In the world of tube audio people often talk about “class A”, “class AB” or “class B” as if these terms were synonymous with sound quality or absolute power. In reality, the operating class only defines how the tube works relative to the signal: how long during the cycle of a wave it remains conductive, how close it is pushed to the limit, and how efficiently electrical power is converted into useful audio power.

Behind these seemingly simple acronyms lie precise physical concepts, but also many widespread myths and simplifications. Some claim that class A “sounds better because it is pure,” others believe that a push pull in class AB is automatically inferior, or they declare fanciful power figures that defy the laws of thermodynamics.

In this article we will clearly see what each operating class really means, how performance varies depending on the type of tube and its connection (triode, tetrode or pentode, with or without ultralinear), and above all which power values are realistic for a given circuit. The goal is not to defend one “class” against another, but to provide concrete tools to separate technique from fluff, and to understand what to really expect from a well designed tube amplifier.

Operating classes explained without too much philosophy

In a tube amplifier the “operating class” indicates how long the tube conducts current during the audio signal cycle. Simply put, in class A the tube always conducts, even when the signal is zero. In class AB it conducts for more than half a cycle, but not always. In class B it conducts only for half a cycle, and the other half is handled by the complementary tube.

In modern Hi Fi, true class B is not used, and should not be used, because it introduces very noticeable “crossover” distortion. There are also other special classes: class C, which concerns only RF transmitters and has no audio applications, and class D, typical of switching solid state amplifiers, not tube amps.

A single ended amplifier can only be class A, for a very simple reason: there is only one tube that amplifies the entire signal, so it must always remain conductive, even at the lowest point of the wave, otherwise part of the signal would disappear.

A push pull, on the other hand, can operate either in class A or in class AB. If it is in class A, both tubes always conduct for the entire cycle, but unlike a single ended, the DC currents cancel out in the transformer. The power obtainable is the same as a parallel single ended, but with a transformer free of DC currents, so it works better and more linearly.

And here we get to the notorious concept of “pure class A”, which is simply a piece of nonsense invented by marketing. There is no such thing as “pure class A,” because the distinction between class A and class AB is already clear and sufficient. Many people confuse the two by saying, “well then you find out it is class A only up to a certain point, then no longer.” In that case it is class AB, and there is nothing strange about it, that is exactly why the concept of class AB exists. That is, many people do not realize that an amplifier that works in class A up to a certain point is, in every respect, a class AB amplifier. There are not two different things: this is exactly how class AB works, part of the signal is amplified in A, then beyond a certain threshold it gradually enters the B region. Saying “it works in A up to a certain point” is simply describing class AB in other words. And that is precisely why it makes no sense to talk about “pure class A.” Saying simply class A already implies that the amplifier never goes beyond the point where one of the tubes stops conducting. The very definition of class A by nature excludes any transition to B, so adding “pure” is only redundant or rather, a marketing trick, a phrase with no technical meaning.

In a push pull, moreover, it is practically impossible to remain 100 percent in class A. As you approach the power limit, just before clipping, there will always be a small zone where the circuit moves into AB, due to the inevitable tolerances of the tubes or slight imperfections in biasing. But this also applies to single ended, which in theory should always be class A. In practice they are never perfect. Because of the nonlinearity of the tube and the characteristics of the transformer, clipping in a single ended is inherently asymmetrical. So even there, near the power limit the operation becomes slightly unbalanced, behaving for a moment as if it were class AB, even though there is no second tube to compensate it.

Worrying about this is a waste of time. No one listens to music with the amplifier under continuous full load, and all the watts delivered before that point are and remain class A, with the sound and characteristics typical of that class. If you have a 20 watt amplifier that works in class A for 19.5 watts and then moves into AB only on the last half watt, it will never sound different from one that stays in perfect A to the end. Among other things, the difference can also depend only on the type of tubes installed. Just change a pair and the transition point shifts.

Once, while explaining this concept over the phone to a client, he told me, “I know the difference between A and AB, but many say A and then you find out it is A only up to a certain point, so saying pure class A serves to say it is perfect A for the whole power.”

My answer was, “Look, from a push pull of KT88 in class A you get about 20 to 22 watts. On Facebook I once saw people talking about a KT88 push pull amplifier of 80 watts in pure class A. That is not physically possible, that was a class AB amplifier, period. And if you tried to discuss it technically, the seller got annoyed and hit you with the classic well you do not know how it is done… yes, of course I know. You make a class AB and then claim it is ‘pure class A’.”

