Bret Ludwig wrote:

"Don't clip the amplifier" is easy to say, and tough to do. Totally
avoiding amplifier clipping under any and all circumstnaces requires
either active power compression control (i.e. "Power Guard") or a
really, really, really big amplifier, the very small signal performance
of which is usually suspect unless the amplifier is made extremely
heavy and hot and has a very high quiescent power draw. A 20 watt tube
amp that pulls 50 watts at full power is more efficient in practice
than a 250 watt solid state amp that pulls 80 watts quiescent and 500
at full power if either provides the same _subjective_ performance-even
though the solid state amp is more efficent for each watt it puts out.

Do what? First of all, there will be many conditions where the 20 watt
amplifier will be well into gross clipping while the 250 watt amp is
loafing along. Next, even assuming 50% final stage efficiency, which
is a stretch for most tube implementations, there's no way in hell the
tube amp will draw merely 50 watts at full output.

Let's assume 6BQ5s - they'll deliver 17 watts wound out, but that's in
the ball park and we want to use the lowest-dissipation tube for this
examination: each tube's filament eats 6.3 volts at 0.76 amps, so a pair
takes 1.5 amps at 6.3 volts. That's another 10 watts there, plus another
10 or so for the driver and input stage heaters, not to mention 20 more
watts for all the quiescent currents. Things are starting to heat up,
which is pretty obvious to anyone who's spent much time with tube amps.
At this point we're talking close to 40 watts quiescent for this 20 watt
tube amp, so it's about 80 watts maxed out. Now let's look at a 250 watt
transistor amp: the rails will be at +- 50 volts (full power @ 4 ohms), so
comparing apples to apples and setting quiescent dissipation at 40 watts
(less a third for the drivers et al) means you get a output stage bias of
250 milliamperes. Not huge, but not bad either.


On the other hand, even assuming Hamm's paper is still relevant 32 years
after publishing and several generations of amplifier designs since,
and comparing amplifiers which he claims would have more-or-less equal
resistance to overload, 60 watt tube amps have a quiescent dissipation
around 70 to 90 watts: that would bias the SS output stage at 0.7 ampere.
Note this is enough to maintain class-A operation into several watts.
The extra dissipation all goes into the output stage, as opposed to
warming up an armada of tube heaters; one could argue that a transistor
amplifier dissipating as much heat as a tube amp half its size uses
that heat more efficiently!


In any event, the core of the argument here is not so much the thermal
efficiency of various amplifiers (which is a hell of a way to defend
tubes), but how they _sound_, and I can come up with any number of
scenarios for which a 20 watt amp, no matter how liquid its midrange*,
will be Not Enough.

* which I suspect is a function not of transfer characteristics but
more of non-zero output impedance changing frequency response when
driving a loudspeaker.


Francois.