Well, here is the first cut at a two stage design.
I went and read the entire STC Application Report on the 807 (June 1954) and looked at the recommendations provided in that report. The report contains SE curves for triode and tetrode operation, but not UL. Normally under these circumstances I would pick something about 30% of the way from triode toward tetrode and see what I get. However, as luck would have it, the Automatica web site (
http://www.audiomatica.it/tubes/807.htm) has some published curves for 807 UL operation. So I decided that I'd do a derived design from scratch.
The resultant design is trade off between supply current, power, and distortion. I wanted to preserve as much output power as possible without getting the AC swing into the compression region before tube cutoff. As such I decided to keep the operating point at a lower voltage and let the bias current run higher. As a result, I preserved almost all the output power of the tetrode mode, kept the 2nd harmonic distortion at about the same level but drove the 3rd harmonic distortion way down. The trade off is that the bias current is about 25% higher then in tetrode mode. This should produce a very smooth sounding amplifier with plenty of detail. See the attached operating curves along with the associated design data.
The bias point requires only a 20 volt drive so a 6SN7 makes a good choice for a driver stage. The circuit topology I chose is one I prefer for two stage SE amps. It is entirely auto biased with no feedback of any kind. It places a master gain control in between the two stages to keep the control noise out of the driver stage and also controls the frequency roll off points for the amplifier. All the important design information is included on the schematic. This design requires about 1.58v peak input to produce the full 6.3W output per channel. This level is well within the capabilities of most CD player and iPods so the overall dynamic range should be ok.
Attachment:
UL Plate Curves.png
Now for some specifics. I recommend the Edcor GXSE25-8-2.5k output transformers. These are a relatively expensive transformer but I am concerned about the bias current level (~103mA). I know that the GSXE10-8-2.5k can't quite handle that kind of bias current. The GSXE15-8-2.5k might be able to handle it but a chat with a design engineer at Edcor would probably be required to make sure. The low end roll offs due to cathode bypass are pretty benign. The low end roll off is dominated by the coupling capacitor and the volume control/grid leak resistor. I would have put in a bigger bypass cap to get better low end performance but I am concerned about bias excursions if the 807 is temporally overdriven. The RC time constant is 10.7ms giving about a 54ms bias excursion recovery time. This is about a third of a beat at 400 beats per minute (similar to a relatively fast guitar lick or fast classical flute solos). I am afraid that any larger a cap and occasional excursions would become noticeable at high drive levels. The high end roll off due to the 807 Miller capacitance shouldn't be a problem even at low volume levels. If it is a concern, the volume control can be replaced with a fixed 100k grid leak resistor, and then have the volume control moved to the input. The 5k grid stop resistors are recommended in the application report to prevent RF oscillation. This value should prevent any oscillation problems and also reduce problems with bias excursion.
The power supply requirements are a 300v B+ supply capable of at least 220ma. I didn't design a power supply yet because I didn't know whether this would be done in mono-blocks or in a single chassis. I also don't know what is the rectifier preference. Personally I like a big 5U4 in an amp like this. The 807 is a big tube and I think the other tubes should reflect that large size if possible.
Attachment:
Schematic.png
NOTE: Please see this SE ATS25/807 two stage amp - SCHEMATIC WARNING!So this is my first cut. Take a look at the overall design, poke holes in it, recommend any changes you think would improve the design, etc. This is just the first cut at a workable design so comments, suggestions, changes, etc. are more then welcome.
Let me know what you think.