The conclusion is simple. Tube amplifiers can be class A or class AB, and “pure class A” does not exist. It is just a label invented by crafty sellers and fairground gurus to fool those who lack the tools to understand the difference.

Watts: RMS, peak and… the “musical” kind

When talking about power in amplifiers, the term “watt” is often used casually, and not always knowingly. In reality there are several ways to measure or state power, and understanding the difference is essential to avoid being fooled.

RMS power (Root Mean Square) is the only serious and technically correct measure. It represents the average power actually delivered by the amplifier on a resistive load, without distortion and under continuous operating conditions. This is what really matters, because it indicates how much the amplifier really “pushes” in normal use.

Peak power or peak to peak (PeP) refers to the instantaneous maximum values of the signal, the highest point of the positive wave and the lowest of the negative wave. It is a value much larger than RMS power, about eight times higher but it does not represent the real continuous power of the amplifier, it represents only the maximum excursion of the signal before clipping.

Saying “this amplifier delivers 160 watts PeP” looks fancier than “20 watts RMS,” but it is basically the same thing. Only the way the number is expressed changes, not the actual power.
The PeP value has its technical meaning in certain measurement contexts, so as a definition it exists and is correct, but as usual, marketing has appropriated it to inflate the specs and make everything look more powerful than it really is.

And then there are the infamous “musical watts”, a marketing invention of the seventies and eighties to make everything look more powerful. In practice, “musical watts” are simply RMS watts multiplied by two. That is all. No formula, no scientific measurement, just a doubled number to write more on the label. So an amplifier that actually delivers 10 watts RMS magically becomes a “20 musical watt” unit. A fair trick, good only to confuse those who do not have the tools to understand the difference.

In short:

• RMS watts: real continuous power.
• Peak watts: instantaneous maximum before distortion.
• Musical watts: a marketing gimmick with no technical basis, just inflated numbers to sell more.

Invented power figures and datasheet miracles

Another classic in the tube audio world is the statement of completely invented power figures, often based on fanciful calculations or on a naive reading, if not deliberately crafty, of tube datasheets. How to defend yourself from unfair commercial practices by certain manufacturers of tube equipment, who often take advantage of the limited technical experience of some users to spread confusing or misleading information. Recently I received yet another email from an enthusiast who asked me for clarification on exactly this kind of situation:

Hi my name is *** recently I asked on a forum about a schematic with EL34s, a user recommended your SE project with EL34s, the one driven by 6SL7 in totem, because he built it and said it works very well, but you write that it is a 7 watt and that is not enough for me. I instead found this schematic (censored schematic) and it is the only 15 watt one I found and I built it with EL34s instead of KT88s, but it does not satisfy me. Since they spoke highly of you, can you explain the difference, maybe because this schematic uses ultralinear and you do not? Can you make me a transformer with a 43 percent tap or can you modify your project for more than 15 watts since I need more? Thanks.

After reading his email, the question was clear. How come my project with a single ended EL34 is 7 watts, while another similar schematic claimed 15 watts? Obviously those “15 watts” were completely made up, but it is worth clarifying why, so we can understand how such fanciful claims arise.

First of all, I have already talked about the use of ultralinear in single ended in another article, which you can read by clicking here. The widespread idea that ultralinear increases power is false. In single ended configuration, power decreases, it does not increase.

Second, I always state power per channel, so my 7 watts are 7 + 7, not 7 total. In the project that was reported to me, however, it said “15 watt stereo,” which suggests this is the sum of the two channels, a crafty but improper practice, done only to write a bigger number on paper and confuse the less experienced.

Out of curiosity, and to check how realistic that value was, I decided to simulate the circuit. I used LTSpice, a free analog simulation software developed by Analog Devices, and the tube models by Norman Koren, a reliable name in the field. In the test I simulated an output stage with KT88, using an ideal 3k transformer with a 43 percent tap. An ideal transformer means no losses, so already under the most favorable possible conditions. If even in this “dream” simulation the result does not come close to the stated 15 watts, in reality it is impossible to reach them.

