Tubes Do Something Special Letters
Tubes do sound louder
Editor: I can directly confirm Peter van Willenswaard's observation that tubes sound louder than solid-state amplifiers (September 2000, pp.55-59). The excellent graphs Peter provided allow me to make an educated guess as to what is going on below the surface.
Back in 1981, when MOSFET amplifiers were all the rage, I borrowed a 75W Nikko from my employer. Its competition was a Leak Point One with EL84 output tubes, weighing in at about 15W. I hooked up the Nikko, put on a record, and went about my business.
Suddenly, the Nikko went AWK! and shut down. I turned it off and checked for damage. Negative, so I gingerly turned it back on, and started the record again. Same thing. Finally, I sat down to watch the amplifier. At exactly the same place that the amp had failed, both power meters lit up like flares and the amp went into protection. That amp, at five times the specified power of my little Point One, couldn't approach the level of loudness that the Leak produced!
Since then I have found that many MOSFET amplifiers often appear much less dynamic than tube or bipolar designs. I attribute this to problems with the driver stage, which, in MOSFETs, is difficult to design.
Turning to PvW's curves, it is obvious from figs.2 and 4 that the tube amp will sound much louder than the transistor unit. Our ears integrate sound levels; they are far more sensitive to power than to amplitude. Thus, the amp with more area between the zero line and the curve will sound louder.
Now here's the zinger: The positive swing of the 300b amplifier is obviously grossly distorted. There are compensating mechanisms at work here, but still, why doesn't Mr. van Willenswaard notice the distortion? I simply have no explanation.
The distortion itself comes from overdriving the control grid. This grid is normally maintained negative by a special power supply in the amplifier. However, if the input signal goes sufficiently positive, it is possible to counteract, and even reverse, this voltage. As long as the voltage stays negative, the grid can control the action of the tube while consuming no power of its own. The horizontal lines in the graph begin at the exact moment the voltage turns positive. The driver stage, which is not normally designed to deliver power, can no longer drive the grid more positive. Notice the efficiency of the process: As the driver stage tries to increase the voltage, the net reaction of the grid is to go still more negative, and the line actually tilts down.
The solution to this problem is to design a driver stage capable of supplying considerable power to the input circuit of the output tube. This is known as class-A2 operation.
The reason Mr. van Willenswaard found different peak outputs depending on whether the amp was connected to a resistance load, to speaker A, or to speaker B lies in the amount of energy stored in the system. A resistor is incapable of storing energy, so Mr. van Willenswaard found few differences in the amplifiers he checked with a resistor. Add a speaker, and the picture changes radically. Speakers possess considerable inductance; thus they can store a fair amount of energy. You would not expect speakers of different design to store identical amounts of energy, and that is what Mr. van Willenswaard found: The more inductive the speaker, the more energy was stored.
The second factor to be considered is how closely the amplifier can control the speaker. The property that determines this is the damping factor. The source impedance is merely the inverse of this number. As soon as Mr. van Willenswaard equalized the source impedance (by adding resistance) of all the amplifiers, there was little difference in the graphs they produced.
I hope this clears up some of the mystery surrounding vacuum-tube electronics. It was once a dynamic, vital technology, but now is mostly relegated to owners of high-end equipment, guitar amplifiers, and microwave ovens. Good listening!—Bob McIntyre, Toledo, OH, firstname.lastname@example.org
Thank you Mr. McIntyre, and I agree with most of your observations. But could we conclude then, that amplifiers that have near to zero output impedance, waste energy that the loudspeaker has in store and is ready to use for building up extra sound pressure, by shorting this energy out?—Peter van Willenswaard
Editor: The transistor amplifier used in the tests described in Peter Van Willenswaard's "Tubes Do Something Special" (September 2000) is not a 25W amplifier. An amp that clips at 17V peak into 8 ohms (fig.1) is delivering 18W. Since the amp is clipping, they are definitely not "clean" watts. I would conservatively rate this amp at 16W.
Referring to fig.4 of that article, I see a 300B amp that is clipping asymmetrically at 20V positive and 36V negative. Yet PvW says that he "could hear nothing at all problematic"!!! Either Peter's eardrums are made of concrete or he has a (very!) highly developed imagination.—J.A. Arbona, email@example.com
Mr. Arbona's calculation is correct. I hadn't used this solid-state amplifier for more than 15 years. I seemed to remember it was 25W, but it proved to be less than that. Anyway, most of the comparisons I did in the article were relative and in voltages, so it changes nothing about the point I was trying to make.
