As I said in the introduction, I decided that I wanted to go with a single ended 6V6 design run in ultra linear mode. To do this I simply took the design point for a single ended class A stage directly off the 6V6 data sheet. This called for a -12.5v bias, a 250v plate and screen voltage, and a 5kΩ plate load. I then chose an Edcor output transformer that had a UL tap, calculated a resistor for cathode bias, and my preliminary output stage design was done. Now what about the driver?

I only needed 12.5v peak to drive this output stage. I also wanted a good performer sonically. So I decided to use a 6SN7. With it’s low µ and very low noise it fits the needs for this driver very well. The 6V6 family (6V6, 7408, 6AQ5, 6005, etc.) can be a little cool in the UL mode and I wanted to preserve some warmth as this amp will be used mostly for soft jazz and light classical music. As such, I decided to use a fairly small plate load on the 6SN7. This preserves some of the 2nd order harmonics and keeps the sound both warm and very smooth. Here is my load line for the driver stage.

I decided that the unbypassed cathode gain of -8 was just about perfect. This means that I only need a little over 1.5v peak (1.1v RMS) input to fully drive the amp. This is well within the capabilities of most line level outputs, iPods, CD player, etc. Also, the -4.25v bias point of the input stage means that the probability of every overdriving the input stage is very low.

I decided to put the volume control after the input stage to minimize it’s effects on amplifier noise figure. This led to the following schematic for the amplifier.

Now it was time to worry about the power supply. As I said, I wanted a very quiet B+ supply. I also wanted to use tube rectification so that I could avoid using a “standby” switch. This necessitated adding an additional stage of LC filtering to the supply as well as additional filtering for the driver stage. (This is the 2.58K / 47µf pair in the schematic above.) Here is the resultant power supply schematic.

My goal for the supply was less than -90dBv ripple at the inputs to the power stage. This equates to ≈ 8.6mV at 272vdc. Ripple at the output of the rectifier is approximately 5.5v so my filter section needed a ripple reduction factor of approximately 640.

The primary ripple frequency into the filter is 120Hz. As a rule of thumb, for any series-shunt LC pair, if the reactance of inductor is greater than about 20 times the capacitor reactance, then the filter stages can be treated separately. This is clearly the case here. For the first stage, the ripple reduction factor is ~181. This puts the ripple at the output of the first filter stage at 30.4mV. The ripple reduction factor for the second filter stage is 283 for a total ripple output of 0.1mV. But the voltage was still too high at this point and a dropping resistor was required. Here the resultant RC filter stage has a ripple reduction factor of 27.4 resulting in a final ripple into the power stage of ≈ 7.5µv (for a total ripple level of -171dBv @ 272vdc). Clearly this B+ supply will not be introducing any extraneous hum into the audio path. The additional RC pair on the driver stage supply gives an additional ripple reduction factor of 92 giving a theoretical ripple voltage of 81nV at 250vdc.

Obviously these theoretically small ripple voltages are not realized in the actual circuit. But the values are small enough to virtually guarantee no B+ induced hum on the audio outputs. Now with the electrical design complete, it was time to turn to the layout.

It took a while playing with all the components to arrive at an acceptable layout. The top plate of the chassis would carry the major components with the point to point wiring underneath. The only compromise on this decision was to place the 5H power supply choke inside the chassis. This was done for two reasons; one, putting it on the top plate resulted in a much larger build and, two, I decided to use an open frame choke I had on hand so for aesthetic reasons I wanted to hide it. Here is the layout for the top plate showing all the major component locations.

This layout has several advantages over the others I tried. First, it places the power transformer and the rectifier tube as far as possible from the main driver tube. This minimizes interaction there. It also puts the driver tube close to the last power supply filter stage. This minimizes the possibility of additional hum being induced in the run from supply to driver. You’ll also notice the dashed gray line across the center of the amp. This represents a metal shield between the power supply and audio sections of the amplifier. This is to help minimize any interactions inside the amplifier.