The cathode current is 110 mA. Considering that the cathode is raised from ground by 40 volts and the anode voltage is 450 volts, the drop across the tube is 450 – 40 = 410 volts. Multiplying by the current we get 410 × 0.11 = 45.1 watts, so the tube is dissipating 45 watts, already exceeding the rating limit of a KT88, which is 40 watts total (35 for plate + 5 for screen grid).

It is obvious that in reality an output transformer is not ideal. It has a DC resistance that causes a small voltage drop. So the actual plate voltage will be lower than the theoretical 450 volts. If we assume a loss of about 10 volts on the primary, we can correct the simulation by setting the supply voltage to 440 volts, so it reflects a more realistic behavior of the circuit.

Now there are 400 volts across the tube with a current of 100 mA, so dissipation is exactly 40 watts, the maximum allowed limit for a KT88. The transformer in the simulation is still ideal, so without losses or ohmic resistance. Since the cathode is raised from ground by 40 volts, the maximum signal applied to the control grid before the tube enters clipping will be about 40 volts peak. At this point we can observe what voltage is obtained on the transformer secondary.

The signal measured on the secondary shows a positive peak of 12.29 V and a negative peak of 13.57 V, for a total of 25.86 V peak to peak. Applying the basic formulas:

• I = V / R = 25.86 / 8 = 3.23 A peak to peak
• W = V × I = 25.86 × 3.23 = 83.52 watts peak to peak
• WRMS = Wpp / 8 = 83.52 / 8 = 10.44 watts RMS

This is power calculated with an ideal transformer, therefore without any losses. In a real circuit, however, the transformer always has some leakage. On average around 30 percent, depending on build quality and core size. Taking this efficiency into account, the actual power drops to about 8 real RMS watts, perhaps 10 in the best cases. Even being optimistic, it will never be 15 watts, as stated by certain “creative” schematics.

Repeat a lie a hundred, a thousand, a million times and it will become truth. Cit: Joseph Goebbels

Now let us see what changes when using an EL34. In this case the plate voltage is 340 volts, with the cathode raised from ground by 21 volts. The plate current is 77.8 mA, for a total dissipation of 24.8 watts, practically at the limit of the maximum 25 watts allowed for the tube. The signal generator applied to the control grid provides an amplitude of 21 volts peak, a value sufficient to bring the tube close to the clipping point.

The signal on the secondary shows a positive peak of 9.69 V and a negative peak of 10.32 V, for a total of 20.01 V peak to peak. Skipping the calculations already shown before, the corresponding power on an ideal transformer is about 6.25 watts RMS. Subtracting the real transformer leakage, which inevitably reduces performance, the actual power drops further. So even under the most favorable conditions, you certainly do not reach 15 watts per channel, and not even 7 real watts with that circuit. In my project, however, the EL34 is used as a pentode, a configuration that offers higher power efficiency than triode or ultralinear connection. So let us see what happens in simulation with the tube connected as a pentode.

With the EL34 connected as a pentode, the signal measured on the secondary shows 11 V positive peak and 15.86 V negative peak, for a total of 25.86 V peak to peak. This corresponds to about 10.44 watts RMS on an ideal transformer, therefore without any loss. Considering a realistic efficiency, with an average leakage around 30 percent, the actual power becomes about 8 real RMS watts.

Personally, I prefer to state power honestly, measuring it just before one of the two half waves starts to flatten, that is, just below the clipping point. For this reason my project is stated as 7 watts RMS per channel, so 14 watts total in stereo, a real and repeatable power that still exceeds that of many schematics that declare impossible numbers, physically unreachable without heavy distortion. For those who want to replicate the simulations, I am providing the LTSpice library with Norman Koren’s models, available here: Koren_Tubes.zip.

Power and impedance: why tube amps do not behave like solid state

One client asked me a curious question:

Hi Stefano, I am ***. Last year I built the PP2010 by Ciuffoli with your transformers. I want to ask, if I connect 4ohm speakers to the 8ohm tap, can I reach 100 watts?

After a brief exchange it turned out that someone, one of the usual social “gurus,” had suggested this theory to him. I thought he was talking about solid state amplifiers, but he was referring to a tube amp.

The confusion arises because tube amps and transistor amps work in completely different ways. In solid state amplifiers, which are current amplifiers, the power delivered increases if you reduce the load impedance. For example, an amp that does 40 watts on 8ohm can reach 80 on 4ohm, because it can supply more current.