Regarding Mr. Arbona's second remark, we know very little about the relationship between what a waveform on a 'scope looks like and what we perceive when the ear hears it. Maybe it would have been different with a soprano voice, but with a drum stick hitting a tambourine, it was just as I reported.—Peter van Willenswaard
Editor: After so many years of hearing the same misconceptions again and again, I feel a clarification is in order.
The first misconception is that the particular sound of an SET (single-ended triode) amplifier is due to its tube's linearity. Nothing is further from the truth. The pentode was developed exactly because of the triode's nonlinearity and its excessive second-harmonic distortion. The kinetic energy produced by the acceleration of electrons toward the plate produces a secondary emission of those electrons repelled from the plate, which are thus attracted by the positively charged screen grid, creating a reverse current to the screen. Those electrons with less kinetic energy remain by the plate, attracted back to it, and produce a variation in plate current that is not voltage- (signal-)change-related, and that, my friends, is distortion.
The pentode's third suppressor grid solves that problem. So much for the triode's linearity. The design and other problems and characteristics of tubes are more technical, but can be found in any good book on tubes. The sound so beloved by SET owners is probably due to its very large second-harmonic distortion ingredient. This makes it sound "warm," whereas in the pentode, second and third harmonics are almost identical in level, and usually banished by that dreaded negative feedback.
To say that a triode sound is "high fidelity" is therefore an oxymoron. The distortion may reinforce the fundamental, creating the sense of a dynamic sound, openness, and all the other accolades and apotheoses, but that is a personal liking; it is certainly not how the original music sounded. A triode makes its own music. It's like a musical instrument, and some like it a lot, but it's neither hi-fi nor linear.
The other "concept" or "definition" aired so many times is that of push-pull class-A operation. There's no such animal. The term is probably used by designers/manufacturers because class-A operation holds some fascination with audiophiles, but the truth is being stretched by close to 180 degrees. By definition, class-A involves a single-ended tube. You may have two or more paralleled, but each of those handles the whole signal, the entire 360 degrees of it. This is how the classification was made, the basis of which is called the plate conduction angle. That angle is the duration of plate current flow relative to the input sinewave cycle. So in a class-A design, the plate current will flow throughout the 360 degrees of the input (signal) sinewave.
There are different designs in push-pull amps. Some will make the current flow for longer than 180 degrees on each tube of the pair, but always much less than the full 360 degrees. This is class-AB by definition. (Class-B is 180 degrees exactly for each tube in a push-pull design.) Period. In some push-pull designs, the current—but not the input—can be applied for a longer duration or made constant; in others, at less than half power the push-pull amp will almost attain class-A operation, but it is still not class-A. These are variations on the AB theme. The same holds true for solid-state designs.
Until we discover a different Ohm's Law, that's how it is.—Shalom Noury, Cagliari, Italy, firstname.lastname@example.org
Screen grids don't happen in triodes, Mr. Noury. You'll find them in tetrodes, though, which is where they do capture the stray electrons you mention and hence cause the "kink" in the tube characteristic. The solution was the addition of a suppressor grid, which led to the pentode. So the bounced-off electrons have nothing to do with the triode's nonlinearity, which itself is a direct consequence of the Child-Langmuir space-charge law by which the triode operates.
Pentodes certainly are a lot more linear than the "kinky" tetrodes they descend from, but hardly more linear than triodes when their input characteristics are compared. Comparing their output characteristics, one can see that triodes come equipped with a healthy internal (local) feedback mechanism that pentodes lack. Pentode circuits tend to favor odd-order distortions, triode SE circuits display (almost) only second-order harmonics. So if one chooses between triodes and pentodes, one primarily chooses between types of nonlinearity; presenting such a choice as one between nonlinear and linear is demagogy. And the main reason most designers choose pentodes over triodes is that pentodes are at least twice as efficient as triodes—more power for the same money, a sales argument.
Second, push-pull output stages can most certainly be class-A. If, for a given load impedance, the bias current through the output devices is chosen so that neither output device ever runs out of current even at maximum voltage swing (defined by the power supply), the output stage, by definition, operates in class-A, as both output halves are conducting at all times.—Peter van Willenswaard