5687 Aikido Kits and More Circlotron Power Amplifiers

07 April 2008

More Circlotrons

More Circlotron Circuits
I had a good time revisiting the Circlotron topology last time, but there was much more that I hoped to cover. So let’s start at the beginning again and work our way up to something more complex. Below is a simple Circlotron circuit. The two 1k resistors define a two-resistor voltage divider that sets the signal reference at the midpoint between both cathodes.

If both cathodes sit at 0V, then no current flows through the two resistors (or through an external load). When this output stage is putting some power into the load impedance, these two resistors will see a voltage differential and will conduct current and dissipate heat, although not much heat. For example, with a constant 100W into an 8-ohm load, each 1k resistor will dissipate only 0.2W. On the other hand, if one of the triodes shorts from cathode to plate, the full 100V will be developed across both resistors in series, so each resistor would then dissipate only 2.5W. Interestingly enough, I have seen commercially made Circlotron power amplifiers that used huge, high-wattage 1k resistors. Why?

My guess is that there is some confusion over what these resistors do in the circuit. Ready for this? I once saw a circuit for creating phi (magic) energy from a Circlotron-like topology, as shown below.

The idea was that the 1-ohm resistor actually did see a current flow, even in the absence of a voltage differential across the resistor. It was believed that because two opposing streams of current flowed through the 1-ohm resistor, they cancelled out any voltage drop across the resistor; and that having two 1A streams of current flow against each other created a bunch of excess “phi” energy, which would radiate out to the benefit of the pot-smoking world. (Don’t laugh, it’s got be class-A, remember what the glossy ads say.)

The same source also showed a passive phi-energy generator that consisted of a thick sheet of copper rolled and flexed into an arc and held in place by thick nylon belts. The idea was that the stressed metal generated phi energy and the arc shape allowed it to be focused much as a parabolic mirror would. When I stopped laughing, I pointed out that the bottom floor of every skyscraper must be chockfull of phi energy, as the weight of the building produced a compression stress on all the steel beams, and most so at the bottom of the building.

Sad to say, but many audiophiles are much closer to New-Age space cadets than they imagine. Remember Peter Belts and his fantastic discoveries. Speaking of which, it would make a great book (or website) that chronicled the history of whacked audio nonsense. For example, I remember reading in the late Hi-Fi Answers magazine from England that a book’s number of pages influenced its sonic contribution to the listening room it found itself in and that either an odd or even number of pages made all the difference.

Okay, here on planet Earth, let’s get back to tube circuits. The two-resistor voltage divider is not essential, as it could be replaced by a center-tapped inductor, as shown below.

Grounding the center-tap also places the signal reference at the midpoint between the two outputs. Moreover, if a perfect inductor were used, no heat would be dissipated by the inductor, even if the amplifier were putting out 1kW into an 8-ohm load. Unfortunately, real inductors use real wire and hold real metal cores, so the wire’s resistance and the core’s losses due to hysteresis and eddy currents will generate heat. But our aim here is not to find a more-expensive substitute for the two 1k resistors, but a conceptual steppingstone to the next circuit shown below.

The Circlotron shown above uses an output transformer to couple to the external load and it still uses the two-resistor voltage divider to reference the input signals, although a center-tapped primary would also work. Below, we see the AC voltage relationships in the output stage.

Now let’s make a big jump to using two output transformers, each holding the same winding ratio, as both will see the same AC voltages and current flows and impedances. The two secondaries are wired in parallel, so as far as the triodes are concerned, the same load is presented as in the first transformer-coupled example (assuming the same output transformer is used).

Now we get rid of the two floating power supplies and replace them with two large-valued capacitors and a single, ground-referenced power supply. (Center-tapped output transformers are necessary.)

Notice that the AC voltage relationships are the same within the output stage as they were with the single output transformer and the two power supplies.

On the plus side, we now can use a single power supply (perhaps tube-rectified) for not just one channel, but for as many channels that we wish to add: two, three, four…. In addition, we can now use first-rate output tubes, such as the 2A3, 6L6, 12B4, 211, 845, 300B, EL34, EL84, KT88…rather than being stuck with the clumsy 6AS7. Nor do we have to fear losing our expensive loudspeakers when a tube shorts out, as the output transformer will save the speakers from catching on fire.

