Power output of tube amps compared to solid states

The why rather than the how:

1) instantaneous power as it relates to distortion. In a very short increment of time, tubes can put out a lot more power than transistors or FETs for a given distortion level. This has mainly to do with electron mobility being a zillion times higher in a vacuum than in doped silicon, not to mention asymmetrical drift velocity issues in silicon. But I'll digress here. This is why 'soft clipping' is possible.

2) overload recovery of tubes is much better than solid-state. This is a follow-on to #1. After a transient, there is no reactive impedance to charge recovery in a vacuum like there is in semiconductors. This prevents frequency domain disruptions in tubes (the messing up of harmonics). This is the main point Lavardin has tried to resolve with their own solid-state devices.

There are other more-minor factors mired in the nitty gritty of device physics but these two are the most important, as I see it.

When it comes to steady-state signals, the power levels ARE the same if the Watts are the same. But music is most definitely NOT a steady-state signal. Nearly all measurements and graphs you will see on the performance of an amplifier are based on steady-state. These cannot be used to accurately explain transient phenomena. They can give an overall indication of performance ability, but nothing more. This is why many people believe measurements do not describe how a component actually sounds.

Getting back to the post, efficiency is a separate issue from power output. For Forests, a 50 or 60W tube amp will be great. We had a 55W Cayin on a pair for a long time and it was wonderful. Their impedance is high so they do not demand lots of current, making tube amps the way to go with Forests IMO.

Arthur

Hey Joe - thanks! I knew only a few would care but I wasn't expecting a full realization. I could have gone on for 10 pages. Your anecdotes and thoughts are right on! It really is the difference between algebra and calculus.

Arthur

Oh Marakanetz. I'm not sure I can help. Believe it or not, I have several oscilloscopes. I also have impedance analyzers, spectrum analyzers and electronic loads. I work in a state-of-the-art electronics laboratory. The reason I said what I did is because I realize these measurement devices are so we can observe simplifications of the real thing. I know it all too well.

To think that a human can build something that would have the capability of human hearing is an ego trip that no one should succumb to. In my opinion. The day measurements can beat the human brain, our demise, and our goal, will have been reached.

1kW of power from a tube as you imagine it, has no physical foundations. Heat generation alone would preclude that possibility. I was talking about what are called "time scales." We have many scales, as much in engineering as in philosophy, ultimately, but all of them fall short of truly representing what we can witness because time effects extend ad infinitum in both directions and thus there is no end to splitting them up into comprehensible slices. Algebraic formulas and measurements occupy only a small part, but calculus covers them all. This was the point Joe so succinctly and eloquently outlined above. One of the tools available to see the remaining parts is called "frequency domain analysis;" the repercussions of which are called "harmonics," which also extend ad infinitum. There is no end to the formulas in the frequency domain. The day I realized all this was mentally liberating.

Classical formulas exist so that we can comprehend what is going on. This is done by throwing out all variables that we deem "unnescessary" because we would otherwise be overwhelmed. The Taylor Series Expansion in calculus points it out very clearly. But are those variables we throw out truly unnecessary? Would they exist if they weren't? Of course not. That is the crux between art and science.

You are free to believe as you like. If your definition of power has no total derivative, then you win. Otherwise, consider power as subject to time scales and you will know what I am talking about. If you would like to understand the mechanisms I speak of in lots more detail, I very-highly recommend Papoulis' book, aptly called "Signal Analysis."

Arthur

Great post there Joe. About your comment that Class A operation reveals lower-power figures that expected: You are absolutely correct that theoretically Class A should be able to yield higher power because there is no zero crossing. There are three main aspects, as I see it, as to why in reality it doesn't work out that way - and also how they impact tone because your comments about differences in tone between single-ended and push-pull are actually related.

1) Accurate low frequency reproduction is more challenging in tube circuits that slide into Class B because the power continuity gets broken when no tube is operating. This particularly impacts the low frequencies because they are the ones that need the most power. And they are the ones that mainly have to do with tone.

If I may use an automotive analogy, it is like comparing a manual transmission with a dual-clutch transmission, such as the ones in the latest Audis and BMWs. There is no interruption of power in the dual-clutch setups since there is always a set of plates that are transmitting full power. This is like Class A. But in the other case, you have to put the clutch in and the RPMs drop like a rock as you disconnect the engine (amp) form the transmission (speakers). This is like Class B. The end result is that tone (mainly bass) is much more easily conveyed in Class A operation since the power is always on tap. There is no denying in the world of automobiles that constant power leads to higher performance. Guess which mode of transmission is used in race cars.

