I lived with my old single feed headphone amplifier, in one form or another, for a while. However, two events caused me to start rethinking the design.
The first was that, after designing the Less-pressivo, I started working on upgrades to it. One of these upgrades was the procurement of a pair of Electra-Print parafeed transformers to use on the output to replace the cheapies. And, the addition of this new iron was, quite simply, transformative. Even without a lot of tinkering or changing of parts, it sounded better than my single feed amp. This was a problem for a couple of reasons: the first being that I had a lot of money and time and effort sunk into the single feed amp so it felt like a real failure to have it bettered by something so inexpensive. The second is that I am not really a fan of many of the evangelists of parafeed. For various reasons which don't deserve mention, I am rubbed the wrong way by some of them.
The second event, which was frankly more serious, was that there was a very slight low level hum to the single ended amp that I couldn't get rid of. It was low to the point that I could only hear it late at night when the world was silent, but it was just enough to be annoying. After a lot of experimentation, I finally traced it down to the proximity of the power transformer to the output transformers. While I found that moving the power transformer a mere 2cm further away eliminated the hum, the chassis didn't have 2cm of extra space. So, since I was going to need to redo the chassis anyhow, I thought it might be a good opportunity to begin redesigning.
The first step was to get rid of the solid state regulator. The reasons for doing this were in part that it had, at times, proven a little unstable in the past. Additionally, since there was only one regulator, it was effectively feeding both channels and, in theory, leading to cross channel interaction. I decided to replace it with a constant current source plus regulator tube, i.e., a shunt regulator. I'd seen this done in push pull and parafeed amps, but never in single feed. Plus, I had a box of regulator tubes, so I figured I should try them out.
In a single feed amp, there are two primary weaknesses. The first is that there is an air gap in the output transformer. This keeps the transformer from saturating from the DC current through it, but it also means that the windings are not as well coupled as they might be. Without extreme measures, there isn't much that can be done about this. The second issue is that the final power supply capacitor, which is often a large electrolytic, as well as the cathode bypass capacitor, which is an even larger electrolytic, are both directly in the signal path. One of the common remedies for this problem is the use of a so-called ultrapath capacitor which connects directly from B+ to the cathode effectively bypassing both of these. This design, in parafeed form, goes back to Western Electric, but in a single feed design was more recently championed by Jack at Electra-Print. I decided to try this here as I had had good luck with it in other places.
One of the problems with Ultrapath is that any power supply noise is injected into the cathode which amplifies this noise by mu. In an output stage with a low mu power tube, this isn't an issue, but here I was using a 7788/E810F which is a very high mu tube. Interestingly, the shunt regulator had cut the ripple down to almost 0 so it wasn't an issue, but the regulator itself had a small amount of wideband noise which caused the amp to hiss ... loudly.
A common solution in ultrapath designs is to add a cathode bypass cap which shunts this noise to ground but which stays, effectively, out of the signal path. I tried this, adding a 1000uF cap.
Two things happened which were both important. The first is that the noise was indeed more or less eliminated. The second, however, is that the character of the amplifier changed completely. Indeed, it sounded like there was a big cheap electrolytic capacitor in the signal path, and the bass and gain increased considerably. As an experiment, I replaced the generic electrolytic with a 680uF Blackgate NX. The character of the amp changed again, this time sounding like a Blackgate. This suggested that two important things were going on -- The first, and more obvious, is that the cathode bypass cap was in the signal path. Second, by implication, the Ultrapath cap was not doing it's job. That is, rather than the AC signal being shorted from the cathode to B+, the signal was finding another path which was through the regulator tube.
So, as a first step, I simply removed the Ultrapath cap which left only the regulator and its 0.1uF bypass cap for the return signal. This was a good step. Second, I replaced the cathode bias resistor as well as the cathode bypass cap (the Blackgate) with a green LED. My experience is that this is generally a superior, and less expensive, way to bias a tube, and this case did not prove any different. The upside of this is that the regulator's bypass cap is effectively the only cap in the signal path. (This is true in theory -- in practice, the CCS and shunt regulator do not work perfectly, so the power supply does still have some influence on the sound, though that influence is considerably reduced from other designs. For instance, since the final PS cap is located pretty far from the CCS, I found I needed a 0.1uF cap on the anode of the CCS. Additionally, a Wondercap (the predecessor to the current Dynamicap) sounds better here than did a ceramic cap -- though the difference is very subtle, and anything more than this is likely unnecessary)
This was certainly a good enough amp to live with. It solved the hum issue of it's predecessor and was a considerable step up from the Less-pressivo with the iron upgrade. However, after a suggestion from rdf on DIYAudio.com, it seemed there was one additional refinement to be made. This was to move the base of the regulator and its bypass to the cathode, effectively creating a short circuit around the biasing LED. The LED was also replaced with a resistor. Because the current through the resistor is constant, regardless of changes in current through the tube, a resistor acts as a fixed bias mechanism. Use of a wirewound enabled a fairly precise bias setting.
This made this a very good amplifier. For me, at this point, moving to better performance is going to require some real essoterica.
|Inside (more caps were ultimately added)|
|Power Supply (Rectifier and CLC Filter)|
So, I think this circuit deserves a good name. Many of the circuits people build get cute names: Ultrapath, Monkey on a Stick, Free Lunch, Aikido, Red Light District, etc. My first thought was that since this is kind of like an Ultrapath, but since it uses a glow tube for the path, that Shining Path would be funny ... then I looked up who the Shining Path were/are, and nixed this as being in very bad taste. The next idea was the Electric Avenue. The problem here is that while this kind of gives the impression of a brightly lit path, it isn't really that descriptive of the circuit. Anyway, until I (or you -- please email suggestions) come up with something better, it will be named for the great Eddy Grant ...