Power output of tube amps compared to solid states

This particular paper is interesting, but very dated . . . and no disrespect to the author or his work in context of the time. But it's important to understand that this was published some fifteen years before that of Thiele and Small, and the T/S equations predict exactly every aspect of the driver/source-impedance interactions that he gives these approximate, experimental methods for optimizing.

It also bears mention that at this time, typical loudspeakers were designed with a very small back-enclosure and a horn (using Webster's equations), or a very large infinite-baffle . . . reflex designs were extremely rudimentary and typically more like an infinite-baffle as far as the system Q is concerned. Let's also not forget that liberal application of tone controls was also not frowned upon like it is today . . . a great way to compensate for all kinds of unanalyzed factors relating to speaker design.

Hi Unsound . . . it's not so much a matter of whether or not an amplifier is "DC coupled" so much as what its output impedance is -- Atmasphere manufactures "DC coupled" amplifiers with highish-to-very-high output impedances. I would summarize his position on this (sorry Atmasphere if I'm imprecise) as that speaker manufacturers have a responsibility to keep their impedance curves fairly smooth, and/or that the end user needs to be aware of what speakers produce good results with his amplifiers when making purchasing decision.

My point in the previous post is that with the mathematical tools available to the modern speaker designer, one need not use that very crude experimental methods outlined in the paper Atmasphere linked to . . . it's been possible for some time to accurately predict the loudspeaker "damping" behavior for any given amplifier output impedance. And of course different loudspeaker designers have different goals for this criteria.

BTW, there are very few amplifiers that could truly be called "DC coupled", and very few of these actually have a closed-loop frequency response that extends to DC (which IMO isn't necessarily a good idea anyway). Most conventional "DC coupled" amplifiers may indeed have the output stage DC-coupled to the positive side of the loudspeaker, but on the negative side, the speaker current is actually returned through the main filter capacitors . . . meaning that with the exception of any residual DC-offset current (or if the amplifier has a fault), they are effectively capacitively-coupled amplifiers, with the main filter caps inside the feedback loop.

Much as I like prosound woofers, most of them have too much electrical damping (too low Qes) to give good bass extension with a low output impedance (high damping factor) amplifier.

Excellent point - they also tend to have lower mechanical mass for sensitivity, and lower compliance for better power handling, resulting in a higher resonant frequency. The problem with increasing the amplifier source impedance for this type of driver (especially in a domestic application) is that while this raises the Q and the bass output, it leaves the resonant frequency unaffected . . . the result is then simply wooly, boomy mid-bass and no increase in bass extension.

An interesting exception is the JBL 2235/2245, which have mass-rings (weights) attached to the voice-coil former to both increase Q and to lower Fs. But this also makes them much more prone to over-excursion . . . so it's not uncommon for a reconer to remove the weights if there's a problem bottoming-out the voice-coil against the magnet, again raising Fs and lowering Qms in the process.

Pubul57, there's no shortage of material on the web that contains Atmasphere waxing lyrical about his approach to building amplifiers . . . I suggest you look at his website or peruse his posting history on Audiogon if you haven't caught it by now.

You're precisely correct Shadorne . . . and that's why in the pro world, a 12" driver is considered small for a woofer, and usually used for midrange/midbass. Even most pro 18" drivers have a pretty high Fs and low Q compared to a consumer 12" . . . and why most of those pro 18" bass drivers are in W-bins, sugar-scoops, tapped-horns, etc.

But also keep in mind that distortion is very SPL-dependent, and the onset comes on very quickly indeed! So for a smaller cabinet size and better bass extension in the typical domestic environment, having a very clear idea of the maximum intended SPL is crucial for the loudspeaker designer to achieve the requisite performance.

A bit off topic here, but, if a speaker manufacturer needs to veer from the criterion you use to describe a "decent speaker design" and needs use something other than a higher impedance amp to achieve better results, what's the harm?

Ah, Unsound . . . this is the crux of the matter. Atmasphere and I have actually beaten this horse pretty dead into the ground on other threads . . . again, I'll try to summarize:

On one hand, I feel that one of the most important hallmarks of good engineering is to carefully consider the application, and design the equipment to work at its best within it. So a loudspeaker designer should consider the types of amplifiers that are likely to be driving the speaker, and design so as to acheive the most consistent and best results over the greatest possible number of situations. An amplifier designer should do likewise in consideration of the types of loudspeakers that are likely to be connected to it. If this approach is followed, then if a person buys a "great-sounding amp", and a "great-sounding pair of speakers" . . . then there's the highest likelihood of getting a "great-sounding system".

I also personally feel that if an engineer's preferences for certain circuit topologies get in the way of these goals . . . then the engineer should maybe reconsider their preferences. Expecting a consumer to anticipate the effects of non-standardized technical equipment interface criteria is unreasonable, or at least very unlikely.

On the other hand, Atmasphere feels that benefits of his particular circuit design approach are so great as to tolerate the fact that there will be some inconsistencies in performance depending on the particular application. To me, this is like making wines that can be wonderful and complex, but are inconsistent between vintages . . . the requirement is then on the purchaser to have some sommelier skills to compensate for it.

And it's not that I necessarily have a great many circuit preferences that are contrary to Atmasphere's . . . I love the elegance of good vacuum-tube circuits, I think that there are good applications for well as open-loop topologies, etc. etc. It's just that I would never build and sell an amplifier with such a high output impedance, knowing that inevitabily a certain percentage of the time somebody would connect it to a loudspeaker that's largely incompatible . . . and I wouldn't be happy with the way it would perform under those conditions.

