Tubes vs Solid State - Imaging, Soundstaging, 3D

Al and Learsfool wanted to know more about the low level detail thing. Here it is.

This has to do with the rules of human hearing/perception. Apparently we have the ability to hear into a noise floor and perceive detail which is interesting as normally the human ear has a masking rule wherein louder sounds mask quieter sounds. Apparently hiss is some sort of exception.

I personally suspect that this is because wind sounds very much like hiss, and it was important to us as a species (for survival reasons) to be able to hear sounds that are not as loud as the wind.

Anyway, hiss seems to be the exception to the masking rule. However not all amplifiers have a noise floor composed of hiss, although this is something that does not show up well on our test instruments. This has to do with the fact that if you apply negative feedback in large quantities to an amplifier, the noise floor becomes that of harmonic and inharmonic low level distortions. The inharmonic distortions (InterModulations) are caused by non-linearlities at the feedback node. The harmonics generated by this practice can go up to the 81st harmonic! (see Norm Crowhurst, who wrote about this phenomena a good 50 years ago).

You might look at it another way- that by adding feedback to an amplifier, you don't rid yourself of the energy of the distortion- instead it gets spread out across the spectrum. Of course with many amplifiers those upper harmonics will fall outside the amplifier's passband, but the point here is that the noise floor is not that of hiss. It is that of distortion.

The ear can penetrate a noise floor of hiss but if the noise floor is composed of distortion in this manner, the ear will find that to be the lowest level of output from the amp, IOW the ear cannot penetrate that noise floor to retrieve information below it. Hence, amps that apply large amounts of feedback will seem to be less detailed.

A further complication comes in when we try to amplify a signal that is in a state of constant change- that is to say does not repeat itself, as in real music. The noise floor of the amplifier is not nearly so low under these circumstances, nor is the over distortion as low as it is with a steady-state signal.

Chaos Theory shows that an amplifier with negative feedback is a Chaotic system ('Chaos' being a defined term, not the same as the street meaning of the word). To that effect it may appear to be predictable under certain conditions (steady state signal) but may have other results in other conditions (music). The formula for negative feedback (Crowhurst) is identical to that that we see in classic Chaotic systems (N+1 et.al. if you care to read up on it).

What this teaches us BTW is that no application of feedback will solve the problem (Nelson Pass points this out in his excellent article about feedback and distortion on his website). IBM engineers learned this long ago, which led to the invention of the parity bit in digital communications. A different field and application, but the underlying principle is the same.

Do I need to elaborate more?

I'm late to this thread, but:

my perception has been that the prevailing viewpoint among experienced audiophiles is that a particular strong point of tube amplification tends to be imaging and dimensionality. And that has certainly been my experience, and the experience of several of the others who have responded.

We can only speculate as to why your experience has been the opposite.

The reason solid state can seem to have sharper images in the sound field is due to the fact that generally speaking, solid state amps tend to have less low level detail (the why of this is a topic for a different thread but in a nutshell has to do with how the human ear interacts with the noise floor of the amplifier; if anyone is interested I can go into that in greater detail, if you will pardon the expression).

How this affects imaging is that without as much low level detail, the images in the sound field will seem to be in sharper relief. However upon careful listening you will find that in comparison to most tube amps, the images have a 2D/poster quality, as the air (ambiance) around the individual performers is removed; this makes the images 2D but more distinct.

Add the low level detail and the effect is more like what you find live- that there is ambiance created by the sounds of each instrument that tends to take away the stark relief but also adds a 3D quality, which of course is what it is supposed to be...

So hearing hiss is better than not hearing harmonics that can't be heard? I must be missing something?

Correct- you are missing something. The hiss which is the noise floor of a zero feedback circuit is at a low level, just as the noise floor of an amplifier with feedback is at a low level. The point is that the noise floor of a circuit with natural hiss will seem to have more detail, as our ears can hear information below then noise floor; in a circuit with feedback they can't.

Regarding your other comments re class D and tubes... you could get a tube amp to drive your speakers- you might want to use a set of ZEROs. Worth a try if you ever consider it. IMO Class D is a technology that may well prove to be superior. The answer is told by price/performance curves and the question is where is class D on its curve? If somewhere in the middle it may yet surpass existing technologies.

Since Transistors and Tubes are both mature technologies and Transistors being arrived at a lower performance level than Tubes, the fact is that our grandchildren will be arguing over the differences. But what this also means is that class D will overtake Transistors before it can overtake tubes. In some cases we have already seen this, so the theory is being proven in practice.

