27 May 2006

Hybrid Promise

Tubes amplify sweetly; solid-state output devices slam and punch. Combined, the tube input stage bestows the delicate signal to the solid-state output stage, which then robustly relays the signal to the loudspeakers or headphones. So the proposition compels us: the notion of using vacuum tubes only for voltage gain, and solid-state devices only for current gain.

It’s a division of labor that lets each device do what it does best. At least that’s the seductive promise. But can it be fulfilled? Remember, the two technologies work in series, not in parallel. Ask yourself which would taste best, wine from a bottle filled by a glass pipe or by a garden hose or by a glass pipe connected to a garden hose or from glass pipe and garden hose simultaneously? How much would it matter if the glass game first or the garden hose? Series is series.

Here’s a metaphor: Imagine that a fierce debate whirled in telescope circles over glass lenses versus plastic lenses, with one group arguing that glass lenses revealed more and corrupted less than plastic lenses, which were said to rendered a plastic-like sheen to all that was viewed and induce headaches after an hour of gazing through them; while the other group claimed that plastic lenses were not only much cheaper, more modern, and durable than glass lenses, they were superior in that really large lenses could readily and cheaply be made from plastic, which would be impossible or obscenely expensive with heavy, brittle glass, which couldn’t be injection molded and had to be ground and which always held some slight tint that colored the view.

While this debate raged, 99.99% of lenses would be made from plastic. Yet, many ads in the telescope press would portray their plastic lenses as being as clear and as enjoyable to look through as glass; none would claim that their glass lenses viewed like plastic. Then one day, someone comes up with the idea of making a telescope whose small eyepiece was made from glass, but whose final large lens was made from plastic; the best of both worlds, the ads proclaim. But would it be?

To switch metaphors one last time, would you still like the good cop, if all that he said to you had to be relayed through the bad cop?

A little paranoia isn’t always bad and, in fact, it can be helpful.

So are hybrid amplifiers always a bad idea? (In college, I had friend who divided the whole of existence with a guillotine-like slice into either great or garbage; there was nothing in between for him. Only the purest white; the purest black. Oddly enough, he was seldom happy.) Since I have listened to a few good-sounding hybrid power amplifiers and hybrid headphone amplifiers built on this premise, I know it is possible to make a wonderful-sounding hybrid amplifier for a fraction of the cost of an all-tube amplifier of comparable power output. Will it be perfect? I doubt that anything made by human hands can be perfect, but it might be quite enjoyable to listen to music through it.

Amplifier clipping
I have been thinking about how the solid-state output stage differs from the tube output stage and my thoughts turn to clipping. All amplifiers clip, but some amplifier clip egregiously. Square waves, with their bristly high-frequency harmonics, never soothe the savage breast. On the other hand, because most tube-based power amplifiers use an output transformer, their clipping is much kinder and gentler than most solid-state power amplifiers. Why? The transformer functions like a bandpass filter, excluding the lowest and the highest frequencies. A square wave may go into the transformer, but it is not going to come out square. And yes, in spite of what too many falsely believe, a tube can create a square wave. Moreover, the output transformer limits the amount of negative feedback that can be safely applied to the transformer-coupled amplifier, which in turn limits the severity of the clipping. The feedback loop does not know that the amplifier has run out of steam; it only knows that the amplifier’s output no longer matches its input and it does all it can to bring the output in line, even if the output signal takes on a sharp square edge in trying.

In other words, we can feed the solid-state output stage the sweetest signal, but when the output stage clips, all our much-loved sweetness will be sawed off, leaving only a jagged end.

At first it seems there are only two possible ways out: either build a solid-state output stage so powerful that it can never clip or turn down the volume so that clipping never occurs. Or is there a third way? What if we pre-clip the tube’s output signal before it reaches the solid-state output stage? In other words, what if we softly round off the peaks of the signal before the solid-state output stage relays it to our loudspeakers? In this way, the solid-state output stage would be faithfully passing on a softly distorted signal, rather than it sharply distorting a large pure signal. Sounds like a great idea, so where’s the catch?

The catch lies in the sacrificed potential power output from the solid-state output stage. For example, a solid-state output stage that might have cleanly passed a 100W signal might only put out a maximum of 80W under this scheme. What—give up 20 watts!? Are you mad? Yes, I know that for many audiophiles this suggestion would be as welcome as penis-reduction surgery. Still, there is much merit to this idea: the ear can barely notice the difference between 80W and 100W, but it is supremely sensitive to the harsh harmonics generated by a clipped signal.

Soft-clipping circuit
First of all, not all solid-state output stages clip in the same way. Some clip much more gracefully. How so? 99.9% of all solid-state amplifiers use three stages: an input stage that provides gain and a feedback input, a second stage that provides gain (VAS, voltage amplification stage) and current gain, and a unity-gain buff output stage. Now, if the second stage holds an output-referenced cascode topology, this stage will help make for much softer clipping, as it runs out of gain as the output approaches clipping. Thus, even if the input stage tries to force a severe square edge on the clipped output, the second stage cannot pass that intention to the output stage. So, why isn’t this topology more popular? Well, it entails a reduction in potential power output, which is as popular as… well, you remember what kind of surgery.

Rather than make the tube circuit softly clip, we can pre-clip its output passively. In the schematic below, we can see how the signal cleanly passes through a 10k resistor at low amplitudes, as none of the signal diodes are forward biases, but gets rounded off as the signal approaches the rail voltages, as the diodes start to conduct.

As long as the input signal falls within the +/- 28V window, it is passed without rounding. once it goes outside this window, the diodes become progressively conductive and the signal's peaks are rounded off. In the graph below, we see a SPICE generated graph of the above soft-clipping circuit under test.

The blue trace is the result of a 28V input signal; the red trace, a 32V signal. Assuming that the unity-gain power buffer output stage could deliver a maximum of 30V of signal into an 8-ohm load, the output wattage would equal 56W. On the other hand, 28V of peak output signal equals only 49W. How important are those missing 7 watts of power? Not nearly as important as getting rid of the sharp clipping shown below from a 32V input signal.

Next time
More unity-gain buffer circuits—including an all-tube version.

//JRB