If I may steer this thread a bit back to the original question . . . there's are two characteristics typical of tube amps that I think make the perception common that "tube watts" are more powerful than "solid-state watts":

First, as others have alluded to . . . tube amps (as a group) have less offensive overload behavior, so it's possible for a slight bit of clipping to be much less noticeable than for a typical solid-state amp.

But there's also the fact that the overwhelming majority of both solid-state and tube amplifiers use unregulated power supplies, meaning that the voltage available to run the circuitry goes down the harder the amplifier is driven. And exactly how much it goes down is a function of the power supply's total capacity and its ability to store energy . . . in relation to the energy peaks required by the music being played.

And the particular set of energy-storage dynamics between a typical tube-amp power-supply and that of a typical solid-state amp are VERY different. ("Typical" here means push-pull outputs operating in Class AB or B, SS direct-coupled, and tubes transformer-coupled with C-L-C filtering.) The tube amp generally has significantly longer time-constants in its filtering in relation to the currents required by the output stage . . . meaning that whatever "dynamic-headroom" power is available (short-term peak power above the maximum available steady-state power) can be delivered over a longer period of time.

There are several mechanisms at work here. First, a push-pull tube amp reflects its impedance back to the power-supply as two full-cycles for every output cycle. Second, the output transformer primary inductance acts as effective energy storage for signal waveform asymetry. And third, the impedance transformation of the output transformer works backwards as well, drastically reducing the peak current demand on the power supply.

So a hypothetical 40-watt push-pull vintage tube amp may have 40uF of capacitance on the main plate supply and 5K transformer primary, and let's say we're using a 4-ohm output tap. 40uF seems chincy by today's standards, but this is equivalent to more like 100,000uF for a direct-coupled solid-state amp . . . and for comparison, a good quality 150w/4-ohm solid-state amp usually has something like 25,000uF. And while the SS amp does have two caps, since the output current is half-wave rectified (rather than frequency-doubled as in the P-P tube amp), their effective capacitances don't add together. And then the tube amp usually has another 40uF of capacitance and a choke in front of it, which probably gives about 2-1/2 times again the total energy storage for the power-supply.

So it's entirely possible that this hypothetical 40-watt tube amp may have similar (amount in dB) of dynamic headroom to the hypothetical 150-watt SS amp, but is able to maintin its dynamic power rating for ten times as long . . . let's say 50 milliseconds instead of 5 milliseconds. With a typical music waveform, this is a dramatic difference.

Trelja, thanks for the kind words.

Unsound, are you referring to the Threshold S/500? If so, there must be some inconsistencies around what is meant by "double its rated power for several minutes" . . . because I remember this amplifier having rather conventional headroom characteristics (but it's been at least 15 years since I've worked on one, so I may be wrong). To clarify, what I'm terming "dynamic headroom" will generally manifest itself as the ability to generate additional unclipped short-term voltage and current beyond the steady-state clipping power.

If "short term" means several minutes . . . then the limitation that keeps short-term capacity from being long-term capacity is almost surely thermal dissapation, not energy storage or output device current-limiting. The only SS amps that I can think of that exhibit this characteristic are what I would call "asymmetrical class H" operation, like the old NAD "power envelope" design. This works by having the amplifier operate from lower-voltage rails most of the time, but for large-signal peaks there is another set of commutating transistors (like conventional class G or H) that pulls the voltage up to a higher rail, and then keeps it there for a fairly long time-constant. The reason why it doesn't operate at the higher rail all the time is purely thermal . . . so I would actually consider it this design a higher-powered amp with thermal limitations, rather than a lower-powered amp with lots of headroom. But that's purely terminological I guess . . .

Do tubes have the same "advantage" in preamplifiers?

No. Tube preamp stages almost universally operate in Class A, so they have a steady-steady current draw regardless of the output level. Consumer preamps also usually operate into high-impedance loads (little current required), and virtually never have output transformers (to perform the power-supply impedance transformation).

In a headroom sense, tube preamps do operate from much higher-voltage rails . . . and even though they tend to have more linearity problems as they approach their voltage limits, there's still usually a certain headroom advantage. But this is likely to only be an issue for something like a professional mic preamp, seeing the raw feed from a large-diaphragm dynamic mic that's three inches from a kick-drum head.

Atmasphere wrote:

I have an additional comment about distortion that Joe mentioned- that of, shall we say, 'dynamic distortion'. Its my opinion that we need some sort of distortion test that uses a non-repeating waveform similar to what you see in real music.

I take it you're familiar with Matti Otala's 1972 AES paper that started the whole TIM measurement thing, and the huge number of papers and followup over several years in response? Do you feel that there are aspects of this issue that need further research?

BTW Norman Crowhurst pointed this out 50 years ago in his writings about negative feedback.

I take it you're referring to Crowhurst's two AES papers from 1957 and 1969. There are several historical things to keep in mind with how amplifiers were rated in those days, and most of what Crowhurst seems to have been concerned with in the 1957 paper is practical problems with the use of negative feedback as a method for reducing costs. He also makes limited to "regenerative distortion" (feedback making distortion of the distortion), but much of the theory here is very vague.

If we look at the perception of simply how powerful the amplifier is . . . I think it breaks down over tube/solid-state lines much more than global-NFB/no-global-NFB lines, and I personally don't feel that negative feedback has a whole lot to do with it.

Now for subjective sound quality and measured performance, feedback is a huge topic, and I'd love to get some discussion on these papers and their theory!! But let's start a new thread for it . . .