[2004-08-23]

Having designed and built a couple basic RC coupled vacuum tube amps with inexpensive parts, I wanted to take things to the next level, both in design and parts quality. I wanted to work with some new tubes and employ some of the many advanced circuit design techniques I have been studying. I also came across some big surplus iron, and wanted to put it to use along with good Magnequest outputs and plate chokes.

First I was going to build a Morrison Micro 2a3, having slowly figured out how to design a direct coupled circuit. Direct coupling of course is where the output tube sits at a relatively high voltage, such that the cathode is at a voltage equal to the plate voltage of the driver, minus the bias voltage of the output tube. This eliminates the need for a coupling capacitor. Then I saw the various Monkey designs, where the voltage supply for the driver comes from the cathode circuit of the output tube. This reduces the amount of current the circuit draws because the driver recycles some of the current that would normally be dissipated in the cathode resistor. This recycling effect is what had led to Jeremy Epstein calling his design "Free Lunch". In a circuit like this, the amount of voltage dropped across the plate load of the driver is exactly equal to the bias voltage of the output tube. Usually you would want some margin of voltage to allow the driver to swing more voltage than the output tube would require to drive it to full output, but in this design it is not possible. This led me to believe that a choke was the only appropriate plate load for the driver tube. A resistor would need to be much larger than the small amount of resistance needed to drop the bias voltage, and a current source would need a few volts of headroom, further limiting the drive voltage available. I hoped that a choke would allow the full driver swing, and perhaps a bit more on transients due to the choke's ability to store and release energy.

After settling on a topology, I just needed to pick tubes and parts and stuff. I decided that this would be a good time to try out the 6c45pi driver tube. This tube is very popular due to it's low plate resistance, high gain, and huge power handling due to it's burly construction. I wanted to use good output iron, so I looked at the Magnequest stuff. I would have done parafeed output, but at the time it was much more expensive. Now, with the Robin Hood line of outputs, it would have been affordable. I chose the TFA-204 output transformer and the BCP-15 plate choke for the 6c45.

With almost all of the parts on the top plate, the insides of this amp are pretty lonely.

There were a few things I wanted to try in the power supply too. I wanted to use 6cj3 damper diodes as rectifiers because of thier long delay, and sonic abilities as described by Lynn Olson. I also wanted to use a large potted milspec power transformer, and large potted milspec chokes I got at a local surplus shop. My favorite transformers on this amp are the small potted dual 2.5V filament transformers. There is something very satisfying about a brick of coiled wire :) Finally, I wanted to use lots of oil caps. I could not find an inexpensive source of paper-in-oil caps, but I found a very good deal on polypropylene-in-oil motor run caps at an online HVAC shop. These are rated for the voltage I needed, and are supposed to be very sturdy. I used many 10uF caps instead of single larger caps because they were cheaper per microfarad :)

The power supply is a basic CLCLC supply using a 1uF Solen as the first cap, then 30uF oil caps for the other 2. Chokes are 20H potted surplus used slightly above thier rating of 110mA. Rectifier is a pair of 6cj3 damper diodes forming a full wave bridge.

2a3 socket with Ohmite adjustable resistors and Allen Bradley hum balance pot with cool recessed mounting hardware scavenged from a Tektronics scope module.

One thing that was quite handy was the use of adjustable wirewound Ohmite power resistors in the cathode circuit. This allowed me to adjust the resistance to whatever I wanted. I got the idea from old Loftin-White amplifier schematics where there appears to be a single large multi-tapped adjustable resistor providing all of the needed resistance in the circuit. I used a stack of 3 of these, adjustable resistors above and below a fixed resistor. It is nice to stay below 2K or so on the adjustable resistors so that they are not so sensitive to small adjustments.

After reading Lynn Olson's paper on loop distortion, I decided that it made a lot of sense to tie the cathode of the 2a3 to the B+ supply in order to reduce the components in the output loop signal path. This is commonly known as Ultrapath output, and is certainly nothing new. It feels like it is getting increasingly difficult to explore new territory in tube amplifier topology.

Chassis construction was pretty straightforward. I wanted to mount as many of the components as possible on the top plate. This proved a bit difficult with the capacitors. I had to punch 2 holes per cap for the terminals, then use screws to catch the rim of the cap to hold them to the top plate. It was a lot of trouble since there were so many of them, but the result is very sturdy.

The beauty of cheap surplus iron is that you can build power supplies that are total overkill. This amp, while only 3 watts per channel, weighs a healthy 55 pounds - yes I weighed it :) It was difficult to work with on the bench because it is so damn heavy.

So how does it sound? Pretty damn good I think. The bass is suprizingly deep, probably because of the high quality output transformers. When IO say "deep bass" I mean that within the modest limits of my Fostex TQWT speakers. I would love to hear this amp on something with real bass, like... Edgarhorns maybe :) I don't notice any harshness due to the 6c45 oscillating at high frequencies, but I guess it is hard to be sure without a good scope. The soundstage and imaging are very good and it is incredibly detailed. I am listening to the radio a lot with it, and it is really fun to listen to the ambient sound change as the DJ switches between the studio, taped reports, etc. When there is no one speaking, and no music being played, and you can still tell when the switch occurs between different ambient acoustics, that is a cool effect.