In addition to wanting this amp to look good, I also really wanted this amp to be quiet and fairly well shielded. Since I had already decided that the base would be lacewood, this meant lots of metal. In addition to the top plate and base plate, it has a metal divider, a hum shield inside the front wall, metal plates for the inputs and outputs, and a metal mounting plate for the AC power input, fuse, and switch. This translated to a large amount of 80mil 6061-T6 aluminum plate. Here are all the aluminum plates cut and marked for drilling.

I had already done a fair amount with the 6SN7 and the 6V6 design point came off the data sheet so I didn’t feel like I really needed to prototype the amplifier circuit itself. The power supply on the other hand was something I really needed to test out.

I needed to confirm that I could get the very low ripple numbers I needed from my design. So parts in hand, I turned to my prototyping station and went to work. Below is a picture of the entire power supply connected to a 2kΩ load resistor.

The only thing omitted from this prototype is the 470kΩ bleeder resistor and the 5v filament supply for the 5U4GB is coming from the prototyping station. The total ripple measured at the resistor in this setup was about 100µV or -128dBv. Not quite the -171dBv calculated above but quite remarkable for a prototype of this nature. In fact, so remarkable that I checked the reading several times. It was however VERY sensitive to lead placement.

So my power supply design was good. Time to layout the amplifier.

To assemble the amp, the top plate was secured in place so that the wiring could be accomplished from below. With the top plate installed all the major components were mounted. Here is a picture from the underside.

Once drilled, painted, and installed on the wood chassis it was time to actually start the build. Here is a picture of all the smaller metal pieces installed on the chassis.

From this point, assembly was fairly straight forward. Just follow the schematics and do the wiring point to point. Every metal plate in the chassis has a point to tie all the parts together. This ensures a continuous electrical shield. The signal ground for the amp is tied to the chassis ground at only one point to avoid ground loops. Here is the finished wiring.

There are several features that help make this amp work well. The first is the bare copper wire ground bus. This serves as the main signal ground for the amp and connects to the power supply behind the shield. By tying all signal grounds to this bus, it helps to prevent ground loops. The second feature is the filament wiring. This is solid 20 AWG twisted pair that is very carefully laid out. Note how it follows the corners of the chassis whenever possible and when it feeds the sockets it actually loops up and comes down on the socket. This places it largely near perpendicular to the other socket wires helping to minimize interactions. Third, the signal inputs and the interstage signals are carried by twisted pair wring. Again, this helps to minimize interactions.

With the amp all wired up it was time for some testing. It turned out even better than I thought it would. First, the amp is dead quiet. With my iPod connected and paused, at full volume I could not detect any audible hum on the output. An oscilloscope hooked across the speaker terminals confirmed this as well. The amp produces 3.2W at full power into an 8Ω load. This is about three quarters of a dB below the data sheet number for the pentode wiring, but I expected a little loss when going to UL mode.

The amp was designed for flat response from 20Hz to 20kHz (even thought my hearing is nowhere hear that good anymore) but the Edcor output transformers are only rated down to 40Hz. They are still very clean at 20Hz but the signal rolls off below 40Hz and is about 4.5dB down at 20Hz.

I think this amp was a success for me on several different levels.

First, the amp sounds great. Whether playing soft jazz, swing, classical, opera, or blues, it all sounds excellent. The amp sits in my work area and is alternatively fed by my computer, my iPod, or my portable CD player all with great results. This is exactly what I wanted for this space.

Second, the aesthetics of the amp work well. It is still traditional and understated in design, but the materials and theme work well together. The black and brushed silver create a nice feel and the lacewood is definitely something out of the ordinary. Here is a close up of one corner of the amp showing how it all comes together.

Lastly, I think I have sufficient evidence to support my hypothesis that AC heater supplies on unipotential cathode tubes do not induce hum in the audio chain. This amp uses 6.3 volt AC heaters with the transformer winding center tap tied directly to the single point ground. If there was going to be any interaction between the signal path and the heaters, I would have heard it in this amp.

This is a success. It looks good, it sounds good, and if does exactly what it was intended to do. With several hundred hours of listening now behind me I find myself liking this amp more and more each time I listen to it. I’m finding things in old familiar music that I never knew were there and I find myself exploring new music with ever greater enthusiasm. I couldn’t ask for more from an amplifier; nor anything I’ve built with my own hands.