2) The actual power ratings are lower than Class AB because of physical power dissipation limits. To operate with a constant bias requires high heat dissipation from the tube's plate. This would be fine except that the cooling medium inside a tube is vacuum, which is as bad as it gets! A vacuum is a terrific thermal insulator since there can be no convection cooling and thus all the cooling is due to "black body raditiation." I have better stop there before I get into the Second Law of Thermodynamics! lol. So yeah, to keep the tube from overheating you have to cut back on the output power. Basically you trade power for bias, which has of course its known merits in addition to this inconvenience.

3) The output transformers in a Class B amp are not designed the same as in a Class A amp. If the amp is Class AB, I am sorry to break the news to some of you but that means it is a Class B amp as far as circuit design goes. Many people feel it is better to say AB rather than B but as far as the electrical design is concerned, there is Class A and then there is Class B.

In a Class A amp, the output transformers have to have an air gap. This means that the core is not made of one continuous piece like it is in a Class B amp. This is due to the fact in Class A you have constant positive DC current in the primaries of the output transformers that should be equal to half the peak output current. In order to prevent the core from "overloading," technically called "saturation," you have to literally cut a slit into the core somewhere so that the excess magnetic field (that goes above and beyond satisfying the core's inherent magentic self-inductance which is what allows it to work in the first place) gets trapped in the air gap. Since air doesn't magnetically saturate, it is a stabilization method for the core.

In addition, this means that a transformer for a Class A amp must be a lot larger than a Class B one because you can't maximize the use of the core since the polarity is always positive, and, the air gap adds its own detriment to the performance of the core.

In a Class B (aka Push-Pull), you don't have to worry about these problems because of what I pointed out in 1) above: the net DC current in the output transformer is essentially zero because the tubes each turn off at the end of their respective cycles. They aren't "on" all the time like they are in Class A.

I feel certain that all these differences account for the changes in sound and tone between a Class A amp and a Class B amp, in addition to the differences in circuit topology of course. If you can live with reduced output power, higher temperatures, higher cost, and higher weight, then Class A has definite advantages. :)

I have two push-pull amps now and I am dying to get a Class A SET because I finally have speakers that can live with their downsides. I’m still in the process of figuring out which one to get. As far as differences between different push-pull amps, that lays in the gray area of performance overlap between designs. Which is better will depend on the type of tube, the power supply design, circuit design, personal preferences, speakers, and room.

As I said before, the sonic results of all these technical details can only be adequately assessed if there is an experienced human in the feedback loop. Only then are all the requisite variables fully taken into account.

Arthur

Oh yes, sorry for the confusion, Paul is right. I had SE Class A in mind. That's the type of amp I've been on the prowl for so that's what I've been thinking about lately.

My comments still apply for Class A push-pull however because they have the same ungapped core design as Class B push-pull since the two halves of each channel remain in opposite polarity and similarly sum in the secondaries. Hence my generalization from the circuit design point-of-view.

Arthur

I found very interesting work that was done by a group of Russians a while back. I am pretty sure I printed it out but I originally found it online. I'll have to see if I can find it again to post a link to it here, but at any rate, they had actually done measurement studies of how different amplifiers amplify Gaussian white noise (much closer to music than a sine wave is). I believe they may have also used music too.

They did Fourier analysis of the outputs of each amplifier and compared them. They found very real differences in the low-frequency reproduction, not much difference in the midrange, and once again some differences in the treble. Then they repeated the tests using the "best" amp of the group with various cables hooked up to its output. I think they used like 10 meter long cables to magnify the effects. Once again, the responses were different with different cables.

I'll see if i can find it again. Last time I looked for it I wasn't able to. But I found it extremely interesting and have not seen anything like it since. Maybe I'll just take some amps to the lab and redo the experiments myself! That's what I should do if I can find the time.

*PS* I just did a search to see if I could find it and ran across this paper. I haven't read it yet but it looks interesting. Too bad the input signals are still sine waves though....

www.apiguide.net/04actu/04musik/AES-cableInteractions.pdf

Arthur