Origin of il Monstro, an 845:845 Power Amplifier

View Sakuma 845:845 Schematic

Original Schematic at Japanese 'Direct Heating' site

The size of the parafeed coupling capacitor is influenced by the whichever of the other two parallel elements in the parafeed network has the lowest impedance at low frequencies. So no matter how high the inductance of the output transformer, if you use a plate load with a low frequency impedance/inductance lower than the inductance of the OT (which would also be much lower in comparison to something like the C4S load), the plate load determines the cap size, not the OT.
 
It's really pretty simple to use a reactance/frequency chart to play around with values and see what will work.
 
As an example consider a case where you wish to keep your resonance frequency of a parafeed circuit low enough to create a knee that extends the natural rolloff of the iron, more or less the standard design criterion for a parafeed design. With a 2A3 or 300B with roughly 800 to 1000 ohms Rp and a using parafeed OT that is interleaved and thus has a high inductance and a typical primary impedance of 2.5K to 3K, you will find that the numbers work well with a plate choke inductance of about 30-40H and a parafeed cap value around 3uF.
 
Now if you go to a 45 with an Rp of about 1700 ohms you want a primary impedance of about 5K. This is great, because you can get more turns on the OT and your inductance can be even higher. And yes, you can use a bit smaller cap, around 2uF. But- you also have to increase the value of the plate load to around 40-50H to maintain your knee frequency.
 
Go to something like a 15K line transformer and you can imagine that the inductance might go up even more, thus the cap value might get smaller, and the plate load will need to be even higher.
 
So talking about a transformer that allows for using a smaller coupling cap without considering the plate load doesn't fit the math of parafeed ciruit design. If the inductance of the OT goes up, the plate load has to go up too, to allow for using a smaller cap. If the inductance of the OT goes down, yes, the cap value can get smaller, but you may create a situation where you will lose power bandwidth and bass power will suffer.
 
One advantage of the cascode CCS we use is that the load impedance is at least 5meg so the inductance of the OT becomes the main variable that influences the size of the parafeed coupling cap. An airgapped choke or load resistor will not begin to approach this impedance value in the critical low frequency range where the inductance and capacitance values matter. (source: http://www.audioasylum.com/forums/tubediy/messages/13565.html)

 

                              ----------------- Choke - 845 Output
                              |                  48H    377VDC
                              |                 304ohm  177VAC
Rec                           |
83 -- C --- L --- C --- L --- C --- L --- C --- Choke - 845 Driver
     24uF  10H   24uF  10H  120uF  12H    |      48H    355VDC
      |   82ohm   |   82ohm   |  155ohm   |     304ohm   44VAC
    492VDC      467VDC      438VDC      412VDC   
     23VAC     1.025VAC    0.237VAC    0.396VAC 

The circuit does not show any common ground share by the two filament supplies.
 
If the filament supplies are unloaded and separated, there may be a large voltage between them due to capacitive coupling from the filament power transformer primary winding. In practice, as shown on the diagram, there is 2kohms between them (the two 1k cathode bas resistors). Normally the coupling capacitance is small, perhaps 200pF. This should be thoroughly swamped out by a 2k resistance. Assuming I have correctly understood the circuit, and especially in light of the peculiar item above, I would suspect some magnetic coupling into the chokes from nearby power transformers.
 
It appears from these [B+] numbers that you are drawing huge currents through the tubes, 200mA in the output and 175mA in the driver. I would suspect that the filaments are actually being grounded somewhere, not elevated to the (nominally) 50v bias point. You should be drawing more like 50mA per tube, and your plate voltages would then rise quite a bit, providing most of that 100v you were asking for.

1.0VAC   --+
           |  Peerless 15095 input
5.0VAC   --+
          ( ) 845 Driver
21.81VAC --+
           |  EXO-173ni 
21.67VAC --+ 
          ( ) 845 Output
81.7VAC  --+ 
           |  EXO-50 output XFMR
32.12VAC --+
         (( )) 8 ohm Pioneer 51W fullrange

 
                                ----------------- Choke - 845 Output
                                |                  48H    507VDC
                                |                 304ohm .056VAC
                                |
                                |                        45V Bias
                                |
Rect -- C --- L --- C --- L --- C --- L --- C --- Choke - 845 Driver
83     24uF  10H   24uF  10H  120uF  12H    |      48H
        |   82ohm   |  82ohm    |  155ohm   |     304ohm
      547VDC      537VDC      528VDC      519VDC          497VDC
      8.85VAC    .079VAC     .026VAC     .017VAC         .048VAC
 
                                                         50V Bias

845_pfeed_sep02.gif Schematic

• copper chassis plates (onlinemetals.com)
• handles for cabinets (Home Depot Expo)
• final cathode resistors (?)
• tin foil/film blocking caps (angela.com)
• silver signal wire (22ga Myron Toback)
• silver power wire (16ga Myron Toback)

• Round voltmeters for filaments
• Move hum pots and cathode bias to side of 845 sockets
• Use single point ground plate for signal chassis

• Clean end bells, primer, sand, primer, sand, primer, sand, primer, paint!
• Mask/paint stacks - I used the blue painter's tape to mask off the wires (coiled at the side). Some light sanding before painting, but be careful not to drop the stacks or loosen the laminations.
• Install brass or other non-magnetic hardware - Used brass allthread and acorn nuts. Find Jamestown manufacturing online (a boatbuilding supply house) for a good selection of stainless, brass, and bronze hardware.

• + driver fil 18ga
• shield - driver fil
• + output fil 18ga
• shield - output fil
• + high voltage 20ga
• shield chassis GND
• + signal GND 18ga
• shield chassis GND

• Revise grounds
• Replace final stage decoupling 10H L with 12H Hammond 193B
• Remove wiring for filament voltage meters
• add safety bleeder 50k 25W WW resistor on final C - switch on GND leg (amp sounds better off)
• Change B+ capacitor sizes
  83 rectifier - 3uf - 10H - 10uF - 10H - 120uF (2x60uF) > to signal chassis

• Hammond sends replacement 300BX
• I order new filament XFMRS
• Re-fit PS for new filament transformers
• Add RC to PS stages - 12uF+47R with 10uF - 33uF+110R for 120uF

• Add .022 Sequa paper/foil cap - better but not good
• Add another .022 Sequa cap + .22 AuriCap in parallel
• Add .47 tin foil SCR and R (RC impedance correction) per posts by Joppa and V2 in AA

• Add 33uF SCR bypassed with .22uF AuriCap to cathode bypass (better)
• Move EXO-50 output tranny primary GND to top of cathode R (still better)