In this design, we depart from the standard Williamson circuit or even the direct "phase-splitter as driver" type often used for the 807.
Instead, in a more modern look, we use a constant current source in the preamplifier stage, and a longtailed pair for the driver stage in an attempt to lower distortion even further before any application of feedback. The concept is to have as low a distortion overall design before feedback. The feedback requirement is a given for any PP amp that will spend at least part of its life in class AB and in PP amps in general to lower 3rd THD.

The output stage is more or less conventional, with triode strapped 807's in pushpull arrangement.
The chosen load is 6K6 plate-to-plate. While this load represents a decrease in maximum power output, it also reduces distortion, especially 3rd harmonic. Later there will be a picture, schematic and discussion for the output section.

The output stage uses a combination of cathode and grid bias, allowing the user a wide selection in tubes, and maximizing the power output for the power supply voltage I had to use (450V).

Orignally I conceived the design as a pair of long-tailed amplifiers, with the first LTA providing most of the voltage gain, and the second LTA providing some power as well as some voltage gain. Also in the beginning, I arbitrarily chose to use 12AU7's for all these stages, as the tube is easily obtained at reasonable prices (so far) as compared to alternatives of 6SN7 or 12DW7. The 12AU7 also has better drive capacity against the 807t (t-triode) Miller capacitance. Also the 6CG7 was considered.

As compard to the 6SN7, all these other tubes would provide better drive as the second stage. However, since the primary design goal was to minimize pre-feedback distortion, the 6SN7 was the clear winner in all testing. Thus the 6SN7 is chosen.

Building the two-stage input/driver and testing the same introduced an interesting side issue that the 6SN7 as an input tube, especially run from a CCS can be a bit noisy. Most of this noise was in heater-cathode interaction even though the tube's heaters for both stages were properly biased. I tried several 6SN7's and flavors, but was unhappy with the residual noise. Also a few, while not exceedingly so, were slightly microphonic.

The solution came in dumping the original LTA input arrangement and going to a one-tube input fed by a CCS plate load. For this solution, the ultra-quiet 5879 pentode was chosen, operated in direct triode connection (no screen resistor). This tube requires both the screen and the suppressor grids be connected to the plate. Performance is very good. During later testing, both the screen and suppressor grids were tied to the plate via 100 ohms carbon resistors. This seemed to give more genteelity to the amp that distortion measurements did not pick up. Thus the final version incorporated these resistors.

Below is a pix of a test using a single 6SN7 section as an input amp that is CCS loaded, driving a 6SN7 LTA with a CC sink in the tail. This worked well, but the following pix showing the 5879 was both quieter and allowed a lower Vk at the LTA tail giving more headroom and saving some CCS2 disappation. Current in the input stage is about 3.25mA, while the LTA is running 8mA per section for a 16mA tail.

In the final tests, the combination of single CCS-loaded 6SN7 driving the 6SN7 LTA produced the following results:
RMS OUT / THD (second + third + fifth) fourth/seventh = not measurable
10V / 0.09%
20V / 0.6%
35V / 1.3%


The combination CCS-loaded 5879 driving the 6SN7 LTA:
RMS OUT / THD (second + third + fifth) fourth/seventh = not measurable
10V / 0.05%
20V / 0.3%
35V / 1%


Gain is about 30db, not 35db as in diagram.
The feedback point will be the cathode of the 5879 taking usual care with phasing etc.

Power this monster

The main amp operates from about 450V and the output stage idles around 150mA - 160mA depending on bias settings. In addition, the 6SN7 LTP's each draw a nominal 16mA, then there is the CCS stabilization voltage as well as the input stage current. I was fortunate to have a power tranny up to meeting these high demands. In addition, this tranny provided 3 separate heater windings, one for each stage thus allowing each stage to be properly optimized and heater-biased as needed.

The power supply is solid-state rectified, and in order to provide a safe margin for the filter caps, they were placed in series. Voltage alignment resistors are placed across each cap to ensure proper voltage division, with the resistors doubling as bleeders at turn-off. A separate winding is bridge rectified to provide the negative bias for the output stage.

The main supply is choke filtered in a CLCLC type arrangement, providing B+ for the output stage and driver stage. The input stage and CCS get it's voltage from a final RC filter. The entire supply is very robust and of a low imedance type which I prefer. Nothing gets really hot in this amp (except the tubes of course) and the power tranny and power resistors that are used all run at normal warm temperatures as expected with the power tranny passing the pinkie test even if a bit warm. The power tranny also has a shielded core. This is conected to AC chassis ground, while the audio ground is lifted above this to aid in interconnection noise reduction.

The switch is for speaker/headphone jack selection.

This was an interesting project from the beginning, and one of the largest I have undertaken. Was it worth it? Tons of time and labor went into this creation. Sound wise it is a extremely clear with super localization. Bass at first seemed weak, but then I was comparing to an SE amp that was naturally warmer. When called on to deliver real bass, the amp worked fine and provided excellent detail. Frequency response is well extended from about 10Hz to above 35KHz. In the end, power is about 12W/ch. although peaks to 15W are easily achieved. Running nominally at 1W-1.5W/ch. playing audio, the amp has the habit of disappearing on better CD's, while poor CD's fare badly. One CD in particular which we had only listened to on my Loftin-White SE amp was clearly not the quality recording we originally thought. So the amp pretty much shows things for what they are, and that's really the idea, isn't it?