Put another way, will class D then overtake tubes?? That remains unknown at this time. What we *do* know is that it has not done so yet.

I maintain that you don't have to know anything about technology to understand that tubes are in fact superior. The simple fact is that at this time, tubes have been obsolete for longer than they were the only game in town! Think about that. Normally when a succeeding art appears, it supplants the prior art easily and there is not much looking back- the prior art becomes a cottage industry at best; cars and horse-drawn buggies are a good example.

But tubes are still vibrant- the market somehow continues to demand them half a century on after their declared 'obsolescence' (there are more manufacturers of tube gear in the US now than there was in 1958...). That's really all you need to know- if the succeeding art were really better, tubes would be long gone. You don't see them in TVs anymore and they've not been there for a long time. That says that tubes don't work as well as transistors in TVs- or computers! But they seem to do just fine in audio, and new types (KT150 being one of the latest) are being introduced. That's not a description of a technology that is obsolete.

Assuming the noise floor is low and can't be heard anyhow, what is there to hear below it?

Al's comments above are the correct answer to this.

Chaos Theory points to the 'why' of it- the amplifier with feedback is operating in a chaotic fashion where the noise-distortion floor is much higher with a constantly changing non-repetitive signal.

Csontos makes a good point in his post above. IMO this has more to do with feedback and less to do with whether its tube or transistor in theory. In practice though it is much easier to build a zero feedback circuit using tubes than it is with transistors. Ninety-nine and 44/100% of all transistor amps run a considerable amount of feedback, hence the generalization.

For the time being designers have yet to universally recognize the importance of human hearing rules in audio design, so we are likely to see considerable differences in opinion for quite some time :)

In this conversation I've not discussed preamps although I have alluded to them by using the word 'circuit' as opposed to 'amplifier'. The same rules of human hearing of course apply to preamps, and you do deal with the same issues of feedback. However in preamps it is very easy to build a zero feedback circuit with vanishingly low distortion and wide bandwidth, so the argument in favor of feedback weakens considerably, especially in the face of the damage it can do (if you want to build the preamp with opamps though you are kind of stuck- that can only be done with feedback). If the preamp looses information, it really does not matter how good the amp or speakers are- you can't recover lost information downstream.

Few things are ever "universally" realized.

Does not make sense to me that such things would be ignored by the experts whose products can benefit.

Its all about the dollar. People often purchase audio gear with their eyes rather than their ears. Then once they are saddled with it they have to justify why so they can feel good about it. Its much easier to design with feedback than without. In the case of amplifiers, the use of feedback is common as it is part of a standard of how the speaker is driven. But the standard ignores human hearing/perceptual rules. I am sure many have seen this link before but here it is again:

http://www.atma-sphere.com/Resources/Paradigms_in_Amplifier_Design.php

In a nutshell the more your equipment obeys human hearing/perceptual rules the better it will sound, the involving it will be and the less likely that it will wind up in a closet or being resold. But if you are only interested in how good the equipment looks on paper you might think global NFB is a good thing.

if I hear something that sounds good but does not measure well ( to some reasonable reference standard that it probably should) a little question mark will likely go off in my head asking "why".

Now you know why, the technical specs don't jive with human hearing/perceptual rules, which also answers the question below:

In general, technical standards are a good thing. Not sure why audio should be any different?

I am sure its possible to develop test/measurement techniques that *do* correlate with human hearing rules, but right now the industry has had no will to do so and has not for the last 45 years so don't hold your breath.

I suppose the standards that are applied do correlate to some extent with human hearing rules. It would be hard to see how products could be sold if not.

What I am saying here in the last few posts is that for the most part, the measurement standards don't correlate all that well at all; about the only thing that gets much lip service is frequency response, based on the ear having response from about 20Hz-20KHz.

There is nothing in the measurements, for example, that acknowledges that the ear treats distortion as a tonality, other than 'low' distortion is supposed to be good. In practice though, it turns out that certain distortions must be very low, and others the ear does not seem to care about so much. There is no nuance in the specs!

I submit that there isn't[sic] any rules for human hearing. It's clear that very few audiophiles can agree on a regular basis when it comes to which piece of gear sounds the best, or even better.

You are confusing taste (for which there is no accounting) with actual perceptual rules, which vary over the entire population by about 1%. IOW there is a big difference here.

There are numerous examples, for example mp3s take advantage of the ear's masking rule (although not 100% successful in that regard) to reduce the amount of data storage.