300b lovers

Unfortunately, revising the driver away from the usual single 12AU7 means another hole in the chassis, and a pair of octal sockets at that. And split power supplies means another power transformer and rectifiers, although it relaxes the endless search for Holy Grail rectifiers. When an amp is that sensitive to rectifier choice, frankly, there is a design error lurking in there somewhere.

Somewhat counter-intuitively, splitting the power supplies front to back sounds better than isolated stereo power supplies, where the gain is fairly minor. The front to back isolation is not minor. Another counter-intuitive thing is PP amps benefit more from power supply isolation than SE amps.

As for inverters, the split-load inverter plus input tube uses two sections, while Mullard and the floating paraphase use three sections. And the last two sections can be octal medium-power tubes, not a single 12AU7.

This will change the forward gain of the amplifier, not by by much, but some. Probably a few dB less gain thanks to different and lower-mu drivers. Maybe a small trim in the feedback network, but less gain is easy to adjust for (unlike more gain).

In practical terms, a new chassis. Sorry. Those old amps were engineered down to the last inch, with no room to spare, and not one part wasted. Flip one upside-down and the parts in there are really, really tight. Too tight, and prone to overheating, which does no favors to the capacitors.

Most of the engineering effort in the Golden Age, even for Marantz and McIntosh, was simply watts-per-dollar, and keeping chassis size down as a secondary goal. In the late Sixties, the Crown DC300 and Phase Linear 700 blew the watts-per-dollar tube amps out of the water, and forced them into the audiophile market, where they remain today.

As for the tech, none of this is difficult. They will know how to connect and bias the floating paraphase phase inverter ... plenty of old schematics out there ... and the rest is a matter of chassis space and heater current for the new drivers.

"Floating Paraphase" sounds more scary than it is. The secret of any phase inverter is finding what the lower grid is connected to.

In a Mullard circuit, the lower grid is grounded through a coupling cap to ground, but is connected to the other grid through a high-value resistor. This sets the bias point the same as the other side, but the grid does nothing and is simply a reference. The common cathode does the phase splitting. The long-tail is either a high-value resistor or a current source .., yes, I know, transistors. The main drawback of the Mullard circuit is restricted peak current delivery to the power tubes. (But the split-load inverter is much worse in that respect. Split-load inverters do not like to deliver current ... they go out of balance. Ideally, they should be buffered with cathode followers.)

The floating paraphase has the upper tube drive the upper power tube through a totally normal RC coupling. Nothing to see here, sir, move along. The lower tube is often drawn in an opaque way, but what’s going on is the lower grid is connected to a pair of resistors midway between the power tube grids, after the RC coupling, What this weird-looking connection does is local feedback that forces the lower, phase-inverting tube to act as a unity-gain inverter, or plate follower. Phase inversion isn’t quite as pretty as the other two methods, but ... more power is available to drive the power tubes, which is what really counts. And you can ditch the 12AU7 and use real power tubes, because, why not? Just a matter of another socket and heater power.

I have a question to @lynn_olson 
I have 300B SET 6sn7 input, 6f6g in triod driver with RC coupling. One friend of mine who reads DIY forums and tries all this thing by himself, recommended me to enlarge cathode capacitors up to 100,000uF in input and driver tubes.
So I did 10,000uf in input first and because I like the results I did 100,000uf on input and 50,000uf on driver. So I get even more improvement - deeper and more controlled bass, bigger soundstage, more low level details, transparency, bigger more dynamic sound.
Then I increased driver capacity to 87,000 uF and it gave a big improvement in dynamics and sound quality. Despite intuitively it shouldn't do too much!
In all cases I use a bunch of 10000uf-15000uf (CDE, Nichicon KA) medium quality capacitors bypassed by one 50uf AN Kaisei NP.
Can you explain why these huge value cathode capacitors work?

For example, floating paraphase phase inverters instead of split-load inverters or Mullard long-tail pairs. The phase division isn’t as precise, which is why they dropped out of favor, but the drive capability is much stronger than the other two types.

@lynn_olson A simple way to improve this circuit is to use the bias supply for the KT88s as a B- voltage and then use a 2 stage constant current source to set the operating point of the differential driver. This improves the differential effect quite a lot and gets rid of the need for slightly different plate load resistors- they can be matched instead.

To this end we use bipolar supplies in our amps; B+ and B- have the same absolute value. No balancing is required in the differential amplifier and often the plate voltages are within 3-4 volts of each other. This improves the CMRR quite a lot which in turn reduces distortion (there is the additional benefit of more gain and wider bandwidth as well). If the input stage is built in a similar manner, even orders will cancel throughout the circuit, resulting in a dominant 3rd harmonic which can mask higher orders.

The harmonics will be found to fall off at a faster rate (than seen in SET circuits) as the order of the harmonic is increased; they will follow an exponential decay based on a cubic function. This works really well for the human ear (smoother sound and greater detail, both on account of reduced open loop distortion)!

To eliminate a frequency pole caused by a coupling capacitor, we’ve been using a differential cascode circuit as the sole source of gain in our OTLs. Because the gain is increased in that single stage of gain, so is the CMRR and overall differential effect. This allows one to use a cathode follower driver direct coupled to the output section. In our OTLs the power tubes are in turn direct coupled to the loudspeaker. If you were to use an output transformer, a pair of DHTs are easily driven- linearity is such that no feedback need be used. Bias is obtained from the driver circuit, so if bias controls are used, they are in the grid of the driver tube rather than the power tubes.

If feedback is desired, it can be wrapped around the circuit and applied in a manner identical to how its done in opamps- using resistor divider networks that mix the feedback with the incoming signal at the grids of the input stage. This technique vastly reduces distortion that the feedback signal would otherwise encounter, which in turn means the feedback is more effective at its job, generating less higher ordered harmonics (caused by non-linearities in the feedback nodes traditionally used in both tube and solid state amplifiers). By doing this a wider range of speakers can be used.

Here's a question for @atmasphere - Though I'd welcome the comments from anyone else on this thread -  these are all related to the question of what causes a power tube to wear out - 

1. Does it matter what volume a power tube is played at? Does that effect tube life?

2. Does it matter if a tube is cooled, say, by a small fan nearby?

3. If a tube is powered up but not making music does that "cost" tube life just as if you were playing music through it?

4. What is harder on a power tube? Turning it on and off, let's say twice during a day (two listening sessions totaling three hours) or letting it stay on, let's say for an eight hour period?