In tube amplifiers, things are completely different. Here we have power amplifiers that work through an output transformer, which matches the speaker impedance to the one seen by the tubes. The output tubes do not work only in current but also in voltage, and to obtain maximum efficiency and minimum distortion they must operate on a precise load line.

For this reason each transformer has multiple output taps (4, 6, 8, 16 ohm), and the speaker must always be connected to the tap matching its nominal impedance. Connecting a 4ohm speaker to the 8ohm tap does not increase power, in fact it causes an impedance mismatch, reduces performance, alters frequency response and increases distortion.

In short, with solid state lowering the load impedance increases power, with tube amps it worsens operation. The transformer already keeps the reflected impedance on the tubes constant, so power remains the same regardless of the output used, as long as you connect the right speaker to the right terminal.

Practical vademecum of realistic power

For those who want to navigate between real numbers and urban legends, here is a small non exhaustive vademecum with the approximate RMS power obtainable from the most common tubes, both in single ended (SE) and push pull (PP) configuration. The values shown should be considered realistic, referring to well designed circuits and operation within acceptable dissipation limits.

• 300B: In SE, if properly driven, it reaches 9 to 10 watts RMS. If it is less, the driver is weak. In PP, although uncommon, in class AB it can reach 30 watts.
• 2A3: In SE it delivers about 2.5 to 3 watts RMS, while in PP class AB you can reach 12 to 15 watts.
• EL34: In SE it provides 7 to 8 watts RMS, and in PP class AB up to 25 to 30 watts. Connected as a triode in class A it delivers about 5 watts.
• KT88: As a triode in class A it produces 6 to 6.5 watts RMS. In single ended as a pentode up to 12 watts. In PP ultralinear class AB 70 to 75 watts, while in PP as a pentode about 50 watts. This is a special case. In theory the pentode should give more power, but with KT88s the screen grids are delicate and beyond 50 watts they risk melting. In ultralinear configuration the problem does not occur, which is why from this tube you get more power in ultralinear than as a pentode.
• EL84 / 6V6: Very similar in performance. 3 watts in SE, up to 15 watts in PP class AB as a pentode. The EL84 connected as a triode in SE provides less than 1 watt, because its very high transconductance makes it inefficient in this configuration.
• 211: In SE, with drive only on negative grid, about 15 watts RMS. With drive also on positive grid, up to 25 watts.
• 845: In SE, with drive only on negative grid, about 25 watts RMS.

These values should be taken as realistic references, not as absolute limits. Every project has its nuances, but if you find someone stating figures very different from these, you can already sense that they are telling more fairy tales than electronics. If you have doubts or want to know the typical power of other tubes, write it in the comments. I will be happy to add them to the list.

The “atomic amplifiers”

To conclude this article it is worth mentioning a particular category of devices that I affectionately call “atomic amplifiers”. These are those projects, often dating back to the late seventies, born from the minds of slightly visionary designers who tried to squeeze monstrous power out of common tubes. On their datasheets they proudly stated absurd figures, such as 100 watts from a pair of KT88 in push pull, something that on paper may even seem possible, but in reality means driving the tubes beyond every physical and safety limit. Many of these devices, still circulating today, are real time bombs, with out of scale dissipations, crazy voltages and components always beyond the limit. Those who want to delve into this category of “brave” amplifiers can read the article dedicated to the atomic amplifiers that I published some time ago.

Final postscript: the fable of the brilliant plumber

Once upon a time there was a plumber with a passion for tubes. One day he decided that designing amplifiers had to be as easy as changing a washer, and so he got to work. After weeks of mental gymnastics and trials with the soldering iron, he triumphantly declared that his single ended KT88 made 100 watts. Yes, that is right, one hundred watts from a single KT88.

I, who obviously do not have the same imagination, was stunned. Because if he could get 100 watts from a single tube, then with the power of a fart I should be able to fly like Superman. But nothing, SuperMario Unhinged, as his friends called him, did not want to reveal the secret to me. He said that I was not capable, that he was, but he would not explain to me how to do it. Maybe one day I will write the full fable too, but it will need the label “not suitable for those lacking critical thinking”, because the story of SuperMario Unhinged and his 100 watts from a KT88 is so absurd that it risks lowering the IQ of those who read it.