fuse can go bad as well. Many years of current thru a small conductor. Think light bulbs. It can happen and how often have you changed the fuse!

If it comes back, track it down and let us know.

WLT wrote: [cut] If it comes back, track it down and let us know.

"When it comes back", not "If". It's never the fuse... unless you have something in there other than a 5A. If you aren't going to do further debug you might at least swap your tube sets left for right and see if the problem follows the tubes.

Well, it WAS less than a 5A.  

It was a 3 1/2A slow-blow.  Stock MK II had a 4A standard blow.

I had the day off and listened for the rest of the day. No issues. Still
going now.

>>ANYWAY, today I was listening to a Rachmaninoff symphony (as if that mattered)
and heard a couple of buzzy "barks" from the left channel loudspeaker. Then
the audio in that channel resumed but quickly faded a-w--a---y.

The left channel amp fell silent.<<

As already noted:

This is a classic symptom of fuse failure. If one has the means, try loading a regular glass fuse nearly up to its rating and watch the element. As one gets closer and closer to the rating, it will start to 'dance with the music'. The sequence of events, at a guess, was as follows:

a) Already weak fuse starts to fail, and resistance greatly increases = first bark.
b) Fuse fails = second bark
c) Residual capacitance drives the speakers until its gone (fading).

RANT WARNING:

Slow-Blow fuses (those with the element wound around a thin ceramic bar) are the invention of the devil and are designed only to protect real-estate and not the equipment. If one looks at the failure tables for a slow-blow device, one will find out that it will tolerate a 10% overload very nearly indefinitely. And a 100% overload for far too long (multiple of enough seconds to fry a power-transformer).

What one wants is a Dual-Element fuse - one that tolerates a brief overload to handle turn-on surges, but reverts to behaving like a standard fast-blow fuse after such a surge. These fuses may be sized very much closer to the actual operating load as they are capable of managing the turn-on surge when all the tube filaments are cold. Note that original Dynaco amps were supplied with DE fuses when new. And even their pre-amps and solid-state devices were supplied with fractional-amp DE fuses.

All and at the same time, I would get that amp onto a fine-pitch ammeter and compare it to the 'other' one under actual load. They should be within a few % of each other. This would be the belt-suspenders-Velcro approach. Generally, anything that crosses my bench spends at least an hour on my metered Variac before being allowed out into polite society.

http://cdn.ipernity.com/200/56/84/45315684.29c1fbad.640.jpg?r2

I have had this tool for very nearly 32 years, purchased new as a kit (attic stock, not from Heath). probably the most useful diagnostic tool I have.