On the negative side, we retreated from OTL, which holds a small marketing cachet, which is somewhat offset by the fear of losing one’s loudspeakers in a mishap. Second, we have two output transformers instead of one to deal with. This is actually an advantage in my eyes, as we can get 40W from two 20W output transformers in parallel, but it is easier to find great low-power output transformers than high-output transformers and it will be easier to hide inside the chassis two smaller output transformers than one large one. Third, we have to use separate heater power supplies for the output tubes, as each has a big-voltage-swinging cathode, which would run up against the fairly low cathode-to-heater voltage limit (with DHTs, that voltage limit is 0V). Still, small 6.3V/2A power transformers are readily available, so this is not a big negative. Finally, we have to come up with two large-valued cross-coupling capacitors to take the place of the two power supplies that would otherwise be used.

Floating Power Supplies
Once again, I view this as a feature. One of the many big problems with using two floating power supplies is the temptation to treat them as something other than a power supply. In other words, many simply assume that the floating power supply is a different kind of beast from the typical clunky, noisy power supply that normal push-pull amplifiers use. It’s not. All the problems that come with a power supply are still there and they still make a big difference in the sonic fingerprint left by the amplifier. At least with the two large-valued cross-coupling capacitors, we recognize them as being critical and we are not likely to skimp on their quality. In Addition, floating power supplies are only floating in term of DC voltages, but because of the capacitance between transformer windings, they are always slightly capacitively coupled to the primary and core; something that the two large-valued cross-coupling capacitors bypass.

Now we get to the interesting part: Why bother with the added complexity of two output transformers and the two large-valued cross-coupling capacitors? Why not just stick to the conventional transformer-coupled push-pull amplifier?

If you plan on running your conventional transformer-coupled push-pull amplifier in pure, honest-to-goodness class-A, then the added complexity is probably not worth it. On the other hand, if you plan on getting a few years of life out of your output tubes and you don’t want to add to tube-induced-living-room warming, then class-AB is the best choice and this variation on the Circlotron is a good idea. Why?

With conventional, class-AB, transformer-coupled, push-pull amplifiers, we have the problem of snapping primary windings. In other words, as one output tube turns off, it effectively lets go of its side of the primary, which causes half of the primary to left flapping in the void—never a good idea with any inductive circuit.

This is why we have to be careful where and how we use a fuse with a choke-filtered power supply, as a blown fuse can instantly unload an energized inductor. Now, think about it, a capacitor stores an electrical charge, which can easily prove dangerous, if shorted; whereas an inductor stores energy in the form of flowing current and can prove dangerous, when the circuit is opened, as the stored energy must be released somehow. That somehow usually takes the form of a big spark, as the voltage must climb towards infinity because the current has fallen to zero. Sometimes the spark occurs inside your output tube or other amplifier parts, depending how the circuit is arranged.

Well, returning to power amplifiers, in the two transformer Circlotrons, both primaries always have something to bite on, as the crisscrossing capacitors effectively place both primaries in parallel with both output tubes, so neither winding is ever left flapping, even as one tube ceases to conduct. Additionally, the output stage is configured as a modified cathode follower output stage, so we can expect the secondary signal to be both clean and very low in output impedance. Unfortunately, nothing comes without a price, so we must pay in the form of a fairly large drive signal for the output stage.

Precisely how much drive voltage? A good starting point is the cathode voltage swing and the absolute value of the negative bias voltage added together. And this drive signal must be balanced. In other words, we need to build an electrostatic headphone amplifier of sorts to drive the output stage.

Dynaco ST-70
I think that the Dynaco ST-70 would make an excellent platform, as it comes with two output transformers and four octal tube sockets. In other words, we could convert this stereo amplifier into a mono-bloc amplifier. Instead of using four EL34s, we could use two KT88s; instead of the silly stock driver PCB, we could use one 6SL7 and one 6SN7 and use the PCB space for a board that would hold the supporting parts. With triode-connected KT88s, we could expect to get about 20W of sweet output. In other words, it would take some work, but not a near-infinite amount and no chassis cutting or drilling would be required; in addition, I assume that you, like me, have a few just sitting in your garage.

With triode-connected EL34s, we would get about 15W of super-sweet output; ultra-linear-pentode connected EL34s, we should expect more watts, but not the full 35W of the original—we will come to this aspect later. This raises the issue of what to do with the ultra-linear taps on the output transformers. One idea would be to use them, but not in the way that Dynaco had intended, as shown below. Leaving out the the two large-valued cross-coupling capacitors, we can see the UL connections more clearly.