Next Audio Project
Once I get the Wolverine project completed, I've decided what to do after that one, once RL stops interfering with my plans. As much as I'd like to do one, I can not justify a hollow state OTL regardless of how elegant the Circlotron topology is. Using four or five 6BQ6s per phase to get a lousy ten watts or so makes no sense whatsoever. Make a Class AB2 807 amp? Is 80W really all that much better than 30W? The 30 WPC 807 Amp I already have is plenty loud enough, and I still haven't pushed that one to the max yet. an increase of 3.0db is just barely noticeable anyway. So why bother?
I'm still not all that interested in doing a solid state design just yet, and you still can't get those SiC power JFETs very easily and inexpensively either.
So how about really kicking up the output to some truly speaker-poofing levels? Another thing: the finals are surprisingly inexpensive.
The Sakuya
A Mega Level Vacuum Tube Amp
The finals used here are a PP pair of 814s. I can get these for about $20.00, and considering what they can do, that's dirt cheap. This particular type was designed with the same philosophy behind the development of the 807: operate at very low screen voltages, "shadow" the screen grid with the control grid to minimize screen current, improve efficiency since all screen grid power contributes nothing to output power. Getting the screen voltage down also contributes to better linearity in an audio final, since the plates can swing further towards AC ground before dropping below the screen voltage.
The 814 is a directly heated beam former with a Th/W filament that can stand up to ion back bombardment better than an oxide cathode can. This allows operation at significantly higher plate voltages (and adds glowey bottle coolness points). Forch, the best loadlines come in at voltages under 1.0KV, even though you could pull even more power from them. The finals are by no means being unduly stressed in this design.
If this type has the same snivets problem that you sometimes see with the 807, the alternative to screen stoppers is to put a positive bias on the beam formers. This is a possibility since these have their own pin, and aren't internally connected to the cathode.
Loadlines
Here, you see that you can get just a bit under 150W. This is a much better than the usual solution to the 100W+ problem: paralleling up lower powered audio finals. Have a look at this design. If you have but the one pair of PP finals, you can dispense with the need for matched pairs (which won't stay matched for very long) and the necessity for some sort of servo network to keep all those VTs pulling together. Simplify your design, improve the sonic performance.
The THD estimate for this loadline are outstanding, and will measure a good deal lower since this graphical method of estimation tends to produce "high side" errors due to the simplifying assumptions inherent in the process. If the 814 is an 807-oid, then expect that there will be considerable higher order harmonics. This can be cleaned up by the application of local NFB, as you would do with an 807 design. Including gNFB can clean up whatever pentode nastiness remains, as well as improving the damping factor.
The one deficiency (which is what probably keeps the 814 cheap) is that it was intended for Class B and Class C RF use. As it is, it can't pull enough current at Vgk= 0 to give much output. So that means swinging the grids into grid current territory. This was a major problem back in "the day", and so you almost never see Class AB2 audio amps, except where the premium was placed on power and efficiency over sonic performance. These days, it's not a problem any more since you can use power MOSFETs as grid drivers. The power MOSFET, having way more gm than any VT, has the highly desirable Lo-Z output. Since the rdon of the power MOSFET is significantly lower than rp, there is no lack of grid current sourcing. This grid driver comes much closer to the ideal than anything that could have ever been done with VTs. Class AB2 need not come with the sonic penalty that it once did.
Since the 814 requires a bit more grid swing, the front end will have to be a two stage design, as cascodes don't have very much output swing as compared to the DC rail voltage. The front end will definitely consist of an LTP first preamp/phase splitter. The second stage will be a DC coupled differential to preserve both AC phase-to-phase balance and harmonic balance between phases. For implementation, those dissimilar dual triodes used for vertical deflection circuits are looking good. An LTP made from the 6GK5 UHF triode is also a possibility, as these UHF triodes are designed to minimize internal capacitance, and so have less CMiller than the usual audio high gain triodes. That'll make it play better with a high resistance volume control. A second stage of a differential 6SN7 can provide enough gate drive to a MOSFET source follower to prevent slew limiting at the upper end of the audio band.
Bias current and balance can easily be set by varying the gate bias of the grid drivers.
Since there is a big difference between the plate and screen voltages, the "economy" type power supply looks attractive here. The main DC rail can be the choke input type for superior voltage regulation under the wildly varying current demand of Class AB2. The secondary DC rail, since the current demand is a good deal lower, and swings much less, can be supplied by means of a high voltage MOSFET operating as an active decoupler. The main PS will almost certainly be a solid state, with separate heater power and HV. That way, all heaters can be brought up to operating temp before you turn on the HV. That way, everything will be good to go at HV power up.
This design will also include active screen voltage regulation. This could either be a hollow state design, or a solid state design. Either way, it's best to supply the reference voltage by means of a gas discharge tube. These voltage regulator tubes typically produce less noise, and what noise they do produce is a good deal less annoying, than what Zeners produce. They are also a good deal more stable than high voltage Zeners. Good screen regulation, and operating the screens into a Lo-Z AC source are important for linearity.
The next big question is the availability of OPTs that can handle this level of power without sounding mediocre. That could be a make or break for this project.